Vision 2020 Collection

Quality in radiation oncology
Todd Pawlickia and Arno J. Mundt

Department of Radiation Oncology, University of California, San Diego, La Jolla, California 92093
Received 19 January 2007; revised 7 March 2007; accepted for publication 15 March 2007;
published 17 April 2007
A modern approach to quality was developed in the United States at Bell Telephone Laboratories
during the first part of the 20th century. Over the years, those quality techniques have been adopted
and extended by almost every industry. Medicine in general and radiation oncology in particular
have been slow to adopt modern quality techniques. This work contains a brief description of the
history of research on quality that led to the development of organization-wide quality programs
such as Six Sigma. The aim is to discuss the current approach to quality in radiation oncology as
well as where quality should be in the future. A strategy is suggested with the goal to provide a
threshold improvement in quality over the next 10 years. 2007 American Association of Physicists in Medicine. DOI: 10.1118/1.2727748
Key words: quality, variation, process improvement, quality management
I. BACKGROUND AND INTRODUCTION
The industrial revolution resulted in notable worldwide contributions including mass production. In the early part of the
20th century, the emphasis was to bring as much product to
market as possible in the shortest period of time, often with
little attention to quality. The primary method to ensure quality was by inspection. Each component that created the final
product was inspected to verify it was within the specifications set by product engineers. If a component was outside
specifications, then it was either scrapped or sent back for
rework and subsequent reinspection.
During this period, American Telephone & Telegraph
AT&T was hard at work building facilities to support a
universal telephone service. Western Electric Company created by AT&T as a constant source of supply was struggling
with problems of mass production, including the interchange
of parts, precision, and reliability.1 Scientists and engineers
at Bell Telephone Laboratories Bell Labs, a subsidiary of
AT&T, were engaged in product research as well as research
on problems of quality associated with mass production. A
major emphasis was on the application of probability and
statistics to quality.
By the mid 1940s, significant progress was being made on
new approaches to quality. Researchers at Bell Labs realized
that product inspection was not adequate to ensure a highquality product.1 It was much more important to control the
variation in the process that created the product. A significant
invention was the control chart where data acquired by inspection was plotted sequentially to characterize process
variation.2 Based in part on this work, W. Edwards Deming,
Joseph M. Juran, Kaoru Ishikawa, and others were creating
new approaches to quality that extended beyond the manufactured product.3 Unfortunately for AT&T, much of this research did not translate into their production efforts.
After entering World War II in December 1941, the U.S.
enacted legislation to gear the civilian economy to enhance
military production. The armed forces also helped suppliers
by sponsoring training courses on ways to improve quality.
1529
Med. Phys. 34 5, May 2007
In early 1942, Deming was invited to the Stanford University

Statistics Department to lecture on potential ways they could
contribute to the war effort. Deming gave a series of lectures
on using probability and statistics to provide a basis for action in creating high-quality manufactured products.4 Attendees at these lectures and at others that followed included
engineers from U.S. industries contributing to the war effort.
Most quality programs based on probability and statistics
were terminated once the government contracts ended. Deming and others realized that quality can be no better than the
intent and commitment of senior management.
After the war, Deming went to Japan in 1950 to lecture on
quality to the Union of Japanese Scientists and Engineers
JUSE.5 Juran was to follow in 1954 where he raised the
issue of quality from the factory to the total organization.
Japanese industry readily adopted these new quality practices
and Japanese researchers played an integral role in the development of quality. For example, Ishikawa of Tokyo University emphasized seven quality tools now used in most
quality programs3 Table I. In 1976, the JUSE developed
and promoted a list of seven quality tools for management
and planning3 Table II.
By the late 1970s, several U.S. industrial sectors including automobiles and electronics had been hit hard by Japans
high-quality competition. Japanese products were flowing
into the U.S. that were less expensive and of much better
quality. This situation was highlighted to the U.S. public in
the 1980 NBC television documentary, If Japan canwhy
cant we?6 In many ways, the start of the quality movement
in U.S. industry came as a direct response to the quality
revolution in Japan. Senior managers of U.S. companies
were now eagerly learning quality techniques introduced to
the Japanese, many first developed in the U.S. 40 years earlier.
During the past half-century, progress in health care has
been made by medical science and technology breakthroughs. While this has led to improvements in radiation
oncology, a new focus on quality will continue to provide
0094-2405/2007/345/1529/6/$23.00
2007 Am. Assoc. Phys. Med.
1529
1530
T. Pawlicki and A. J. Mundt: Quality in radiation oncology
1530
TABLE I. Seven basic tools of quality. Optimal use of these tools is within a comprehensive quality program.
Tool
Description
a
Cause-and-effect diagram
Check sheet
Control chartb
Histogram
Pareto chart
Scatter diagram
Stratificationc
Identify many possible causes of a problem

Structured form to collect data on frequency of events
Quantify and predict variation in output of a process
Document or communicate the distribution of data
Data analysis to show which situations are more
significant
Plotting method to determine if two variables are related
Data separation so that patterns or conditions can be
identified
Also known as the fishbone diagram or Ishikawa diagram.

Also known as the process behavior chart.
c
Sometimes replaced with the flowchart or run chart.
b
opportunities to improve patient care. In the next section, we

present the current approach to quality in radiation oncology,
which is followed by where we believe quality practices
should be over the next decade with suggested pathways to
achieve these goals.
II. WHERE ARE WE NOW?
Quality in healthcare has two dimensions: 1 high-quality
decision making and 2 high-quality performance.7 One aspect of high-quality decision making is consistency of practice. Professional judgment and peer review are current
methods to achieve consistency.8 Consistent decision making
is also achieved by studying patterns of care.9 and by clinical
trials.10 However, results of clinical trials can be confounded
by inconsistent interpretation and implementation of the trial
criteria.11 A related problem with results published in the
literature is that the reported efficacy may not be generally
applicable if different levels of quality are used at different
institutions. With regard to high-quality performance, reported error rates for significant events in radiation oncology
are low.12,13 The actual error rates are probably higher with
even more nonsignificant events that are never reported.14
There is theoretical evidence that even those nonsignificant
events can play a role in the degree that radiotherapy is a
benefit to patients.15
The focus of research on quality in radiation oncology is

on breakthrough innovation. The emphasis is largely on new
hardware or software products or on methods to utilize such
new equipment. Over the last several years, a limited number
of research papers have been devoted to adapting modern
quality
techniques
to
the
radiation
oncology
environment.1620 The consensus seems to be that automation, new machinery, and new computer systems will ensure
optimal quality. It has been shown, however, that record and
verify systems do not remove all possible sources of error.13
In fact, entirely new failure modes are possible using such
systems.21 Effective use of record and verify systems for
improving quality requires a change in staff functions within
the new electronic environment. As a general rule, continued
efforts on new equipment and software alone will not improve quality beyond its current level.22
The current approach to radiation oncology quality is to
investigate incidents once they have occurred rather than investigating processes for potential problems. Within some
departments or staff, the primary aim is making it through
the day, with no thought about the system or the future. This
approach to quality is summed up in the old adage, If it
aint broke, dont fix it. Such a philosophy can lead to latent
errors in a process that can be manifested long into the
future.23 Hard work and best efforts in radiation oncology are
TABLE II. Seven quality tools for management and planning. These tools are used for abstract analysis as well
as detailed planning.
Tool
Description
Affinity diagram
Organizes a large number of ideas into

natural relationships
Understand cause and effect relationships
Move the thought process from generalities to specifics
Understand how groups of items relate to one another
Compares options to criteria in order to choose
the best option
Schedule and monitor tasks within a complex project
Systematically identify what might go wrong in a plan
under development
Relations diagram
Tree diagram
Matrix diagram
Matrix data analysisa
Arrow diagram
Process decision
program chart
a
Sometimes replaced with the prioritization matrix.
Medical Physics, Vol. 34, No. 5, May 2007
1531
the main mode of operation to improve quality. An example

of this is the requirement to perform a measurement for every IMRT case prior to treatment. Whether this actually improves quality is still a point of debate.24 However, independent anthropomorphic-based IMRT plan verification can
bring user errors to light and thus lead to improvements in
quality.25 When quality is in doubt, the response is to give
best efforts and check more parameters. This approach is
also applied to people and processes when a mistake occurs.
But, simply paying closer attention to people can create what
is known as the Hawthorne effect.2 That is, by just paying
closer attention, people do a better job. Nevertheless, it is
impossible to check everything. There are limited resources
and limited time to do all the work. Current appropriate staffing levels for physicists can be found in the Abt Report,26 but
may need to be modified with new requirements and approaches to quality.
III. WHERE SHOULD WE BE IN 10 YEARS?
The International Organization for Standardization ISO
created a set of international standards for quality known as
ISO 9000. Originally issued in 1987, a major revision was
presented in 2000.27 Elements of ISO 9000 include modern
quality techniques such as the responsibility of senior management and requirements of continuous improvement,
which can be implemented using a comprehensive quality
program.28 The ISO standards have already been suggested
for use in radiation oncology.16 The Joint Commission
www.jointcommission.org is partly responsible for encouraging radiation oncology departments into action. The quality mandate from the Joint Commission is specific to describing a comprehensive quality program rather than a detailed
prescriptive emphasis focused on capability.29 A first step
should be to take action on those recommendations. Modern
quality tools and organization-wide quality programs are already used in other hospital environments.30,31
Quality should depend more on the assessment of process
data and less on the assessment of a patients health status
post treatment outcomes data. Patient outcome is a misleading quality metric because differences in patients characteristics may lead to different results, even for the same
delivery of care. A system view of department processes
should be adopted with a focus on reducing variation through
the people who use them. Within a radiation oncology department, each job or group of workers is not simply additive; their efforts are interdependent. One process of the system, in achieving some numeric goals, may inhibit the
function of another process. Some processes may operate at
a loss to optimize the system as a whole. An example of this
is in treatment planning where a dosimetrist may have down
time waiting for the physician to draw contours on a computed tomography scan. Working to understand variation allows one to first control a process and then to achieve real
process improvement. High-quality means minimizing process variation and moving the average closer to the optimum
value. But what is the optimum value and what are acceptable limits of variation? These questions are answered by a
1531
consistent and up-to-date set of specifications for our procedures and equipment. Task Group reports that contain specifications e.g., TG-40,32 TG-53,33 etc. can be several years
behind technology implementation and not easily updated.
Furthermore, independent checks of subsystems are becoming more difficult as newer technologies, such as Tomotherapy HiArt and Accuracy CyberKnife, are self-contained
planning and delivery systems. Specifications are needed for
these new systems as well.
Error reporting mechanisms e.g., the ROSIS database,
www.clin.radfys.lu.se need continual development to allow
optimal sharing of information, which will help to understand clinical processes, equipment, and their failure modes.
Quality and error reduction should go hand-in-hand. When a
process is predictable, one can consider an error as being
built into the process, i.e., a given number of errors are guaranteed to occur. If high-quality processes with minimal
variation are developed and implemented, then fewer errors
should result. Variation, for example in a repeated set of
measurements, is an easy concept to appreciate. What may
not be so obvious is that clinical processes also contain
variation. Without an appreciation of variation, it is difficult
to predict the future of process operation or understand past
performance. It is also easy to blame others for errors over
which they have little control.
There are, however, situations where even a single error
may result in catastrophic loss, for example a miscalibration
of a new linear accelerator during acceptance or delivery
errors in a single-fraction radiosurgery treatment. A different
set of modern quality and error reduction tools are required
in these cases.34 Tools that can be useful are repeatability and
reproducibility studies, checklists, mistake-proofing, and the
usual independent verification checks.
High-quality patient service is a necessity. A goal should
be to understand and also believe what patients want and use
that as a compass to define optimal quality. Quality and error
reduction programs should consist of process design, process
implementation with statistical evaluation, and design improvement based on the data the process provides. Vague
statements on quality should no longer be acceptable such as
I think we are doing OK or things seem to be getting
better. The new requirement should be the evaluation and
documentation of process behavior based on the data.
IV. HOW DO WE GET THERE?
A modern approach to quality as found in most industries
must be implemented in radiation oncology. Modern quality
techniques developed in industry have already been suggested as important to healthcare.35,36 The concept of measurement and an appreciation of a system are essential to
achieve optimal quality. Measures of improvement should be
created together with a focus on specific projects that are
critical to a departments success. Quality should be restricted to problems of loss caused by variability of function
and related harmful side effects, or it will slip out of the
domain of medical physics into the psychological domain of
cultural or personal values.37 Facets of a modern approach to
1532
quality include senior management leadership, employee involvement and empowerment, patient defined quality patient satisfaction, a view of work as a system consisting of
different processes, and continuous improvement. Such models should help with gathering information and taking action
efficiently and effectively. In-depth education on quality is
needed along with skills transfer that will give a team and
individuals the power to lead sustainable change. Relevant
projects must be incorporated with broad training that is tied
to an overall vision. Picking out a single quality tool or technique will lead to unimpressive results and ultimately not
work. We believe the following six objectives are necessary
to achieve the goal of optimal quality.
A. Encourage research on quality
All technology research in radiation oncology is geared

toward improving the efficacy of radiation therapy. The vast
majority of that research, which is specific to quality, is focused on new equipment and new procedures using the
equipment, i.e., breakthrough innovation. For example, in the
November 2006 issue of Medical Physics, there are seven
papers devoted to this type of research.3844 There are no
papers in that issue devoted to modern industrial approaches
to quality. Research sessions at the annual American Association of Physicists in Medicine AAPM and American Society for Therapeutic Radiology and Oncology ASTRO
meetings that are specific to techniques of quality and error
reduction need to be instituted. This will provide an academic incentive for physicists and physicians to focus on
implementing and demonstrating the benefit of modern industrial quality techniques. Discussions at the annual meetings on quality and error reduction are inappropriately relegated to the professional nonacademic tracks at the annual
meetings. Research on modern approaches to quality may be
undervalued because it can be seen as not scientific or only
incrementally innovative and lacking hard results. Incremental innovation, however, can be just as important to our field
as breakthrough innovation.45 This type of research on quality needs to be done with the same academic requirements as
any scientific study.46 It will take a focused effort to sift
through the many quality techniques for applicability to radiation oncology. This includes metrics for process behavior
charts, appropriate staffing levels, financial implications, and
organization-wide approaches such as Total Quality Management, Quality Function Deployment, Six Sigma, Lean, etc.
Lean thinking, for example, is a system of methods that is
geared toward identifying and eliminating all non-valueadding activities. How Lean a process can be made without
making it prone to errors is an area that requires research.
Research is also necessary to determine which metrics or
criteria are reliable to characterize a process and minimize
variation. Interesting quality techniques such as situational
awareness are themselves being investigated for optimal application to medicine and this type of work needs to be
encouraged.47 Novel approaches such as Web-based audit
systems may improve high-quality decision making in the
future.48 Process control techniques may even play a role in
1532
the physician care of patients.49 All of these aspects argue for

dedicated research sessions on quality and error reduction at
the AAPM and ASTRO annual meetings.
B. Educate radiation oncology leadership
Cultural change and strong management are necessary to

achieve optimal quality.50 Educating leadership does not refer to continuing education but rather an initial education on
what optimal quality means and what methods and resources
are needed to achieve it. Success of new quality programs
will hinge on the ability to obtain endorsement of and commitment to the process from departmental senior managers
and hospital leadership. A delineated vision and goals for the
department and commitment of resources to support the process is necessary along with department-wide communication to spread the initiative and tools. Quality committees
should report directly to departmental leadership. Early physician and physicist buy-in to any program will be essential
for success. Department chairs, physics directors, and administrators need to be informed on modern quality and error
reduction techniques. Successful continuation of new quality
programs will then depend on staff education. This should
not be left to the annual meetings but rather be Web based
and easily accessible to anyone. Medical and physics residency programs should teach modern industrial quality techniques and these should be tested for at the ABR board exams. The most efficient way education can be achieved is
through AAPM and ASTRO initiatives.
C. New approach for AAPM QA task groups
Specification of operating limits for equipment and procedures is an essential part of developing a quality program. A
consistent, up-to-date specification document for all aspects
of radiation oncology is required. At the rate of new technology development and implementation, specifications will be
changing rapidly. The first step is to collate existing specifications e.g., Table II of TG-40 into one place. These documents should be condensed to the minimum necessary specifications rather than an exhaustive list of all possible things
to check. The specifications should be contained in a single
document or Web pages for all aspects of radiation oncology. This document should be live and readily available to
departments and vendors. A system of regular updates can be
accomplished by a joint organization with appointments
from the AAPM and ASTRO that has the authority to make
necessary changes. The organization or committee will discuss specifications for new equipment and procedures or revisit antiquated specifications. For brand new equipment, the
committee can recommend temporary specifications based
on the experience and input from early adopters. A start in
this direction is the AAPM Therapy Committees assignment
of a Quality Assurance and Outcome Improvement Subcommittee that is working to fast track recommendations for new
devices in the form of short Task Group reports. A subsequent initiative, after considerable research has been done, is
for the AAPM Task Groups to make recommendations on
organization- or department-wide quality programs and the
1533
use of modern quality tools. This step has begun with the
formation of TG-100, which has taken the approach to develop a framework for designing quality management activities. Specifically, TG-100 will investigate and present the
technique of Failure Modes and Effects Analysis with specific examples to IMRT and HDR brachytherapy. Lastly,
clear guidelines are needed for what constitutes quality, what
constitutes an error, and how to report errors. Other fields
have also been plagued with confusing nomenclature.51
D. Close collaboration with vendors
A closer and more formal collaboration with vendors

should be fostered to provide equipment and procedures to
customers. A first step is to facilitate data transfer between
competitive vendors systems, which has begun to be addressed by the Integrating the Healthcare Enterprise IHE
Radiation Oncology RO Task Force. Vendor and user collaboration should also be more than mere financial support.
The vendors expertise of their product should be combined
with the users clinical knowledge to create optimal products
and processes. New equipment should be delivered with process behavior limits rather than specification limits. It is the
physicists responsibility to set specification limits and the
vendors responsibility to produce a product that meets those
specifications. Long-term relationships should be developed
with vendors to reduce variation in product output and create
a steady stream of incremental innovations. Vendors should
be encouraged to work closely with clinicians and reward
their employees for participating in peer-reviewed publications. Collaborative research, even vendor initiated research,
should be encouraged and receive the same level of academic scrutiny and standards as any other type of research
work.52 This type of collaborative research should not be
seen as inferior to pure academic pursuits. After all, discoveries or innovations that never make it to market are of little
benefit to patients.
E. Utilize resources outside radiation oncology
Collaboration with other professional societies that have

expertise in quality and error reduction is needed. For example, the American Society for Quality www.asq.org, the
National Quality Forum www.qualityforum.org, the Institute for Healthcare Improvement www.ihi.org, and the National Patient Safety Foundation www.npsf.org provide
valuable information and contacts on how to improve quality. These societies may provide the opportunity to identify
researchers interested in improving quality in healthcare.
Similarly, contact with colleagues in schools of management
or departments of industrial and systems engineering may
provide needed expertise from fields outside medical physics
and medicine. Collaborating with these researchers will provide new opportunities to learn about current quality opportunities that may be applicable to radiation oncology. There
is also much to learn from other journals such as Quality
Management Journal, Quality Engineering, Quality and
Safety in Healthcare, and Quality Progress just to name a
few.
1533
F. Adopt a patient view of quality
The highest bar one can set is to treat patients as customers. It is easy for physicians and physicists to discount the
patients expectations of health quality. The science of quality is making its way into healthcare and patients may be the
greatest resource to determine what needs to be improved.53
Patient requirements should be related to quality characteristics of the service being provided. A way to achieve this is to
have patient advocates on departmental quality committees
as well as on AAPM and ASTRO quality committees. Quality characteristics identified with patient input should be targeted for improvement and translated into necessary functions of staff and equipment. Patient survey data will be an
important tool to learn about the patients perception of care.
But surveys must be used carefully as the potential for error
and bias can be significant.30 Quality improvement capacity
needs to be aligned with professional receptiveness, leadership, technical expertise, and survey data.54 It is important to
remember that the patient is the greatest beneficiary of an
optimal quality program.
V. CONCLUSION
Medical physicists and radiation oncologists have readily
harnessed new technologies and have made many significant
contributions to radiation oncology over the years. Much like
medicine, quality is also an art and there is a need for investigation. Scientific training leads physicists and oncologists
to wait for convincing evidence for the effectiveness of new
techniques before incorporating any change of procedures to
treat patients. However, our field should not quickly dismiss
approaches to quality that have roots in probability and statistics with many years of practical experience and demonstrated benefit in other fields.
ACKNOWLEDGMENTS
The authors would like to thank the Associate Editor and
referees for many constructive comments that made their
way into the final version of this manuscript.
a
Electronic mail: [email protected]

J. M. Juran, Early SQC: A historical supplement, Qual. Prog. 30,
7381 1997.
2
M. Best and D. Neuhauser, Walter A Shewhart, 1924, and the Hawthorne
factory, Qual. Saf. Health. Care 15, 142143 2006.
3
N. R. Tague, The Quality Toolbox, 2nd ed. ASQ Quality, Milwaukee, WI,
2005.
4
N. R. Mann, Keys to Excellence: The Deming Philosophy Prestwick
House, Clayton, 1989.
5
C. S. Kilian, The World of W. Edwards Deming SPC, Knoxville, 1992.
6
If Japan can why cant we? ManagementWisdom.com, http://
www.managementwisdom.com/ifjapcanwhyc.html, 8 January 2007.
7
D. Blumenthal, Quality of careWhat it is?, N. Engl. J. Med. 335,
891894 2006.
8
R. H. Brook, E. A. McGlynn, and P. D. Cleary, Measuring quality of
care, N. Engl. J. Med. 335, 966970 2006.
9
J. F. Wilson and J. Owen, Quality research in radiation oncology: A
self-improvement initiative 30 years ahead of its time?, J. Am. Coll.
Radiol. 2, 10011007 2005.
10
T. Kron, C. Hamilton, M. Roff, and J. Denham, Dosimetric intercomparison for two Australasian clinical trials using an anthropomorphic
phantom, Int. J. Radiat. Oncol., Biol., Phys. 52, 566579 2002.
1
1534
11
C. A. Perez, P. Gardner, and G. P. Glasgow, Radiotherapy quality assurance in clinical trials, Int. J. Radiat. Oncol., Biol., Phys. 10Suppl 1,
119125 1984.
12
R. M. Macklis, T. Meier, and M. S. Weinhous, Error rates in clinical
radiotherapy, J. Clin. Oncol. 16, 551556 1998.
13
G. Huang, G. Medlam, J. Lee, S. Billingsley, J. P. Bissonnette, J. Ringash,
G. Kane, and D. C. Hodgson, Error in the delivery of radiation therapy:
Results of a quality assurance review, Int. J. Radiat. Oncol., Biol., Phys.
61, 15901595 2005.
14
C. Szerlag, Best practices, ADVANCE 16, 8085 2006.
15
A. L. Boyer and T. Schultheiss, Effects of dosimetric and clinical uncertainty on complication-free local tumor control, Radiother. Oncol. 11,
6571 1988.
16
J. W. H. Leer, R. Corver, J. J. A. M Kraus, J. Ch. Togt, and O. J. S.
Buruma, A quality assurance system based on ISO standards: experience
in a radiotherapy department, Radiother. Oncol. 35, 7581 1995.
17
K. Holli, P. Laippala, A. Ojala, and M. Pitkanen, Quality control in
health care: an experiment in radiotherapy planning for breast cancer
patients after mastectomy, Int. J. Radiat. Oncol., Biol., Phys. 44, 827
833 1999.
18
M. A. Pitkanen, K. A. Holli, A. T. Ojala, and P. Laippala, Quality assurance in radiotherapy of breast cancervariability in planning target volume delineation, Acta Oncol. 40, 5055 2001.
19
B. Tomadsen, S. W. Lin, P. Laemmrich, T. Waller, A. Cheng, B. Caldwell,
R. Rankin, and J. Stitt, Analysis of treatment delivery errors in brachytherapy using formal risk analysis techniques, Int. J. Radiat. Oncol.,
Biol., Phys. 57, 14921508 2003.
20
T. Pawlicki, M. Whitaker, and A. L. Boyer, Statistical process control
for radiotherapy quality assurance, Med. Phys. 32, 27772786 2005.
21
G. A. Patton, D. K. Gaffney, and J. H. Moeller, Facilitation of radiotherapeutic error by computerized record and verify systems, Int. J. Radiat. Oncol., Biol., Phys. 56, 5057 2003.
22
W. E. Deming, The New Economics for Industry Government, Education,
2nd ed. MIT, Cambridge, MA, 1993.
23
P. Scalliet, Risk, society and system failure, Radiother. Oncol. 80, 275
281 2006.
24
A. Niemierko, Quality assurance and validation of individual IMRT
treatment plans before delivery, Med. Phys. 31, 421422 2004.
25
G. S. Ibbott, A. Molineu, and D. S. Followill, Independent evaluations of
IMRT through the use of an anthropomorphic phantom, Technol. Cancer
Res. Treat. 5, 481487 2006.
26
Abt Associates Inc., Washington DC, 2003, pp. 174.
27
ISO 9000 and Other Standards, American Society for Quality, http://
www.asq.org/learn-about-quality/iso-9000/overview/overview.html,
8
January 2007.
28
R. Lupan, I. C. Bacivarof, A. Kobi, and C. Robledo, A relationship
between Six Sigma and ISO 9000:2000, Qual. Eng. 17, 719725 2005.
29
M. R. OLeary, Lexikon, Joint Commission on Accreditation of Healthcare Organization JACHO, Oakbrook Terrace, 1994.
30
R. G. Carey and R. C. Lloyd, Measuring Quality Improvement in Healthcare: A Guide to Statistical Process Control Applications ASQ Quality,
Milwaukee, WI, 2001.
31
W. H. Ettinger, Six Sigma: adapting GEs lessons to health care,
Trustee 54, 1015 2001.
32
G. J. Kutcher, L. Coia, M. Gillin, W. F. Hanson, S. Leibel, R. J. Morton,
J. R. Palta, J. A. Purdy, L. E. Reinstein, and G. K. Svensson, Comprehensive QA for radiation oncology: report of AAPM Radiation Therapy
Committee Task Group 40, Med. Phys. 21, 581618 1994.
33
B. Fraass, K. Doppke, M. Hunt, G. Kutcher, G. Starkschall, R. Stern, and
J. Van Dyke, American Association of Physicists in Medicine Radiation
Therapy Committee Task Group 53: Quality assurance for clinical radio-
1534
therapy treatment planning, Med. Phys. 25, 17731829 1998.
D. Hutchison, Chaos theory, complexity theory, and health care quality
management, Qual. Prog. November, 6972 1994.
35
P. P. Harteloh and F. W. Verheggen, Quality assurance in health care.
From a traditional towards a modern approach, Health Policy 27, 261
270 1994.
36
G. Laffel and D. Blumenthal, The case for using industrial quality management science in health care organizations, JAMA, J. Am. Med. Assoc. 262, 28692873 1989.
37
G. Taguchi, Introduction to Quality Engineering American Supplier Institute, Dearborn, MI, 1986.
38
O. Hupe and J. Brunzendorf, A novel method of radiochromic film dosimetry using a color scanner, Med. Phys. 33, 40854094 2006.
39
A. L. McNiven, M. Mulligan, T. Kron, and J. J. Battista, The response of
prototype plane-parallel ionization chambers in small megavoltage x-ray
fields, Med. Phys. 33, 39974004 2006.
40
A. S. Meigooni, S. B. Awan, and K. Dou, Feasibility of calibrating
elongated brachytherapy sources using a well-type ionization chamber,
Med. Phys. 33, 41844189 2006.
41
A. Piermattei, A. Fidanzio, G. Stimato, L. Azario, L. Grimaldi, G.
DOnofrio, S. Cilla, M. Balducci, M. A. Gambacorta, N. Di Napoli, and
N. Cellini, In vivo dosimetry by an aSi-based EPID, Med. Phys. 33,
44144422 2006.
42
C. Talamonti, M. Casati, and M. Bucciolini, Pretreatment verification of
IMRT absolute dose distributions using a commercial a-Si EPID, Med.
Phys. 33, 43674378 2006.
43
S. Yoo, G. Y. Kim, R. Hammoud, E. Elder, T. Pawlicki, H. Guan, T. Fox,
G. Luxton, F. F. Yin, and P. Munro, A quality assurance program for the
on-board imager, Med. Phys. 33, 44314447 2006.
44
O. A. Zeidan, S. A. L. Stephenson, S. L. Meeks, T. H. Wagner, T. R.
Willoughby, P. A. Kupelian, and K. M. Langen, Characterization and use
of EBT radiochromic film for IMRT dose verification, Med. Phys. 33,
40644072 2006.
45
D. Thwaites, Quality assurance into the next century, Radiother. Oncol.
54, viiix 2000.
46
J. E. Ibrahim, Translating quality into research: Do we need more research into quality or should quality activities be conducted using the
principles and methodological rigour of scientific research?, J. Qual.
Clin. Pract. 20, 6364 2000.
47
H. Singh, L. A. Petersen, and E. J. Thomas, Understanding diagnostic
errors in medicine: a lesson from aviation, Qual. Saf. Health. Care 15,
159164 2006.
48
P. Dixon and B. OSullivan, Radiotherapy quality assurance: time for
everyone to take it seriously, Eur. J. Cancer 39, 423429 2003.
49
P. B. Boggs, Peak expiratory flow rate control chart: a breakthrough in
asthma care, Ann. Allergy Asthma Immunol. 77, 429432 1996.
50
T. Kehoe and L.-J. Rugg, From technical quality assurance of radiotherapy to a comprehensive quality of service management system, Radiother. Oncol. 51, 281290 1999.
51
K. H. Yu, R. L. Nation, and M. J. Dooley, Multiplicity of medication
safety terms, definitions and functional meanings: when is enough
enough?, Qual. Saf. Health. Care 14, 358363 2005.
52
P. Pronovost and R. Wachter, Proposed standard for quality improvement research and publication: one step forward and two steps back,
Qual. Saf. Health. Care 15, 152153 2006.
53
B. L. Sadler, To the class of 2005: Will you be ready for the quality
revolution?, J. Qual. Pat. Saf. 32, 5455 2006.
54
E. Davies and P. D. Cleary, Hearing the patients voice? Factors affecting the use of patient survey data in quality improvement, Qual. Saf.
Health. Care 14, 428432 2005.
34
The future of PACS

Paul G. Nagy
Department of Diagnostic Radiology, University of Maryland School of Medicine, 22 South Greene Street,
Baltimore, Maryland 21201
Received 26 February 2007; revised 2 May 2007; accepted for publication 2 May 2007;
published 11 June 2007
How will the future of picture archiving and communication systems PACS look, and how will
this future affect the practice of radiology? We are currently experiencing disruptive innovations
that will force an architectural redesign, making the majority of todays commercial PACS obsolete
as the field matures and expands to include imaging throughout the medical enterprise. The common architecture used for PACS cannot handle the massive amounts of data being generated by
even current versions of computed tomography and magnetic resonance scanners. If a PACS cannot
handle todays technology, what will happen as the field expands to encompass pathology imaging,
cone-beam reconstruction, and multispectral imaging? The ability of these new technologies to
enhance research and clinical care will be impaired if PACS architectures are not prepared to
support them. In attempting a structured approach to predictions about the future of PACS, we offer
projections about the technologies underlying PACS as well as the evolution of standards development and the changing needs of a broad range of medical imaging. Simplified models of the history
of the PACS industry are mined for the assumptions they provide about future innovations and
trends. The physicist frequently participates in or directs technical assessments for medical equipment, and many physicists have extended these activities to include imaging informatics. It is hoped
that by applying these speculative but experienced-based predictions, the interested medical physicist will be better able to take the lead in setting information technology strategies that will help
facilities not only prepare for the future but continue to enjoy the benefits of technological innovations without disruptive, expensive, and unexpected changes in architecture. A good PACS strategy can help accelerate the time required for innovations to go from the drawing board to clinical
implementation. 2007 American Association of Physicists in Medicine.
DOI: 10.1118/1.2743097
Key words: PACS, information systems, Moores law, technology assessment
I. INTRODUCTION
Why make predictions about the future of picture archiving
and communication systems PACS? Our ability to deliver
quality care is increasingly dependent upon information technologies IT that drive all aspects of image acquisition,
communication, interpretation, and storage. A radiology department can no longer operate with film and expect to effectively process and read a computed tomography CT
study from a modern scanner. Architectural decisions about
todays PACS technology can lead to limitations where new
imaging modalities and techniques are not supported or image data cannot be economically stored and retrieved adequately for medical management decisions. The ability to
grow and innovate relies on having an, open, flexible, and
scalable architecture. This paper presents six predictions that,
if taken together, present a scenario for the next generation of
PACS.
We are taught as scientists that all extrapolations including, perhaps especially, predictions are fraught with uncertainty, so that it is important to clearly state the assumptions
on which such extrapolations are based. One of the key assumptions in looking to the future of PACS is that the industry has a relatively small market size compared with other IT
industries. Despite its mission-critical clinical application,
2676
Med. Phys. 34 7, July 2007
this market is still being driven by medical device manufacturers. Those factors are important in the PACS industrys
somewhat later adopter path in the classic pattern of technologic innovations when compared with other computer
industry sectors. By looking at other domains in the computer industry, therefore, we can make modest predictions
about the future of PACS with some confidence. Noted futurist William Gibson is often quoted as saying, The future
is here. Its just not widely distributed yet.1 These predictions, too, may not occur at precisely the forecast intervals
but, barring extraordinary intervening circumstances, will almost certainly prove valid.
Prediction 1. Moores law will hold true for at least
another 10 years
The fundamental predictor often cited in the computer

industry when planning for future growth is Moores law.2
Gordon Moore, cofounder of Intel, predicted in 1965 that
the number of transistors on a microprocessor would continue to double every 24 months with the cost remaining the
same.3 Moore was conservative in his original forecast, predicting that this trend would continue for a decade. Instead,
Moores law has held true for more than 40 years, from a
time when a microprocessor had 1-2000 transistors to to-
0094-2405/2007/347/2676/7/$23.00
2676
2677
Paul G. Nagy: The future of PACS
days standard model dual-core 64-bit processors with more

than 200 106 transistors.4
Moores law has also proven accurate for other technologies in the computer industry, such as storage. September
2006 marked the 50th anniversary of the magnetic hard
drive.5 The first hard drive, invented by IBM, was made up
of 50 24-in. aluminum platters holding 5 MB worth of data
and weighing more than a ton.6 Todays hard drive capacity
is 1 TB 1000 gigabytes,7 within a 3.5-in. enclosure weighing less than 1 kg 1.5 lb.
Debate has long focused on when and how Moores Law
will either meet its quantum limits,8 user needs will be far
outpaced by capacity,9 or the computer fabrication industry
will no longer be able to support rapid continuous growth.10
Most experts including Moore himself believe that Moores
law will continue to be valid for at least another 10 years.11
Another popular notion is that Moores law has become a
self-fulfilling prophecy driving consumer expectations and
marketing.12 In 2004, the International Technology Roadmap
for Semiconductors, a large consortium of semiconductor
vendors including Intel, performed technical analyses on
the limitations of technology and issued a detailed consensus
report suggesting that Moores law will be pertinent for the
next 15 20 years13 as more nanotechnology is incorporated
into the fabrication process. One important but not always
appreciated aspect of Moores law is that it does not represent a constant linear growth rate of one technology but,
instead, a succession of overlapping sigmoids of different
technologies.14 As one technology matures and flattens out, a
new technology takes its place to drive the industry. With
hard drives this effect can be seen in differing technologies
such as antiferromagnetically coupled media, perpendicular
disk heads, and paramagnetic materials. In semiconductors,
this can be seen in a continuous stream of smaller fabrication
methodologies that require ramp-up time before achieving
economies of scale.
If Moores law continues to hold true, in 2012 the average
personal computer will be equipped with more than 109 transistors the rough equivalent of a 40-GHz system and will
have a 3-TB hard drive. In 2017, by the same predictive
metrics, the average $1000 desktop computer will have more
than 6 109 transistors the rough equivalent of a 320-GHz
system, with a 24-TB hard drive.
Prediction 2. PACS storage and archives will be all
spinning serial advanced technology attachment
SATA archives with high-density storage
farms driven by the performance requirements
of clinical applications for just-in-time delivery
The high-density online market hosts a growing industry

focusing on backup storage systems and systems with large
files, such as PACS. Todays enterprise-class storage systems
can house up to 42 1000-GB, 3.5-in., 7200-rpm SATA drives
in a 6-in. rack enclosure 4U and can maintain 42 TB of
storage with subsecond access times.15 The key to this market is using slower drives of 7200 rpm rather than Fibre
Channel 15 000-rpm drives. The slower drives generate less
Medical Physics, Vol. 34, No. 7, July 2007
2677
FIG. 1. Growth in number of images per study and number of CT studies

performed between 1998 and 2006.
heat and consume less power so that more can be aggregated

in a single enclosure. Applications such as e-mail and database servers with more than 100 input/output requests per
second require higher speed storage systems, because they
are driven by the seek time needed for the cylinder to rotate
under the drive head. For applications such as PACS, the key
metric is in throughput and does not depend on seek times
but, instead, on throughput rates in tens to hundreds of megabytes per second. Putting more than ten hard drives into a
redundant array of inexpensive disks configuration will combine nearly linearily the throughput speeds of the individual
disks and exceed a gigabit network performance
100 MB/ s needed for fast image delivery in a PACS
system. The other primary benefit of using 7200-rpm SATA
drives is that they are less expensive, currently available at
less than $2000 per terabyte of enterprise class storage, with
8 times the market volume of small computer system interfaces SCSIs and Fibre Channel high-speed drives.
This is encouraging news but not sufficiently encouraging
to warrant reducing next years operating budget for storage.
A study performed at the Mayo Clinic in 2002 indicated that
medical imaging data growth was outstripping the computer
expansion capabilities predicted by Moores law.16 Experience at the University of Maryland over the past 8 years has
seen rapid growth in storage needs, at a pace only slightly
slower than Moores law. As shown in Fig. 1, the number of
CT studies increased from 10 165 in 1998 to 57 713 in 2005
14.7% and 24.6% of total study volumes in those respective
2678
2678
Prediction 3. The commoditization of the server

hardware market will allow the utilization
of high-density processing farms, such as blade
architectures, which can provide a new level
of fault tolerance, processing horsepower,
and scalability
FIG. 2. Storage usage per year has seen growth predominantly from an
escalation in the number of CT studies and CT study volume.
years. The number of images per study in the department is

also increasing, from an average of 69 in 1998 to 234 images
per study in 2006. Figure 2 shows the 8-year growth in storage demand resulting from these two trends. This escalating
growth is equal to a doubling in storage every 2.5 years, only
slightly slower than Moores predicted doubling of computer
capabilities every 2 years.
Two trends will continue to increase imaging data storage
use. One is the archiving requirements associated with submillimeter thin-slice CT data. The University of Maryland
does not routinely archive thin-slice data but is planning to
begin doing so in the third quarter of 2007.17 This change in
practice will double our study volume and synchronize our
growth with that predicted by Moores law. The other factor
is the potential move of CT manufacturers to 1024 1024
matrix sizes, an innovation that could quadruple dataset
sizes.
High-density spinning-disk storage devices simplify the
older hierarchical storage model used in PACS, in which
images are available on line for a period of time typically 6
months or so and then go to a DVD or tape library.
Spinning-disk storage systems are displacing tape systems as
backups in other markets because of their competitive pricing and superior performance. Retrieving data from backup
on tape for either migration or recovery from a failure takes
significant time. One study on migration suggested that migration of PACS data from a tape archive takes roughly a
third of the time that original acquisition required.18 Six
years of data backed up on tape might realistically take up to
2 years to transfer to a new system.19
Spinning disks themselves are on a limited lifetime as a
technology. Solid state disks SSDs are a type of nonvolatile
memory that retains information unpowered popular in portable media players. SSDs have no moving parts, consume
less power, and are faster than spinning disks in seek time
and latency. SSDs are currently six times as expensive per
gigabyte as spinning hard drives. They are predicted to replace the spinning disk by 2012.20
A blade farm is a server processing farm in which the

equivalent of 16-20 servers can be placed in the same 8 in.
of rack space that once could house only one or two servers.
Procuring the power and network cabling for staging a normal server can be time consuming and expensive. In a blade
configuration, a chassis provides a common housing for the
blades as well as power and networking. These servers are
typically dual-processor, dual-core systems that can provide
the equivalent of 64 processors in a single-chassis unit.
Blade server sales accounted for 6% of all server sales in
2006, but, with an increase of more than 30% over 2005,
represented the fastest growing market segment.21 Another
advantage of blade systems is that they contain their own
network backbones, which can run at terabit transfer rates
within and between blade enclosures, accelerating server-toserver communications.22
An inexpensive server processor farm provides useful
functions. The first and most important for the PACS industry is that using multiple servers for the same function within
a PACS provides fault tolerance by removing single points of
failure. In an August 2006 survey of data centers, hardware
failure was the leading cause 49% of unplanned outages.23
Moving parts, such as spinning hard drives, fans in power
supplies, CPUs, and robotic arms are generally the first to
fail in a server. The potential costs in labor and disruption
with unplanned outages can provide a reasonable investment
justification for additional servers, each of which should cost
less than $5000. Most institutions running PACS as missioncritical applications will have uptime guarantees equal to or
in excess of 99.9%, equating to less than 9 h of downtime
scheduled or unscheduled within a year. Such an uptime
rate is not realistically achievable with single-hardware dependencies. In 5 years, the use of fault-tolerant distributed
systems, such as blade farms, should be predominant, and, in
10 years, few or no single-server systems will be running in
critical environments.
Another advantage of a multiple or distributed server environment is scalability. An enterprise running with a single
server is not only vulnerable to single points of failure but to
performance bottlenecks. As PACS systems grow to meet
enterprise demands in other departments, it is important to
incorporate the capability to quickly scale server requirements.
The final advantage of large, inexpensive computing
server environments is their ability to facilitate the introduction into clinical use of a new class of applications that require massive number crunching, such as computer-aided diagnosis CAD, volumetric organ segmentation, contentbased image retrieval, and real-time registration and
subtraction of deformable volumetric datasets.
Distributed systems require a redesign in architecture to
take full advantage of the fault tolerance, scalability, and
2679
processing power these systems can provide. Consider, as an

example, the performance and scale of the Google Internet
search engine, which provides subsecond querying of a catalog of more than 25 109 Web pages. This single application
is distributed among 450 000 servers around the globe.24
Prediction 4. The next generation of PACS will rely

heavily upon server-side video rendering for large
datasets. The first three generations of
architecture simply cannot handle the large
volumetric datasets being generated by todays CT
and magnetic resonance imaging scanners
To better understand this prediction, it is important to look

at the history of PACS and the evolving shift between the
client and server. The housing and processing of images has
been consistently moving from the workstation client to the
server over the past 20 years. It should be noted that this is a
very simple model, and that significant differences characterize different vendors even within the same generation of
technology. Nevertheless, the general trend is away from the
client and toward the server for image storage, processing,
and visualization.
The first generation of PACS, from the mid-1980s to the
early 1990s, were essentially screen-capturing systems that
fed images off the modalities into workstations using proprietary networking and file format protocols. These systems
cost more than $20 million, and fewer than 25 such systems
were deployed. Systems based on this architecture are no
longer in production.
The development of the second generation, from the early
to mid-1990s, was driven by the militarys Digital Imaging
Network-PACS proposals that required implementation of
the Digital Imaging and Communications in Medicine
DICOM standard for image file formats and communication. Images were routed to workstations from the modalities
via DICOM. Studies were routed to several workstations in
an effort to anticipate the reading location of the radiologist.
Systems were limited to radiologists, with no enterprise distribution.
The third generation of PACS introduced the use of ondemand central architectures, in which images were centralized into a primary storage pool and could be pulled quickly
from any diagnostic workstation. Only requested studies
were sent to workstations. Because images were stored centrally, lightweight Web-based image viewers were deployed
to begin to address the imaging needs of referring physicians
within the enterprise.
The fourth generation began in the late 1990s and was
defined by two significant contributions. The first is a single
client for users outside of radiology as well as inside. This
innovation was enabled by extending Web browsers or by
using portable byte code engines such as the Java virtual
machine or the C sharp common language infrastructure instead of thick clients. This approach minimized the support
costs incurred by installing the thick applications on hundreds of computers throughout an enterprise. This also delivered a full-featured application to assist physicians outside of
2679
radiology. The second innovation was the use of dynamic

compression to spare the network bandwidth in delivering
images to the client.
The fifth generation of PACS will be distinguished by
server-side rendering. There are two factors that will drive
the industry towards server-side computing. One is the
growth in volumetric data generated in a single study; the
larger the study, the longer the time required to deliver those
images to the client. The second is the limitations in growth
of network speeds resulting from high labor costs associated
with re-laying cabling throughout a medical institution. Although storage growth may keep up with imaging volume
growth, network transfer performance will not. The result is
that images can be stored, but it may take longer and longer
to send specific sets of images to the requesting physician.
The solution will require leaving the images on the server
and rendering them from there.
A server-side rendering farm is a group of servers dedicated to displaying volumetric data and providing a remote
session to the end users. This allows all the large volumetric
datasets to stay in the data center and spares the network
connection to the enterprise by sending only a presentation
of the data and not the data itself. Physicians throughout the
enterprise have considerable use for and are coming to depend on three-dimensional visualization of volumetric data.
The current model of deploying specialized workstations
throughout a hospital is unsupportable and far too expensive.
Server-side rendering systems can deploy lighter clients
throughout a hospital with lower support costs and greater
accessibility to the tools through site licensing instead of
fixed seat licensing, which forces the physician to go to the
workstation to use the tools. Within 5 years, the majority of
PACS vendors will either develop or support server-side rendering systems for volumetric data.
One trend that the PACS industry should explore lies in
recent innovations made in the computer game industry. As a
result of the massive processing power required to create
realistic, real-time, immersive graphical environments for
computer gamers, the processing power on a video card frequently exceeds the number of transistors on the computers
CPU.25 Silicon Graphics developed and released a common
low-level library called OpenGL to display and render geometric shapes.26 Several vendors specialize in server-side visualization of volumetric data through OpenGL libraries,
software CPU algorithms, or proprietary hardware techniques. These systems will continue to evolve, and new
PACS systems will likely be designed around this visualization methodology.
Prediction 5. The next-generation desktop client
software applications will combine functionality,
performance, and deployability
The client application of newer generations of PACS will

be thinner. Traditional PACS applications had to be written
with direct access to the Windows application program interface to provide file access and Windows response time in
scrolling and manipulating images. These applications,
2680
known as thick clients, have traditionally been written in

C + +. The downside to direct access to low levels within the
Windows operating system is the potential for shared library conflicts in versioning, known colloquially as DLL
hell. These clients can be deployed within the limited confines of dedicated workstations in the radiology department
but require hands-on access for support. Support for deployment beyond radiology is usually extremely challenging.
Thinner clients, such as Web clients, are easy to deploy
throughout thousands of workstations, because they provide
a standard browser application framework that limits access
to low-level libraries. This unfortunately also limits functionality and performance on the client. The common middle
ground that most PACS systems use for writing their clients
is through byte code engine-based programming languages
such as Java and C#. These clients provide better deployability than thick clients and better functionality than thin
clients.27
One exciting technology that appears to provide the best
of both worlds is Asynchronous Javascript and XML
AJAX. Used in applications such as Google Maps, AJAX
methodologies bring a thick-client look and feel through a
Web browser. Combining AJAX browser clients with serverside rendering systems has the potential to move performance requirements off the network and the client onto the
server while providing full functionality to the client. A complete rewrite of applications is required to take advantage of
AJAX, so the industry will not move to this new technology
for several years. Another clear advantage of AJAX and
some byte code engines is their ability to be operating system independent and perform on Windows as well as other
platforms such as Macintosh OS X and Linux. Although
Microsoft Windows will still be around in 5 years, there
will be an increasing heterogeneity in operating systems, especially in the mobile platform market.
Prediction 6. Third-party diagnostic workstations and
specialized systems will be seamlessly integrated
into clinical workflow. This will result from PACS
becoming more flexible through better standardsbased interoperability and service-oriented
architectures
Volumetric imaging has brought both progress and advantages to visualization and parameterization of organ systems
and pathologies in the past 10 years, thanks in no small part
to the medical physics community. At the same time, the
range of visualization techniques has expanded rapidly along
with new modalities, hybrid imaging technologies, and innovative procedures. No single PACS vendor can innovate in
all the subspeciality areas of image visualization and analysis, including but not limited to mammography, ultrasound,
nuclear medicine, cardiology, chest imaging, perfusion imaging, functional imaging, registration and fusion imaging, CT
colonoscopy, not to mention all the new tomosynthesis and
CAD applications coming to market.
One disturbing byproduct of these advances has been the
increasing proliferation of specialty workstations with limited integration into clinical workflow. It is becoming inMedical Physics, Vol. 34, No. 7, July 2007
2680
creasingly untenable for radiologists to sacrifice productivity

while tethered to specific locations needed to gain access to
sets of specialized tools.
The solution will be an increasing trend toward opening
up the business logic of a PACS. One popular view of software development sees three encapsulated layers: the application layer, business logic layer, and data layer. The data
layer, also known as the model, is at the bottom, where images and persistent data reside in the database. The business
logic layer, also known as the controller, is the brains of the
system and includes all the rules for the application, such as
autorouting, hanging protocols, worklists, and study status.
The application layer, also known as the view, is the direct
human user interface, where the focus is on usability. These
layers are encapsulated to allow for rapid and flexible software development. Opening the business layer will allow
third-party vendors to query unread worklists, route images
on demand, and set dictation status remotely from independent workstations.
This business logic can be exposed to third-party workstations in several ways. The most likely will be through
DICOM general purpose worklist and service-oriented architectures. DICOM has a standard for the general purpose
worklist, which is intended for postprocessing steps and radiology interpretation workflow, much as DICOM modality
workflow is used for acquisition workflow. A serviceoriented architecture SOA exposes a discoverable machine
interface known as Web Services, to the business logic of
the application. A Web Service is a set of XML documents
transmitted over Hypertext Transfer Protocol. This process is
behind the much-advertised Web 2.0 applications. SOAs are
useful in that they provide better adaptability to changing
business needs as well as flexibility to accommodate future
technologies.
Regardless of how such change is accomplished, the trend
is clearly toward an escalating need for integration and transparency of workflow in radiology. Today, DICOM modality
performed procedure steps can transmit the status of a patient, including arrival, initiation of scanning, completion of
scanning, etc. DICOM storage commitment ensures the safe
delivery of the images to the PACS system. This transparency allows other systems throughout the hospital to better
utilize radiology services by tracking patients. The challenge
is to reduce the cost and speed of implementing these interfaces so that they become plug and play utilities. The Integrating the Healthcare Enterprise initiative is working to
take existing standards and accelerate their adoption into the
clinical environment by creating common use cases and
demonstrating them in real time at Connectathons, forums
in which vendors showcase their integration capabilities.
Open standards have the longest shelf life in technology
lasting decades. The entire reason we develop and adhere to
standards is to take advantage of their consistency in an environment where all the other technology is changing rapidly. Today the market is already becoming saturated with
the majority of radiology operations using PACS. As sites
have learned their lessons with their first PACS system and
begin to migrate to newer systems over the next 5 years, the
2681
emphasis will become increasingly important on open standard integration. Although the DICOM transfer syntax might
evolve, the main DICOM object model will be relatively
unchanged in 10 years and will be one of the few things
recognizable to systems used today.
II. CONCLUSION
Institutions should develop strategies for future proofing their technology buying decisions. The most important
factor is to ensure that the purchasing model allows the institution to enjoy the market benefits of Moores law. The
most direct way to accomplish this is to move to a softwareonly model in which hardware purchases are made independently of the PACS vendor, with only minimum hardware
requirements for servers, storage, and workstations. This allows a site to regularly replace hardware at market value and
enjoy benefits in better performance, reliability, and cost.
Another recommended task is to hold discussions with the
PACS vendor on ways to incorporate third-party workstations into the clinical workflow. The lowest cost, long-term
solution for customers is integration through openly published standards and not piecemeal vendor partnerships with
proprietary interfaces. Vendors should also provide a roadmap detailing the ways in which they plan to incorporate
volumetric datasets for visualization and archiving without
compromising the productivity of the radiologists.
Physicists have a real opportunity to provide the technical
leadership that will help to shape strategies as radiology departments meet the challenges of the future. Physicists are
ideally suited to provide informatics leadership because of
our proficiency with technology, intimate domain knowledge
of imaging, and close partnership with clinical staff and physicians. Informatics is a natural extension for many physicists and an avenue for both career growth and contributions
to operations and productivity. Important discussions in the
medical physics community have focused on the role of the
physicist in implementing and supporting information systems in a hospital.28,29 Both sides agree that IT is certainly an
opportunity; the debate is over whether changes should be
made to the requirements of an already large curriculum of
physicist training. It is clear that more discussion is necessary on the new and challenging opportunities in the practice
of medical physics. What is undisputed is that unique to the
medical discipline, the radiologist and the physicist have had
a mutually supportive relationship for more than a century.30
The radiologist needs IT guidance and support in information
systems now more than ever before.
ACKNOWLEDGMENTS
I would like to thank Dr. Nancy Knight from the University of Maryland School of Medicine and Dr. William
Hendee from the Medical College of Wisconsin for their expert assistance in preparing this manuscript.
1
William Gibson, www.wikipedia.com, http://wn.wikiquote.org/wiki/

William_Gibson, 29 May 2007.
2
Moores Law Diagram, Wikipedia, the Free Encyclopedia, http://
en.wikipedia.org/wiki/Image:Moore_Law_diagram_%282004%29.png,
2681
15 December 2006.
G. Moore, Cramming more components onto integrated circuits, Electronics
38, 114117 1965; ftp://download.intel.com/museum/
Moores_Law/Articles-Press_Releases/Gordon_Moore_1965_Article.pdf,
8 January 2007.
4
V. Freeman, Athlon 64 X2 4800+ processor review, Sharkey Extreme,
June 22, 2005. http://www.sharkyextreme.com/hardware/cpu/article.php/
3514901, 9 January 2007.
5
S. Levy, The hard disk that changed the world, Newsweek, August 7,
2006, http://www.msnbc.msn.com/id/14096484/site/newsweek/, 8 January 2007.
6
A brief history of the hard disk drive, DUX Computer Digest, May 19,
2005. http://www.duxcw.com/digest/guides/hd/hd2.htm, 8 January 2007.
7
S. Olsen, Hitachi Global Storage announces 1TB Hard Drive, Daily
Tech., January 5, 2007. http://www.dailytech.com/article.aspx?newsid
5590, 8 January 2007.
8
M. Kanellos, Intel scientists find wall for Moores Law, ZDNet News,
December 1, 2003. http://news.zdnet.com/2100-9584_22_5112061.html,
9 January 2007.
9
D. Mohney, Moores Law meets market saturation, The Inquirer, December 5, 2002. http://www.theinquirer.net/default.aspx?article6559, 9
January 2007.
10
Rocks law of the doubling of fabrication costs, Wikipedia, the Free
Encyclopedia, http://en.wikipedia.org/wiki/Rock%27s_law, 15 December
2006.
11
M. Kanellos, Moores Law to roll on for another decade, CNET News,
February 10, 2003, http://news.com.com/2100-1001-984051.html, 9 January 2007.
12
B. Schaller, The origin, nature, and implications of Moores Law, Microsoft Research, September 26, 1996; http://research.microsoft.com/
~gray/Moore_Law.html, 9 January 2007.
13
P. Gargini, Sailing with the ITRS into nanotechnology, PowerPoint
presentation,
ftp://download.intel.com/technology/silicon/
Paolo_Semicon_West_071904.pdf, 15 December 2006.
14
C. M. Christenson, The Innovators Dilema: The Revolutionary Book that
Will Change the Way You Do Business Collins Business Essentials, New
York, 2003.
15
K. Schwartz, Nexan gives ATA more muscle with SATABeast, eWeek.com,
October
18,
2005,
http://www.eweek.com/article2/
0,1895,1872607,00.asp, 9 January 2007.
16
B. J. Erickson, K. R. Persons, N. J. Hangiandreou, E. M. James, C. J.
Hanna, and D. G. Gehring, Requirements for an enterprise digital image
archive, J. Digit Imaging 14, 7282 2001.
17
C. Meenan, B. Daly, C. Toland, and P. Nagy, Use of a thin-section
archive and enterprise 3D software for long-term storage of thin-slice CT
datasets. J. Digit Imaging 19suppl 1, 8488 2006.
18
F. M. Behlen, A DICOM document-oriented approach to PACS infrastructure, J. Digit Imaging 113 suppl. 1, 3538 1998.
19
F. M. Behlen, R. E. Sayre, J. B. Weldy, and J. S. Michael, Permanent
records: Experience with data migration in radiology information system
and picture archiving and communication system replacement, J. Digit
Imaging 13 2 suppl. 1, 171174 2000.
20
P. Thibodeau, Flash memory not ready to explode as PC storage technology,
Computer
World,
April
3,
2007,
http://
www.computerworld.com/action/article.do?command
viewArticleBasic&articleId9015478, 15 April 2007.
21
M. Fontecchio, Blade server growth drives revenue Q3 06, SearchDataCenter.com,
November
22,
2006.
http//
searchdatacenter.techtarget.com/originalContent/
0,289142,sid80_gci1231203,00.html, 9 January, 2007.
22
Hewlett-Packard, Inc, HP BladeSystem c-class servers blades
interconnects, News Release, http://h18004.www1.hp.com/products/
blades/components/c-class-interconnects.html, 9 January 2007.
23
DM Review Editorial Staff, The 24 7 challenge: IOUG 2006 survey on
high availability trends, DM Review, August 29, 2006, http://
www.dmreview.com/editorial/newsletter_article.cfm?articleId1061768,
9 January 2007.
24
Google, Wikipedia, the Free Encyclopedia, http://en.wikipedia.org/
wiki/Google#Technology, 31 January 2007.
25
T. Wilson and J. Tyson, How graphics cards work, HowStuffWorks,
http://computer.howstuffworks.com/graphics-card.htm, 9 January 2007.
26
N. D. Terry, Introduction to OpenGL, Tulane University Class CPEN
3
2682
461/661, Spring 1997, http://www.eecs.tulane.edu/Terry/OpenGL/

Introduction.html, 9 January 2007.
27
C. Toland, C. Meenan, M. Toland, N. Safdar, P. Vandermeer, and P. Nagy,
A suggested classification guide for PACS client applications: The five
degrees of thickness, J. Digit Imaging 19suppl 1, 7883 2006.
28
J. R. Halama and W. Huda, Medical physicists should assume PACS
responsibilities, Med. Phys. 29, 19131915 2002.
2682
29
G. C. Nikiforidis, G. C. Kagadis, and C. G. Orton, It is important that

medical physicists be involved in the development and implementation of
integrated hospital information systems, Med. Phys. 33, 44554458
2006.
30
P. R. Almond, Foreword, Radiation Physics, in A History of the Radiological Sciences, Vol. III, edited by P. R. Almond Radiology Centennial,
Reston, 1996, pp. vix.
BGRT: Biologically guided radiation therapyThe future is fast

approaching!
Robert D. Stewart
School of Health Sciences, Purdue University, 550 Stadium Mall Drive, West Lafayette,
Indiana 47907-2051
X. Allen Lia
Department of Radiation Oncology, Medical College of Wisconsin, 8701 Watertown Plank Road,
Milwaukee, Wisconsin 53226
Received 3 January 2007; revised 9 August 2007; accepted for publication 10 August 2007;
published 11 September 2007
Rapid advances in functional and biological imaging, predictive assays, and our understanding of
the molecular and cellular responses underpinning treatment outcomes herald the coming of the
long-sought goal of implementing patient-specific biologically guided radiation therapy BGRT in
the clinic. Biological imaging and predictive assays have the potential to provide patient-specific,
three-dimensional information to characterize the radiation response characteristics of tumor and
normal structures. Within the next decade, it will be possible to combine such information with
advanced delivery technologies to design and deliver biologically conformed, individualized therapies in the clinic. The full implementation of BGRT in the clinic will require new technologies and
additional research. However, even the partial implementation of BGRT treatment planning may
have the potential to substantially impact clinical outcomes. 2007 American Association of
Physicists in Medicine. DOI: 10.1118/1.2779861
Key words: biologically guided radiation therapy, biological imaging, outcome modeling, LQ
model
I. INTRODUCTION
The efficacy of radiotherapy RT is ultimately determined
by our ability to deliver doses of radiation sufficient to eradicate a tumor, while adequately sparing nearby normal structures. Until now, the practice of RT has been based on the
premise that effective treatments require the delivery of uniform radiation doses to all target volumes. The uniform-dose
theorem1 provides some justification for this treatment strategy. However, the uniform-dose approach only yields the
best possible tumor control for the implausible situation in
which all regions of the tumor have exactly the same biological characteristics and sensitivities to radiation. Biological factors that modulate treatment outcomes include such
factors and processes as a cells capacity for repair, acute and
chronic hypoxia, cell proliferation, the spatial distribution
and density of tumor clonogens, and the differential sensitivity of cells to radiation during the cell cycle. Genetic and
other patient-specific factors that impact on radiation responses are seldom considered in radiation treatment planning for specific patients.
Theoretical studies and the accumulated clinical data214
strongly suggest that the treatment plans designed to exploit
patient- and tumor-specific biological features will substantially improve treatment outcomes. However, the complexity
of biological systems and our incomplete knowledge of the
mechanisms underpinning tumor and normal-tissue responses, pose many challenges for the clinical implementation of biological criteria into the treatment planning process.
To advance patient-specific treatment planning without unduly jeopardizing patient care, we propose a three-phase
3739
Med. Phys. 34 10, October 2007
strategy to integrate biological information into treatment

plans for the next generation of RT, i.e., biologically guided
radiation therapy BGRT. Although the full implementation
of BGRT will undoubtedly require new technologies and additional research to advance our understanding of biological
mechanisms, we believe that even a partial implementation
of BGRT treatment planning has the potential to substantially
impact clinical outcomes within the next ten years.
II. MODELS FOR THE PREDICTION OF TREATMENT
OUTCOMES
Implementation of BGRT requires the identification of
suitable biological objectives to facilitate the intercomparison and ranking of treatment plans. The two main biological
objectives are to maximize local tumor control probabilities
TCPs while simultaneously minimizing normal tissue complication probabilities NTCPs. As an alternative to specifying biological objectives in terms of TCP and NTCP values,
concepts such as the biologically effective dose BED15 and
equivalent uniform dose EUD16 are sometimes used for
treatment plan ranking and intercomparison. Regardless, effective tumor control presumably requires the reproductive
inactivation of all aberrant target cells. The direct killing of
cells by radiation also contributes to normal tissue complications, although intercellular communication, cell-matrix interactions, and other factors influence the collective response
of normal tissues to radiation.1719
A representative NTCP model is briefly reviewed to highlight some of the challenges and research needs associated
with the important dose-limiting issue of normal tissue tox-
0094-2405/2007/3410/3739/13/$23.00
3739
3740
R. D. Stewart and X. A. Li: Biologically guided radiation therapy
icity. A method to identify isoeffective prescription doses is

introduced. The putative relationship between cell death
mechanisms and the linear-quadratic LQ survival
model2023 is also examined to highlight the need for better
outcome models and as a cautionary note about the potential
strengths and weaknesses of using the LQ model to guide the
identification of equivalent prescription doses and patientspecific dose distributions. Examples illustrating the impact
of uncertainties in biological parameters, including the effects of interpatient heterogeneity in tumor radiosensitivity
parameters, are presented and discussed within the context of
patient-specific treatment planning and the future of outcome
modeling.
II.A. LymanKutcherBurman NTCP model
Much of our knowledge of normal-tissue tolerances

comes from the seminal publication of Emami et al.24 The
authors used data from the literature and personal experience
to compile tolerance-dose values for uniform irradiation of
twenty-eight critical structures. In the accompanying article
by Burman et al.25 the tolerance data were used to determine
parameters for the phenomenological NTCP model proposed
by Lyman.26 Because the Lyman model is defined for uniform irradiation, and normal tissues are rarely irradiated uniformly, several algorithms to convert a heterogeneous dose
distribution into a uniform dose distribution resulting in the
same NTCP have been proposed. The effective volume
method for dose-volume histogram DVH reduction proposed by Kutcher and Burman27 is the most commonly used
approach. The combination of the DVH reduction algorithm
and Lymans NTCP model, which is often referred to as the
LymanKutcherBurman LKB model, provides a phenomenological method to estimate NTCP values for uniform or
nonuniform irradiation of many critical tissue structures.
The LKB model can be described by three equations:28,29
NTCP =
t=
ex
2dx
Deff TD50
mTD50
Deff =
,
viD1/n
i
where Deff is the dose that, if given uniformly to the entire

volume, will lead to the same NTCP as a nonuniform dose
distribution Deff is sometimes referred to as equivalent uniform dose, EUD, TD50 is the uniform dose given to the
entire organ volume that results in a 50% risk of complications, m is a measure of the slope of the sigmoid curve represented by the integral of the normal distribution, n is a
parameter that describes the magnitude of the volume effect,
and Di and vi are the dose and fractional volume, respectively, of each bin used to tabulate the differential DVH.
The LKB model has been widely used as an auxiliary tool
to evaluate and compare treatment plans,3032 guide doseescalation studies,33,34 and select tumor doses for new
Medical Physics, Vol. 34, No. 10, October 2007
3740
regimens.35 However, with the exception of a few

studies33,36,37 that use parameters derived from more recent
complication data, the original TD50 estimates of Burman
et al.25 are still nearly always used for treatment planning
applications. The tolerance doses reported by Burman et al.25
need to be updated based on the new clinical data, e.g., more
accurate patient-specific volumetric dosimetric data for
newer treatment planning and delivery technologies. In addition, differences exist in delivered dose patterns between
older and newer modalities, such as high dose gradients with
intensity modulated RT IMRT, large doses to smaller volumes versus small doses to large volume when comparing
fixed-beam with rotational delivery. The accuracy of the effective volume method for DVH reduction is an open question that raises many questions about the accuracy of NTCP
estimates derived from the LKB model. To improve confidence in the prediction of normal tissue complications for
partial-volume irradiation, the 3D dose distribution delivered
to each patient needs to be recorded.
II.B. Poisson TCP model
The most commonly used TCP model is based on the

Poisson distribution.38,39 With the Poisson model, the probability that all tumor cells are inactivated is given by
N
TCP = TCPi = exp viiSDieiTTlag .

i=1
i=1
Here, i denotes the expected number of tumor clonogens

per cm3 in the ith tissue region, vi is the volume of the ith
tissue region cm3, SDi is the fraction of the tumor cells in
the ith tissue region that survive total treatment dose Di, and
T is the overall time to complete the treatment. The effects of
subclinical disease on tumor control can be incorporated into
Eq. 4 by decreasing the clonogen density parameter in
regions outside the gross tumor volume. For the special case
when i = 0 i.e., the tissue region does not contain any tumor
cells, TCPi equals unity regardless of dose. After time interval Tlag, accelerated tumor repopulation becomes effective
and is characterized by rate constant, . The tumor repopulation rate is conveniently expressed as = ln2 / Td, whereTd
is the effective tumor doubling time. Regional differences in
cell proliferation within the tumor can be accommodated by
adjusting the parameter. Other effects associated with differences in tumor radiation sensitivity parameters among patients interpatient heterogeneity and within the tumor intratumor heterogeneity are implicit in Eq. 4, i.e., within
the SDi term.
The estimation of radiobiologic parameters from a patient
population with heterogeneous radiation response parameters
is notoriously challenging.40 Also, the Poisson TCP model
does not provide a fully adequate representation of the statistical aspects of tumor growth kinetics during a course of
radiation therapy.41,42 The Logistic or Gompertz growth kinetic models4345 may be more appropriate than the simple
exponential growth kinetics used in many tumor control
models, including Eq. 4.
3741
Despite the limitations and approximations, the Poisson

TCP model provides some potentially useful constraints to
guide the selection of dose distributions needed to achieve
tumor control. For example, in treatment planning, we are
seeking the distribution of Di values such that the overall
TCP1. Because the overall TCP is the product of the TCP
values for all N tissue regions, the Poisson TCP model predicts that treatment failure will occur if the TCP for any
tissue region is zero. These so-called biological cold spots
may arise because some regions of tissue are underdosed or
because some tumor cells are particularly resistant to radiation or are rapidly dividing. Equation 4 implies that the
equality constraint TCPi = TCP1/N may be imposed without
decreasing the overall chance of achieving local tumor control. For the special case when 1 i = , vi = v, and i = for
i = 1 to N and 2 all tumor cells are equally sensitive to
radiation, the Poisson TCP model predicts that the optimal
dose distribution is a uniform dose distribution, i.e., Di = D
for i = 1 to N see also Webb et al.1. The observation that the
overall TCP is a product of TCP values for a series of subvolumes also provides some justification for the use of the
EUD concept16 as a surrogate dosimetric quantity for the
assessment and intercomparison of treatment plans.46 That is,
uniform and nonuniform dose distributions are biologically
equivalent, as long as they produce the same overall TCP.
II.C. Cell death and the double strand break
Models for local tumor control and, to a lesser extent,

normal-tissue complications, are intimately linked to the reproductive death or outright killing of cells by radiation. Reproductive death may be defined as some form of molecular
or cellular damage that suffices to permanently inhibit cell
division, whereas cell death encompasses apoptosis, necrosis, mitotic catastrophe, and other more immediate modes of
cell dissolution. Reproductive death suffices to achieve local
tumor control, although apoptosis and other cell dissolution
modes are certainly an effective way to permanently inhibit
cell division. Decades of research on the mechanisms of formation and consequences of chromosome aberrations21,47,48
provide a conceptually appealing framework to link double
strand break DSB induction and processing to reproductive
cell death. A DSB is formed when the passage of radiation
through a cell creates a cluster of DNA lesions that effectively cuts the DNA double helix into two smaller pieces.
Experiments have shown that the yield of many types of
clustered DNA lesions, including DSB, are proportional to
absorbed doses up to several hundred Gy for low and high
LET radiation.4952 Experiments such as these imply that
single-hit mechanisms are responsible for the induction of
DSB and other types of clustered damage by radiation.
Once formed, the break-ends associated with one DSB
may be rejoined to their correct partner correct rejoining or
to a break-end associated with a different DSB incorrect
rejoining. A third alternative is that one or both break-ends
may never be rejoined to another break-end incomplete rejoining or an unrepairable DSB. The breakage and incorrect
reunion process, which has also been referred to as binary
3741
misrepair,53,54 gives rises to intra- and interchromosomal and

chromatid aberrations, some of which render cells unable to
produce viable progeny.55
The two major evolutionarily conserved pathways available to rejoin DSB in eukaryotes are homologous recombination HR and nonhomologous end joining NHEJ.56,57 In
the HR pathway, a damaged chromosome is aligned with an
undamaged chromosome that has sequence homology. The
genetic information stored in the undamaged chromosome is
then used as a template for the repair of the damaged chromosome. Because the HR pathway has an undamaged template to guide the repair process, DSB may be rejoined without any loss of genetic information i.e., correct repair. In
contrast, the NHEJ pathway does not require a template with
sequence homology. Instead, a series of proteins bind to the
break-ends of the DSB, some of which have exo- and endonuclease activity that is responsible for the removal of damaged nucleotides, and directly rejoin ligate the break. In
CHO-K1 cells, the rejoining of DSBs formed by the I-SceI
endonuclease produces deletions of 1 to 20 base pair bp.58
Rare deletions up to 299 bp as well as the aberrant insertion
of 45 to 205 bp also occur.58 Ionizing radiation can create
DSBs composed of more than 10 or even 20 lesions,59 and
the NHEJ pathway is unlikely to correctly repair complex
DSBs such as those formed by ionizing radiation.57 That is,
correct rejoining of the DSB formed by ionizing radiation is
unlikely to be synonymous with correct repair restoration of
the original base pair sequence. Instead, correct DSB rejoining most likely produces small-scale mutations deletions,
insertions, or other sequence alterations. A subset of these
small-scale mutations may be lethal. Also, unrepairable or
incompletely repaired DSB may contribute directly or indirectly to cell killing.
Although DSB are widely viewed as the most critical
form of DNA damage produced by radiation, other types of
clustered damage are still potentially significant. For example, several experiments have shown that complex single
strand breaks SSBs or other forms of nonDSB clustered
damage can sometimes be converted into a DSB as a result
of unsuccessful excision repair.60,61 The formation of additional DSB through the unsuccessful excision repair of nonDSB clustered damage may result in additional binary misrepair and, ultimately, increase the yield of lethal exchanges,
or at the very least, increase the yield of small-scale mutations and unrepairable or incompletely repaired DSB. Observations such as these suggest that complex clustered damage,
and repair mechanisms other than NHEJ and HR e.g., excision repair, impact on a cells ability to tolerate radiation
damage see also the review by Wallace62.
Despite the potential appeal of attributing cell death exclusively to the induction and processing of DNA damage,
other biological processes are known to modulate the apparent survival response of cells exposed to ionizing radiation.
For example, low dose,63 microbeam,6466 and medium
transfer experiments67 provide compelling evidence that cells
damaged by radiation communicate with undamaged by
stander cells through factors transmitted through the extracellular matrix68 and through gap junctions.69,70 Intercellular
3742
signals communicated to bystander cells initiate many of the

same effects as direct cellular irradiation,71,72 including
signal-mediated death. It is likely that radiation-damaged
cells also respond to intercellular signals, although their response to these signals may or may not be the same as the
ones exhibited by bystander cells.
II.D. Linear-quadratic survival model
In the LQ survival model, the fraction of cells that survive

absorbed dose D Gy may be conveniently written as
SD = exp D 1 +
GD
/
When interpreted in terms of the induction and processing of

DSB, the D term represents the expected number of lethal
damages arising from DSB formed by the same track hit,
including possibly unrepairable DSB, small-scale mutations
correct but error-prone DSB rejoining and complete or incomplete intratrack binary misrepair. This latter process may
occur when multiple DSB are formed by the same track. The
GD2 term represents the formation of lethal exchanges
through the pairwise interaction of break-ends associated
with DSB created by different tracks, i.e., intertrack binary
misrepair. Because DSB rejoining by HR requires an intact
chromosome with sequence homology to act as a template
for repair, the intra- and intertrack binary misrepair reactions
responsible for the formation of most exchanges are likely
accomplished by NHEJ rather than HR.
The Lea-Catcheside dose-protraction factor G is given
by23
G=
2
D2
t
dtD
tett ,
dtD
t is the instantaneous absorbed dose rate Gy h1

where D
at time th. This dose rate function captures the temporal
pattern of radiation delivery in its entirety, including splitdose and multifraction irradiation schemes. The biophysical
interpretation of Eq. 6 is that break-ends associated with a
DSB created at time t may, if it is not first rejoined to its
correct partner, interact with break-ends associated with a
second DSB produced at time t. The rate of DSB rejoining is
characterized by or, alternatively, the effective half-time
for DSB repair, ln 2 / . In the LQ model, cellular radiation sensitivity is characterized by three biological parameters: Gy1, / Gy, and h1 or h. For the special
case when 1 time to deliver each fraction is short compared
to the repair half-time and 2 the time interval between fractions is large compared to the protraction for n daily fractions is G = 1 / n.22 For situations in which the fraction deliver
time is not short compared to the repair half-time e.g., see
Wang et al.73, the formula
G=
2 x
G1
=
e + x 1,
n nx2
is more appropriate. Here, x ln2T1 / and T1 is the fraction delivery time time to deliver one fraction. The time to
3742
deliver a fraction in IMRT has the potential to have a significant impact on tumor control for tumors with a low / ratio
and a short repair half-time.73
Although the mechanistic basis of the LQ has been challenged and remains open to vigorous debate,7477 the interpretation of the LQ survival model in terms of DSB induction and processing provides an explanation for the linearquadratic dependence on dose of the surviving fraction and,
equally important, provides a conceptual framework to understand the interplay between damage repair processes and
dose protraction dose rate effects. Studies showing that defects in DSB repair mechanisms reduce cell survival directly
implicate DSB processing in survival responses.78,79 In addition, positive correlations among chromosomal damage and
clonogenic survival provide strong circumstantial evidence
that DSB induction and processing is an important cell killing mechanism. In one study, Cornforth and Bedford80 demonstrated that, in plateau-phase cultures of AG1522 normal
human fibroblasts, asymmetric exchanges exhibited a near
one-to-one correlation with the logarithm of the surviving
fraction, as would be predicted by Eq. 5 when D
+ GD2 is interpreted as the expected number of lethal aberrations per cell. Studies such as this one suggest that the
conversion of DSB into lethal exchanges through intra- and
intertrack binary misrepair are an especially important cellkilling mechanism and provides some justification for considering the LQ as more than a purely phenomenological
model.
The mathematical simplicity of the LQ survival model is
both its strength and weakness. Low-dose hyperradiosensitivity and adaptive responses notwithstanding,8183
the strength of the LQ is that trends in clonogenic survival
can often be adequately predicted in vitro and in vivo using a
minimum number of adjustable parameters , / , and or
. However, when contrasted to the underlying biology, the
LQ model can only be viewed as a highly idealized representation of a very complex, time-dependent sequence of
biophysical events and processes. That is, the aggregate effects of many time- and cell-cycle dependent processes are
represented by a single set of time- and cell-cycle independent radiosensitivity parameters, , / , and or . This
oversimplification of the underlying biology implies that the
collective radiation response of a population of cells, even if
this cell population has uniform and stable genetic traits, is
determined by a distribution of values for , / , and
rather than one true set of values. This inherent variability
in cellular radiosensitivity is further amplified by a host of
other patient- or tumor-specific factors.8486
Our ability to accurately estimate a TCP using the LQ
model is ultimately limited by our lack of knowledge of the
most appropriate parameter distributions for , / , and
and by our incomplete knowledge of how well the LQ model
relates dose to clonogenic survival reproductive death. To
reduce the possibility that treatment plans are inappropriately
overoptimized for a specific combination of radiosensitivity
parameters, distributions of values for , / , and should
be used to predict treatment indicators e.g., EUD or BED
and outcomes rather than point estimates.8790 The use of
3743
3743
FIG. 1. Prescription doses for the treatment of prostate cancer equivalent to 37 daily fractions of 2 Gy filled circle. A fraction delivery time T1 parameter
of 10 min is used to compute G1 for all treatments refer to Eq. 7. Solid lines denote prescription doses estimated using the Fowler et al. Ref. 91
radiosensitivity parameters. Dashed lines indicate prescription doses estimated using the Wang et al. Ref. 92 radiosensitivity parameters.
parameter distributions in treatment planning applications

can help quantify uncertainties in treatment outcomes arising
from the inherent variability in radiosensitivity parameters,
as well as from the use of an oversimplified or incomplete
model, such as the LQ, to predict cell killing.
II.E. Biologically guided prescription dose and
individualized treatments
For many reasons, estimates of radiobiological parameters

derived from clinical data often have substantial levels of
uncertainty. To illustrate, an analysis of clinical data for prostate cancer by Fowler et al.91 suggests the following radiosensitivity parameters: = 0.039 Gy1 0.0196 Gy1 ,
0.0587 Gy1, / = 1.49 Gy 1.25 Gy, 1.76 Gy, = 1.49 h
1.42 h , 2.86 h, and Td is effectively infinite. The values in
parentheses are the estimated 95% confidence interval for the
parameters. Because the effective tumor doubling time is effectively infinite, the value of Tlag has no impact on tumor
control calculations. A similar analysis of clinical data for
prostate cancer by Wang et al.92 suggests the alternate
radiosensitivity parameters: = 0.15 0.04 Gy1, /
= 3.1 0.5 Gy, = 0.267 h but less than 1.5 h, Td = 42 days,
and Tlag = 0 days. The observation that two different groups
can analyze similar datasets and arrive at very different sets
of radiosensitivity parameters undermines confidence in the
direct application of biological models for the prediction of
treatment outcomes. For a representative treatment consisting of 37 daily fractions of 2 Gy with T1 = 10 min, the surviving fraction obtained with the Fowler et al. parameters is
1.36 103 and the surviving fraction obtained with the
Wang et al. parameters is 1.53 107. Calculations such as
this one further undermine confidence in our ability to reliably predict treatment outcomes.
Although uncertainties in radiosensitivity parameters appear an overwhelming obstacle for the direct application of
biological models in treatment planning, many issues associated with the inadequacy of biological models and radiosensitivity parameters can be circumvented through the use of
isoeffect calculations. Consider as an example the task of
identifying biologically equivalent prescription doses for the
treatment of prostate cancer. To identify isoeffective total
treatment doses, correct Eq. 5 for repopulation effects and
then take the logarithm and apply the quadratic formula to
obtain
D=
/
1+
2G
/
1+
2G
4Gln S T
/
1+
4G
T
BED +
/
To identify isoeffective prescription doses, S or equivalently

BED ln S / are held constant e.g., S = 1.36 103 for the
Fowler et al. parameters and S = 1.53 107 for the Wang et
al. parameters in Eq. 8 while G 1 / n protraction factor
and T total treatment time are adjusted to reflect other fractionation schedules of interest. Figure 1 shows the biologically equivalent isosurvival prescription doses derived using Eq. 8. Despite the seemingly large differences in
Fowler et al. and Wang et al. parameters and predicted clinical outcome S = 103 or 107, both sets of radiosensitivity
parameters predict quite similar prescription doses for n = 1
to 50. For n = 23 to 50, the isoeffective fraction sizes predicted with both sets of radiosensitivity parameters differ
from each other by less than 3%.
The results shown in Fig. 1 demonstrate that the seemingly large differences in the population-averaged radiosen-
3744
3744
FIG. 2. Potential impact of interpatient heterogeneity on the determination of isoeffective prescription doses. All prescription doses are equivalent to a
reference treatment composed of 37 fractions of 2 Gy filled circles. A fraction delivery time T1 parameter of 10 min is used to compute G1 for all
treatments refer to Eq. 7. The solid and dotted lines are the mean and 95% confidence interval, respectively, derived from estimated prescription doses for
104 patients. Patient radiosensitivity parameters sampled from a uniform probability density function using Monte Carlo methods: 0.1 Gy1 to 0.3 Gy1,
/ 1 Gy to 10 Gy, 0.1 h to 6 h, and Td 30 d to 365 d.
sitivity parameters reported by Fowler et al.91 and Wang

et al.92 have a modest impact on estimates of isoeffective
prescription doses. However, patients treated in the clinic
exhibit different responses to the same prescription dose. To
illustrate the potential impact of this interpatient heterogeneity on estimates of isoeffective prescription doses, consider a
scenario in which radiosensitivity parameters for a hypothetical patient population are randomly distributed in the following range: 0.1 Gy1 to 0.3 Gy1, / 1 Gy to 10 Gy,
0.1 h to 6 h, and Td 30 d to 365 d. This range of parameters encompasses a large portion of the radiosensitivity parameters typical of mammalian cells. To identify an isoeffective prescription dose, a patients radiosensitivity parameters
are sampled using Monte Carlo methods and used to compute S for a reference treatment i.e., 37 fractions of 2 Gy.
Equation 8 is then used to compute isoeffective prescription doses for that patient by changing nG 1 / n and T. A
distribution of isoeffective doses can be constructed by repeating this algorithm for the entire patient population. Figure 2 shows the mean solid line and 95% confidence interval dashed lines on isoeffective prescription doses for 104
patients. For small changes in the number of fractionation
e.g., n = 30 to 40, isoeffective prescription doses are nearly
the same for all patients despite substantial levels of interpatient variability in radiosensitivity parameters. For more radical changes in the number of fractionation e.g., n = 5 or 10,
differences in radiosensitivity parameters among the hypothetical patient population give rise to substantial differences
in the predicted isoeffective prescription doses.
An appealing aspect of using isoeffect methods for treatment planning is that clinical judgment and experience, in
the form of a clinically accepted treatment, is explicitly considered e.g., the filled circles in Figs. 1 and 2. In addition,
Eq. 8 suggests an easy way to manage the inherent risks
associated with altering an accepted treatment, i.e., manage
risk by making small rather than large changes in the number
of daily fractions. The results shown in Figs. 1 and 2 suggest
that a change from 37 fractions of 2 Gy to n = 30 or to n
= 50 may be accomplished with a nominal risk for altered
tumor control. Large changes in an accepted treatment e.g.,
n = 37 to n = 5 or 10 are potentially risky because of interpatient heterogeneity. Large changes in an accepted treatment
are also risky because altering the fractionation schedule may
substantially alter normal tissue toxicity. Small changes in an
accepted treatment plan are also advisable to minimize the
potential for unanticipated consequences associated with the
approximations and simplifications inherent in the LQ model
and other biological factors, e.g., bystander effects and lowdose hyper-radiosensitivity.
In addition to interpatient heterogeneity, radiosensitivity
parameters vary within a tumor, because, for example, of
genomic instability and developmental or functional disturbances in tumor vasculature. Developmental and functional
disturbances in the tumor vasculature produce varying levels
of acute and chronic hypoxia93 and are often associated with
more advanced stages of disease.94,95 Cells irradiated under
hypoxic conditions are much less sensitive to radiation, and
substantially higher doses may be necessary to inactivate tumor cells in regions of a tumor with intermediate or extreme
levels of hypoxia. Intratumor variations in radiosensitivity
parameters enhance the risks associated with making large
changes in clinically accepted treatment plans. However, in-
3745
tratumor heterogeneity is also a potential opportunity to improve treatment outcomes through techniques such as dose
painting by numbers.9698 The objective of dose painting by
numbers is to use biological information to identify a dose
distribution that achieves biological rather than physical
dose conformity.
The attempt to identify a biologically conformal dose distribution can be cast in terms of isoeffect calculations. To
illustrate, imagine that a technique is developed to noninvasively estimate oxygen pO2 levels within a tumor. The
measured pO2 level can then be converted to an oxygen enhancement ratio OER using the formula99 OERi = mK
+ O2i / K + O2i, where m = 2.7, K = 0.25, and O2i denotes the oxygen concentration in the ith region of the tumor.
This OER may be interpreted as the ratio of the dose to the
hypoxic cells to the dose to the aerobic cells required to
produce the same number of DSB per cell,99 and it ranges
from one for fully aerobic cells to about three for extreme
levels of hypoxia. The analyses of Carlson et al.99 suggest
that OER values are nearly independent of cellular radiation
sensitivity i.e., and / .
For cells irradiated in vitro, Carlson et al.99 have shown
that radiosensitivity parameters for hypoxic H and aerobic
A cells are well approximated by H = A / OER and
/ H = OER / A. The relationships among radiosensitivity parameters for aerobic and hypoxic cells imply that the
dose to regions of the tumor with reduced oxygen levels
needs to be increased by a factor equal to the OER to achieve
the same surviving fraction as well oxygenated regions of the
tumor.99,100 Hence, a biologically conformal dose distribution
might be achieved by designing a treatment to deliver dose
Di = OERi D to the ith region of the tumor. Here, D is a
clinically accepted prescription dose.
The suggested approach to individualize a treatment to
correct for the effects of hypoxia will necessitate a validated
technique to determine pO2 levels within the tumor. In addition, the application of the suggested strategy is not without
risk because of uncertainties associated with the formula to
determine the OER. However, the observation that isoeffect
calculations are relatively insensitive to uncertainties in biological parameters e.g., Figs. 1 and 2 suggests that the approach may still provide useful guidance for treatment individualization. The example also illustrates how future
technological developments and knowledge of the molecular
and cellular mechanisms underlying radiation sensitivity
might be translated into the clinic and used to guide the
treatment planning process.
II.F. Research needs for outcome modeling
Outcome modeling is challenging in part because biological parameters used in these models are patient, tumor, and
organ specific101,102 and in part because TCP and NTCP models are highly nonlinear. That is, seemingly small, patientspecific differences in physical or biological factors can result in very different clinical outcomes for treatment plans
that are otherwise identical. As illustrated by the very different radiosensitivity parameters reported by Fowler et al.91
3745
and Wang et al.92 for prostate cancer, estimates of biological

parameters derived from clinical data are averaged over a
patient population and often have a high degree of uncertainty. However, as Figs. 1 and 2 illustrate, uncertainties in
biological parameters can be substantially mitigated through
the use of isoeffect calculations. Regardless, the path to improved modeling of treatment outcomes will most likely require a multifaceted, iterative research program that includes
better documentation of dosimetric and biological information for individual patients as well as additional effort to
improve the accuracy of dose delivery for a variety of treatment modalities. As suitable clinical datasets accumulate, it
may become possible to extract patient-specific radiosensitivity parameters for use in outcome modeling. Even incomplete biological information e.g., information on pO2 levels
has the potential to substantially enhance the treatment planning process by guiding the selection of appropriate, patientspecific dose distributions.
III. CLINICAL IMPLEMENTATION OF BGRT
Although isoeffect calculations provide useful techniques
to guide the RT planning process even without patientspecific information, the full benefits of BGRT can only be
realized in the clinic when new methods and technologies
become available to assess and predict patient-specific radiation responses. The effort to implement biologically optimal
treatment plans within the clinic will also require research to
address practical issues associated with the accurate delivery
of spatially and temporally complex dose distributions. In
this section, we briefly discuss using biological images to
acquire patient-specific biological information for BGRT. A
three-phase strategy for the clinical implementation of BGRT
is proposed. Other new technologies not discussed here, such
as DNA microarrays,103,104 also have substantial potential to
aid in the clinical implementation of BGRT.
III.A. Extracting radiosensitivity parameters from
biological imaging
Recent advances in biological and functional imaging

have significantly increased the likelihood that BGRT will be
implemented in the clinic in the next decade. In particular,
positron emission tomography PET and functional magnetic resonance imaging fMRI have become widely used in
the clinic in recent years. In addition to 18F-labeled fluorodeoxyglucose FDG PET imaging, which is the most commonly used tracer and is likely associated with local radiosensitivity including tumor burden and hence of clonogen
density i parameter in Eqs. 4 and 7, tracers of hypoxia
misonidazole, angiogenesis AV3 integrin, glucose metabolism labeled FDG, amino acid metabolism labeled tyrosine, methionine, and tumor cell proliferation labeled
thymidine are being developed.105107 Several fMRI techniques under development are expected to provide tumor
physiological and micro-environmental information, such as
blood flow perfusion or diffusion, angiogenesis, hypoxia,
and pH, and to identify radiation sensitive regions in normal
tissues,105,108110 for several tumor sites including brain,
3746
3746
sion 20/20 paper will be devoted to the important role that

biologic and functional imaging will undoubtedly play in
BGRT.
III.B. Three-phase strategy for the clinical
implementation of BGRT
FIG. 3. Simplified relationships between relative cerebral blood volume

rCBV and solid line or / dashed line for malignant glioma Ref.
114.
prostate, and cervix. Regions of tumor burden can also be

identified through the use of MR spectroscopy MRS.111,112
A growing trend in biological imaging is the incorporation of
multiple imaging modalities into the same system.113
To design biologically conformal, patient-specific treatments, methods are needed to relate information from
patient-specific functional and biological images to radiosensitivity parameters used in NTCP and TCP models e.g., ,
/ , , and . Biological imaging typically provides numerical information that reflects some regional biological
characteristic of the tumor or a critical structure relative to a
well-defined standard. One way to use this information to
guide the treatment planning process is to separate patients
into several groups of patients with similar tumor and/or normal tissue response characteristics. An empirical trial and
error approach could then be used to identify a biologically
optimal treatment regimen for each patient group. An alternative, potentially faster and more conceptually appealing
method is to develop rules to transform regional voxelbased signal information into patient-specific estimates of
the parameters used in outcome modeling e.g., , / , ,
and . Then, isoeffect methods such as the ones outlined in
Sec. II E can be used to guide the selection of patientspecific dose distributions.
Although the effort to correlate information from functional and biological imaging with the radiosensitivity parameters is nontrival, substantial progress in this area is
likely in the next ten years. For example, as reported
recently,114 relative cerebral blood volume rCBV maps can
be obtained with dynamic susceptibility contrast-enhanced
MRI techniques and combined with tumor grade and radiosensitivity information112,115 to delineate spatial distributions of the and / radiosensitivity parameters. Figure 3
shows a simplified relationship between rCBV and radiosensitivity parameters for malignant glioma.114 Additional research is needed to develop and, especially, validate rules for
the conversion of information from functional and biological
imaging into patient-radiosensitivity parameters. A future ViMedical Physics, Vol. 34, No. 10, October 2007
Technologies for RT delivery, such as IMRT, imageguided IMRT IG-IMRT, novel brachytherapy, targeted RT,
and proton and heavy ion therapy, have rapidly advanced
over the last ten years, and this trend will most likely continue for the next decade. Advances in RT delivery, particularly in IG-IMRT, will enable the direct integration of not
only anatomic, but also biological images into treatment
units to guide the delivery process. The 4D space+ time
anatomic and biological imaging using CT, MRI, PET and
other modalities will enable the treatment of moving targets
and substantially reduce or eliminate the inadvertent irradiation of normal tissues. Techniques for charged particle
e.g., protons, helium, and carbon ions therapy will advance
and provide new capabilities in the clinic for delivery of
biologically conformed RT.
Although the physical and biological methods and technologies needed to implement BGRT are advancing quickly
and hold great promise for the future of RT, the potential
costs of implementing some technologies may be found prohibitive for routine use in the clinic. The integration of the
multiple technologies required for BGRT will require extensive clinical validation. In particular, the use of incomplete or
inaccurate biological information to preferentially target selected regions of a tumor at the expense of other tumor regions may jeopardize treatment outcomes. For reasons such
as these, we believe an incremental approach is advisable for
the clinical implementation of BGRT, as illustrated in Fig. 4
and briefly explained below.
III.B.1. Phase I population based treatments

Use anatomic and biological images to define targets
and important organs.
Use historical and empirical knowledge of tumor and
normal tissue responses to select population-based prescriptions corrected for tumor type, stage, location, and
prior and/or concurrent treatment modalities. The prescription may use dose as surrogate and may require
uniform dose within the target.
Identify single or combined RT modality e.g., IMRT,
3DCRT, brachytherapy based on their dose volume
merits and generate plans according to the prescription.
Evaluate the plan using dose volume parameters as the
primary criteria and outcome prediction e.g., EUD,
TCP, and NTCP or the method outlined in Sec. II E as
secondary criteria.
Deliver the planned treatment with IGRT. Document 3D
dose maps delivered to the tumor targets and all important organs for outcome analysis.
3747
3747
FIG. 4. Flowchart illustrating a proposed three-phase

strategy to implement BGRT in the clinic.
Periodically follow up the patient after the treatment to

document outcome e.g., local tumor control and normal tissue toxicities and to acquire anatomic and biological images, and/or tissue assays. Correlate the outcome with information from imaging and/or assays with
the dose delivered. Use these data to refine models
and/or model parameters for radiation response
prediction.
III.B.2. Phase II boost dose to selected tumor

regions
Generate the plan according to the population-based
prescription Phase I.
Identify target regions of the tumor that are potentially
radioresistant based on biological images and/or tissue
assays.
Use outcome modeling and related methods to estimate
the size of the dose boost needed to enhance local tumor
control in radioresistant regions of the tumor.
Develop a treatment plan using proper RT modalities
e.g., IMRT, brachytherapy capable of delivering the
boost close to the required doses without decreasing the
dose to any other region of the tumor and with no or
minimal change in the dose to all the important organs.
Evaluate the plan with outcome prediction models e.g.,
EUD, BED, TCP, and NTCP to estimate any gain in
local tumor control.
Deliver the plan with IGRT and document the 3D dose
delivered to each patient.
Acquire biological images and/or assays periodically
during and after the treatment.
Follow up the patient and analyze the data to refine the

outcome prediction models and/or model parameters.
As the clinical implementation of Phases I and II of
BGRT practice proceeds, other more aggressive treatment
approaches may be considered as advances in imaging, predictive assays, delivery technologies, and confidence in radiosensitivity parameters and outcome-prediction models increase. Examples of more aggressive strategies include
treatment strategies to boost local tumor control, such as
boosting the dose to the selected tumor regions with moderate decreases in the dose to other tumor regions. Or alternatively, the dose to the entire tumor region might be increased
if patient-specific modeling predicts only a nominal chance
normal tissue complications will increase. Potential strategies to enable boosting the entire tumor dose include increasing the dose to selected regions of a treatment-limiting organ
or allowing a smaller increase in the dose to large sections of
one or more organs. Biological imaging or assays may be
used to identify these regions of the organ that are functionally less important.
III.B.3. Phase III individualized adaptive BGRT
Use the validated outcome prediction models developed
in Phases I and II to specify nonuniform dose plans
using patient-specific information from imaging, predictive assays, and outcome prediction models, taking into
account molecular genetics and other prior and concomitant therapy.
After one or more treatment fractions, perform additional imaging or other biomarker studies to assess the
change in radiation response of the tumor and normal
3748
tissue. If changes e.g., regression in tumor volume,

change in the levels of hypoxia, accelerated repopulation in certain regions, and/or deteriorated normal tissue
injury are observed, the treatment plan should be reoptimized to account for these changes. The new plan
should be delivered subsequently. This re-imaging and
replanning process should be repeated as needed considering both benefit and practicality.
Document 3D dose delivered to each patient.
Acquire additional information about treatment outcomes for each patient using biologic imaging and/or
tissue assays.
Relate prior-, during-, and post-treatment data to biological mechanisms.
With such an iterative three-phase process, we believe
that BGRT may be safely introduced into the clinic. The
success of BGRT will ultimately hinge on three major milestones: 1 the further development and validation of outcome models for tumor response and normal tissue complications, 2 the accuracy of patient-specific radiosensitivity
parameters derived from biological imaging and functional
assays, and 3 advances in RT planning and delivery methodology and technology.
IV. SUMMARY
The ongoing and rapid development of RT technology,
such as IMRT, IG-IMRT, novel brachytherapy, targeted RT,
proton and heavy ion therapies, technologies for the characterization of the human genome, and the dazzling advances
in biological and functional imaging modalities have significantly increased the potential usefulness of implementing
BGRT treatment planning in the clinic. Biological image modalities, which are currently being introduced into the clinic,
are expected to improve the definition of tumor targets and
normal structures and provide patient-specific information
that can be used to guide the treatment planning. New imaging modalities will also provide new opportunities to quantify tumor and normal-tissue responses during or after a
course of radiation therapy. The ability to rapidly assess
treatment outcomes with surrogates for long-term diseasefree survival will in turn speed our ability to extract useful
information about the biological effectiveness of alternate
and refined radiation treatments.
The integration of biological imaging, radiation response
relationships, and knowledge of the molecular basis of cancer has the potential to dramatically improve RT outcomes
by tailoring the treatment process to the biological characteristics of specific patients. Combining advanced delivery
techniques, such as IG-IMRT, with the selection of patientspecific, biologically conformal prescription doses is the next
frontier in RT and substantial progress is likely within the
coming decade. We believe the three-phase strategy to implement BGRT in the clinic outlined in this paper has the potential to substantially advance patient-specific treatment
planning while minimizing the risks associated with the introduction of new technologies for patient care. The introMedical Physics, Vol. 34, No. 10, October 2007
3748
duction of patient-specific BGRT into the clinic can be accelerated through a multifaceted research program to
improve biological and functional imaging, predictive assays, and biophysical modeling of treatment outcomes. New
treatment planning tools that incorporate biological objectives into the treatment planning process are also needed.
The accumulation and analysis of clinical data for modern
treatment modalities, and the preclinical and clinical evaluation of new tools and techniques for patient-specific treatment planning is highly desirable.
Although the technologies and techniques needed for the
full implementation of patient-specific treatment planning
may require more than a decade to develop and validate,
even the partial implementation of BGRT treatment planning
has the potential to enhance local tumor control without an
undue increase in damage to normal tissues. The validation
and implementation of robust tools and technologies for the
prediction of patient-specific outcomes also has the potential
to improve the efficiency and costs of delivering RT. Why
deliver a treatment consisting of six to eight weeks of daily
fractions if an alternate plan with the same treatment outcome can be achieved with a plan, such as stereotactic RT or
accelerated hypofractionated RT, that only requires one or
two weeks to deliver? The coming decade will surely be an
exciting time for the practice of BGRT.
ACKNOWLEDGMENT
This work is supported in part by a research grant from
Siemens Oncology Care Systems.
a
Author to whom correspondence should be addressed. Electronic mail:

[email protected]
1
S. Webb, P. M. Evans, W. Swindell, and J. O. Deasy, A proof that
uniform dose gives the greatest TCP for fixed integral dose in the planning target volume, Phys. Med. Biol. 39, 20912098 1994.
2
R. I. Mackay and J. H. Hendry, The modelled benefits of individualizing
radiotherapy patients dose using cellular radiosensitivity assays with inherent variability, Radiother. Oncol. 50, 6775 1999.
3
S. Levegrun, A. Jackson, M. J. Zelefsky, E. S. Venkatraman, M. W.
Skwarchuk, W. Schlegel, Z. Fuks, S. A. Leibel, and C. C. Ling, Risk
group dependence of dose response for biopsy outcome after threedimensional conformal radiation therapy of prostate cancer, Radiother.
Oncol. 63, 1126 2002.
4
K. S. C. Chao, W. R. Bosch, S. Mutic, J. S. Lewis, F. Dehdashti, M. A.
Mintun, J. F. Dempsey, C. A. Perez, J. A. Purdy, and M. J. Welch, A
novel approach to overcome hypoxic tumor resistance: Cu-ATSM-guided
intensity-modulated radiation therapy, Int. J. Radiat. Oncol., Biol., Phys.
49, 11711182 2001.
5
I. Turesson, J. Carlsson, A. Brahme, B. Glimelius, B. Zackrisson, and B.
Stenerlow, Biological response to radiation therapy, Acta Oncol. 42,
92106 2003.
6
A. Brahme, J. Nilsson, and D. Belkic, Biologically optimized radiation
therapy, Acta Oncol. 40, 725734 2001.
7
A. Brahme, Biologically optimized 3-dimensional in vivo predictive
assay-based radiation therapy using positron emission tomographycomputerized tomography imaging, Acta Oncol. 42, 123136 2003.
8
S. M. Bentzen, Radiation therapy: Intensity modulated, image guided,
biologically optimized and evidence based, Radiother. Oncol. 77, 227
230 2005.
9
B. Vanderstraeten, W. Duthoy, W. De Gersem, W. De Neve, and H. Thierens, 18Ffluoro-deoxy-glucose positron emission tomography
18FFDG-PET voxel intensity-based intensity-modulated radiation
therapy IMRT for head and neck cancer, Radiother. Oncol. 79, 249
258 2006.
10
J. O. Deasy, Partial tumor boosts: Even more attractive than theory
3749
predicts? Int. J. Radiat. Oncol., Biol., Phys. 51, 279280 2001.

W. A. Tome and J. F. Fowler, On the inclusion of proliferation in tumour
control probability calculations for inhomogeneously irradiated tumours,
Phys. Med. Biol. 48, N261N268 2003.
12
Y. Cao, C. I. Tsien, V. Y. Nagesh, L. Junck, R. Ten Haken, B. D. Ross, T.
L. Chenevert, and T. S. Lawrence, Clinical investigation survival prediction in high grade gliomas by mri perfusion before and during early stage
of RT, Int. J. Radiat. Oncol., Biol., Phys. 64, 876885 2006.
13
M. Alber, F. Paulsen, S. M. Eschmann, and H. J. Machulla, On biologically conformal boost dose optimization, Phys. Med. Biol. 48, N31N35
2003.
14
C. C. Ling and X. A. Li, Over the next decade the success of radiation
treatment planning will be judged by the immediate biological response
of tumor cells rather than by surrogate measures such as dose maximization and uniformity, Med. Phys. 32, 21892192 2005.
15
H. R. Withers, H. D. Thames, and L. J. Peters, A new isoeffect curve for
change in dose per fraction, Radiother. Oncol. 1, 187191 1983.
16
A. Niemierko, Reporting and analyzing dose distributions: A concept of
equivalent uniform dose, Med. Phys. 24, 103110 1997.
17
E. L. Travis, Organizational response of normal tissues to irradiation,
Semin. Radiat. Oncol. 11, 184196 2001.
18
S. M. Bentzen, Preventing or reducing late side effects of radiation
therapy: Radiobiology meets molecular pathology, Nat. Rev. Cancer 6,
702713 2006.
19
J. M. Allan and L. B. Travis, Mechanisms of therapy-related carcinogenesis, Nat. Rev. Cancer 5, 943955 2005.
20
G. W. Barendsen, Dose-survival curves of human cells in tissue culture
irradiated with alpha-, beta-, 20-kV.x- and 200-kV.x-radiation, Nature
193, 11531155 1962.
21
K. H. Chadwick and H. P. Leenhouts, A molecular theory of cell survival, Phys. Med. Biol. 18, 7887 1973.
22
H. D. Thames, An incomplete-repair model for survival after fractionated and continuous irradiations, Int. J. Radiat. Biol. 47, 319339
1985.
23
R. K. Sachs, P. Hahnfeld, and D. J. Brenner, The link between low-LET
dose-response and the underlying kinetic of damage production/repair
misrepair, Int. J. Radiat. Biol. 72, 351374 1997.
24
B. Emami et al., Tolerance of normal tissue to therapeutic irradiation,
Int. J. Radiat. Oncol., Biol., Phys. 21, 109122 1991.
25
C. Burman, G. J. Kutcher, B. Emami, and M. Goitein, Fitting of normal
tissue tolerance data to an analytic function, Int. J. Radiat. Oncol., Biol.,
Phys. 21, 123135 1991.
26
J. T. Lyman, Complication probability as assessed from dose-volume
histograms, Radiat. Res. 104, S13S19 1985.
27
G. J. Kutcher and C. Burman, Calculation of complication probability
factors for non-uniform normal tissue irradiation: The effective volume
method, Int. J. Radiat. Oncol., Biol., Phys. 16, 16231630 1989.
28
R. Mohan, G. S. Mageras, B. Baldwin, L. J. Brewster, G. J. Kutcher, S.
Leibel, C. M. Burman, C. C. Ling, and Z. Fuks, Clinically relevant
optimization of 3-D conformal treatments, Med. Phys. 19, 933944
1992.
29
X. A. Li, J. Z. Wang, R. D. Stewart, and S. D. DiBiase, Dose escalation
in permanent brachytherapy for prostate cancer: Dosimetric and biological considerations, Phys. Med. Biol. 48, 27532765 2003.
30
B. Sanchez-Nieto and A. E. Nahum, BIOPLAN: Software for the biological evaluation of radiotherapy treatment plans, Med. Dosim. 25,
7176 2000.
31
M. M. Miften, S. K. Das, M. Su, and L. B. Marks, A dose-volume-based
tool for evaluating and ranking IMRT treatment plans, J. Appl. Clin.
Med. Phys. 5, 114 2004.
32
B. Warkentin, P. Stavrev, N. Stavreva, C. Field, and B. G. Fallone, A
TCP-NTCP estimation module using DVHs and known radiobiological
models and parameter sets, J. Appl. Clin. Med. Phys. 5, 5063 2004.
33
C. J. McGinn, R. K. Ten Haken, W. D. Ensminger, S. Walker, S. Wang,
and T. S. Lawrence, Treatment of intrahepatic cancers with radiation
doses based on a normal tissue complication probability model, J. Clin.
Oncol. 16, 22462252 1998.
34
K. E. Rosenzweig, J. L. Fox, E. Yorke, H. Amols, A. Jackson, V. Rusch,
M. G. Kris, C. C. Ling, and S. A. Leibel, Results of a phase I doseescalation study using three-dimensional conformal radiotherapy in the
treatment of inoperable nonsmall cell lung carcinoma, Cancer 103,
21182127 2005.
35
D. Y. Song, S. H. Benedict, R. M. Cardinale, T. D. Chung, M. G. Chang,
11
3749
and R. K. Schmidt-Ullrich, Stereotactic body radiation therapy of lung

tumors: Preliminary experience using normal tissue complication
probability-based dose limits, Am. J. Clin. Oncol. 28, 591596 2005.
36
L. A. Dawson, M. Biersack, G. Lockwood, A. Eisbruch, T. S. Lawrence,
and R. K. Ten Haken, Use of principal component analysis to evaluate
the partial organ tolerance of normal tissues to radiation, Int. J. Radiat.
Oncol., Biol., Phys. 62, 829837 2005.
37
A. J. Hope, P. E. Lindsay, I. El Naqa, J. R. Alaly, M. Vicic, J. D. Bradley,
and J. O. Deasy, Modeling radiation pneumonitis risk with clinical, dosimetric, and spatial parameters, Int. J. Radiat. Oncol., Biol., Phys. 65,
112124 2006.
38
S. Webb and A. E. Nahum, A model for calculating tumour control
probability in radiotherapy including the effects of inhomogeneous distributions of dose and clonogenic cell density, Phys. Med. Biol. 38, 653
666 1993.
39
W. A. Tom and J. F. Fowler, On the inclusion of proliferation in tumour
control probability calculations for inhomogeneously irradiated tumours,
Phys. Med. Biol. 48, N261N268 2003.
40
S. Levegrun, A. Jackson, M. J. Zelefsky, M. W. Skwarchuk, E. S. Venkatraman, W. Schlegel, Z. Fuks, S. A. Leibel, and C. C. Ling, Fitting
tumor control probability models to biopsy outcome after threedimensional conformal radiation therapy of prostate cancer: Pitfalls in
deducing radiobiologic parameters for tumors from clinical data, Int. J.
Radiat. Oncol., Biol., Phys. 51, 10641080 2001.
41
S. L. Tucker, H. D. Thames, and J. M. Taylor, How well is the probability of tumor cure after fractionated irradiation described by Poisson
statistics? Radiat. Res. 124, 273282 1990.
42
S. L. Tucker and J. M. Taylor, Improved models of tumour cure, Int. J.
Radiat. Biol. 70, 539553 1996.
43
Z. Bajzer, Gompertzian growth as a self-similar and allometric process,
Growth, Dev. Aging 63, 311 1999.
44
C. P. Calderon and T. A. Kwembe, Modeling tumor growth, Math.
Biosci. 103, 97114 1991.
45
H. McAneney and S. F. ORourke, Investigation of various growth
mechanisms of solid tumour growth within the linear-quadratic model for
radiotherapy, Phys. Med. Biol. 52, 10391054 2007.
46
M. A. Ebert, Viability of the EUD and TCP concepts as reliable dose
indicators, Phys. Med. Biol. 45, 441457 2000.
47
J. R. Savage, A brief survey of aberration origin theories, Mutat Res.
404, 139147 1998.
48
L. Hlatky, R. K. Sachs, M. Vazquez, and M. N. Cornforth, Radiationinduced chromosome aberrations: Insights gained from biophysical modeling, BioEssays 24, 714723 2002.
49
M. Frankenberg-Schwager, Induction, repair and biological relevance of
radiation-induced DNA lesions in eukaryotic cells, Radiol. Technol. 29,
273292 1990.
50
K. Rothkamm and M. Lbrich, Evidence for a lack of DNA doublestrand break repair in human cells exposed to very low x-ray doses,
Proc. Natl. Acad. Sci. U.S.A. 100, 50575062 2003.
51
D. Frankenberg, H. J. Brede, U. J. Schrewe, C. Steinmetz, M.
Frankenberg-Schwager, G. Kasten, and E. Pralle, Induction of DNA
double-strand breaks by 1H and 4He ions in primary human skin fibroblasts in the LET range of 8 to 124 keV/ m, Radiat. Res. 151, 540549
1999.
52
B. M. Sutherland, P. V. Bennett, H. Schenk, O. Sidorkina, J. Laval, J.
Trunk, D. Monteleone, and J. Sutherland, Clustered DNA damages induced by high and low LET radiation, including heavy ions, Phys.
Medica 17Suppl 1, 202204 2001.
53
S. B. Curtis, Lethal and potentially lethal lesions induced by
radiationA unified repair model, Radiat. Res. 106, 252270 1986.
54
C. A. Tobias, The repair-misrepair model in radiobiology: Comparison
to other models, Radiat. Res. Suppl. 8, S77S95 1985.
55
J. S. Bedford, Sublethal damage, potentially lethal damage, and chromosomal aberrations in mammalian cells exposed to ionizing radiations,
56
S. P. Jackson, Sensing and repairing DNA double-strand breaks, Carcinogenesis 23, 687696 2002.
57
P. A. Jeggo, The fidelity of repair of radiation damage, Radiat. Prot.
Dosim. 99, 117122 2002.
58
F. Liang, M. Han, P. J. Romanienko, and M. Jasin, Homology-directed
repair is a major double-strand break repair pathway in mammalian
cells, Proc. Natl. Acad. Sci. U.S.A. 95, 51725177 1998.
59
V. A. Semenenko and R. D. Stewart, A fast Monte Carlo algorithm to
3750
simulate the spectrum of DNA damages formed by ionizing radiation,

Radiat. Res. 161, 451457 2004.
60
V. A. Semenenko, R. D. Stewart, and E. J. Ackerman, Monte Carlo
simulation of base and nucleotide excision repair of clustered DNA damage sites. I. Model properties and predicted trends, Radiat. Res. 164,
180193 2005.
61
V. A. Semenenko and R. D. Stewart, Monte Carlo simulation of base and
nucleotide excision repair of clustered DNA damage sites. II. Comparisons of model predictions to measured data, Radiat. Res. 164, 194201
2005.
62
S. S. Wallace, Enzymatic processing of radiation-induced free radical
damage in DNA, Radiat. Res. 1505 Suppl, S60S79 1998.
63
H. Nagasawa and J. B. Little, Induction of sister chromatid exchanges by
extremely low doses of -particles, Cancer Res. 52, 63946396 1992.
64
T. K. Hei, L. J. Wu, S. X. Liu, D. Vannais, C. A. Waldren, and G.
Randers-Pehrson, Mutagenic effects of a single and an exact number of
a particles in mammalian cells, Proc. Natl. Acad. Sci. U.S.A. 94, 3765
3770 1997.
65
J. M. Nelson, A. L. Brooks, N. F. Metting, M. A. Khan, R. L. Buschbom,
A. Duncan, R. Miick, and L. A. Braby, Clastogenic effects of defined
numbers of 3.2 MeV alpha particles on individual CHO-K1 cells, Radiat. Res. 145, 568574 1996.
66
K. M. Prise, O. V. Belyakov, M. Folkard, and B. D. Michael, Studies on
bystander effects in human fibroblasts using a charged particle microbeam, Int. J. Radiat. Biol. 74, 793798 1998.
67
C. Mothersill and C. Seymour, Medium from irradiated human epithelial
cells but not human fibroblasts reduces the clonogenic survival of unirradiated cells, Int. J. Radiat. Biol. 71, 421427 1997.
68
C. Mothersill and C. B. Seymour, Cell-cell contact during gamma irradiation is not required to induce a bystander effect in normal human
keratinocytes: Evidence for release during irradiation of a signal controlling survival into the medium, Radiat. Res. 149, 256262 1998.
69
E. I. Azzam, S. M. de Toledo, and J. B. Little, Direct evidence for the
participation of gap junction-mediated intercellular communication in the
transmission of damage signals from alpha-particle irradiated to nonirradiated cells, Proc. Natl. Acad. Sci. U.S.A. 98, 473478 2001.
70
E. I. Azzam, S. M. de Toledo, and J. B. Little, Oxidative metabolism,
gap junctions and the ionizing radiation-induced bystander effect, Oncogene 22, 70507057 2003.
71
C. Mothersill and C. Seymour, Radiation-induced bystander effects: Past
history and future directions, Radiat. Res. 155, 759767 2001.
72
E. J. Hall, The bystander effect, Health Phys. 85, 3135 2003.
73
J. Z. Wang, X. A. Li., W. DSouza, and R. D. Stewart, Impacts of
prolonged dose-delivery time on tumor control: A note of caution for
IMRT, Int. J. Radiat. Oncol., Biol., Phys. 572, 543552 2003.
74
R. K. Sachs and D. J. Brenner, The mechanistic basis of the linearquadratic formalism, Med. Phys. 25, 20712073 1998.
75
M. Zaider, There is no mechanistic basis for the use of the linearquadratic expression in cellular survival analysis, Med. Phys. 25, 791
792 1998.
76
M. Zaider, Sequel to the discussion concerning the mechanistic basis of
the linear quadratic formalism, Med. Phys. 25, 20742075 1998.
77
D. J. Brenner and D. E. Herbert, The use of the linear-quadratic model in
clinical radiation oncology can be defended on the basis of empirical
evidence and theoretical argument, Med. Phys. 24, 12451248 1997.
See also the comment in Med. Phys. 24, 1329 1997.
78
M. Khne, E. Riballo, N. Rief, K. Rothkamm, P. A. Jeggo, and M. Lobrich, A double-strand break repair defect in ATM-deficient cells contributes to radiosensitivity, Cancer Res. 64, 500508 2004.
79
N. Chavaudra, J. Bourhis, and N. Foray, Quantified relationship between
cellular radiosensitivity, DNA repair defects and chromatin relaxation: A
study of 19 human tumour cell lines from different origin, Radiother.
Oncol. 73, 373382 2004.
80
M. N. Cornforth and J. S. Bedford, A quantitative comparison of potentially lethal damage repair and the rejoining of interphase chromosome
breaks in low passage normal human fibroblasts, Radiat. Res. 111, 385
405 1987.
81
B. Marples, B. G. Wouters, S. J. Collis, A. J. Chalmers, and M. C. Joiner,
Low-dose hyper-radiosensitivity: A consequence of ineffective cell cycle
arrest of radiation-damaged G2-phase cells, Radiat. Res. 161, 247255
2004.
82
M. C. Joiner, Induced radioresistance: An overview and historical perspective, Int. J. Radiat. Biol. 65, 7984 1994.
83
3750
M. C. Joiner, P. Lambin, and B. Marples, Adaptive response and induced

resistance, C. R. Acad. Sci. III 322, 167175 1999.
84
S. M. Bentzen, H. D. Thames, and J. Overgaard, Does variation in the in
vitro cellular radiosensitivity explain the shallow clinical dose-control
curve for malignant melanoma? Int. J. Radiat. Biol. 57, 117126 1990.
85
S. M. Bentzen, Steepness of the clinical dose-control curve and variation
in the in vitro radiosensitivity of head and neck squamous cell carcinoma, Int. J. Radiat. Biol. 61, 417423 1992.
86
H. Suit, S. Skates, A. Taghian, P. Okunieff, and J. T. Efird, Clinical
implications of heterogeneity of tumor response to radiation therapy,
Radiother. Oncol. 25, 251260 1992.
87
A. Dasu, I. Toma-Dasu, and J. F. Fowler, Should single or distributed
parameters be used to explain the steepness of tumour control probability
curves? Phys. Med. Biol. 48, 387397 2003.
88
B. Warkentin, P. Stavrev, N. A. Stavreva, and B. G. Fallone, Limitations
of a TCP model incorporating population heterogeneity, Phys. Med.
Biol. 50, 35713588 2005.
89
V. Moiseenko, Effect of heterogeneity in radiosensitivity on LQ based
isoeffect formalism for low alpha/beta cancers, Acta Oncol. 43, 499502
2004.
90
J. Lian and L. Xing, Incorporating model parameter uncertainty into
inverse treatment planning, Med. Phys. 31, 27112720 2004.
91
J. F. Fowler, R. Chappell, and M. Ritter, Is alpha/beta for prostate tumors
really low? Int. J. Radiat. Oncol., Biol., Phys. 50, 10211031 2001.
92
J. Z. Wang, M. Guerrero, and X. A. Li, How low is the alpha/beta ratio
for prostate cancer? Int. J. Radiat. Oncol., Biol., Phys. 55, 194203
2003.
93
B. G. Wouters and J. M. Brown, Cells at intermediate oxygen levels can
be more important than the hypoxic fraction in determining tumor response to fractionated radiotherapy, Radiat. Res. 147, 541550 1997.
94
M. Hckel, K. Schlenger, B. Aral, M. Mitze, U. Schaffer, and P. Vaupel,
Association between tumor hypoxia and malignant progression in advanced cancer of the uterine cervix, Cancer Res. 56, 45094515 1996.
95
K. Sundfr, H. Lyng, and E. K. Rofstad, Tumour hypoxia and vascular
density as predictors of metastasis in squamous cell carcinoma of the
uterine cervix, Br. J. Cancer 78, 822827 1998.
96
D. Thorwarth, S. M. Eschmann, F. Paulsen, and M. Alber, Hypoxia dose
painting by numbers: A planning study, Int. J. Radiat. Oncol., Biol.,
Phys. 681, 291300 2007.
97
M. Alber, F. Paulsen, S. M. Eschmann, and H. J. Machulla, On biologically conformal boost dose optimization, Phys. Med. Biol. 482, N31
N35 2003.
98
S. M. Bentzen, Theragnostic imaging for radiation oncology: Dosepainting by numbers, Lancet 62, 112117 2005.
99
D. J. Carlson, R. D. Stewart, and V. A. Semenenko, Effects of oxygen on
intrinsic radiation sensitivityA test of the relationship between aerobic
and hypoxic linear-quadratic LQ model parameters, Med. Phys. 33,
31053115 2006.
100
J. Z. Wang, X. A. Li, and N. A. Mayr, Dose escalation to combat hypoxia in prostate cancer, a radiobiological analysis of clinical data, Br. J.
Radiol. 79, 905911 2006.
101
M. Guerrero and X. A. Li, Deriving radiobiological parameters from a
large number of clinical studies for radiotherapy treatment planning of
early-stage breast cancer, Phys. Med. Biol. 48, 33073326 2003.
102
X. S. Qi, C. J. Schultz, and X. A. Li, An estimation of radiobiologic
parameters from clinical outcomes for radiation treatment planning of
brain tumor, Int. J. Radiat. Oncol., Biol., Phys. 64, 15701580 2006.
103
Y. F. Wong, D. S. Sahota, T. H. Cheung, K. W. Lo, S. F. Yim, T. K.
Chung, A. M. Chang, and D. I. Smith, Gene expression pattern associated with radiotherapy sensitivity in cervical cancer, Cancer J. 12, 189
193 2006.
104
T. Watanabe, Y. Komuro, T. Kiyomatsu, T. Kanazawa, Y. Kazama, J.
Tanaka, T. Tanaka, Y. Yamamoto, M. Shirane, T. Muto, and H. Nagawa,
Prediction of sensitivity of rectal cancer cells in response to preoperative
radiotherapy by DNA microarray analysis of gene expression profiles,
Cancer Res. 66, 33703374 2006.
105
S. Apisarnthanarax and K. S. C. Chao, Current imaging paradigms in
radiation oncology, Radiat. Res. 163, 125 2005.
106
S. Bujenovic, The role of positron emission tomography in radiation
treatment planning, Semin Nucl. Med. 34, 293299 2004.
107
H. Schoder and S. M. Larson, Positron emission tomography for prostate, bladder, and renal cancer, Semin Nucl. Med. 34, 274292 2004.
108
J. L. Ulmer, C. V. Salvan, W. M. Mueller, H. G. Krouwer, G. O. Stroe, A.
3751
Aralasmak, and R. W. Prost, The role of diffusion tensor imaging in

establishing the proximity of tumor borders to functional brain systems:
Implications for preoperative risk assessments and postoperative outcomes, Technol. Cancer Res. Treat. 3, 567576 2004.
109
F. Z. Yetkin and D. Mendelsohn, Hypoxia imaging in brain tumors,
Neuroimaging Clin. N. Am. 12, 537552 2002.
110
R. J. Gillies, N. Raghunand, G. S. Karczmar, and Z. M. Bhujwalla, MRI
of the tumor microenvironment, J. Magn. Reson Imaging 16, 430450
2002.
111
J. Kurhanewicz, M. G. Swanson, S. J. Nelson, and D. B. Vigneron, Combined magnetic resonance imaging and spectroscopic imaging approach to
molecular imaging of prostate cancer, J. Magn. Reson Imaging 16, 451
463 2002.
112
S. J. Nelson, E. Graves, A. Pirzkall, X. Li, A. Chan, D. B. Vigneron, and
T. R. McKnight, In vivo molecular imaging for planning radiation
3751
therapy of gliomas: An application of 1H MRSI, J. Magn. Reson Imaging 16, 464476 2002.
113
W. R. Hendee, Biomedical imaging research opportunities workshop II:
A summary of findings and recommendations, Med. Phys. 32, 2484
2486 2005.
114
G. P. Chen, E. Ahunbay, C. J. Schultz, and X. A. Li, Development of an
inverse optimization package to plan non-uniform dose distributions
based on spatially inhomogeneous radiosensitivity extracted from biological images, Med. Phys. 34, 11981205 2007.
115
K. M. Schmainda, S. D. Rand, A. M. Joseph, R. Lund, B. D. Ward, A. P.
Pathak, J. L. Ulmer, M. A. Baddrudoja, and H. G. J. Krouwer, Characterization of a first-pass gradient-echo spin-echo method to predict brain
tumor grade and angiogenesis, AJNR Am. J. Neuroradiol. 25, 1524
1532 2004.
Radiological interpretation 2020: Toward quantitative image assessment

John M. Boonea
University of California, Davis, UC Davis Medical Center, 4860 Y Street, Ellison Building Suite 3100,
Sacramento, California 95817
Received 6 July 2007; revised 30 August 2007; accepted for publication 31 August 2007;
published 15 October 2007
The interpretation of medical images by radiologists is primarily and fundamentally a subjective
activity, but there are a number of clinical applications such as tumor imaging where quantitative
imaging QI metrics such as tumor growth rate would be valuable to the patients care. It is
predicted that the subjective interpretive environment of the past will, over the next decade, evolve
toward the increased use of quantitative metrics for evaluating patient health from images. The
increasing sophistication and resolution of modern tomographic scanners promote the development
of meaningful quantitative end points, determined from images which are in turn produced using
well-controlled imaging protocols. For the QI environment to expand, medical physicists, physicians, other researchers and equipment vendors need to work collaboratively to develop the quantitative protocols for imaging, scanner calibrations, and robust analytical software that will lead to
the routine inclusion of quantitative parameters in the diagnosis and therapeutic assessment of
human health. Most importantly, quantitative metrics need to be developed which have genuine
impact on patient diagnosis and welfare, and only then will QI techniques become integrated into
the clinical environment. 2007 American Association of Physicists in Medicine.
DOI: 10.1118/1.2789501
Key words: quantitative imaging, cancer, computed tomography CT, positron emission tomography PET, single photon emission computed tomography SPECT, magnetic resonance imaging
MRI
I. CURRENT STATUS
Radiological diagnosis performed by radiologists in the current era is almost completely qualitative. Qualitative is not a
bad thingthe diagnostic utility of medical imaging, with
qualitative interpretation by an experienced radiologist, represents one of the greatest advances in modern medicine1 of
the 20th century. In the 1970s, when currently practicing
senior radiologists were trained, the practice of radiology
focused primarily on screen-film radiography. For example,
in 1975, computed tomography CT was fledgling, rectilinear scanners were only slowly being replaced by projection
nuclear scanners using Anger logic, B-mode ultrasound imaging was grainy and analog, and magnetic resonance imaging MRI was a crude sketch in Lauterburs laboratory
book. During this era, the hallmark of a radiologists talents
lay in his or her diagnostic skills in radiographic interpretation, fundamentally a qualitative experience. While some basic quantitative measures such as tumor width or anatomically relevant angles were and are still made on radiographs
using a ruler or protractor in scoliosis, for example, the fact
is that screen-film radiography does not lend itself naturally
to robust quantitative methods.
II. ROLE OF QUANTITATIVE IMAGING IN THE
FUTURE
The practice of radiology today has completely changed
relative to 20 years ago. Radiography, even digital radiography, is declining as a fraction of radiologists workload in the
United States, with tomographic modalities such as CT,
4173
Med. Phys. 34 11, November 2007
MRI, ultrasound, SPECT, and PET becoming the predominant modalities used in imaging todays patients. The potential of quantitative imaging QI from tomographic data sets
is clearly superior to that of projection images. The modern
radiology department, in the academic or private practice
settings, is completely digital and PACS based. Digital tomographic volume data sets lend themselves more readily to
quantitative analysis than screen-film radiographs.
Radiology residents and fellows in training now have
spent considerable time during their youth with a joystick in
their hands playing video games, defending earth from the
Covenant, high jacking sports cars, or shooting up Liberty
City. This generation of radiologists is primed and ready to
do real time fly-bys of patients high resolution 3D anatomy.
Unfortunately, the performance of current $100,000 medical
PACS workstations in radiology departments pale in comparison to the graphical capabilities of modern high-end
video game hardware, which typically incorporate both parallel and pipeline processing techniques to accelerate graphic
visualization.
Modern imaging systems produce images which are rich
in quantitative information. While most tomographic modalities, including ultrasound imaging and MRI, possess solid
quantitative spatial information, CT arguably is at least today the modality which produces image data sets which are
the richest in quantitative anatomical information. The reasons for this are threefold: 1 CT in typical clinical practice
has the highest spatial resolution smallest voxel volume
with the MTF to back it up of tomographic modalities, mak-
0094-2405/2007/3411/4173/7/$23.00
4173
4174
John M. Boone: Quantitative imaging 2020
ing distance, area, and volume measurements fundamentally

more precise. Furthermore, linear distances and thus areas
and volumes due not suffer from geometrical distortion like
ultrasound speed of sound or MRI susceptibility images
doindeed, if the dimensional fidelity of a CT scan is compromised by something like patient motion, artifacts result
which are easily recognized by radiologists. 2 The gray
scale data in CT Hounsfield units HUs is quantitative,
unlike any other modality in typical clinical use. The normalization of the CT projection data to incident x-ray beam intensity prior to backprojection results in image data which
are more reflective of the patients anatomy and less a function of the particulars of the modality acquisition characteristics. Basic tissue characterization24 comes free in CT.
3 Finally, the injection of iodinated or other contrast
agents in CT provides functional information pertinent to
blood flow, perfusion, vascular volume, and permeability,
and this functional data also lends itself readily to quantification a corollary to 2, because the change in HUs is
linearly proportional to the concentration of the iodine and
volume of contrast agent in each voxel. This makes quantitative assessment of things like time density curves or volume flow more straightforward and fundamentally quantitative. CT is currently being exploited for clinical applications
for its gray scale accuracy. CT is capable of quantifying the
fat content in the liver,5 which is useful in determining transplant appropriateness. CT is used routinely to assess the calcium content in single pulmonary lesions, to determine likelihood of malignancy.6,7 Bone mineral density can be
evaluated accurately using CT to determine fracture risk.8,9
While CT currently may have a small edge in terms of
anatomical spatial quantitative integrity, positron emission
tomography PET and to a lesser extent single photon
emission computed tomography SPECT have sensitivity
unrivaled in modern human imaging. The fundamental detection event in both PET and SPECT records the decay of a
single atom inside the patients body. While it is true that
many thousands PET or millions SPECT of such decays
are needed to produce a diagnostic image, metabolic activity
18FDG PET or tracer uptake SPECT is not only extremely sensitive, but can also be quantitative when imaging
protocols are standardized and calibration techniques are adhered to. The history of nuclear medicine has been rich with
quantitative imaging procedures,10,11 dating back to the
1960s,12,13 and modern nuclear imaging will almost certainly
improve in terms of diagnostic utility as quantitative methods are developed further e.g., attenuation correction methods in PET/CT are just the start. The sensitivity and functional aspects of nuclear imaging suggest that quantitative
imaging techniques will have a vast role to play in a number
of disease areas.
MR imaging already provides excellent anatomical information, and for some applications such as neuroimaging, the
superior contrast resolution of MR over CT makes it useful
for quantifying spatial length, area, and volume parameters
of tumors or anatomical structures. The volumetric capabilities of MRI have already been exploited for a number of
clinical applications.1417 In addition, for serial follow-up of
Medical Physics, Vol. 34, No. 11, November 2007
4174
anatomical metrics, the lack of ionizing radiation that MRI

offers makes it the modality of choice for many clinical applications, especially outside of cancer imaging. Depending
on the pulse sequence used in MRI, the images are quantitatively but nonlinearly a function of T1 and T2 relaxation
phenomenon and are also dependent on the three dimensional proton density distribution in the patient. Machine dependencies such as coil sensitivity also impact the magnitude
of the gray scale value in MRI images. With some calibration
work and judicious use of pulse sequences, MRI techniques
can be devised in which the gray scale is quantitatively
meaningful.18,19
MR also has the unique ability to accurately measure vascular flow, without the use of contrast agents, by capitalizing
on the phase component of the RF signal. Whereas there is
ample use of flow effects to delineate the anatomical distribution of blood vessels magnetic resonance angiography
MRA, MRA for quantitative assessment of vascular flow
is not used routinely for clinical diagnosis. Calibration procedures using fluid flow phantoms may be necessary to calibrate and validate blood flow ml/s metrics.
The physics of ultrasound imaging is complicated and images produced by this modality are characterized by directional dependencies, due to the angular dependency of sound
reflection and refraction at tissue interfaces. Attenuation correction as a function of depth of penetration is in many cases
ad hoc e.g., time gain control curve. However, bulk tissue
properties such as the Youngs modulus can be inferred with
sufficient calibration, and elastic20 as well as viscoelastic21
properties of tissue can be measured. In cardiac and obstetric
ultrasound imaging, use of distance measurements are frequently used, with the classic example of assessing fetal age
by use of the crown-rump length.22,23 Doppler techniques are
also capable of measuring vascular flow profiles, but in current usage it is the relative flow that is used for diagnosis.
Doppler flow metrics are highly dependent upon the angle
that the sound beam makes with the moving medium, and
thus calibration techniques for absolute flow measurements
would need to accommodate geometrical parameters made in
each in vivo measurement.
Despite the theoretical potential of imaging systems to
produce quantitative images, current scanner development
has focused mostly on delivering image contrast and high
signal to noise performance.2427 With the era of quantitative
imaging upon us, scanner manufacturers will likely want to
include quantitative assessment in the design criteria for
scanner hardware and software, and these new criteria may
pose new challenges for the medical imaging industry. For
example, proprietary postprocessing techniques are used
across the modalities to produce better looking images, but
in many cases these algorithms compromise the quantitative
accuracy of the image data. It will be a challenge for equipment vendors to reveal or at least suppress their image processing techniques such that consistent quantitative results
can be achieved across manufacturers platforms. This is an
important concern because in hospital settings where more
than one model of scanner exists, it is very difficult to sched-
4175
ule the patient to be imaged in a follow-up procedure

months after a previous scan on exactly the same scanner.
III. THE FUTURE IS RICH WITH QI
APPLICATIONS
The future of quantitative imaging is approaching, and
approaching fast. While qualitative radiological diagnosis
will always be a significant part of what radiologists do in
the future, the growing requirement of evidence based medicine will create an environment where diagnoses are necessarily reinforced with quantitative data, especially when the
diagnosis leads to expensive therapeutic options. The greatest driver of quantitative imaging today is in cancer therapy
trials,28,29 where quantitative measurements of tumor characteristics e.g., growth, metabolism, uptake are needed to determine tumor response to chemotherapeutic or other therapeutic actions. The cost of bringing a new drug to market is
very high, and therefore routine imaging costs are small in
comparison. It has already been demonstrated30 that in some
cases quantitative imaging techniques can accurately predict
tumor response and therapeutic outcome, much earlier than
longer term outcome measures such as tumor recurrence or
death. Early predictors of a drugs efficacy can reduce the
time needed for clinical trials, which in turn can substantially
reduce the cost of the trial.
While quantitative measurement of tumor response is now
used principally in patients who are on clinical trials, ultimately the quantitative imaging tools and communication
strategies developed between radiologists and oncologists31
for clinical trials will be exploited for the standard care of
most cancer patients, even those not participating in clinical
trialsbut why? Drug companies typically bear the cost of
clinical trials, and thus it is in their own self interest to use
tools such as QI techniques to save the costs of clinical trials.
But if the early determination of whether a patient is benefiting from a course of therapy is both cost effective and beneficial to the patient, then will not those who pay for health
care want the receive the same efficiencies? Absolutely. Accurate quantitative data, including that provided by imaging
procedures, represents a rich source of objective clinical data
in a world in which evidence-based procedures3236 are reimbursed. Therefore, imaging has a huge potential role to
play in a future landscape where medical diagnostic and
treatment decisions are governed by this higher standard.
Under payer mandates such as pay for performance,37,38
radiologists will become accustomed to performing routine
quantitative assessments on digital image data in the cancer
setting, and this gradual change in mindset will be accompanied and reinforced by ever increasing algorithmic
sophistication,39,40 which will make the generation of quantitative metrics more efficient clinically, but also will make
these quantitative metrics more accurate. The most basic
metric in cancer imaging is the assessment of the change in
tumor volume between scans. Beyond cancer, the use of
quantitative imaging techniques will extend to other disease
settings. In musculoskeletal radiology, kinetic parameters
which describe joint motion will help to quantify joint
4175
health, and direct causal estimates of bone strength e.g.,

relying on bone density and architecture,4143 coupled with
an engineering model will replace indirect correlative assessments such as bone mineral density analysis.4446 Quantitative imaging procedures will have an impact in the assessment of neurological disorders and neurodegenerative
diseases, for example by quantifying plaque dynamics in
multiple sclerosis,4750 measuring amyloid plaque burden in
Alzheimer disease,51,52 and assessing cognitive response in
dementia measured using PET, fMRI, or MRS
techniques.53,54 Imaging will play an enormous role in the
quantitative assessment of abdominal organ function-islet
cell metabolism in the pancreas, glomerular filtration rate in
the kidneys,19 assessment of organ volume55 with bilateral
comparisons for paired organs will prove useful in the diagnosis of metabolic and infectious diseases, as well as cancer.
There are numerous quantitative parameters that can be
assessed on images which relate to heart function and vascular performance, such as wall motion analysis, stroke volume, ejection fraction, and percent stenosis. The recent advent of cardiac CT scanners suggests that some of these
parameters can be characterized using the four dimensional
x, y, z, and t capabilities of this modality.56
The above examples are but the tip of the iceberg, and the
use of the quantitative capabilities of medical imaging is
limited only by the creativity of the physicists and physicians
involved in radiology. It should be noted that QI is really a
superset of something already quite familiar to radiologists
and medical physiciststhat being computer aided detection
CAD. With CAD, a computer searches the mathematical
topography of the digital image, and when it finds specific
mathematical signatures, it responds by setting a flag or
painting an arrow. Fundamentally, the detection event occurs
owing to specific quantitative properties in that region of the
image that was flagged. In CAD, however, usually a number
of quantitative metrics are determined spatially e.g., edge
gradient, jaggedness, texture, etc. and these many intermediate numerical parameters are fed into a higher order discriminator such as a neural network, which processes the
inputs and computes an output, and if that output exceeds
some threshold value, then a detection event is flagged.
Therefore, CAD is fundamentally a self directed quantitative
imaging procedure, however, the overall metric a detection
event is based upon statistical classifiers or a training experience over a large database of image regions with known
classes. Although CAD is a fundamentally quantitative tool
that radiologists especially in mammography are familiar
with, for brevity, I am limiting the scope of this essay to
quantitative metrics that have physical meaning to the interpreting physician, such as area, volume, uptake, or metabolism.
IV. INTEGRATING QI INTO RADIOLOGICAL
DIAGNOSIS
Medical physicists are ideally situated to assist, promote,
and develop techniques for increasing the utility of quantitative imaging techniques in radiological interpretation. Physi-
4176
cists can and will play an important role in understanding the

quantitative potential of medical imaging procedures, but
converting quantifiable metrics to medically useful diagnostic information will in general require collaboration with radiologists or other physicians. Because of the visually based
manner in which radiologists-in-training learn their craft, in
many cases they may not immediately realize that an interpretive task could benefit from quantitative data taken from
the image.
The ultimate utility of quantitative imaging depends on a
number of important factors: a The clinical application for
which quantitative information may be useful in patient care,
and how well the quantitative metric correlates to disease
status, b the standardization and consistent use of acquisition protocols, c the logistics of calibrating the scanners to
deliver the required quantitative information, d the image
analysis tools that are used to determine the quantitative metrics, e the willingness of radiologists to work with these
tools and adopt the use of quantitative metrics into their diagnostic workup, and f the ability of the radiology information system RIS or picture archive and communication
system PACS to archive and distribute quantitative metrics
in specific data formats for use in other databases. I will
discuss each of these factors in the context of a straightforward and concrete clinical example, that of using CT for the
evaluation of tumor growth. Although the simple determination of tumor dimensions on CT should in principle be
straightforward, the state of the art implementation of this in
the majority of radiology practices is surprisingly primitive.
IV.A. Clinical application
Cancer patients with solid tumors undergoing chemotherapy are often subjected to serial CT studies with or without contrast injection to evaluate whether or not the size of
the lesion is changing over time. This type of study requires
that tumor size be compared at different time points during
the treatment. While some would argue that CT is a crude
metric to determine tumor viability compared to PET or
MRS or other imaging biomarkers, it is nevertheless the
most ubiquitous QI measure used today and therefore serves
as a good example.
IV.B. Acquisition protocol
The imaging physicist can and should be involved in the

development of CT protocols for quantitative assessment.
Radiation dose issues should be considered in the context of
the age of the patient and the type and location of the tumor.
Knowledge of the patients treatment plan is important, in
particular whether or not the patient is considered palliative
and whether they will be receiving radiation therapy at the
tumor site, after chemotherapy. Obviously, the CT protocols
for tumor assessment in pediatric patients who are considered curable and who are not receiving radiation therapy
should be as low in dose as reasonable, while the protocols
for CT evaluation of tumors in older patients who will be
undergoing radiation therapy may be less stringent in terms
of radiation dose. For the evaluation of the response of a
4176
specific tumor type e.g., hepatocellular carcinoma to a specific drug, the time between the base line scan and a given
follow up scan should be controlled fixed to a specific value
with small leeway, such as 3 months plus or minus 1 week.
A CT protocol may scan the entire abdomen of the patient
using thicker CT sections at lower mA s values and therefore lower radiation levels for the same noise. For better
accuracy, the scan protocol could use thinner sections and
higher mA s near the anatomical site of known tumors to
produce lower noise, higher fidelity images. Contrast injection, if used, should be consistent in terms of type, volume
injected, injection rate, and delay between injection and imaging.
There are obviously a whole host of other parameters that
can and should be considered and set, because consistency is
needed between scans to achieve accurate quantitative results. Only a few acquisition parameters are discussed above,
in the interest of brevity.
IV.C. Image calibration
What is the variability in CT HUs over a period of

months? Is it possible for your institution to really control
which CT scanner is used to image the patient for follow up
studies? If not, how does the HU and spatial accuracy pixel
size vary between the brand X scanner in the main hospital
and the brand Y scanner in the basement? Does this matter?
Well, of course you will not know unless you measure it, and
this then suggests the use of a phantom. If the protocol involves the injection of iodine, then this suggests that the
phantom should have iodinated regions for evaluation as
well. Iodinated regions have the potential for greater variability in HU response than differences in tissue density due
differences in spectral properties and beam hardening corrections between scanner types. Why does HU accuracy matter
if you are just measuring edges of the tumor? The visual
perception of the edge of a tumor is dependent on the
Window/Level setting in CT, and thus if constant W/L values
are to be used, then the HU of the tumor and its surround
should be reproducible as well.
IV.D. Image analysis tools
The RECIST criteria57 call for unidimensional measurements of a tumor at its widest point in-plane and there are a
set of other criteria required when multiple tumors are
present. Although the RECIST standard should be used short
term on NCI sponsored clinical trials, with the thin section
CT imaging capabilities of modern multislice CT scanners,
there is a strong sentiment in the QI community based on
simple reasoning that volume measurements are more accurate than linear measurements.
Most PACS workstations have basic measurement tools
such as distance and area, but these are often tedious and
time consuming to use. While there are many efforts underway to build easier and more automatic tumor segmentation
tools, getting these tools onto a given PACS system is not
always easy, and if these tools are to be used to affect patient
care, then only FDA approved software packages can be
4177
used. Radiologists will only embrace image analysis tools if

they are easy to use, intuitive, allow margin editing, and have
a high degree of automation to make them quick. Clearly,
there is a lot of work that is necessary to bring these features
in line with the isotropic capabilities of modern CT58 and
with the increasing clinical caseload that most radiologists
are experiencing. Adding to the challenges of consistent
quantitative assessment, the variability between scanners,
workstation hardware, and PACS decision support tools is
large. The imaging community has begun the process of developing benchmarking tools for radiologist interpretation.59
The National Cancer Institute has led efforts to develop databases which can be used for validating and standardizing
the performance of image analysis software,60 and this is a
necessary first step along the long road of demonstrating
consistent quantitative performance across analysis packages
and the physicians who use them.
IV.E. Radiologist attitude
Radiologists typically reject new mandates which slow

them down and the evaluation and reporting of quantitative
results in their dictation typically takes longer than the free
flow of words which is the hallmark of subjective radiological interpretation. It would be easy to say that the onus is on
radiologists to adapt to the increasing number of QI techniques that are reported in the literature, but this simply
avoids the reality of modern radiology practice in the United
States. I contend, in fact, that the onus is on us, medical
physicists and other researchers, to develop QI techniques
which are clinically meaningful and genuinely improve radiological diagnosis, and not just complicate it. The use of
quantitative measures needs to be obviously better to the
clinical radiologist for adoption. Some quantitative metrics
will be perceived to be better than their subjective counterparts on face value, such as the quantitative assessment of
tumor growth30 or kidney function.61 Most QI techniques
will require comprehensive evidence to convince practitioners that their use will benefit patient care and thus their
practice. Therefore, I encourage all those investigating QI
techniques to design experiments either prospective or retrospective which demonstrate superior, clinically relevant
outcomes with statistical significance. To encourage QI practices into the evidence-based medicine environment, we need
to promote and deliver evidence-based research. Of course
all research should be based on evidence, but in the context
of QI techniques, the evidence should be in clinical terms
that physicians understand and consider meaningful. A new
QI technique may be 50% more accurate in the assessment
of 18FDG uptake, but a more meaningful metric to practitioners is to demonstrate that this new technique predicts 5 year
survival 50% better. It will be easy for a radiologist to adapt
a new QI technique if he or she is convinced that this quantitative approach is better for patients.
Furthermore, quantitative information determined either
automatically or semiautomatically from images needs to be
integrated smoothly into the interpretive environment. Instead of simply providing a quantitative value void of conMedical Physics, Vol. 34, No. 11, November 2007
4177
text, the quantitative result which could be tumor volume,

adrenal size asymmetry, or kidney glomerular filtration rate
GFR, etc. should be displayed graphically with other contextual information. For example, tumor volume could be
automatically plotted as a function of time incorporating
previous volume measurement information, adrenal organ
sizes could be bar-charted next to each other to demonstrate
asymmetry graphically, and GFR could be plotted with the
90% and 99% normal ranges shown on the graph. Placing
quantitative information into context to seamlessly aid in the
interpretation is an approach that will accelerate and promote
the incorporation of these numerical data into routine diagnosis. Quantitative information can and should also be displayed in real time during interventional procedures
percent reduction in perfusion after a chemoembolization
procedure, real time MRI assessment of tissue volume ablated by high intensity ultrasound, or increase in absolute
blood flow after placing a vascular stent.
The ultimate incentive for radiologists to incorporate QI
into their diagnostic report is financial. The U.S. government
has used financial incentives to encourage the use of new
technology in the past, as evidenced by the increased Medicare reimbursement for digital mammography compared to
screen film mammography, and for the additional reimbursement in mammography when computer aided diagnostic
tools are used. Similar incentives would expedite the use of
QI in the appropriate disease settings.
IV.F. PACS and RIS interfaces
It is incumbent upon software engineers and PACS vendors to provide hooks into the display software that comprises a PACS program which allow researchers and others
to add QI software packages to the standard PACS display
capabilities. Only if QI techniques can be bundled into an
intuitive interface on the PACS console will QI techniques
become adopted by radiologists. The software will need to be
robust and operate in a time-efficient manner. Obviously,
FDA approval for these packages is required for routine
clinical use, but hooks into PACS software for QI research is
an important start toward the long term adoption of quantitative measures in imaging.
Reporting quantitative metrics such as SUV or tumor volume in the free text portion of a radiologists report is inadequate and impractical. Radiology information system RIS
providers need to provide the ability to define special fields
in the radiologist report, where specific quantitative information can be located. Examples include tumor volume or tumor SUV value, for an array of tumors a patient might have.
These data fields need to be accessible by other databases, so
that tumor growth rate etc. can be automatically computed
for a given patient by the push of a button. For example, one
cancer patient could have five distinct tumors and four different CT scans over time. Hand computing 15 different volume changes from the free text reports is too time consuming
and is simply not feasible for routine care. This is what databases and computers are supposed to do, convert menial
labor to a push button procedure, but before this happens in
4178
the clinical setting, the data need to be placed in specific

fields so software knows where to find it. This proposition
may require the development of standards.
V. SUMMARY
There are a number of activities in place toward enhancing the development of quantitative imaging techniques.
Most of these efforts are technical in nature, as part of an
individual scientists research programs. There are also a
number of national initiatives which are focused on the development of calibration techniques and protocol standardization for the various modalities, led by scientists at the
American Association of Physicists in Medicine, the National Institute for Standards and Technology,62 and the National Institutes of Health.31 These efforts, taken collectively,
have only gotten the ball rolling slowly. It is anticipated that
over time, and with the recognition and acceptance of the
medical imaging community, the snowball effect will occur
and QI will be a part of every radiologists practice.
a

D. L. Wahner-Roedler, S. S. Chaliki, B. A. Bauer, J. B. Bundrick, L. R.
Bergstrom, M. C. Lee, S. S. Cha, and P. L. Elkin, Who makes the
diagnosis? The role of clinical skills and diagnostic test results, J. Eval.
Clin. Pract. 13, 321325 2007.
2
T. R. Johnson, B. Krauss, M. Sedlmair, M. Grasruck, H. Bruder, D.
Morhard, C. Fink, S. Weckbach, M. Lenhard, B. Schmidt, T. Flohr, M. F.
Reiser, and C. R. Becker, Material differentiation by dual energy CT:
Initial experience, Eur. Radiol. 17, 15101517 2006.
3
M. M. Goodsitt, H. P. Chan, T. W. Way, S. C. Larson, E. G. Christodoulou, and J. Kim, Accuracy of the CT numbers of simulated lung nodules
imaged with multi-detector CT scanners, Med. Phys. 33, 30063017
2006.
4
E. O. Crawley, In vivo tissue characterization using quantitative computed tomography: A review, J. Med. Eng. Technol. 14, 233242 1990.
5
Y. Kodama, C. Ng, T. Wu, G. Ayers, S. Curley, E. Abdalla, J. Vauthey,
and C. Charnsangavej, Comparison of CT methods for determining the
fat content of the liver, AJR, Am. J. Roentgenol. 188, 13071312 2007.
6
J. J. Erasmus, J. E. Connolly, H. P. McAdams, and V. L. Roggli, Solitary
pulmonary nodules: Part I. Morphologic evaluation for differentiation of
benign and malignant lesions, Radiographics 20, 4358 2000.
7
K. Shaffer, Role of radiology for imaging and biopsy of solitary pulmonary nodules, Chest 116, 519S522S 1999.
8
C. E. Cann, Quantitative CT for determination of bone mineral density:
A review, Radiology 166, 509522 1988.
9
T. F. Lang, J. Li, S. T. Harris, and H. K. Genant, Assessment of vertebral
bone mineral density using volumetric quantitative CT, J. Comput. Assist. Tomogr. 23, 130137 1999.
10
C. Marchal, R. Itti, L. Philippe, E. Boissinot, J. Moline, and M. Lavandier, The radiospirometric flow-volume diagram recording technique
and comparative study with conventional spirometry, Eur. J. Nucl. Med.
2, 165171 1977.
11
B. C. Yung, S. Sostre, and J. P. Gearhart, Normalized clearance-touptake slope ratio: A method to minimize false-positive diuretic renograms, J. Nucl. Med. 34, 762768 1993.
12
H. S. Winchell, Hematology. I. Quantitative studies of iron and heme
metabolism using radioisotope kinetics, J. Nucl. Med. 8, 375 1967.
13
F. J. Pircher, E. A. Carr, Jr., and M. E. Patno, Evaluation of quantitative
aspects of the radioisotope renogram, J. Nucl. Med. 4, 117131 1963.
14
R. H. Lazeron, L. M. de Sonneville, P. Scheltens, C. H. Polman, and F.
Barkhof, Cognitive slowing in multiple sclerosis is strongly associated
with brain volume reduction, Mult. Scler. 12, 760768 2006.
15
D. Horinek, A. Varjassyova, and J. Hort, Magnetic resonance analysis of
amygdalar volume in Alzheimers disease, Curr. Opin. Psych. 20, 273
277 2007.
16
J. G. Fine, M. Semrud-Clikeman, T. Z. Keith, L. M. Stapleton, and G. W.
Hynd, Reading and the corpus callosum: An MRI family study of vol1
4178
ume and area, Neuropsychology 21, 235241 2007.
F. P. Macmaster, R. El-Sheikh, A. R. Upadhyaya, J. Nutche, D. R. Rosenberg, and M. Keshavan, Effect of antipsychotics on pituitary gland volume in treatment-naive first-episode schizophrenia: A pilot study,
Schizophr. Res. 92, 207210 2007.
18
R. W. Katzberg, C. L. Dumoulin, M. A. Buonocore, L. R. Opsahl, R. D.
Darrow, and T. W. Morris, Noninvasive measurement of renal hemodynamic functions using gadolinium-enhanced magnetic resonance imaging, Invest. Radiol. 29, S123S126 1994.
19
M. H. Buonocore and R. W. Katzberg, Estimation of extraction fraction
EF and glomerular filtration rate GFR using MRI: Considerations derived from a new Gd-chelate biodistribution model simulation, IEEE
Trans. Med. Imaging 24, 651666 2005.
20
A. Thomas, M. Warm, M. Hoopmann, F. Diekmann, and T. Fischer, Tissue Doppler and strain imaging for evaluating tissue elasticity of breast
lesions, Acad. Radiol. 14, 522529 2007.
21
M. Sridhar, J. Liu, and M. F. Insana, Viscoelasticity imaging using ultrasound: Parameters and error analysis, Phys. Med. Biol. 52, 24252443
2007.
22
J. E. Drumm and R. ORahilly, The assessment of prenatal age from the
crown-rump length determined ultrasonically, Am. J. Anat. 148, 555
560 1977.
23
J. E. Drumm, J. Clinch, and G. MacKenzie, The ultrasonic measurement
of fetal crown-rump length as a method of assessing gestational age, Br.
J. Obstet. Gynaecol. 83, 417421 1976.
24
P. F. Judy, R. G. Swensson, and M. Szulc, Lesion detection and signalto-noise ratio in CT images, Med. Phys. 8, 1323 1981.
25
R. M. Nishikawa and M. J. Yaffe, Signal-to-noise properties of mammographic film-screen systems, Med. Phys. 12, 3239 1985.
26
W. A. Edelstein, P. A. Bottomley, and L. M. Pfeifer, A signal-to-noise
calibration procedure for NMR imaging systems, Med. Phys. 11, 180
185 1984.
27
D. P. Chakraborty and G. T. Barnes, Signal-to-noise ratio considerations
in radiographic imaging, Med. Phys. 10, 467469 1983.
28
B. Zhao, L. H. Schwartz, C. S. Moskowitz, M. S. Ginsberg, N. A. Rizvi,
and M. G. Kris, Lung cancer: Computerized quantification of tumor
responseinitial results, Radiology 241, 892898 2006.
29
S. Rohde, B. Turowski, J. Berkefeld, and A. F. Kovacs, CT-based evaluation of tumor volume after intra-arterial chemotherapy of locally advanced carcinoma of the oral cavity: Comparison with clinical remission
rates, Cardiovasc. Intervent. Radiol. 30, 8591 2007.
30
C. H. Holdsworth, R. D. Badawi, J. Manola, M. F. Kijewski, D. A. Israel,
G. D. Demetri, and A. D. Van den Abbeele, Use of CT and FDG-PET as
early prognostic indicators of response to imatinib mesylate Gleevec in
patients with gastrointenstinal stromal tumors GIST, AJR, Am. J.
Roentgenol. in press.
31
The IRAT network, http://www.aaci-cancer.org/irats/about network.asp,
accessed 2007.
32
J. Beuth and J. M. Schierholz, Review. Evidence-based complementary
oncology. Innovative approaches to optimize standard therapy strategies,
In Vivo 21, 423428 2007.
33
S. M. Erturk, S. Ondategui-Parra, H. Otero, and P. R. Ros, Evidencebased radiology, J. Am. Coll. Radiol. 3, 513519 2006.
34
D. L. Sackett, W. M. Rosenberg, J. A. Gray, R. B. Haynes, and W. S.
Richardson, Evidence based medicine: What it is and what it isnt.
1996, Clin. Orthop. Relat. Res. 455, 35 2007.
35
J. P. Ioannidis, Molecular evidence-based medicine, Eur. J. Clin. Invest.
37, 340349 2007.
36
G. W. Sledge, Jr., Evidence-based medicine in breast cancer, Clin.
Breast Cancer 7, 435 2007.
37
A. Fine, Medicare evaluates pay-for-performance, Manag. Care Q. 14,
2223 2006.
38
T. P. Meehan, Pay-for-performance: A new trend in health care, Conn.
Med. 71, 225226 2007.
39
R. Lu, P. Marziliano, and T. C. Hua, Liver tumor volume estimation by
semi-automatic segmentation method, Conf. Proc. IEEE Eng. Med. Biol.
Soc., 2005, Vol. 3, pp. 32963299.
40
P. J. Yim, A. V. Vora, D. Raghavan, R. Prasad, M. McAullife, P. OhmanStrickland, and J. L. Nosher, Volumetric analysis of liver metastases in
computed tomography with the fuzzy C-means algorithm, J. Comput.
Assist. Tomogr. 30, 212220 2006.
41
M. C. Ashe, K. M. Khan, S. A. Kontulainen, P. Guy, D. Liu, T. J. Beck,
and H. A. McKay, Accuracy of pQCT for evaluating the aged human
17
4179
radius: An ashing, histomorphometry and failure load investigation, Osteoporos. Int. 17, 12411251 2006.
42
B. C. Khoo, T. J. Beck, Q. H. Qiao, P. Parakh, L. Semanick, R. L. Prince,
K. P. Singer, and R. I. Price, In vivo short-term precision of hip structure
analysis variables in comparison with bone mineral density using paired
dual-energy x-ray absorptiometry scans from multi-center clinical trials,
Bone 37, 112121 2005.
43
L. J. Melton III, T. J. Beck, S. Amin, S. Khosla, S. J. Achenbach, A. L.
Oberg, and B. L. Riggs, Contributions of bone density and structure to
fracture risk assessment in men and women, Osteoporos. Int. 16, 460
467 2005.
44
P. D. Ross, Osteoporosis. Frequency, consequences, and risk factors,
Arch. Intern. Med. 156, 13991411 1996.
45
H. Kroger, M. Tuppurainen, R. Honkanen, E. Alhava, and S. Saarikoski,
Bone mineral density and risk factors for osteoporosisa populationbased study of 1600 perimenopausal women, Calcif. Tissue Int. 55, 17
1994.
46
M. E. Ooms, P. Lips, L. A. Van, and H. A. Valkenburg, Determinants of
bone mineral density and risk factors for osteoporosis in healthy elderly
women, J. Bone Miner. Res. 8, 669675 1993.
47
D. H. Miller et al., MRI outcomes in a placebo-controlled trial of natalizumab in relapsing MS, Neurology 68, 13901401 2007.
48
F. Fazekas, P. Soelberg-Sorensen, G. Comi, and M. Filippi, MRI to
monitor treatment efficacy in multiple sclerosis, J. Neuroimaging 17,
50S55S 2007.
49
M. A. Rocca and M. Filippi, Functional MRI in multiple sclerosis, J.
Neuroimaging 17, 36S41S 2007.
50
N. De Stefano and M. Filippi, MR spectroscopy in multiple sclerosis, J.
Neuroimaging 17, 31S35S 2007.
51
L. J. Thal, K. Kantarci, E. M. Reiman, W. E. Klunk, M. W. Weiner, H.
Zetterberg, D. Galasko, D. Pratico, S. Griffin, D. Schenk, and E. Siemers,
The role of biomarkers in clinical trials for Alzheimer disease, Alzheimer Dis. Assoc. Disord. 20, 615 2006.
52
M. Yoshita, E. Fletcher, D. Harvey, M. Ortega, O. Martinez, D. M. Mungas, B. R. Reed, and C. S. DeCarli, Extent and distribution of white
matter hyperintensities in normal aging, MCI, and AD, Neurology 67,
4179
21922198 2006.
D. E. Huddleston and S. A. Small, Technology insight: imaging amyloid
plaques in the living brain with positron emission tomography and MRI,
Nat. Clin. Pract. Neurol. 1, 96105 2005.
54
T. den Heijer, P. E. Sijens, N. D. Prins, A. Hofman, P. J. Koudstaal, M.
Oudkerk, and M. M. Breteler, MR spectroscopy of brain white matter in
the prediction of dementia, Neurology 66, 540544 2006.
55
E. M. Geraghty, J. M. Boone, J. P. McGahan, and K. Jain, Normal organ
volume assessment from abdominal CT, Abdom. Imaging 29, 482490
2004.
56
S. V. Raman, M. Shah, B. McCarthy, A. Garcia, and A. K. Ferketich,
Multi-detector row cardiac computed tomography accurately quantifies
right and left ventricular size and function compared with cardiac magnetic resonance, Am. Heart J. 151, 736744 2006.
57
P. Therasse, S. G. Arbuck, E. A. Eisenhauer, J. Wanders, R. S. Kaplan, L.
Rubinstein, J. Verweij, G. M. Van, A. T. van Oosterom, M. C. Christian,
and S. G. Gwyther, New guidelines to evaluate the response to treatment
in solid tumors. European Organization for Research and Treatment of
Cancer, National Cancer Institute of the United States, National Cancer
Institute of Canada, J. Natl. Cancer Inst. 92, 205216 2000.
58
J. M. Boone, Multidetector CT: Opportunities, challenges, and concerns
associated with scanners with 64 or more detector rows, Radiology 241,
334337 2006.
59
C. R. Meyer et al., Evaluation of lung MDCT nodule annotation across
radiologists and methods, Acad. Radiol. 13, 12541265 2006.
60
S. G. Armato III, G. McLennan, M. F. Nitt-Gray, C. R. Meyer, D.
Yankelevitz, D. R. Aberle, C. I. Henschke, E. A. Hoffman, E. A. Kazerooni, H. MacMahon, A. P. Reeves, B. Y. Croft, and L. P. Clarke, Lung
image database consortium: Developing a resource for the medical imaging research community, Radiology 232, 739748 2004.
61
M. H. Buonocore and R. W. Katzberg, Estimation of extraction fraction
EF and glomerular filtration rate GFR using MRI: Considerations derived from a new Gd-chelate biodistribution model simulation, IEEE
62
Imaging as a biomarker: Standards for change measurements in therapy,
2007.
53
From medical images to multiple-biomarker microarrays

Robert F. Wagner
Center for Devices and Radiological Health, FDA, 10903 New Hampshire Avenue, Building 62,
Room 3126, Silver Spring, Maryland 20993-0002
Received 1 August 2007; revised 9 October 2007; accepted for publication 13 October 2007;
published 28 November 2007
A favorite clich, often attributed to Yogi Berra, is Prediction is difficult, especially about the
future! In this brief review, we shall recall some historical difficulties in predicting the future of
diagnostic medical imagingin particular, the modern technologies of CT and MRIand examine
the analogous situation in the emerging technology of diagnostic microarrays for the multiplebiomarker problem. All of these technologies began their lives as ill-conditioned problems, i.e.,
there was insufficient data to solve the central problem at hand. DOI: 10.1118/1.2805252
I. OVERVIEW
aging and microarray technologies are destined to be teamed

together for the foreseeable future.
In the first half of this article we will look at how well investigators in the past predicted the future of technology in
diagnostic medical imaging, and their vulnerability to a wellknown dictum about forecasting.1 In the second half we will
consider an analogous situation for the new kind of medical
image represented by genomic and proteomic microarrays.
These latter emerging technologies allow for simultaneous
measurementfrom a single tissue or blood sampleof the
levels of expression of many thousands of different gene or
protein fragments using a novel chip design that merges molecular biology and optoelectronics. We will find some natural parallels between diagnostic imaging and microarrays. At
the end we will identify a central theme from the assessment
of conventional medical imaging that can offer insight into
the evolution of contemporary microarray technologies.
The topics covered here were selected from the research
experience of the author and his professional colleagues because they exemplify the difficulty of predicting the course
of the history of these technologies while it was being lived
in real time. No attempt has been made to provide a broad
history of either the very large field of medical imaging or
that of multiple-biomarker microarray technologies.
First, let us remark that a medical image is a random array
or vector: Random because there are multiple sources of
variability or randomness, including detector noise and interpatient and intrapatient variability, and array because an
image is obviously a multidimensional matrix. Multivariate
statistics is therefore an essential tool for understanding the
amount of information that can be captured in a medical
image. The new technology of multiple-biomarker microarrays may be thought of as a special category of medical
imaging. In conventionally processed medical image arrays,
the underlying relational structures are usually obvious because natural morphology and function are typically displayed directly. In the case of microarrays, computationally
intensive statistical techniques are necessary to discover any
relational structure in the data and within and across samples
from nondiseased and diseased patients. An obvious limitation of microarrays is that they do not generally localize their
findings to well-specified sites in the body. So medical im4944
Med. Phys. 34 12, December 2007
II. BEFORE THERE WAS CT

In the early 1970s, computed tomography CT was still a
gleam in the eye of just a few leading researchers. One of the
more vigorous areas of radiologic research at that time involved the adaptation of bright color-TV-CRT phosphors for
use in x-ray intensifying screens. Together with advances in
x-ray tube and focus design, this led to significant improvements in blood vessel visualization in, for example, x-ray
magnification cerebral angiography. However, the clinical
need and relevance of such improvements were debated in
some casesmany radiologists seemed content with available imaging technologies. Few if any anticipated CT, which
was barely a year over the horizon at the time. In fact, a
frequently heard comment in retrospect was that if the temporal and spatial parameters of the coming first generation of
CT 45 min exposures, multimillimeter resolution in crosssectional images had been freely offered to radiologists at
the time, they would have seen no use for such a system!
However, CT was to represent a qualitative change, over and
above its quantitative parameters.
III. MODERN MEDICAL IMAGING
III.A. Some critical historical moments
Let us consider some critical moments during the emergence of CT and magnetic resonance imaging MRI during
the 1970s. Recall that these technologies were both enabled
by giant steps in digital technologies, especially the minicomputer, the computer in a breadbox that for the first time
allowed complex mathematical image reconstruction algorithms to be executed on large format images in a manageable time and space and therefore with manageable expense.
While the mathematical algorithm community has naturally emphasized the open-ended potential for their own specialized contributions to imaging, physicists have tried to
understand not only the potential, but also the limitations,
imposed on these technologies by the laws of physics and
statisticsin particular, the quantity and quality of the
samples of input data required for clinically useful images.
0094-2405/2007/3412/4944/8/$23.00
4944
4945
Robert F. Wagner: From medical images to multiple-biomarker microarrays
4945
differences.47 It took some years before a widespread understanding emerged that CT was in principle quantum limited,
and thus highly efficient as defined above, for low-contrast
tasks. Investigators who wished to push the envelope in this
field then turned their attention to the question of whether
reductions in the number of views could be efficiently taken
for certain high-contrast imaging applications or by the incorporation of prior information and constraints.8
III.C. The early predictions for MRI
FIG. 1. Example of an early CT image of the brain showing a large spacefilling lesion. The conditioning of this inverse problem can be improved to
some extent by simply demanding lower resolution in the image, achievable
most simply by blurring ones vision. Courtesy of Siemens Medical
Solutions.
Both CT and MRI are, in principle, mathematically illconditioned because they attempt to solve an inverse problem of reconstructing a continuous object i.e., the body
from a finite number of discrete and noisy data samples.2 The
original and most common way of dealing with this illconditioning is by relaxing the resolution requirements on
the final result until a useful image, with disguised or hardly
visible artifacts, is obtained see, for example, Fig. 1. For
our present purposes, we define an artifact as a structure in
the image not present in the original object and not due
solely to random photon or detector noise.
One of the principal benchmarks used for assessing these
imaging technologies is the ability to discriminate a region of
a given size and contrast from its surround, versus the radiation dose or exposure and/or imaging time required for this
task. We may refer to this benchmark of image quality
versus dose or exposure and/or imaging time as imaging efficiency. How did the early investigators view the prospects
for imaging efficiency in CT and MRI?
III.B. The early predictions for CT
At the emergence of CT in the early- to mid-1970s, most

investigators recognized this as a relatively high-dose modality but had only vague impressions regarding the prospects
for improving its imaging efficiency. Some early mathematical researchers raised hopes for reducing dose through more
elaborate approaches to image reconstruction.3 On the other
hand, several authors from the physics community, including
our own group, argued that some existing CT designs and
software algorithms were already operating close to the statistical limitations imposed by quantum noise for the task of
discriminating large-area 1 cm low-contrast tissue
Medical Physics, Vol. 34, No. 12, December 2007
Some of the early investigators of MRI or nuclear magnetic resonance NMR imaging as it was then called were
pessimistic about achieving clinically useful levels of resolution in realistic imaging times. Signal-to-noise SNR calculations based on elementary theories of signal detection
and statistical decision making lead to the conclusion that
imaging time in three dimensions depends on the inverse
sixth power of the resolution required or fourth power in
two dimensions, among other variations of the scenario9. I
attended an MRI brainstorming meeting at NIH in the late
1970s. David Hoult, a pioneer in the field who was recruited
by NIH to develop an early clinical scanner for them, emphasized his concern that attempts to reduce the limiting
resolution of MRI would face a formidable obstacle from the
constraint of the power law just cited. In a few words, if
image resolution was to be improved by a factor of 2, the
imaging time might have to be increased by a factor ranging
from 16 to 64, and this would not be clinically practical.
On May 45, 1995 I attended a meeting at the National
Library of Medicine in celebration of the centenary of the
discovery of x rays by Roentgen Radiology Centennial
Commemorative Conference. Pioneers as well as contemporary experts across many fields of medical imaging presented
their experience of the progress in their fields. One of the
presenters was a major player in the development of both CT
and MRI at General Electric Corporate Research and Development Laboratories, Lewis S. Lonnie Edelheit, who had
succeeded the beloved mentor of many of us of the previous
generation, Rowland W. Red Redington. In Edelheits presentation to this commemorative conference he showed a
slide containing a copy of a letter he had written to GE
management early in the development of MRI. The letter
told a story similar to that of Hoult, namely, the outlook for
improved resolution in MRI was not great owing to the
strong dependence of imaging time on required resolution;
he was not able to encourage GE management to increase
their support for development of MRI at that point.
A brief historical summary of the evolution of that view
was presented by Redington in Ref. 10. Redington emphasized that, among other points, the previous discussions on
resolution neglected to distinguish between high- and lowcontrast regimes and ignored the great soft-tissue contrast
available in MRI. His short paper also includes a discussion
of the difference between signal, noise, and resolution characteristics in MRI versus in CT that guided some of our own
work.9 He outlined in vivo NMR spectroscopy as a potential
answer to what comes next after NMR imaging?. And he
4946
observed that it was then only a matter of time until a

straightforward MR pulse sequence would be demonstrated
that would produce blood flow information. Written more
than 25 years ago, this still makes fascinating reading. Redingtons clear vision of the future stood out against an otherwise cluttered background of those times.
Edelheits achievements both before and since the letter to
management mentioned above greatly overshadow this misread of the early situation that he shared with us that day.
Many technological tricks came out of continuing research in
MRI that allowed multiplex methods for collecting
multiple-slice MRI data in about the same time originally
dedicated to single-slice data. Methods for speeding up MRI
imaging continue to evolve to this day, including techniques
for subsampling data space, taking advantage of samples
from high-contrast regions while undersampling potentially
correlated regions of space and/or time. The intrinsic
thermal-noise limitations remain for low-contrast MRI signals, but for applications where there is higher contrast available, such as angiography or other applications that use MR
contrast agents, these tricks can offer substantial reductions
in imaging time.
III.D. Some contemporary efforts toward pushing the
envelope
In the recent decade, Mistretta and colleagues have taken

a small step backward, and then a large step forward, in
speeding up some special tasks in MRI imaging.11 Recall
thataside from the first generation of MRI imagingthe
most popular approach to data collection was in terms of
Cartesian coordinates in k- or spatial-frequency space, which
is natural to MRI, at least in one dimension and extended by
Fourier-based pulse techniques to higher dimensions. Mistrettas step backward was to revisit one of the firstgeneration approaches to MRI, namely, using the radial geometry of projection-reconstruction PR imaging that is
natural and standard in CT, rather than the later-generation
MRI Cartesian geometry.
These investigators then began experimentingin the
context of high-contrast sparse-signal applicationswith
collecting only a small fraction of the typically required
number of projections or angular samples. This offered modest efficiency gains in the context of 2D PR geometry but
remarkable gains when the projections were distributed in all
directions in 3D PR geometry.12 Further gains result from
using training image sets to learn and incorporate the image
signal correlation structure for time-resolved imaging.13 This
work has led to images that are visually comparable or superior to images collected using conventional approaches to
sampling in Cartesian geometry but with a great reduction in
imaging time by reducing the number of views or angular
samples. These investigators are in the process of applying to
time-dependent CT imaging14 the principles learned from
their work in MRI. Some of these more recent developments
have come as quite a surprise to researchers in these fields.
Another somewhat surprising feature of contemporary CT
technology has been the rate ofand to some extent the
4946
demand forsteady miniaturization and increase in the number of CT detectors, both circumferentially around the body
and axially along the body. In the early days of CT, many
investigators thought that such an increase would automatically lead to an increase in required x-ray dose. However, the
large-area low-contrast performance may be maintained independently of the number of detectorsit is not the number
of photons per detector that directly determines the dose/
image-quality relationship in the low-contrast regime, it is
rather the spatial density of detected x-ray counts9as long
as the number of detected counts exceeds the dark noise and
readout noise of the detector.
A major unsurprising and long-awaited development has
been the emergence of technologies foreseen more than three
decades ago but only made practical by advances in digital
detectors, e.g., dual-energy x-ray imaging and limited-angle
tomography or tomosynthesis. One popular prediction gradually being realized is that many if not most 2D or planar
imaging technologies will be supplanted by 3D volumetric
imaging technologies in the foreseeable future. The radiation
dose to the body may increase appreciably for some of these
latter developments.
Finally, an unconventional re-examination of the geometry of CT scanning has led to the investigation of a configuration referred to as inverse-geometry volumetric CT.15 The
name derives from the fact that a large-area, scanned x-ray
source is used with a detector array that is smaller than the
source in the transverse direction but comparable in the
axial direction. The authors claim that noise and resolution
with their micro-CT system are comparable to an existing
volumetric micro-CT system, while avoiding the cone-beam
artifacts that occur in the latter. Additional projected benefits
include reduced scatter and detector cost.
To summarize, the first generations of investigators in
both CT and MRI did not have a clear view of the requirements for the number of image samples, e.g., photons and
radiation exposure or dose, views, and imaging time, needed
for clinically useful images for particular imaging tasks.
Much work advanced on an ad hoc basis, in particular by
means of a beauty contest between competing modalities
or algorithmscontinuing in this mode up to the present. A
foundational approach to many of the issues sketched above
required several decades to formulate and has been published
in the present decade by Barrett and Myers.2 This approach
should appeal naturally to the physicists in our audience. It
depends in part on physical measurement and analysis of the
so-called singular vectors, or natural imaging building
blocks corresponding to a given physical configuration of
image detection system; this analysis leads to a separation
between the concepts of the measurement space and the
null space. The measurement space is that portion of the
object that is accessible to measurement by the imaging system; the null space is that portion of the object that is inaccessible to measurement by the imaging system. A trivial
example of a null space would be the zeroes of the system
transfer function. Thus, objects that differ only in their nullspace component appear the same when imaged. Many artifacts may be thought of as ghosts from the null space,8 i.e.,
4947
trying to solve for information that is not available from

finite sampling of the object, together with noisy detected
data, leads to the generation of unwarranted image
contentin the absence of prior information about the object. The fundamental approach of Ref. 2 combines mathematics, statistics, and physics but still awaits application to
many contemporary imaging scenariosincluding those of
Mistretta and colleagues and Schmidt et al.,15 sketched
abovea challenge previously presented by the present
author,16 among others.
This unresolved situation regarding the fundamentals suggests seeking a practical working solution in the short term.
In a few words: How does one quantify the impact of artifacts that always accompany undersampled imagery? Many
investigators still depend on a performance measure whose
central component is the SNR at the pixel level. This is in
spite of the fact that decades of publications by many of the
authors cited above cf. Ref. 9 have demonstrated that pixel
SNR is a highly ambiguous figure of meritespecially
within the context of correlated noise and/or long-range image artifacts. A practical middle groundbeyond nave SNR
but short of the more formal analysiswas investigated in
the early 1990s by Hanson, Myers, Wagner, and
coauthors.17,18 It is based on measuring the impact of randomized artifacts on diagnostic performance. Although these
references focus on simulated imagery, there is no fundamental reason why the methods cannot be extended to clinical images or a hybrid of clinical images with simulated
inclusionsaside from the expense of collecting and reading the images. The main challenge to the investigator would
be to collect a random sample of realistic examples where
the anatomy challenges the physical source of the artifacts.
IV. MICROARRAYS
IV.A. Background
The biomedical landmark of our time has been the sequencing of the genomes of many organisms, in particular
the sequencing of the human genome. As more genes and
proteins are identified, and those involved in disease initiation and progression become better understood, an increasing
number of candidate diagnostic biomarkers as well as novel
therapeutic agents are expected to emerge.19 The target of
the diagnostic and therapeutic agents may be the general
population, or it might be some specially targeted subpopulations. The targeting of therapeutics based on the biological
make-up of subclasses of patients is known as pharmacogenomics and is popularly referred to as personalized medicine. The promise of the field is great and open-ended,
yetas was the case with the breakthrough technologies in
medical imagingthe scale of the promise and efficiency
requirements for developing clinically useful diagnostic tools
is not yet well understood.
Investigators emphasize that it is essential to distinguish
between genetics and genomics. In genetics, one tests for the
presence or absence or usually the mutation status of one or
typically only a few specified genes, e.g., the gene associated
with cystic fibrosis or the BRCA1 and BRCA2 genes assoMedical Physics, Vol. 34, No. 12, December 2007
4947
ciated with a class of breast cancers. In genomics, proteomics, and related technologies, however, investigators are testing not for the presence or absence, but rather for the level of
expression, of a large number of genes and/or proteins, and
patterns of expression across many genes or protein fragments. The physical measurements in the proteomic approach are obtained from mass spectrometry of specially prepared tissue specimens. The physical measurements in the
genomic approach will be sketched here for interested newcomers. Both technologies are complicated by potential
variations in expression level as a function of time of day or
metabolic or other natural cycle including, of course, the
patients age.
The technology used in genomics to measure the level of
expression of a large number of specified genes simultaneously is referred to as a microarray assay; an individual
microarray is often referred to as a DNA chip and is about
the size of a thumbprint. There are several different contemporary microarray technologies.20 We provide here an introductory summary of the generic approach. A microarray contains a location and spot for each of the specified reference
genes. The spots can be prepared by synthesizing previously
designated sections of known genetic sequence and binding a
quantity of this material to a substrate. The spot, together
with some associated controls, is referred to as a probe. In
contemporary work, the number of probes on a chip is in the
order of tens of thousands. A prepared sample of genetic
material from a patients blood or other tissue of interest, the
target, is labeled with a fluorescent dye, and then washed
over the chip. The goal of the measurement is to quantify the
degree of hybridization between the material in the target
and that in the probe. Hybridization refers to the complementary binding of the bases in the genetic code A to T, G to C,
etc.. The degree of hybridization at each spot is assumed to
be proportional to the number of copies of the specified RNA
section in the patients tissue samplea measure of the level
of expression of the specified gene. After hybridization, the
spots on the microarray are read out using laser light to excite the fluorescence, optics and photodetection to measure
the fluorescence, and finally digitization. Microarrays are
thus a form of molecular imaging of tissue samples extracted
from patients.
As in medical imaging but greatly more so, the genomics and proteomics problems are not generally well conditioned; great variability in data consistent with a wide range
of potential but not real solutions to any particular diagnostic problem can arise at each of several stages of analyzing
the problem. The very first stage involves quality control of
the laboratory conditions and measurement apparatus. There
are a multitude of issues here, well beyond our present scope.
The next principal stagetrying to find those genes that are
biologically significant for a particular task, among many
thousands of candidatesis fraught with the general problems of all tasks of data mining, namely, the fact that many
candidates are expected to stand out in a noisy sample purely
by random chance, so-called false discovery.20 Newcomers
may be surprised to learn that a typical starting point for the
overall procedure is measurement of the level of expression
4948
of many thousands or tens of thousands of genes from the

tissues of only a hundred or at most a few hundred patients.
That is, the sampling of data and the conditioning of the
problem are just the opposite of what one would prefer in an
ideal experiment. For example, in typical problems in
computer-aided diagnosis in medical imaging, one may typically design a diagnostic model with one or two dozen imaging features using the images of several hundred patients.
Investigators attempt to complement the problem of data
mining with the use of any prior knowledge or understanding
they have of the biological role of particular genes. This is
analogous to scientist image processors in computer-aided
diagnosis in medical imaging querying clinicians about the
relative importance of various image features, e.g., spiculations of masses, or pleomorphism of calcifications in mammography.
IV.B. Statistical learning machines
A crucial final stage involves the task of training a classifier how to weight the levels of expression of a modest number of gene or protein fragments that survive the initial data
mining, using a statistical learning machine, e.g., a classical or neural-network classifier, to maximize success in classifying the condition of interest, e.g., cancer versus noncancer, or an aggressive subclass of cancer versus an indolent
one, or a responder versus a nonresponder to a specified
therapy, among others. A natural bottleneck occurs here since
at present the accessible and affordable world provides only
a finite sample of patients with known truth status who can
be used to train the classifier, and next to test it with previously unseen cases. This limitation of samples can lead to
great uncertainty in the classifier itself, as well as attempts to
estimate its performance over the target population using any
modest number of available patients.
In general, the uncertainty in the performance estimate
decreases with increases in either or both of the number of
training samples and the number of testing samples. The issues here are conceptually identical to those one encounters
when studying task-dependent trained model observers when
assessing medical imaging systems.21 As in this latter problem, the uncertainty in performance estimation also decreases with increasing intrinsic separability of the populations. This is analogous to the intrinsic contrast of a
particular diagnostic task in medical imaging. A further issue
for the problem of statistical learning machines is that the
number of training samples required for a given confidence
level also increases with the number of biomarkers features
in the classifier and/or the complexity of the classifier.22
Because of the great opportunity for false discovery of
genes at the early stage of development of a given technology and the limitations of available samples from patients
with known truth status, many early promises have not held
up under later scrutiny, while the jury is still out on many
other high-profile studies.
The work of Golub and colleagues on classification of
human acute leukemias23 has attracted a great deal of attention for several reasons, including the fact that it was one of
4948
the first published results on binary classification based on

gene expression levels. The authors identified 50 informative genes among many thousands of candidates. They then
trained a large number of classifiers for the task of discriminating acute lymphoblastic leukemia ALL from acute myeloid leukemia AML based on the levels of expression of
these informative genes in patients bone marrow samples
and found success in the training set almost independent of
their choice among those genes. A peculiar feature of this
work is that only 38 patients were in the training set. A
celebrated theorem attributable to Cover24 but yet to be celebrated in the biomarker community indicates that the problem they solved at this point would be trivialif the escape
clause those 50 genes were not strongly coexpressed. According to this theorem, almost every possible labeling of
cases in two classes here, ALL versus AML will be perfectly separable by a simple hyperplane unless the number of
cases is a multiple greater than unity when only a few features are used or about two for such high-dimensional data
times the number of features or biomarkers used. An inspection of Fig. 2, however, reproduced here from their
work, makes it visually obvious that there is a strong correlation, or strong level of coexpression, of the genes in that
sample. The escape clause above is now operational: Those
50 genes apparently live on a lower-dimensional subspace.
In that case e.g., if it is only a 5D to 10D space, then they
may have sufficient samples that the discrimination they observed is not simply a consequence of geometry in highdimensional space. We encourage authors to address this geometry problem before making inferences from their work.
Although the authors did not address this issue, they were
aware of the necessity to validate their results on independent test data. They reported that they found good predictability in an independent test set of 34 samples. We return to
this point below.
This particular exampleaside from any issues that can
be raised in its critiqueappears to offer an example of lowhanging fruit in the following sense. In retrospect, if one
looks at the color map of the response of the 50 genes in the
authors original paper in Science Ref. 23 and Fig. 2 here,
one sees that the individual genes on their own have very
high, albeit not quite perfect, discriminating power. For this
reason, this data set has come to be recognized as an easy
one for classifier development, analogous to imaging of
high-contrast phenomena in the technologies discussed
above. Still, it is relatively early in the acquisition of data
and critical review of work on this problem.
Another celebrated data set is that of Petricoin, Liotta and
colleagues for discriminating between the presence or absence of ovarian cancer using spectral lines from mass spectrometry of protein fragments from blood samples.25 After
elaborate feature selection and training, the performance of
their classifier was considered to have been validated by the
original investigators using independent test samples. However, like many other early multiple-biomarker tests, their
results have not been reproduced and have been subjected to
4949
4949
FIG. 2. Levels of expression of 50

most informative genes the labeled
rows from 38 patients the columns
in two kinds of acute human leukemia
ALL and AML. Red represents genes
overexpressed with respect to the
mean level across patients and blue
represents genes underexpressed with
respect to that mean level. Notice the
strong coexpression of the genes in
both major regions top and bottom of
figure. These genes may live on a
low-dimensional subspace of the original 50 dimensions. Figure reproduced
with permission from Science, Ref. 23,
copyright 1999 American Association
for the Advancement of Science
AAAS.
challenges and much debate based on potential biases in the

way the samples were analyzed see, e.g., Ref. 26 among
many others.
In general, there are great uncertainties in the performance assessment of multiple-biomarker classifiers because
of the general structure of the problem and the modest sizes
of available data sets. In addition to trying to reduce the
sources of bias in the assessment process, it is also of great
interest to estimate confidence intervals on the resulting performance estimates. We have been addressing this problem
in contemporary work within our group.2730
IV.C. Levels of generalizability of estimated
uncertainties
A classic question arises at this point. Should one estimate

the uncertainty that one would observe if the classifier is
fixed and on replication of the experiment only a new random sample of testers is drawn, or should one estimate the
uncertainty that one would observe if, on replication of the
experiment, a new random sample of trainers is drawn as
well as a new random sample of testers? Of course, confidence intervals for the latter approach will in general be
wider than those for the former approach.
The point of view of myself and collaborators is that the
second, or more general, uncertainty should be specified.28,29
Classifier designers sometimes argue that their training is
frozen. However, it is almost always the intent of these
investigators to continue to build their database, improve
their statistical learning, and update their algorithms. This is
motivation for the more general approach of Refs. 28 and 29.
The former may be useful for analyzing a pilot study to
guide the sizing of a larger subsequent pivotal study; the
latter can apply equally to the setting of either a pilot or a
pivotal study. An analysis of the uncertainties in a pilot study
may be the only way to design a more definitive pivotal
study.
Further motivation for our more general position can be
drawn from a recent exchange in the multiple-biomarker literature. Dave and colleagues31 trained a model for
prediction-of-survival on 95 biopsy specimens from patients
4950
with untreated follicular lymphoma. They tested the model

by validating it in an independent test set of 96 specimens
from the same population and found a highly significant association of their model with survival in the independent test
set p 0.001.
Tibshirani acknowledged the authors result that their
recipe gave a significant p-value when used on the test set
but went on to point out that the result is extremely
fragile.32 When the training and test sets are swapped, and
the authors recipe for training is applied, the authors model
does not emerge. Additional analyses suggest that their result occurred by chance and is not robust or reproducible.
Thus, even what we refer to as the traditional hygiene of
independent training and test sets fails to probe the robustness of a model and should give us pause regarding our level
of satisfaction with that simple practice.
V. SOME GENERAL FEATURES OF THE LARGER
PICTURE
Many of the issues surrounding requirements for sample
numbers in the classifier problem have begun to be addressed
by investigators of model observers trained for assessing diagnostic imaging technologies21 and in the subfield of statistical pattern recognition known as computer-aided diagnosis
in medical imaging.33,34 The general problem is also intimately related to the so-called multiple-reader, multiple-case
MRMC paradigm used in receiver operating characteristic
ROC analysis in medical imaging,35 where there may be a
great difference in skill training and experience, say among
trained readers i.e., radiologists. Readers in medical imaging are analogous to training sets in the machine learning
problem; cases in medical imaging are analogous to the testing cases in the machine learning problem.36 But a coherent
foundational approach is still at an early stage of development. We therefore expect to see divergence of views and
approaches by investigators in many disparate subfields of
microarray technology, arguably analogous to the situation in
the first few decades of the development of contemporary
high-tech medical imaging systems. The recognition of the
need for a fundamental multivariate statistical analysis for all
of these applications is still not widespread throughout the
land.
In both medical imaging and microarray technologies a
central goal is to reduce the very high dimensional information in the detected data to a few clinically meaningful dimensions or features and ultimately to a diagnostic decision.
The task is complicated by the fact that the high-dimensional
information in either medical image arrays or applications of
microarrays includes at least two principal sources of randomness, namely, random noise from the image-detection
technology and the natural randomness or variability that we
patients bring to the table. The statistical pattern recognition
experience of a radiologist is typically the key to the dimensionality reduction task in medical imaging, but the statistical
patterns in microarray data are not typically visually accessible. The field of microarray technologies does not enjoy
another of the great advantages of medical imaging disMedical Physics, Vol. 34, No. 12, December 2007
4950
cussed at the outset. In imaging, one can often improve the

conditioning of an inverse problem by simply backing off
from the resolution requirements and looking at the resulting
images cf. Fig. 1. To date, no such simple strategy exists
for dimensionality reduction in the field of microarray technology, although many sophisticated alternatives are
emerging.37 It will also require some time to sort out the fact
that some investigators will discard dimensions or biomarkers found to play a critical role by other investigators.
VI. CONCLUSION
An appreciation of the foundational statistical issues is
essential to progress in all these fields and to attempting to
predict their future prospects. Andas in all complex
decision-making tasksmuch more data are required to cull
the fruitful paths from the myriad blind alleys with which
almost all researchers have ample familiarity.
Ultimately, the whole world will be our database. In the
meantime, who will organize and analyze the currently
sparse available data in a resource-efficient way that respects
fundamental scientific and statistical principles? Cf. Ref.
38. It may take some time to converge on a unified foundational approach to the theory and practice of the general
multiple-biomarker problemanalogously to the situation in
medical imaging.
The bottom line of this look at the past, present, and future: Niels Bohr was right all along!1 Predictionespecially
about the futureis difficult. But our predictions about the
future of technologies involving random arrays have a much
better chance of survival if they are based on foundational
principles of multivariate statisticsin addition to the underlying science.
1
Prediction is difficult, especially about the future! Y. Berra or N. Bohr?

Apparently Niels Bohr is ahead in the competition for this attribution.
Other household names are also high on the list, according to searches
conducted by http://letterfromhere.blogspot.com/2006/12/bohr-leadsberra-but-yogi-closing-gap.html
2
H. H. Barrett and K. J. Myers, Foundations of Image Science John Wiley
and Sons, Hoboken, NJ, 2004.
3
R. Gordon, Dose reduction in computerized tomography, Invest. Radiol. 11, 508517 1976.
4
R. F. Wagner, D. G. Brown, and M. S. Pastel, The application of information theory to the assessment of computed tomography, Med. Phys. 6,
8394 1979.
5
K. M. Hanson, Detectability in computed tomographic images, Med.
Phys. 6, 441451 1979.
6
K. M. Hanson, Noise and contrast discrimination in computed tomography, in Radiology of the Skull and Brain: Technical Aspects of Computed
Tomography, edited by T. H. Newton and D. G. Potts C.V. Mosby, St.
Louis, 1981, Vol. 5, Chap. 113, pp. 39413955.
7
R. Stanton and O. Tretiak, Dose reduction through variable dose CT
Scanning: Optimality of the filtered backprojection algorithm, J. Comput. Assist. Tomogr. 7, 10541061 1983.
8
K. M. Hanson, Bayesian and related methods in reconstruction from
incomplete data, in Image Recovery: Theory and Applications, edited by
H. Stark Academic Press, Orlando, 1987, pp. 79125.
9
R. F. Wagner and D. G. Brown, Unified SNR analysis of medical imaging systems, Phys. Med. Biol. 30, 489518 1985.
10
R. W. Redington, Diagnostic NMR, IEEE Trans. Med. Imaging MI-1,
230233 1982.
11
D. C. Peters, T. M. Grist, F. R. Korosec, J. E. Holden, W. F. Block, K. L.
Wedding, T. J. Carroll, and C. A. Mistretta, Undersampled projection
reconstruction applied to MR angiography, Magn. Reson. Med. 43, 91
4951
101 2000.
K. K. Vigen, D. C. Peters, T. M. Grist, W. F. Block, and C. A. Mistretta,
Undersampled projection-reconstruction imaging for time-resolved
contrast-enhanced imaging, Magn. Reson. Med. 43, 170176 2000.
13
C. A. Mistretta, O. Wieben, J. Velikina, W. Block, J. Perry, Y. Wu, K.
Johnson, and Y. Wu, Highly constrained backprojection for timeresolved MRI, Magn. Reson. Med. 55, 3040 2006.
14
C. A. Mistretta, Innovative physics boosts MRI speed, cuts CT dose,
Invited article for Diagnostic Imaging magazine August 2007, pp. 33
42, concluded on p. 54.
15
T. G. Schmidt, J. Star-Lack, N. R. Bennet, S. R. Mazin, E. G. Solomon,
R. Fahrig, and N. J. Pelc, A prototype table-top inverse-geometry volumetric CT system, Med. Phys. 33, 18671878 2006.
16
R. F. Wagner, Lessons from my Dinners with the Giants of Modern
Image Science, Radiat. Prot. Dosimetry 114, 410 2005.
17
K. J. Myers, R. F. Wagner, K. M. Hanson, H. H. Barrett, and J. P. Rolland,
Human and quasi-Bayesian observers of images limited by quantum
noise, object variability, and artifacts, Proc. SPIE 2166, 180190 1994.
18
Details of many applications of the approach of Ref. 17 may be found in
manuscript or abstract form on the website of K. M. Hanson at Los
Alamos, http://public.lanl.gov/kmh/publications
19
G. S. Burrill, Biotech 200620th Anniversary Report on the Industry.
Life Sciences: A Changing Prescription Burrill and Company, San Francisco, CA, 2006.
20
G. J. McLachlan, K.-A. Do, and C. Ambroise, Analyzing Microarray
Gene Expression Data Wiley, Hoboken, NJ, 2004.
21
B. D. Gallas, Estimating the variance of the ideal linear observers
signal-detection performance from a finite number of trainers and testers,
presented at Medical Image Perception Conference X, Duke University,
Durham, North Carolina 2003. Proceedings of the conference on disc
available from Medical Imaging Perception Society MIPS, or file available from [email protected]
22
V. N. Vapnik, The Nature of Statistical Learning Theory Springer-Verlag,
New York, 1995.
23
T. R. Golub et al., Molecular classification of cancer: Class discovery
and class prediction by gene expression monitoring, Science 286, 531
537 1999.
24
T. M. Cover, Geometrical and statistical properties of systems of linear
inequalities with applications in pattern recognition, IEEE Trans. Electron. Comput. EC-14, 326334 1965.
25
E. F. Petricoin et al., Use of proteomic patterns in serum to identify
ovarian cancer, Lancet 359, 572577 2002.
12
26
4951
D. E. Ransohoff, Bias as a threat to the validity of cancer molecularmarker research, Nat. Rev. Cancer 5, 142149 2005.
27
W. A. Yousef, R. F. Wagner, and M. H. Loew, Comparison of nonparametric methods for assessing classifier performance in terms of ROC
parameters, in Applied Imagery Pattern Recognition Workshop, 2004:
Proceedings of the 33rd AIPR Conference IEEE Computer Society, Los
Alamitos, CA, 2004, pp. 190195.
28
W. A. Yousef, R. F. Wagner, and M. H. Loew, Estimating the uncertainty
in the estimated mean area under the ROC curve of a classifier, Pattern
Recognit. Lett. 26, 26002610 2005.
29
W. A. Yousef, R. F. Wagner, and M. H. Loew, Assessing classifiers from
two independent data sets using ROC analysis: A nonparametric approach, IEEE Trans. Pattern Anal. Mach. Intell. 28, 18091817 2006
TPAMI.
30
W. A. Yousef, R. F. Wagner, and M. H. Loew, The partial area under the
ROC curve: Its properties and nonparametric estimation for assessing
classifier performance, submitted to IEEE Trans. Pattern Anal. Mach.
Intell., 2007.
31
S. S. Dave et al., Prediction of survival in follicular lymphoma based on
molecular features of tumor-infiltrating immune cells, N. Engl. J. Med.
351, 21592169 2004.
32
R. Tibshirani, Immune signatures in follicular lymphoma, N. Engl. J.
Med. 352, 14961497 2005.
33
H. P. Chan, B. Sahiner, R. F. Wagner, and N. Petrick, Classifier design
for computer-aided diagnosis: Effects of finite sample size on the mean
performance of classical and neural network classifiers, Med. Phys. 26,
26542668 1999.
34
B. Sahiner, H. P. Chan, N. Petrick, R. F. Wagner, and L. Hadjiiski, Feature selection and classifier performance in computer-aided diagnosis:
The effect of finite sample size, Med. Phys. 27, 15091522 2000.
35
R. F. Wagner, C. E. Metz, and G. Campbell, Assessment of medical
imaging systems and computer aids: A Tutorial Review, Acad. Radiol.
14, 723748 2007.
36
S. V. Beiden, M. A. Maloof, and R. F. Wagner, A general model for
finite-sample effects in training and testing of competing classifiers,
IEEE Trans. Pattern Anal. Mach. Intell. 25, 15611569 2003.
37
R. M. Simon, E. L. Korn, L. M. McShane, M. D. Radmacher, G. W.
Wright, and Y. Zhao, Design and Analysis of DNA Microarray Investigations Springer-Verlag, New York, 2004.
38
D. Gur, R. F. Wagner, and H.-P. Chan, On the repeated use of databases
for testing incremental improvement of computer-aided detection
schemes, Acad. Radiol. 11, 103105 2004.
Hypofractionation in radiation therapy and its impact

Lech Papieza and Robert Timmerman
University of Texas Southwestern Medical Center, Dallas, Texas 75390
Received 26 August 2007; revised 2 October 2007; accepted for publication 1 November 2007;
published 19 December 2007
A brief history of the underlying principles of the conventional fractionation in radiation therapy is
discussed, followed by the formulation of the hypothesis for hypofractionated stereotactic body
radiation therapy SBRT. Subsequently, consequences of the hypothesis for SBRT dose shaping
and dose delivery techniques are sketched. A brief review of the advantages of SBRT therapy in
light of the existing experience is then provided. Finally, the need for new technological developments is advocated to make SBRT therapies more practical, safer, and clinically more effective. It
is finally concluded that hypofractionated SBRT treatment will develop into a new paradigm that
will shape the future of radiation therapy by providing the means to suppress the growth of most
carcinogen-induced carcinomas and by supporting the cure of the disease. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.2816228
Key words: ablative radiation therapy, hypofractionation, SBRT
I. INTRODUCTION
I.A. Historical foundation of conventional fractionation
schemes
Traditional radiation therapy delivery requires multiple fractions of 1.5 to 3 Gy administered daily over a period of 37
weeks. This regimen has been established based on early
experience of treatment with radiation and has been justified
by widely accepted models of radiobiological effects of x
rays on human tissue.13 However, the strategy of traditional
fractionation in external beam radiation therapy has not been
the only one practiced historically. After the discovery of
ionizing radiation, hypofractionated, or even single large
dose fraction treatments, were practiced. It has been observed that large doses per fraction were tumorcidal, especially for epithelial tumors. Furthermore, early clinical experience provided valuable lessons regarding the balance
between tumor control and tissue toxicities. Early evaluations of radiation therapy showed that the delivery of a large
dose per fraction treatments was leading to unacceptably
high toxicities. Therefore, quite early in the development of
radiation therapy these schedules were abandoned because of
complications such as fibrosis, stenosis, and vascular injury.
Data collected at initial stages of the development of radiation therapy were irrefutable. The understanding of the
factors contributing to complications was incomplete. The
ability to manage the parameters underlying the unfavorable
results for large dose fraction therapy was severely limited in
the early days of radiation therapy by immature technologies
of dose delivery. However, these early experiences and their
impact on the acceptance of particular fractionation schemes
have been largely forgotten by later practitioners of radiation
therapy. The paradigm of delivering a dose in small portions
has been taken for granted and disconnected from the reflection that early treatments suffered multiple dosimetric limitations. These limitations included, among others, the inability to decrease exposure of normal tissues due to low beam
112
Med. Phys. 35 1, January 2008
energies utilized. These treatments suffered also from large

targeting uncertainties caused by the unavailability of adequate imaging techniques and the incapacity of even mildly
accurate representations of the dose distribution inside the
patient anatomy.
I.B. The impact of the early technology
When energies of beams were available only in the KeV

range, the dose to healthy tissue, particularly tissue at body
surface, exceeded many times the dose deposited at depth
1020 cm below surface where typical tumors were located. This unfortunate characteristic of early dose delivery
technology was responsible for a huge integral dose being
delivered to the patients skin and other healthy parts of the
body when deeper situated tumors were treated. To make
things worse, the early treatment techniques were also incapable of accurately estimating how the dose delivered was
redistributed to different portions of the patients body. When
we take into further account the inability of earlier treatment
techniques to locate the target relative to x-ray beams, we
realize that only mechanisms of differential effects of radiation on tumors and normal tissues could have made radiation
therapy an acceptable mode of cancer treatment.
This rationale for utilizing the mechanisms of differential
effects between tumors and normal tissues, and ensuing reasons for decreasing the dose per fraction, has been based on
the insights from radiobiology,13 and in particular, from
mechanisms of repair of sublethal damage, repopulation, reassortment within the cycle, and reoxygenation. It has been
inferred from these mechanisms that by fractionating, one
can take advantage of superior repair capabilities between
fractions of normal tissue from the damage caused by radiation. In addition, the normal tissues were understood to be
able to better repopulate between fractions, and thereby heal
or reform mucosa. As far as the tumor was concerned, exposure of different phases of the cell cycle some more sensitive to radiation than others was assumed to be more likely
0094-2405/2008/351/112/7/$23.00
112
113
L. Papiez and R. Timmerman: Radiation therapy hypofractionation
assured under the provision of multiple, conventional fractionation. Finally, since oxygenated tissue was considered to
be more radiosensitive, the better reoxygenation of tumors
during multifractionated dose delivery was thought to be advantageous for the therapy.
The overall mechanism of positive differential effects in
eradication of tumor cells vs normal tissues in radiation
therapy based on multifractionation regimes has been called
the therapeutic window. However, the therapeutic window
benefits have not always been well documented and sometimes actually may not have worked very well for the therapeutic advantage. In particular, all the above-described
mechanisms could, under specific conditions, act to the disadvantage of radiation therapy. For example, tumors have a
capacity to repair and tumors may repopulate too. In turn,
normal tissues can also become more sensitive throughout
the course of radiotherapy. Thus, under certain circumstances
and for certain types of cancers and treatment sites, fractionation may have actually been detrimental rather than advantageous for the success of radiation therapy treatment.
I.C. The conventional fractionation vs Gamma Knife
experience
The first attempt to revive hypofractionation in the era of

modern radiation therapy arose when high-energy beams in
the MeV energy range began to be available and were delivered with Gamma Knife GK radiosurgery. The original
device was conceived by a neurosurgeon4 who was rightly
inspired by the thought that MeV radiation can be used as a
surgical tool for the removal of tumorcidal tissue. In consequence, the idea of focusing radiation on the target from
many sources simultaneously, while keeping the dose small
to the healthy tissues of the brain, was born. The focusing of
a high dose, together with precise targeting of a dose assisted
by stereotactically defined localization of tumor tissue in the
brain, made it possible to realize the idea of tumor ablation
with limited toxicity to other parts of the brain. Despite the
reservations of the radiation therapy community regarding
the application of Gamma Knife technology and one fraction
treatment regime to cranial radiotherapy, the results of the
treatment of brain metastasis and brain tumors in general has
been characterized over time as favorable. Thus, the question
appeared whether the radiosurgical approach in the brain
could also work in other organs of the body. To answer this
question, however, one was obliged to design tools to move
radiosurgical techniques to extracranial sites. In particular,
this transfer of experience would require targeting extracranial sites in the same manner as in cranial therapy stereotactically. These tools appeared first in the early 1990s.
II. METHODS
II.A. Hypothesisthe need for large dose fractions
It has been observed over many years of practice that

conventional radiotherapy is miserably ineffective for gross
disease, especially for common carcinogen-induced carcinomas. The long-term follow-up of nonsmall cell lung cancer
Medical Physics, Vol. 35, No. 1, January 2008
113
NSCLC patients has been showing that only 2030% of

tumor treatments with conventional fractionation techniques
stopped or recessed the growth.57 The focus of the established radiotherapy is that it is better tolerated compared to
surgery, consistent with the physicians oath of first do no
harm. Unfortunately, this may come at a high cost to the
patient. This stress of radiotherapy has an aura of rationality
when the therapy is deemed to be only marginally effective.
When treating cancer one has to assume, however, that the
potential recurrence is deadly and thus a recurrence has to be
treated as the ultimate toxicity. Facing these risks and assuming that radiation therapy may be the treatment of last resort,
the most appropriate paradigm of radiation therapy should be
establishment of the most appropriate balance between benefit and harm from the therapy.
The radical reevaluation of radiotherapy objectives with
the goal of changing it from regional adjuvant therapy e.g.,
breast, rectum, sarcoma to curative therapy based on high
local control rates comparable to surgery requires dramatic
changes in the dose prescription. It is clear that this change
cannot be accomplished through traditional dose escalation.
Far-reaching improvements in local control would evidently
require a massive increase in the number of fractions in traditional dose escalation schemes, extending treatments to
910 weeks. From previous experience, it is known, however, that extended over multiple weeks, treatment leads to
only marginal improvement in therapy results.8 One may
speculate that the increased accumulation of tumor regrowth
and repair in conventional fractionated radiotherapy in fact
negates the effectiveness of increased tumor exposure. It
seems, therefore, that the one practical method of increasing
the efficacy of curative radiotherapy is to apply ablative,
high dose per fraction schemes of therapy delivery. The
question appears, however, if the toxicity of such ablative
therapies may be tolerable. These questions may be resolved
only by clinical data, particularly when new technologies and
new techniques of treatment planning and treatment delivery
are undertaken to minimize the effects of toxicity based on
justifiable hypothesizing from existing data. Fortunately,
some of these data already exist, and a considerable amount
of these data indicate that toxicity of ablative therapies are in
fact well tolerated.
The hypothesis of limiting radiotherapy toxicity at ablative dose levels must first take into account the fact that the
tissue biological response will be different for hypofractionated regimes and for traditional fractionation.911 The hypofractionated mode of treatment has an ablative character, and
as such, leads to direct obliteration of cancerous cells. However, it has to be assumed that the same fate would befall
healthy tissue if it were exposed to equally high levels of
radiation as the tumor itself. The situation is similar to radiosurgical treatments performed with the Gamma Knife.
Thus, the outcome of hypofractionated approach more
closely follows the model of surgery than conventionally
fractionated radiotherapy. The inevitable damage to healthy
tissue may be tolerated only if the volumes of obliterated
cells constitute a small portion of the total volume of the
organ affected, i.e., as long as surviving tissues of the organ
114
that experienced partial ablation are capable of replacing the

function of the lost fraction of the organ. This strategy, when
applied to extracranial targets, is called stereotactic body radiation therapy, or SBRT. The basic assumptions are that:
a
b
The volumes of obliterated healthy tissues are minimal.

In these larger portions of treated organs where a tumor
is absent but where radiation is still delivered, the dose
is below a threshold level, guaranteeing the full recovery of these tissues from radiation impact.
It is expected that different types of tissue will react
differently to injuries caused by ablation resulting from
large doses per fraction treatment. Some small volumes
of tissues parallel functioning tissues can be removed
without dramatic impact on the organ functioning. For
other tissue affected by treatment serially functioning
tissues, the effect of ablation of even a relatively small
number of cells may be critically detrimental for organ
functioning.
The large dose per fraction model assumes that a tumor

and the healthy tissue directly surrounding the tumor will be
eradicated as happens during surgery. This approach to
therapy makes it necessary, however, to differentiate between
types of tissues exposed to injury during irradiation. In particular, it can be expected that irradiation effects may cause
different reactions in parallel functioning tissues versus serially functioning tissues.2,3 Tissues made up of predominantly
structurally defined functional subunits FSU are called parallel functioning tissues peripheral lung, peripheral liver, peripheral kidney, etc. and are characterized by independent
redundancy. Thus, a parallel organs dysfunction is related
mostly to the fraction of the organ volume exceeding the
threshold dose. On the other hand, tissues made up of predominantly structurally undefined FSUs are called serial
functioning tissues mucosa of GI tract, ducts, bronchus, spinal cord, etc. and are characterized by chain functioning.
The serial organ dysfunction is related mostly to the number
of surviving clonagens and the linear distance they must
travel to rescue damaged areas. The consequence for hypofractionated SBRT is that the treatment has to avoid excessive volume irradiation of parallel organs and keep the dose
level in serial organ below the threshold level to preserve a
sufficient amount of clonagens.1246
II.B. SBRT hypofractionated therapyearly clinical
experience
As the goal of SBRT is the ablation of tissue of the target

and its close surroundings, the primary importance is assigned to accurate targeting. Radiobiological distinctions between cancerous and healthy tissue are in this case of secondary concern. Thus, when designing SBRT treatments, one
is led to the conclusion that the target and its close vicinity
should be exposed to high doses while regions of the organ
containing the tumor, not in direct contact with the target,
should be exposed to a level of dose below the threshold.
This irradiation pattern assumes then that the ablation of tissue is restricted to the target and its close surroundings tuMedical Physics, Vol. 35, No. 1, January 2008
114
mor and its margin and large portions of below threshold

dose exposed organs lung, liver, etc. are safe from
radiation-induced toxicity.
Justifications for the hypofractionation and the design of
treatment parameters for extracranial stereotactic radiation
therapy discussed above are not sufficient for clinical practice. There is ample necessity for clinical data to support
these conjectures. Those data were collected in the early
1990s, first in Karolinska Hospital in Stockholm, where
Blomgren and Lax first applied, on a regular basis, extracranial radiosurgery treatments.1219 Following on the experience of Gamma Knife radiosurgery, they aimed to use stereotactic guidance and well-focused dose distributions for
extracranial targets. To this end, they designed the stereotactic body immobilization frame that played a role similar to
the head helmet in cranial radiosurgery. The role of frame
was to
Immobilize the patients body to increase the accuracy
of patient positioning and patient body stability during
treatment;
Decrease the respiratory motions of internal organs in
order to make structures and organs in the body more
static during therapy; and
Stereotactically target tumors.
Radiotherapy treatment parameters required by hypofractionated SBRT were not possible before development of the
relevant technology of radiation delivery. Stereotactic targeting for extracranial targets and the ability to plan and deliver
focused on-target dose distributions are minimal requirements to facilitating SBRT. These provisions were first made
possible around 1990. This was the time when the first treatments of this type were accomplished at Karolinska, and less
frequently, at other centers. The original SBRT treatments
relied on setting the target to Cartesian coordinates of the
frame system and employing non-coplanar beam arrangements for focusing the radiation on the target.16,38 Additionally, the treatment of chest and abdominal areas employed
abdominal compression for the reduction of respiratoryinduced motion of body organs at the time of treatment delivery.
It should be noted that by modern standards of using 4D
CT imaging, IMRT dose shaping, and IGRT-reliant targeting
for radiation therapy delivery, the original tools of SBRT
were not particularly sophisticated. However, the treatment
outcomes from these early days of SBRT therapy were still
very encouraging.1241 This can only mean that recent technologies, though helpful to increase SBRT precision and efficiency, are not compulsory to properly execute hypofractionated treatments for specific classes of cancer therapy.
Essential, however, for the early success of SBRT therapy
was the vigilant attention to detail during simulation, planning, and treatment delivery. Well-executed patient immobilization and accurate repositioning of the patients body in
the frame before each treatment, together with careful transformation of coordinates of the target from the planning
115
stage to treatment setup, were able to adequately compensate

for the unavailability of 4D CT and IGRT tools.
In addition, the management of respiratory organ motions
at times when techniques such as gating, tracking, or breath
hold were unavailable was done simply and efficiently by
suppression of diaphragm motion by abdominal compression. This simple method allowed for the successful decrease
of margins to below 1 cm in any direction, accommodating
all target definition uncertainties and target motions. Observation of the extent of the target motion, or motion of target
surrogate under fluoro, competently allowed verification of
the adequacy of margins. Finally, planning has been judged
by the conformity of the prescribed dose surface with the
PTV volume surface and by the quality of the dose descent
into healthy tissue. The quality of the dose distribution
judged by the dose gradient approach mandated the multiple
non-coplanar beam arrangements.37 The dose quality also
has been judged by the careful avoidance of any beams
passage through serially functioning organs near the target.
The personal involvement of the clinician at each step of the
process was deemed indispensable, as well as committed assistance from trained physicists, dosimetrists, and radiation
therapists. The early practice of hypofractionated SBRT
therapy in Europe,1228 Asia,2932 and North America3343 has
been universally distinguished by this specifically careful attitude to the innovative therapy. As noted above, the overall
commitment of resources to SBRT treatment per fraction
may be considerably larger than in the case of traditionally
fractionated radiation therapy. However, hypofractionated
treatment means that only 15 treatment sessions will be
used per course of therapy, rather than the 2035 treatment
sessions needed for conventionally fractionated treatments.44
The exceptional outcomes of SBRT treatments leave no
doubt, however, that the additional toil per fraction is worth
the effort.
II.C. SBRTtreatment techniques and radiation
therapy technology
The precision of the setup and delivery required for hypofractionated therapies makes technologies of IMRT and
IGRT though not indispensable as explained above well
suited for efficiently carrying out SBRT treatments. In turn,
the SBRT therapy may be specifically well suited for showing the value of these new technologies to cancer therapy.
The clinical gains resulting from IMRT and IGRT may be
difficult to prove for standard fractionations. For traditional
radiation therapy, advances in local control are not expected
to be dramatic when IMRT and IGRT are implemented, and
improved survival data would be notoriously difficult to
demonstrate. However, in the case of SBRT treatment, the
combination of radical dose fractionation and the ability to
contain the toxicity, inured by superior accuracy of technologically advanced delivery, will allow more ready measurement of the improvements in local control and in survival
rates. Therefore, the overall results of SBRT delivery, when
supported by IMRT and IGRT technology, should establish
these technologies clinical credibility. However, as will be
115
explained below, the definitive sophistication of integrated

IMRT and IGRT methods for the delivery of SBRT therapy
has not been achieved yet.
The other aspect of integrated IMRT and IGRT technology in SBRT delivery is the unquestionable benefit that these
technologies bring to the efficiency of this treatment technique. Multiple tasks related to the precision of the patient
body setup, as well as the matching of body structures to
treatment beam positions, are ultimately difficult to achieve
unless automated and software-controlled beam shaping adjustments are performed in real time. This superiority of dose
delivery can be accomplished only through appropriate integration of technologies that allow recognizing patient body
geometries with treatment delivery modifications in real
time. Generally, multiple x-ray, optical, and electromagnetic
detectors are employed to register anatomical structures with
their motion.. Then, software algorithms have to be developed for analyzing data in real time and responding with
continuous treatment adjustments. The active involvement in
treatment modification in real time by equipment operators
therapists and treatment parameter-evaluating personnel
physicians, physicists has to be eliminated in this scenario
to make the treatment efficient and precise at the same time.
III. DISCUSSION
III.A. SBRTvision for future
SBRT is not yet a completely mature treatment technique.

It requires considerable growth to achieve its full potential.
Among the numerous developments needed for the enhancement of the technique, three may have the most immediate
and significant impact:
1 The main obstacle for safe application of the SBRT
treatment technique is the unavailability of data that allow unambiguous determination of the parameters for
fractionation schemes and dose prescriptions;
2 The next impediment for regular application of SBRT is
the imprecision of the dose actually delivered at treatment relative to the dose planned for the therapy. The
tightness of the margins and the intent of precise dose
shaping in the treatment volume makes the dose inaccuracy in SBRT a more concerning problem than in conventionally fractionated therapy. The main source of the
delivery imprecision is the respiratory-induced motion
of tissue at treatment.45 This is particularly significant
for IMRT treatments of chest and abdominal targets.
3 Finally, the most basic barrier for confident employment
of SBRT to a large and diverse patient population and
for diverse cases of cancer therapy is the lack of indepth understanding of the tissue response to treatment
in both short-term and long-term perspective. Interaction
between large dose fractions depends on the individual
characteristics of the tumor, patient genetic make-up,
and immune system stimulation or suppression during
radiotherapy. The information about these mechanisms
is very limited at this time.
116
Point 1 above refers to the determination of the optimal

fractionation scheme, including the level of dose prescribed
to targets and organs, dose uniformity, and dose distribution
gradient in and around structures, and the number of fractions used. The number of fractions in SBRT ranges customarily from 1 to 5 per course of treatment. The dose per fraction varies considerably in different fractionation schemes,
though it almost never exceeds 30 Gy, even for single fraction radiotherapy, and usually does not fall below 7 Gy for
hypofractionated treatments. The fractionation and geometrical parameters of the prescribed dose are customarily
nonuniform.1243 This inconsistent approach derives from the
fact that SBRT has been initiated in few centers around the
world simultaneously and few efforts have been made so far
to harmonize the rationalization and treatment practice. The
available data that could differentiate between the validity of
one approach over the other are still sparse. Therefore, more
studies are currently being designed with the goal of answering questions on optimal SBRT planning and conduct. The
common goals in these studies are
Local control of 90% or more at 5 years, actuarial,
Survival 60 80% at 5 years actuarial,
Grade III or higher toxicity 15 20%,
Ideally less invasive than surgery,
Ideally more convenient than surgery and conventionally fractionated radiotherapy,
Ideally less costly than any alternative treatment.
Moreover, these studies all need to be proven by prospective
testing, built on by experience, and based on guidelines of an
established QA program e.g., RTOG 0236 study43. The
overall purpose of these studies will be to find a balance
between benefit and harm from the range of SBRT therapies.
Generally, the first step will be to start with a phase I dose
escalation toxicity study where the only variable affecting
the outcome will be dose. The escalation will last until exceeding a predetermined level of severe toxicity usually
15 20%, determining in this fashion the highest maximum tolerable dose MTD. Then, in the phase II study the
MTD from the phase I study will designate how more patients are to be treated to confirm the toxicity with a larger
sample 15 20% severe. Finally, at phase III, the studies
will be randomized in order to eliminate the selection bias.
Point 2 has two aspects. The first is related to the uncertainty of the evaluation of the dose that is actually delivered
to the target and the organs at risk during SBRT treatments.
This uncertainty makes it difficult to correlate results of
treatments with the dose. Consequently, the precise recommendations for dosimetry for treatments of various sites with
SBRT are not yet clearly resolved. These difficulties are related to limited control of tissue motion that influenced the
dose delivery in the past, as well as to inferior dosimetry
calculation algorithms, especially those calculations that ignored the scattering contribution from uneven density of tissue in exposed volumes. For dose calculations in the presence of large volumes of different than water-equivalent
density tissue, these algorithms can lead to relatively large
116
errors and make the a posteriori evaluation of relationships

between dose and tissue response unreliable. The second aspect of imprecise imposition of dose clouds over targets and
organs at risk are related to the inability to control the delivery of radiation in the presence of the tissue motion. Recent
developments in the area of dose delivery in IMRT therapy
to moving body anatomy show that tissue motion at treatment can actually improve the quality of the dose distribution in the patient body relative to treatments delivered to
static body anatomy.46 The intriguing possibility of the dose
delivery improvement in the presence of the motion of body
anatomy during treatment creates the exceptional opportunity of enhancement of SBRT treatment delivery and requires further, careful investigations. This feature of motionadapted therapy provides a particularly valuable opportunity
of containing the dose in organs at risk especially, in serially
functioning tissue organs. The potential ability to control the
upper bound of the dose for serially functioning organs may
give the most valuable tool for extending SBRT treatment to
sites that otherwise would not have been approachable due to
unavoidable toxicity in organs bordering on treated targets.
Finally, point 3 can be seen as influencing the range of
application of SBRT as well as its efficacy. By the very nature of hypofractionation, the in-treatment modification of
therapy has limited applications in SBRT. The treatment time
interval is brief and thus, modification of the plan at treatment increase or decrease of dose, reshaping of dose distribution, modification of agents and medicaments applied in
parallel to, or post-, therapy has only limited prospect of
being applied. Recent advancements in molecular imaging
and biometrics indicate that molecular testing as well as biochemical matching before treatment or post-first/second
fraction would allow customization of SBRT to the treatment parameters. Factors such as the dose per fraction, overall dose, dose geometrical shape in the target, and in the
target vicinity, can influence the SBRT technique with a substantial benefit to a patient. This in turn can minimize the
risks of therapy delivery and enhance its local and global
worth in terms of near-term as well as long-term goal.
In conclusion, we can summarize these reflections with
the claim that the great impact of SBRT on the future of
radiotherapy, and cancer therapy in general, seems inevitable. This assertion is based on the already proven success
of the novel paradigm of SBRT hypofractionation. Moreover,
it is based on improved future benefits that may be predicted
from integration and fine-tuning of the latest technologies
with this treatment technique. The expected local control of
tumors with SBRT, in lung and in other chest and abdominal
organs, can be safely predicted from existing data as reaching the range of 80 90%.1243 Moreover, properly executed
hypofractionated radiotherapy should create the opportunity
for stimulating the immune system and allow repeating the
treatments. Thus, SBRT will likely achieve the level of surgical efficiency, without the associated surgery toxicity, making sequential treatment of multiple metastases possible. Ultimately, individualized SBRT therapies based on molecular
imaging diagnostics and genetic predictors, will permit multiple applications of curative and prophylactic SBRT treat-
117
ments with the aim of local control of tumor growth as well

as prevention of metastasis formations. The new radiation
therapy paradigm of hypofractionated SBRT, particularly
when enhanced by other modalities, and new applications of
radiosensitizers and radio protectors, should bring the prospective of turning many chest and abdominal cancers into a
chronic disease.
a

B. G. Douglas and J. F. Fowler, The effect of multiple small doses of x
rays on skin reactions in the mouse and a basic interpretation, Radiat.
Res. 66, 401426 1976.
2
A. B. Wolbarst, L. M. Chin, and G. K. Svensson, Optimization of radiation therapy: Integral-response of a model biological system, Int. J. Radiat. Oncol., Biol., Phys. 8, 17611769 1982.
3
R. J. Yaes and A. Kalend, Local stem cell depletion model for radiation
myelitis, Int. J. Radiat. Oncol., Biol., Phys. 14, 12471259 1988.
4
L. Leksell, The stereotaxic method and radiosurgery of the brain, Acta
Chir. Scand. 102, 316319 1951.
5
B. G. Haffty et al., Results of radical radiation therapy in clinical stage
I, technically operable non-small cell lung cancer, Int. J. Radiat. Oncol.,
Biol., Phys. 15, 6973 1988.
6
L. Kaskowitz et al., Radiation therapy alone for stage I non-small cell
lung cancer, Int. J. Radiat. Oncol., Biol., Phys. 27, 517523 1993.
7
D. E. Dosoretz et al., Radiation therapy in the management of medically
inoperable carcinoma of the lung: Results and implications for future
treatment strategies, Int. J. Radiat. Oncol., Biol., Phys. 24, 39 1992.
8
J. A. Hayman et al., Dose escalation in non-small cell lung cancer using
three-dimensional conformal radiation therapy: Update of a phase I trial,
J. Clin. Oncol. 19, 127136 2001.
9
T. T. Puck and P. I. Marcus, Action of x-rays on mammalian cells, J.
Exp. Med. 103, 653666 1956.
10
B. Fertil, I. Reydellet, and P. J. Deschavanne, A benchmark of cell survival models using survival curves for human cells after completion of
repair of potentially lethal damage, Radiat. Res. 138, 6169 1994.
11
L. B. Marks, Extrapolating hypofractionated radiation schemes from radiosurgery data: Regarding Hall et al., IJROBP 21:819824; 1991 and
Halland Brenner, XJROBP 25:381385; 1993, Int. J. Radiat. Oncol.
Biol. Phys. 28, 829837 1995.
12
H. Blomgren, I. Lax, I. Naslund, and R. Svanstrom, Stereotactic high
dose fraction radiation therapy of extracranial tumors using an accelerator. Clinical experience of the first 31 patients, Acta Oncol. Madr. 34,
861870 1995.
13
H. Blomgren, I. Lax, H. Goranson, T. Kraepelien, B. Nilsson, I. Naslund,
R. Svanstrom, and A. Tilikidis, Radiosurgery for tumors in the body:
Clinical experience using a new method, J. Radiosurgery 1, 6374
1998.
14
H. Blomgren, I. Lax, and I. Nslund, Extracranial stereotactic radiosurgery. Clinical results, Int. J. Radiat. Oncol., Biol., Phys. 36, 253 1996;
I. Lax, H. Blomgren, D. Larson, and I. Naslund, Extracranial stereotactic
radiosurgery of localized targets, J. Radiosurgery 1, 135148 1998.
15
H. Blomgren, I. Lax, and I. Nslund, Radiosurgery for tumors in the
body, Int. J. Radiat. Oncol., Biol., Phys. 39, 328 1997.
16
I. Lax, H. Blomgren, and I. Nslund, Extracranial stereotactic
radiosurgerymethod, Int. J. Radiat. Oncol., Biol., Phys. 36, 258
1996.
17
I. Lax, Stereotactic radiotherapy of liver tumors, J. Radiosurgery 1,
239240 1998.
18
I. Lax, H. Blomgren, I. Naslund, and R. Svanstrom, Stereotactic radiotherapy of malignancies in the abdomen, Acta Oncol. Madr. 33, 677
683 1994.
19
J. Wulf, K. Baier, G. Mueller, and M. P. Flentje, Dose-response in stereotactic irradiation of lung tumors, Radiother. Oncol. 77, 8387 2005.
20
J. Nyman, K. A. Johansson, and U. Hulten, Stereotactic hypofractionated
radiotherapy for stage I non-small cell lung cancermature results for
medically inoperable patients, Lung Cancer 51, 97103 2006.
21
P. Baumann et al., Factors important for efficacy of stereotactic body
radiotherapy of medically inoperable stage I lung cancer. A retrospective
analysis of patients treated in the Nordic countries, Acta Oncol. 45,
787795 2006.
22
F. B. Zimmermann et al., Stereotactic hypofractionated radiation therapy
1
117
for stage I non-small cell lung cancer, Lung Cancer 48, 107114 2005.
P. Fritz et al., Stereotactic, single-dose irradiation of stage I non-small
cell lung cancer and lung metastases, Radiat. Oncol. 1, 30 2006.
24
H. Hof et al., Stereotactic single-dose radiotherapy of stage I non-smallcell lung cancer NSCLC, Int. J. Radiat. Oncol., Biol., Phys. 56, 335
341 2003.
25
U. Haedinger, W. Thiele, and J. Wulf, Extracranial stereotactic radiotherapy: Evaluation of PTV coverage and dose conformity, Med. Phys.
12, 221229 2002.
26
U. Haedinger, J. Wulf, U. Oppitz, B. Olshausen, W. Thiele, and M.
Flentje, Stereotactic treatment of extracranial targets: Evaluation of planning target volume PTV to compensate for target mobility due to target
motion and inaccuracy of patient positioning *1 , Int. J. Radiat. Oncol.,
Biol., Phys. 48, 346347 2000.
27
U. Haedinger, J. Wulf, U. Oppitz, W. Thiele, and M. Flentje, Extracranial stereotactic radiotherapy: Evaluation of the PTV to compensate target
mobility and set-up inaccuracy over the complete target volume measured
by DVH from repeated CT-simulation, Int. J. Radiat. Oncol., Biol., Phys.
51, 4546 2001.
28
M. Nevinny-Stickel, R. A. Sweeney, R. J. Bale, A. Posch, T. Auberger,
and P. Lukas, Reproducibility of patient positioning for fractionated extracranial stereotactic radiotherapy using a double-vacuum technique,
Strahlenther. Onkol. 180, 117122 2004.
29
T. Xia et al., Promising clinical outcome of stereotactic body radiation
therapy for patients with inoperable Stage I/II non-small-cell lung cancer, Int. J. Radiat. Oncol., Biol., Phys. 66, 117125 2006.
30
Y. Nagata et al., Clinical outcomes of a phase I/II study of 48 Gy of
stereotactic body radiotherapy in 4 fractions for primary lung cancer using a stereotactic body frame, Int. J. Radiat. Oncol., Biol., Phys. 63,
14271431 2005.
31
R. Hara et al., Clinical outcomes of single-fraction stereotactic radiation
therapy of lung tumors, Cancer 106, 13471352 2006.
32
H. Onishi et al., Stereotactic hypofractionated high-dose irradiation for
stage I nonsmall cell lung carcinoma: Clinical outcomes in 245 subjects
in a Japanese multi-institutional study, Cancer 101, 16231631 2004.
33
R. Timmerman et al., Extracranial stereotactic radioablation: Results of
a phase I study in medically inoperable stage I non-small cell lung cancer, Chest 124, 19461955 2003.
34
L. E. Gaspar and T. E. Schefter, Analysis of extracranial stereotactic
radiosurgery via equivalent uniform dose and tumor control probability:
Practical ramifications and translational implications, Int. J. Radiat. Oncol., Biol., Phys. 54, 225 2002.
35
B. D. Kavanagh, R. D. Timmerman, S. H. Benedict, Q. Wu, T. E.
Schefter, K. Stuhr, S. McCourt, F. Newman, R. M. Cardinale, and L. E
Gaspar, How should we describe the radioblologic effect of extracranial
stereotactic radlosurgery: Equivalent uniform dose or tumor control probability? Med. Phys. 30, 321324 2003.
36
C. R. Kelsey, T. Schefter, R. Nash, P. Russ, A. E. Baron, C. Zeng, and L.
F. Gaspar, Retrospective clinicopathologic correlation of gross tumor
size of hepatocellular carcinoma: Implications for extracranial stereotactic
radiosurgery, Int. J. Radiat. Oncol., Biol., Phys. 57, S283 2003.
37
L. Papiez, R. Timmerman, C. DesRosiers, and M. Randall, Extracranial
stereotactic radioablation: Physical principles, Acta Oncol. 42, 882894
2003.
38
J. E. Reiff, M. Werner-Wasik, P. R. Anne, R. K. Valicenti, and J. M.
Galvin, Interfraction variations of the isocenter relative to coordinates of
a stereotactic body frame for the treatment of extracranial lesions, Int. J.
39
J. E. Reiff, M. Werner-Wasik, B. E. Garofola, R. K. Valicenti, P. R. Anne,
A. P. Dicker, W. J. Curran, Jr., and J. M. Galvin, The efficacy of a
stereotactic body frame in an adaptive radiotherapy program, Int. J. Radiat. Oncol., Biol., Phys. 51, 172173 2001.
40
R. D. Timmerman, L. Papiez, R. McGarry, L. Likes, C. DesRosiers, S.
Frost, and M. Williams, Extracranial stereotactic radioablation: Results
of a phase I study in stage I non-small cell lung cancer, Int. J. Radiat.
Oncol., Biol., Phys. 57, S280S281 2003.
41
R. Timmerman, L. Papiez, R. McGarry, L. Likes, C. DesRosiers, S. Frost,
and M. Williams, Extracranial stereotactic radioablation: Results of a
phase I study in medically inoperable stage I non-small cell lung cancer,
Chest 124, 19461955 2003.
42
R. Timmerman, L. Papiez, and M. Suntharalingam, Extracranial stereotactic radiation delivery: Expansion of technology beyond the brain,
Technol. Cancer Res. Treat. 2, 153160 2003.
23
118
43
R. Timmerman, R. McGarry, C. Yiannoutsos, L. Papiez, K. Tudor, J.

DeLuca, M. Ewing, R. Abdulrahman, C. DesRosiers, M. Williams, and J.
Fletcher, Excessive toxicity when treating central tumors in a phase II
study of stereotactic body radiation therapy for medically inoperable
early-stage lung cancer, J. Clin. Oncol. 24, 48334839 2006.
44
R. Timmerman, J. Galvin, J. Michalski, W. Straube, G. Ibbott, E. Martin,
R. Abdulrahman, S. Swann, J. Fowler, and H. Choy, Accreditation and
quality assurance for radiation therapy oncology group: Multicenter clinical trials using stereotactic body radiation therapy in lung cancer, Acta
118
Oncol. 45, 779786 2006.

Treating Tumors that Move with Respiration, edited by H. Urschel, L.
Kresl, J. Luketich, L. Papiez, and R. Timmerman Springer, Berlin, 2007.
46
L. Papiez, R. McMahon, and R. Timmerman, 4D IMRT planning and
deliveryFormulation and preliminary results, in The Radiotherapy Dynamic, XVth International Conference on the Use of Computers in Radiation Therapy, Vol. 1, edited by J. P. Bissonnette and D. Bahig University of Toronto, Toronto, 2007, p. 298.
45
Small fields: Nonequilibrium radiation dosimetry

Indra J. Dasa
Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania 19104
George X. Ding
Department of Radiation Oncology, Vanderbilt University School of Medicine, Nashville, Tennessee 37232
Anders Ahnesj
Department of Oncology, Radiology and Clinical Immunology, Section of Oncology,
Uppsala University, S-751 85 Uppsala & Nucletron AB, S-751 47 Uppsala, Sweden
Received 31 July 2007; revised 24 September 2007; accepted for publication 18 October 2007;
Advances in radiation treatment with beamlet-based intensity modulation, image-guided radiation
therapy, and stereotactic radiosurgery including specialized equipments like CyberKnife, Gamma
Knife, tomotherapy, and high-resolution multileaf collimating systems have resulted in the use of
reduced treatment fields to a subcentimeter scale. Compared to the traditional radiotherapy with
fields 4 4 cm2, this can result in significant uncertainty in the accuracy of clinical dosimetry.
The dosimetry of small fields is challenging due to nonequilibrium conditions created as a consequence of the secondary electron track lengths and the source size projected through the collimating
system that are comparable to the treatment field size. It is further complicated by the prolonged
electron tracks in the presence of low-density inhomogeneities. Also, radiation detectors introduced
into such fields usually perturb the level of disequilibrium. Hence, the dosimetric accuracy previously achieved for standard radiotherapy applications is at risk for both absolute and relative dose
determination. This article summarizes the present knowledge and gives an insight into the future
procedures to handle the nonequilibrium radiation dosimetry problems. It is anticipated that new
miniature detectors with controlled perturbations and corrections will be available to meet the
demand for accurate measurements. It is also expected that the Monte Carlo techniques will increasingly be used in assessing the accuracy, verification, and calculation of dose, and will aid
perturbation calculations of detectors used in small and highly conformal radiation beams. 2008
American Association of Physicists in Medicine. DOI: 10.1118/1.2815356
Key words: small field dosimetry, nonequilibrium conditions, Monte Carlo
I. INTRODUCTION
With image guidance and improved treatment delivery techniques at the core of modern radiation therapy, the treatment
fields that traditionally spanned from 4 4 cm2 up to 40
40 cm2 are now being reduced down to a subcentimeter
range in advanced and specialized radiation treatments such
as beamlet-based intensity modulated radiation therapy
IMRT, image-guided radiation therapy IGRT, tomotherapy, stereotactic radiosurgery SRS with high-resolution
multileaf collimator, Gamma Knife, and CyberKnife. In particular, the SRS, Gamma Knife, and CyberKnife rely on very
small field sizes on the order of a few millimeters to treat
tumors and spare normal structures. IMRT requires the addition of small fields with nonequilibrium conditions to treat
target volumes using optimization routines. Several dosimetric challenges due to this trend are lack of charged particle
equilibrium CPE,1,2 partial blocking of the beam source
giving rise to pronounced and overlapping penumbra,36 the
availability of small detectors for sizes comparable to the
field dimensions,7,8 and variations of the electron spectrum
inducing changes in stopping power ratios.9,10
Electronic equilibrium is a phenomenon associated with
the range of secondary particles and hence dependent on the
206
beam energy, the composition of the medium, and particularly the density of the medium. With a large variety of radiation detectors marketed by various manufacturers covering all sizes from mini to micro, types ionization chamber,
semiconductor, chemical, film, etc., and shapes thimble,
spherical, plane parallel, the choice of a suitable detector for
small field dosimetry could be a challenging and rather confusing task without proper guidelines. It is not too uncommon in clinical practice to compare measurements with various detectors and choose the detector that yields the highest
output for a given field size, or to select a measured value
that is common to several detectors, without proper consideration of the possible perturbations and corrections for each
of the detectors. Such approaches do not provide the scientific basis needed to achieve the confidence for dosimetric
accuracy commonly set for clinical practices. To deal with
these difficulties, a growing number of authors1017 have reported the comparison of measured data with Monte Carlo
simulation. However, at the present time Monte Carlo simulations cannot be assumed to invariably provide a gold standard without appropriate experimental validation. These developments challenge the conventional way of performing
dosimetric measurement and treatment planning. The pur-
0094-2405/2008/351/206/10/$23.00
206
207
Das, Ding, and Ahnesj: Small fields: Nonequilibrium radiation dosimetry
207
FIG. 1. With field sizes large enough

to yield charged particle equilibrium
CPE and fully viewed sources, the
full width at half maximum FWHM
of dose profiles yields correctly determined field sizes since the field borders will be approximately at the level
of 50% of the dose level of CPE, as
shown in panel a. When the field size
is of the same order as the charged
particle lateral diffusion distance, the
penumbra from opposing field edges
overlap, causing a small error in field
size determination from FWHM data
panel b, but completely break
down for very small fields since the
resulting curve has a lower maximum
and hence its half value will be pushed
outward from the correct position, resulting in an overestimated field size
as shown in panel c.
pose of this paper is to illustrate the problem, provide possible solutions, and predict future trends in small field radiation dosimetry.
II. PROBLEMS
II.A. What is small?
The definition of a small field in radiation dosimetry is

currently very subjective and ad hoc. There is no clear consensus definition as to what constitutes a small field. Commonly, a field size of less than 3 3 cm2 is considered outside the conventional treatment field size that needs special
attention both in dose measurements and in dose calculations. A more scientific approach is needed to set the criteria
which define a small field condition based on the beam energy and the density of the medium. There are essentially
three equilibrium factors that determine the scale if a radiation field is to be considered as small or not: i the size of
the viewable parts of the beam source as projected from the
detector location through the beam aperture; ii the size of
the detector used in measurements; and iii the electron
range in the irradiated medium. These factors are discussed
below.
II.B. Effects of the radiation source size
By collimating a beam from a source of finite width, it is

clear that below a certain field size, only a part of the source
area can be viewed from a detectors point of view. The
output will then be lower than compared to field sizes at
which the entire source can be viewed from the detectors
field of view.3,4,6,1820 The output changes differently depending on position contributing to the phenomena yielding a
blurred and widened profile as illustrated in Fig. 1. If the
entire source cannot be viewed from the center of the field,
then the geometrical penumbra is extended all over the field
cross section.3,4,19 Under such conditions traditional methods
for field size determination such as full width at half maxiMedical Physics, Vol. 35, No. 1, January 2008
mum FWHM break down, resulting in overestimated field

sizes as shown Fig. 1. It has been demonstrated by Ding et
al.21 that the beam output planar fluence profile can be
significantly influenced by the auxilary collimator jaw settings used to achieve the small field sizes. This is reflected in
Fig. 2, indicating the profile patterns for the 6 MV beam for
6 6 and 24 24 mm2 fields with various jaw settings. For
nonfocused multileaf collimators the light fields are not congruent with the radiation field due to the curved nature of the
leaves permitting variable amounts of radiation through different thicknesses of the leaves.22 This positional dependence, varying from one side of the field to the other, further
complicates the precise metrics of small field sizes.
II.C. Electron ranges and loss of CPE
The electrons produced from megavoltage photon beams

have a considerable range that gets prolonged in a lowdensity medium. Compared to the field size, the lateral range
of the electrons is the critical parameter to the CPE, rather
than the forward range of the electrons. Li et al.23 described
the lateral range of electrons which are energy dependent.
Shown in Fig. 3 are primary dose profiles in water across a
collimating edge for different beam energies, specified by
their quality index TPR20/10.24,25 This provides the information of penumbra ranges in unit density media that set the
dimensions of when small field conditions apply based on
overlapping electron distribution zones from different field
edges.26
II.D. Measurement
For photon beams, measured data are used in dose calculation to provide the absolute dose normalization at a reference field, and through a beam modeling procedure indirectly drive dose calculations based on relative data such as
total scatter factor Scp, tissue maximum ratio TMR, percent depth dose PDD, and off-axis ratio OAR. The refer-
208
208
FIG. 3. The graph shows the primary dose profile in water across a collimating edge for different beam energies, specified by their quality index
TPR20/10, ranging from 0.68 to 0.80 at an interval of 0.2. Ideal spot
source fluence profiles are assumed with only the electron transport that
contributes to the penumbra width.
DtE,r
Dtr
=
Dtref
Dtref
FIG. 2. The effects of source size and beam shaping geometry on the output
of a small field, a 6 6 mm2 and b 24 24 mm2. Both field sizes are
defined by the micro multi-leave collimator mMLC and the variation are
caused by different settings of auxilary jaws. Reproduced with permission
from Ding et al., Ref. 21.
ence dose calibration is performed according to the dosimetry protocols24,25,27 with a well-defined beam geometry,
where beam quality and dosimetric parameters are known to
a high degree of accuracy. Dose measurements with ionization chambers rely on the assumptions of cavity theory.
When the size of the cavity is smaller than the range of
charged particles originated in the medium, the cavity is
treated as nonperturbing. In such a situation, the dose to the
medium is related to the dose to the air in the cavity by the
stopping power ratios of medium to air. However, with decreasing field size, neither the CPE nor the conditions for the
cavity theory can be fulfilled due to the lateral range of the
electrons. For a small field when the CPE does not exist, the
presence of a detector can change the local level of the CPE,
adding more perturbations to complicate the problem.
Under electronic equilibrium, cavity theory describes a
method to calculate the dose D in a medium based on
measured charge in the cavity,
Q
Dt =
m
W
e
,
a
W
e
ref
a ref
where Q is the detector reading of charge, m is mass of the

air in an ion chamber, W / e is ionization potential of air, r is
the field dimension, ref is reference field size, and S / ta is
the mass collision stopping power ratio of tissue t to air
a.2830 All of the parameters in Eq. 2 are energy dependent; hence, the dose in small fields compared to the reference or calibration field 10 10 cm2 is uncertain due to
spectral variations. The measured ionization readings QE , r
are influenced by many factors as shown below:
QE,r = Qm PionPreplPwallPcecPpcf ,
where Qm is measured reading, Pion in the ion recombination,

Prepl is the replacement correction factor, Pwall is wall correction factor, Pcec is the central electrode correction factor, and
Ppcf is the perturbation correction factor as described in
TG-21 Ref. 25 and by Nahum.2 Usually the ratio of these
correction factors as shown in Eq. 2 is ignored in routine
clinical practices where CPE exists, but they cannot generally be ignored for small fields as noted by Seuntjens and
Verhaegen31 and Sauer and Wilbert.8 It was observed that for
a small volume mini-ion chamber Exradin A14P the perturbation factor is larger than unity by about 36%, 30%, and
18% for circular field diameter of 1.5, 3, and 5 mm, respectively. As noted by several other investigators,8,16,32 these
large differences are mainly due to nonequilibrium conditions which are dependent on the type and design of the
detector. The magnitude is significantly larger in low-density
media. In a heterogeneous medium, these factors in particular Prepl are not easy to predict because of the uniqueness of
measurement conditions in small fields.33,34 The Monte Carlo
simulation has been shown to be an effective tool to study
these effects.33,34 A significant amount of work is still needed
to calculate factors for the field size, beam energy, and detector geometry.
209
The second factor in Eq. 2 is usually ignored since W / e

is treated as relatively constant, even though it might depend
slightly on beam energy.35,36 This assumption is based on
small changes in the spectrum versus field size. The third
factor in Eq. 2 is of significant interest. However, it is often
ignored in clinical settings.37,38 The photon spectrum does
change9,39 with field size and location off axis; however, the
stopping power ratio is shown to be relatively insensitive for
low-energy photon beams within the accuracy of the measurements 1%.9,10,1315,4043 However, for high-energy
beams it cannot be ignored.44 With a clear knowledge of the
photon and electron spectra in small fields, the terms 2 and 3
in Eq. 2 can therefore be better estimated in the future
evaluation. Another measurement-related problem is the experimental determination of field size. When small field conditions prevail, overlapping penumbra cause traditional
methods for field size determination based on full width at
half maximum FWHM metrics to break down, resulting in
overestimated field sizes as shown in Fig. 1.
II.E. Corrections and perturbations
The main problem associated with the dosimetry of small

fields is the very presence of the detector itself that produces
a perturbation hard to quantify in a reliable way.33,34 This is
because the detector is normally different from the medium
in both composition and density. The major source of the
effect comes from the perturbation of the charged particle
fluence, which depends not only on the detector geometry
but also on the medium in which the measurement is performed, as well as on the beam energy and field size. Therefore, it is difficult to use standard correction methods in the
dosimetric measurement of the small field in heterogeneous
media because, in addition to other known corrections,25 the
value of Prepl in Eq. 3 is field size- and phantom geometrydependent.
II.F. Absolute dose
For accurate patient treatment, knowledge of the absolute

dose in a reference beam is required. This is performed
through TG21 Ref. 25 or TG-51 Ref. 27 protocols in
North American, and the IAEA protocol24 or other dosimetry
protocols developed in other countries. These protocols provide the methodology to perform dosimetry in a reference
field, usually 10 10 cm2, where the radiological parameters in the reference conditions are available. For some
treatment modalities used in SRS treatments, such as Gamma
Knife, CyberKnife, and tomotherapy, the reference beam
condition field size of 10 10 cm2 does not exist. In such
a situation there is no simple method to provide absolute or
reference dosimetry. It is often indirectly performed by transferring, extrapolating, or intercomparing among various detectors, usually film, TLD, or a small volume ion chamber.
The procedure for such a transfer is usually dependent on the
choice of an individual physicist and the measurement equipment used. The Radiological Physics Center RPC in Houston has undertaken the task of intercomparing doses from
participating institutions. Significant deviations in dosimetry
209
from the RPC values have been observed, indicating the seriousness of the dosimetric problem. The accurate dosimetry
for small field remains a problem until an appropriate and
consistent procedure is available and widely accepted by a
majority of users to calibrate nonstandard beams.
II.G. Low-density inhomogeneity
In low-density medium like the lung, small fields are subject to significant perturbations that are energy- and density
dependent. Treatment planning algorithms that use simple
one-dimensional density scaling fail to provide accurate dose
distributions as noted in several publications.4551 Advanced
treatment planning algorithms in general provide more accurate dose calculations in treatment planning.34,5254 However,
in clinical trials, a consistent approach in radiation dosimetry
is necessary. In order to make the outcome comparison
meaningful, new treatment protocols have yet to take advantage of these more accurate treatment planning algorithms,
which are now available in many commercial 3D treatment
planning systems.
III. PRESENT KNOWLEDGE
III.A. Experiment
With advances in technology, radiation detectors have

evolved and improved in quality. Various manufacturers offer a wide range of radiation detectors including ion chamber, solid-state detector, diode, TLD, scintillator, chemical,
Fricke, film, alanine, and others. These detectors can be categorized as standard, mini-, and micro-detectors depending
on the sizes 101, 102, and 103 cm3, respectively.
An assortment of ion chambers and other radiation detectors is available that can be used for a specific task in dosimetry. Ionization chambers have been widely used in radiation
dosimetry due to their near independence of energy, dose,
and dose rate. They provide a reproducible direct reading and
can be calibrated to a national standard to calculate the dose.
Ion chambers are relatively inexpensive, readily available,
and are manufactured in various shapes cylindrical, spherical, and parallel plate and sizes for various
applications.12,17,39,43,5558 For special procedures such as
SRS, Gamma knife, CyberKnife, tomotherapy, and IMRT,
the wide availability of radiation detectors questions the selection and proper use of detectors, which has been addressed by various investigators.7,10,1215,39,5967 Figure 4
shows the measured dose at central axis in the form of relative dose, Scp , for 6 and 15 MV beams measured from various detectors. A rapid drop in dose with a certain detector is
observed when the field size is decreased. This figure represents only the ratio of reading without any correction, as
noted in Eq. 3. The effect is more magnified for highenergy beams, possibly due to a nonequilibrium condition,
perturbation corrections as noted in Eq. 3, and volume
averaging.61 With various detectors, the small field produces
challenges in dose measurements with a greater probability
of significant error.
210
210
FIG. 4. Total scatter factor Scp versus

field size measured with various available radiation detectors for a 6 MV
and b 15 MV beams. The data are
plotted without the consideration of
any corrections or perturbations and
simply the ratio of reading in a field
size to the reference field.
Semiconductor diode detectors are being widely used for

patient dosimetry for both photon and electron beams. Diode
detectors have small sensitive volumes and are categorized
as mini- and micro-detectors. Characteristics include quick
response time microseconds compared to milliseconds of an
ion chamber, excellent spatial resolution, absence of external bias, and high sensitivity. In addition, stopping power
ratios for diodes are nearly energy independent, but the presence of low-energy photons causes problems due to the increased photoelectron cross sections in silicon compared to
water. The response of the diode detectors depends on temperature, dose rate SSD or wedge, and energy.6870 Depending on the design of the diode, some may have angular
dependence as well. In order to achieve the required accuracy recommended by AAPM,71 these effects should be corrected, or a diode with a minimum dose rate and energy
effect should be used. Often diode detector measurements

are compared with an ion chamber to provide confidence in
small field dosimetry. The stereotactic photon diode SFD
with nearly micron-size sensitive volume has become an attractive choice for small field dosimetry.23,7274
Diamond detectors are solid-state detectors with large signals and the sensitive volume is relatively small 1.0
6.0 mm3, which makes them ideal for small field dosimetry and for beam profile measurements.75 When ionizing
radiation is absorbed, it induces a temporary change in the
electrical conductivity of the material.7680 The response of a
diamond detector is directly proportional to the absorbed
dose rate. Diamond detectors do not exhibit any directional
dependence and they are tissue-equivalent. Diamond detectors do exhibit a small dependence on dose rate, but a cor-
211
rection can be applied for the measurements. The detector is

hard to manufacture and hence more expensive than other
solid-state detectors.
Thermoluminescent dosimetry TLD Ref. 81 has been
used for point dose measurements and in vivo dosimetry. The
TLD material comes in several different forms, such as rods,
chips, and powder. Rods and chips are reusable once they
have been properly annealed. TLD exhibits an energy dependence as well as a dose dependence. The accuracy is limited
to the irradiation and measuring techniques. It is usually suitable for cross reference of point dose in small fields and
IMRT.
Film is used for relative dose measurement. There are two
types of films: silver halides82 and Gafchromic.83 Silver halide films require processing, whereas Gafchromic films are
self-developing. TG-69 provided an overview82 of silver halide films. Gafchromic film has some superior characteristics; however, its use is limited to relative dosimetry.84 Even
though film exhibits strong energy dependence, it does provide a planar dose maps in small fields16,18,85 that is superior
to other detectors.
Metal-oxide silicon semiconductor field-effect transistor
MOSFET dosimeters have been investigated for their use
in clinical dosimetry86 and IMRT verification.87 Due to their
small size, MOSFETs are ideal for small field dosimetry in
low-density medium,33 brachytherapy, and in vivo dosimetry.
The MOSFET detectors are relatively small in size with an
active area of 0.2 0.2 mm2. It is energy independent in
MV beams. Also, it is relatively independent of dose rate and
temperature. It has been noted that MOSFET dosimeters are
similar to conventional dosimeters in reproducibility, linearity, energy, and angular responses.86 However, the MOSET
detector is used mainly for specialized point dose measurements and has a short life. It also requires repeated calibration for accurate dose measurements.
Bang gel detectors88 are tissue-equivalent and provide a
3D dose map with high spatial resolution. They are energy
independent over a wide range of energies, making them
ideal for measuring three-dimensional dose distributions. Pappas et al.89,90 provided satisfactory data for small SRS cone
with gel detectors. The downside to gel dosimetry is that it
takes considerable fabrication time for the proper working
conditions. Bang gel readout is based on imaging techniques
that are susceptible to imaging artifacts. The new research
using MAGIC gel has shown some promising results in its
usefulness in dosimetry of SRS and IMRT.91
Radiophotoluminescent glass plates have been used successfully to measure dose in SRS and Cyberknife.13,66,67
Scintillator detectors in various forms have also provided
dosimetry in small and elongated fields.9294 Alanine pallets
have been attempted to measure dose in small fields using
the electron spin resonance ESR method;95,96 however,
such devices are limited and may not be used clinically except to verify the point dose at a central location or in a
national laboratory.
211
III.B. Theoretical and analytical approach
To derive accurate beam profiles from large volume

chamber measurements, deconvolution and extrapolation
methods have been used.59,61,90,97102 The confidence in dosimetry is greater with larger fields; hence, various mathematical approaches have been proposed for getting either
the relative or absolute dose.103105 Sauer et al.8 and Cheng et
al.72 provided an extrapolation method and mathematical expressions for output in small fields accurately.
III.C. Monte Carlo simulation
Monte Carlo MC approaches are rapidly finding a niche

where measurements are not possible or rather difficult. Various reports have highlighted the feasibility of MC for small
field dosimetry.9,34,106110 Monte Carlo simulation provides
an opportunity to investigate most of the correction factors as
discussed in Eq. 3. It also provides a standard against other
techniques for the measurement of relative dose and possibly
absolute dose within acceptable accuracy. Monte Carlo simulation also provides the opportunity to understand dosimetry
in low-density materials where measurements are difficult
with nonequilibrium conditions.45,46,4851,54 There are two approaches in using Monte Carlo techniques to improve the
accuracy of radiation dosimetry. One is to obtain correction
factors for detectors used in the measurements. The other is
to calculate the dosimetry quantities directly equivalent to
performing a measurement in an ideal condition. However,
the use of Monte Carlo needs to be verified with respect to
beam modeling parameters e.g., source size and energy,
etc..
IV. FUTURE DIRECTIONS
IV.A. Absolute dose
There is a growing trend and availability of specialized

treatment machines that cannot provide the standard 10
10 cm2 field for reference calibration as required by standard dosimetry protocols. With advances in treatment devices delivering subcentimeters treatment fields, there is a
need for absolute dosimetry protocol. The normal use of
some of these machines is to produce dose distributions for
targets large enough to establish CPE, but from the superposition of smaller fields. Hence, to calibrate these machines
under working conditions representative of their normal operating mode, dosimetry protocols need to be reviewed and
formulated to accommodate these new modalities. The International Atomic Energy Agency IAEA and American Association of Physicists in Medicine AAPM have formed a
joint task group to address the absolute dose issue in small
fields. As a result, an international protocol will be provided
for beam quality specification in small fields, detectorspecific correction, and perturbation factors. Any detectorrelated problem with responses to the superposition of timeseparated small field condition subfields should be addressed
and investigated so as to provide beam calibration protocols
with the same accuracy as present standard beam protocols,
i.e., approximately 2%.
212
IV.B. Better detectors
Advances in radiation detectors and specialized treatment

techniques have fueled the need for better and suitable detectors. Many types of detectors have been used in small
fields
and
cross
compared
with
other
detectors;10,13,14,17,41,43,55,56,62,73,74,84,89,90,93,94,96,111113 however, the particular perturbation corrections are not known in
detail. It is expected that calculation-aided dosimetry will be
available where specific correction and perturbation factors
are either precalculated for irradiation geometry or calculated
online using state-of-the-art radiation transport codes, e.g.,
Monte Carlo. With improved manufacturing techniques with
the emphasis on making reproducible detectors, it is likely
that empirical corrections in hardware e.g., energycompensated shielding on diodes will be replaced by calculated correction factors. This type of calculation-aided detector could provide energy, dose, and dose rate independence
suitable for small field dosimetry.
IV.C. Treatment planning, beam modeling, and dose
calculations
Apart from the accuracy in measured data, the treatment

planning systems TPS should provide proper modeling to
adequately predict the dose distribution in nonequilibrium
conditions and with inhomogeneous medium. Properly designed multisource modeling using either accurate measured
data3,4,6 or Monte Carlo derived data, together with accurate
dose calculation algorithms that handle non-CPE conditions,
will provide acceptable dose distribution. To meet the demand, the TPS vendors should take responsibility for providing suitable implementations of either convolution/
superposition algorithms,26 Monte Carlo, or any other
methods known to be capable of handling small field conditions. Collaboration with clinicians and treatment planners
will provide planning methodology for lung tumors with respect to the use of margins and immobilization techniques as
dense tumors increase CPE and hence dose, compared to the
surrounding low-density lung tissue.
IV.D. Monte Carlo techniques and radiation transport
calculations
The accurate determination of dose distributions plays an

essential role in the success of radiotherapy. Explicit radiation transport calculations based on Monte Carlo or other
techniques will increasingly play an important role in nonequilibrium dosimetry. With the aid of calculations in the
future, the measurement accuracy should be greatly improved for these cases. It has been shown that, even when the
experimental measurements were correct, conclusions drawn
from the measured data could be incorrect because the results were misinterpreted due to the lack of knowledge of the
perturbation from the presence of a detector.34,114 In addition,
the Monte Carlo simulated beams can be calibrated against
measurements under controlled conditions where the measurements can be determined accurately. For example, a
Monte Carlo simulated incident megavoltage beam can be
212
calibrated in a reference condition per incident electron at

target in the linear accelerator. Since the geometry of the
accelerator head can be modeled in detail, the output or the
entire realistic beam of any field size can be accurately determined by radiation transport calculations. Therefore, the
nonequilibrium dosimetry for the small field in heterogeneous media can be investigated and well understood. The
Monte Carlo technique will not only play a crucial role in the
small beam output calibration but will also contribute to increased accuracy in patient dose calculations. These types of
calculations will play an increasing role in dose verification,
beam analysis, and for direct calculation of dose distribution
in treatment planning and treatment delivery.
V. CONCLUSIONS
It is expected that the accuracy of small field dosimetry
under a nonequilibrium condition can be significantly improved based on the following developments:
New protocols for absolute dosimetry in small and nonequilibrium condition will be developed to provide procedures for accurate absorbed dose calibrations for new
treatment modalities such as Gamma Knife, CyberKnife, tomotherapy, and other devices that do not
have standard reference field sizes. This will effectively
reduce the uncertainty and variations among different
centers in the absolute reference dose calibration.
Small volume detectors ion chambers, diodes, and others will be developed that have minimum perturbations
due to its presence and composition. Also, such detectors will have minimum energy, dose, and dose rate
dependence.
The accuracy of small field dosimetry will be greatly
improved by Monte Carlo simulations, especially under
extreme conditions, for small fields such as beam size
3 3 cm2 with inhomogeneity. Current measurement uncertainty 5% will be reduced. Monte Carlo
calculations will be able to accurately relate the dose to
a small field irradiated in an inhomogeneous media to
the dose of reference conditions in which the radiation
beam is calibrated. This will ensure that the uncertainty
of the dose determination in a small treatment field under extreme conditions remains similar in magnitude to
the uncertainty of a large reference field where the
beam is calibrated.
Monte Carlo simulations will provide correction parameters, such as correction factors for specific detectors
and for specific measurement conditions, and the stopping power ratio as a function of field size and beam
energy. These additional data will be available for routine use and will greatly reduce the measurement uncertainty under the nonequilibrium radiation conditions.
More accurate implementation of model-based calculation algorithms as well as direct Monte Carlo methods
will be available in commercial treatment planning systems for accurate dose calculation under the nonequilibrium radiation conditions.
213

[email protected];
b
Telephone: 215-662-6472; Fax: 215-349-5978.
1
F. H. Attix, Introduction to Radiological Physics and Radiation Dosimetry Wiley & Sons, New York, 1986.
2
A. E. Nahum, Perturbation effects in dosimetry. I. Kilovoltage x-rays
and electrons, Phys. Med. Biol. 41, 15311580 1996.
3
M. B. Sharpe, D. A. Jaffray, J. J. Battista, and P. Munro, Extrafocal
radiation: A unified approach to the prediction of beam penumbra and
output factors for megavoltage x-ray beams, Med. Phys. 22, 20652074
1995.
4
T. C. Zhu and B. E. Bjarngard, The head-scatter factor for small field
sizes, Med. Phys. 21, 6568 1994.
5
T. C. Zhu and B. E. Bjarngard, The fraction of photons undergoing head
scatter in x-ray beams, Phys. Med. Biol. 40, 11271134 1995.
6
T. C. Zhu, B. E. Bjarngard, and H. Shackford, X-ray source and the
output factor, Med. Phys. 22, 793798 1995.
7
P. Francescon, S. Cora, C. Cavedon, P. Scalchi, S. Reccanello, and F.
Colombo, Use of a new type of radiochromic film, a new parallel-plate
micro-chamber, MOSFETs, and TLC 800 microcubes in the dosimetry of
small beams, Med. Phys. 25, 503511 1998.
8
O. A. Sauer and J. Wilbert, Measurement of output factors for small
photon beams, Med. Phys. 34, 19831988 2007.
9
F. Verhaegen, I. J. Das, and H. Palmans, Monte Carlo dosimetry study of
6 MV stereotactic radiosurgery unit, Phys. Med. Biol. 43, 27552768
1998.
10
F. Snchez-Doblado, P. Andreo, R. Capote, A. Leal, M. Perucha, R. Arrans, L. Nunez, E. Mainegra, J. I. Lagares, and E. Carrasco, Ionization
chamber dosimetry of small photon fields: A Monte Carlo study on
stopping-power ratios for radiosurgery and IMRT beams, Phys. Med.
Biol. 48, 20812099 2003.
11
F. Crop, N. Reynaert, G. Pittomvils, L. Paelinck, W. De Gersem, C. De
Wagter, L. Vakaet, W. De Neve, and H. Thierens, Monte Carlo modeling
of the ModuLeaf miniature MLC for small field dosimetry and quality
assurance of the clinical treatment planning system, Phys. Med. Biol. 52,
32753290 2007.
12
F. Snchez-Doblado, R. Capote, A. Leal, J. V. Rosello, J. I. Lagares, R.
Arrans, and G. H. Hartmann, Micro ionization chamber for reference
dosimetry in IMRT verification: Clinical implications on OAR dosimetric
errors, Phys. Med. Biol. 50, 959970 2005.
13
F. Araki, T. Ikegami, T. Ishidoya, and H. D. Kubo, Measurements of
Gamma-Knife helmet output factors using a radiophotoluminescent glass
rod dosimeter and a diode detector, Med. Phys. 30, 19761981 2003.
14
F. Araki, Monte Carlo study of a Cyberknife stereotactic radiosurgery
system, Med. Phys. 33, 29552963 2006.
15
R. Capote, F. Snchez-Doblado, A. Leal, J. I. Lagares, R. Arrans, and G.
H. Hartmann, An EGSnrc Monte Carlo study of the microionization
chamber for reference dosimetry of narrow irregular IMRT beamlets,
Med. Phys. 31, 24162422 2004.
16
K. A. Paskalev, J. P. Seuntjens, H. J. Patrocinio, and E. B. Podgorsak,
Physical aspects of dynamic stereotactic radiosurgery with very small
photon beams 1.5 and 3 mm in diameter, Med. Phys. 30, 111118
2003.
17
F. Haryanto, M. Fippel, W. Laub, O. Dohm, and F. Nusslin, Investigation
of photon beam output factors for conformal radiation therapy-Monte
Carlo simulations and measurements, Phys. Med. Biol. 47, N133N143
2002.
18
X. R. Zhu, J. J. Allen, J. Shi, and W. E. Simon, Total scatter factors and
tissue maximum ratios for small radiosurgery fields: Comparison of diode
detectors, a parallel-plate ion chamber, and radiographic film, Med.
Phys. 27, 472477 2000.
19
T. C. Zhu and K. Manbeck, CT reconstruction of the x-ray source profile
of a medical accelerator, Proc. SPIE 2132, 242253 1994.
20
P. Munro, J. A. Rawlinson, and A. Fenster, Therapy imaging: Source
sizes of radiotherapy beams, Med. Phys. 15, 517524 1988.
21
G. X. Ding, D. M. Duggan, and C. W. Coffey, Commissioning stereotactic radiosurgery beams using both experimental and theoretical methods, Phys. Med. Biol. 51, 25492566 2006.
22
A. L. Boyer, T. G. Ochran, C. E. Nyerick, T. J. Waldron, and C. J.
Huntzinger, Clinical dosimetry for implementation of a multileaf collimator, Med. Phys. 19, 12551261 1992.
23
X. A. Li, M. Soubra, J. Szanto, and L. H. Gerig, Lateral electron equilibrium and electron contamination in measurements of head-scatter facMedical Physics, Vol. 35, No. 1, January 2008
213
tors using miniphantoms and brass caps, Med. Phys. 22, 11671170
1995.
24
I. A. E. A. Report No. 398, Absorbed Dose Determination in External
Beam Radiotherapy: An International Code of Practice for Dosimetry on
Standards of Absorbed Dose to Water, Technical Reports Series No. 398,
International Atomic Energy Agency, 2000.
25
Task Group-21, Radiation Therapy Committee, AAPM, A protocol for
the determination of absorbed dose from high-energy photon and electron
beams, Med. Phys. 10, 741771 1983.
26
T. Nyholm, J. Olofsson, A. Ahnesjo, and M. Karlsson, Modeling lateral
beam quality variations in pencil kernel based photon dose calculations,
Phys. Med. Biol. 51, 41114118 2006.
27
P. R. Almond, AAPMs TG-51 protocol for clinical reference dosimetry
of high-energy photon and electron beams, Med. Phys. 26, 18471870
1999.
28
P. Andreo and A. Brahme, Stopping power data for high-energy photon
beams, Phys. Med. Biol. 31, 839858 1986.
29
P. Andreo, J. R. Cunningham, K. Hohlfeld, and H. Svensson, Absorbed
Dose Determination in Photon and Electron Beams, An International
Code of Practice, Technical Report Series No. 277 International Atomic
Energy Agency, Vienna, 1987.
30
P. Andreo and A. E. Nahum, Stopping-power ratio for a photon spectrum
as a weighted sum of the values for monoenergetic photon beams, Phys.
Med. Biol. 30, 10551065 1985.
31
J. Seuntjens and F. Verhaegen, Comments on Ionization chamber dosimetry of small photon fields: A Monte Carlo study on stopping-power
ratios for radiosurgery and IMRT beams, Phys. Med. Biol. 48, L4345
2003.
32
S. Li, A. Rashid, S. He, and D. Djajaputra, A new approach in dose
measurement and error analysis for narrow photon beams beamlets
shaped by different multileaf collimators using a small detector, Med.
Phys. 31, 20202032 2004.
33
G. X. Ding, D. M. Duggan, B. Lu, D. E. Hallahan, A. Cmelak, A. Malcolm, J. Newton, D. Mathew, and C. W. Coffey, Impact of inhomogeneity corrections on dose coverage in treatment of lung cancer using
stereotactic body radiation therapy, Med. Phys. 34, 29852994 2007.
34
G. X. Ding, D. M. Duggan, and C. W. Coffey, Comment on Testing of
analytical anisotropic algorithm for photon dose calculation Med. Phys.
33, 41304148 2006, Med. Phys. 34, 3414 2007.
35
ICRU 31, Average energy required to produce an ion pair, ICRU Report 31, International Commission on Radiation Units and Measurements,
Washington, D.C., 1979.
36
R. Huber, D. Combecher, and G. Burger, Measurement of average energy required to produce an ion pair W value for low energy ions in
several gases, Radiat. Res. 101, 237251 1985.
37
I. J. Das, M. B. Downes, B. W. Corn, W. J. Curran, M. W. Wasik, and D.
W. Andrews, Characteristics of a dedicated linear accelerator-based stereotactic radiosurgery-radiotherapy unit, Radiother. Oncol. 38, 6168
1996.
38
S. C. Sharma, J. T. Ott, J. B. Williams, and D. Dickow, Commissioning
and acceptance testing of a CyberKnife linear accelerator, J. Appl. Clin.
Med. Phys. 8, 119125 2007.
39
A. Mack, S. G. Scheib, J. Major, S. Gianolini, G. Pazmandi, H. Feist, H.
Czempiel, and H. J. Kreiner, Precision dosimetry for narrow photon
beams used in radiosurgery-determination of Gamma Knife output factors, Med. Phys. 29, 20802089 2002.
40
P. Bjrk, T. Kns, and P. Nilsson, Measurements of output factors with
different detector types and Monte Carlo calculations of stopping-power
ratios for degraded electron beams, Phys. Med. Biol. 49, 44934506
2004.
41
A. L. McNiven, M. Mulligan, T. Kron, and J. J. Battista, The response of
prototype plane-parallel ionization chambers in small megavoltage x-ray
fields, Med. Phys. 33, 39974004 2006.
42
G. G. Zhang, D. W. O. Rogers, J. E. Cygler, and T. R. Mackie, Effects of
changes in stopping-power ratios with field size on electron beam relative
output factors, Med. Phys. 25, 17111716 1998.
43
H. Bouchard and J. Seuntjens, Ionization chamber-based reference dosimetry of intensity modulated radiation beams, Med. Phys. 31, 2454
2465 2004.
44
A. Wu, R. D. Zwicker, A. M. Kalend, and Z. Zheng, Comments on dose
measurements for a narrow beam in radiosurgery, Med. Phys. 20, 777
779 1993.
45
A. O. Jones, I. J. Das, and F. L. Jones, A Monte Carlo study of IMRT
214
beamlets in homogeneous media, Med. Phys. 30, 296300 2003.

A. O. Jones and I. J. Das, Comparison of inhomogeneity correction
algorithms in small photon fields, Med. Phys. 32, 766776 2005.
47
J. Y. Cheung, B. K. Ng, and K. N. Yu, Dose enhancement close to
platinum implants for the 4, 6, and 10 MV stereotactic radiosurgery,
Med. Phys. 31, 27872791 2004.
48
H. Saitoh, T. Fujisaki, R. Sakai, and E. Kunieda, Dose distribution of
narrow beam irradiation for small lung tumor, Int. J. Radiat. Oncol. Biol.
Phys. 53, 13801387 2002.
49
H. A. Al-Hallaq, C. S. Reft, and J. C. Roeske, The dosimetric effects of
tissue heterogeneities in intensity-modulated radiation therapy IMRT of
the head and neck, Phys. Med. Biol. 51, 11451156 2006.
50
L. Paelinck, N. Reynaert, H. Thierens, W. De Neve, and C. De Wagter,
Experimental verification of lung dose with radiochromic film: Comparison with Monte Carlo simulations and commercially available treatment
planning systems, Phys. Med. Biol. 50, 20552069 2005.
51
T. Krieger and O. A. Sauer, Monte Carlo- versus pencil-beam-/
collapsed-cone-dose calculation in a heterogeneous multi-layer phantom,
Phys. Med. Biol. 50, 859868 2005.
52
M. M. Aspradakis, R. H. Morrison, N. D. Richmond, and A. Steele,
Experimental verification of convolution/superposition photon dose calculations for radiotherapy treatment planning, Phys. Med. Biol. 48,
28732893 2003.
53
A. Ahnesj, Collapsed cone convolution of radiant energy for photon
dose calculation in heterogeneous media, Med. Phys. 16, 577592
1989.
54
I. J. Chetty, P. M. Charland, N. Tyagi, and D. McShan, Photon beam
relative dose validation of the DPM Monte Carlo code in lung-equivalent
media, Med. Phys. 30, 563573 2003.
55
C. F. Serago, P. V. Houdek, G. H. Hartmann, D. S. Saini, M. S. Serago,
and A. Kaydee, Tissue maximum ratios and other parameters of small
circular 4, 6, 10, 15 and 24 MV x-ray beams for radiosurgery, Phys.
Med. Biol. 37, 19431956 1992.
56
P. Francescon, S. Cora, and P. Chiovati, Dose verification of an IMRT
treatment planning system with BEAM, EGS-based Monte Carlo code,
Med. Phys. 30, 144157 2003.
57
J. C. L. Chow, M. Seguin, and A. Alexander, Dosimetric effect of collimating jaws for small multileaf collimated fields, Med. Phys. 32, 759
765 2005.
58
H.-R. Lee, M. Pankuch, and J. Chu, Evaluation and characterization of
parallel plate microchambers functionalities in small beam dosimetry,
Med. Phys. 29, 24892496 2002.
59
R. K. Rice, J. L. Hansen, G. K. Svensson, and R. L. Siddon, Measurements of dose distributions in small beams of 6 MV x-rays, Phys. Med.
Biol. 32, 10871099 1987.
60
B. E. Bjrngard, J.-S. Tsai, and R. K. Rice, Doses on the central axes of
narrow 6-MV x-ray beams, Med. Phys. 17, 794799 1990.
61
W. U. Laub and T. Wong, The volume effect of detectors in the dosimetry of small fields used in IMRT, Med. Phys. 30, 341347 2003.
62
C. Martens, C. De Wagter, and W. De Neve, The value of the PinPoint
ion chamber for characterization of small field segments used in intensitymodulated radiotherapy, Phys. Med. Biol. 45, 25192530 2000.
63
I. J. Das, M. B. Downes, A. Kassaee, and Z. Tochner, Choice of radiation detector in dosimetry of stereotactic radiosurgery-radiotherapy, J.
Radiosurg. 3, 177185 2000.
64
L. B. Leybovich, A. Sethi, and N. Dogan, Comparison of ionization
chambers of various volumes for IMRT absolute dose verification, Med.
Phys. 30, 119123 2003.
65
J. W. Sohn, J. F. Dempsey, T. S. Suh, and D. A. Low, Analysis of
various beamlet sizes for IMRT with 6 MV photons, Med. Phys. 30,
24322439 2003.
66
F. Aaki, T. Ishidoya, T. Ikegami, N. Moribe, and Y. Yamashita, Application of a radiophotoluminescent glass plate dosimeter for small field
dosimetry, Med. Phys. 32, 15481554 2005.
67
F. Araki, N. Moribe, T. Shimonobou, and Y. Yamashita, Dosimetric
properties of radiophotoluminescent glass rod detector in high-energy
photon beams from a linear accelerator and cyber-knife, Med. Phys. 31,
19801986 2004.
68
G. Rikner and E. Grusell, General specifications for silicon semiconductors for use in radiation dosimetry, Phys. Med. Biol. 32, 11091117
1987.
69
A. S. Saini and T. C. Zhu, Temperature dependence of commercially
available diode detectors, Med. Phys. 29, 622630 2002.
46
70
214
A. S. Saini and T. C. Zhu, Dose rate and SDD dependence of commercially available diode detectors, Med. Phys. 31, 914924 2004.
71
TG-62, Diode In Vivo Dosimetry for Patients Receiving External Beam
Radiation Therapy: Report of the AAPM Radiation Therapy Committee
Task Group No. 62 Medical Physics Publishing, Madison, WI, 2005.
72
C. W. Cheng, S. H. Cho, M. Taylor, and I. J. Das, Determination of zero
field size percent depth doses and tissue maximum ratios for stereotactic
radiosurgery and IMRT dosimetry: Comparison between experimental
measurements and Monte Carlo simulation, Med. Phys. 34, 31493157
2007.
73
C. McKerracher and D. I. Thwaites, Assessment of new small-field detectors against standard-field detectors for practical stereotactic beam data
acquisition, Phys. Med. Biol. 44, 21432160 1999.
74
C. McKerracher and D. I. Thwaites, Notes on the construction of solidstate detectors, Radiother. Oncol. 79, 348351 2006.
75
P. Bjrk, T. Kns, and P. Nilsson, Comparative dosimetry of diode and
diamond detectors in electron beams for intraoperative radiation therapy,
Med. Phys. 27, 25802588 2000.
76
M. Heydarian, P. W. Hoban, W. A. Beckham, I. A. Borchardt, and A. H.
Beddoe, Evaluation of a PTW diamond detector for electron beam measurements, Phys. Med. Biol. 38, 10351042 1993.
77
P. W. Hoban, M. Heydarian, W. A. Beckham, and A. H. Beddoe, Dose
rate dependence of a PTW diamond detector in the dosimetry of a 6 MV
Photon beam, Phys. Med. Biol. 39, 12191229 1994.
78
V. S. Khrunov, S. S. Martynov, S. M. Vatnisky, I. A. Ermakov, A. M.
Chervjakov, D. L. Karlin, V. I. Fominych, and Y. V. Tarbeyev, Diamond
detectors in relative dosimetry of photon, electron and proton radiation
fields, Radiat. Prot. Dosimetry 33, 155157 1990.
79
W. U. Laub, T. W. Kaulich, and F. Nusslin, Energy and dose rate dependence of a diamond detector in the dosimetry of 425 MV photon
beams, Med. Phys. 24, 535536 1997.
80
S. Vatnitsky and H. Jrvinen, Application of natural diamond detector
for the measurement of relative dose distributions in radiotherapy, Phys.
Med. Biol. 38, 173184 1993.
81
Y. S. Horowitz, The theoretical and microdosimetric basis of thermoluminescence and applications to dosimetry, Phys. Med. Biol. 26, 765824
1981.
82
S. Pai, I. J. Das, J. F. Dempsey, K. L. Lam, T. J. LoSasso, A. J. Olch, J.
R. Palta, L. E. Reinstein, D. Ritt, and E. E. Wilcox, TG-69: Radiographic film for megavoltage beam dosimetry, Med. Phys. 34, 2228
2258 2007.
83
A. Niroomand-Rad, C. R. Blackwell, B. M. Coursey, K. P. Gall, J. M.
Galvin, W. L. McLaughlin, A. S. Meigooni, R. Nath, J. E. Rodgers, and
C. G. Soares, Radiographic film dosimetry: Recommendations of AAPM
Radiation Therapy Committee Task Group 55, Med. Phys. 25, 2093
2115 1998.
84
W. L. McLaughlin, C. G. Soares, J. A. Sayeg, E. C. McCullough, R. W.
Kline, A. Wu, and A. H. Maitz, The use of a radiochromic detector for
the determination of stereotactic radiosurgery dose characteristics, Med.
Phys. 21, 379388 1994.
85
R. Ramani, A. W. Lightstone, D. L. Mason, and P. F. OBrien, The use
of radiochromic film in treatment verification of dynamic stereotactic radiosurgery, Med. Phys. 21, 389392 1994.
86
R. Ramani, S. Russell, and P. F. OBrien, Clinical dosimetry using
MOSFETs, Int. J. Radiat. Oncol. Biol. Phys. 37, 959964 1997.
87
C. F. Chuang, L. Verhey, and P. Xia, Investigation of the use of
MOSFET for clinical IMRT dosimetric verification, Med. Phys. 29,
11091115 2002.
88
G. S. Ibbott, M. J. Maryanski, P. Eastman, S. D. Holcomb, Y. Zhang, R.
G. Avison, M. Sanders, and J. C. Gore, Three-dimensional visualization
and measurement of conformal dose distributions using magnetic resonance imaging of BANG polymer gel dosimeters, Int. J. Radiat. Oncol.
Biol. Phys. 38, 10971103 1997.
89
E. Pappas, L. Petrokokkinos, A. Angelopoulos, T. G. Maris, M. Kozicki,
I. Dalezios, and V. Kouloulias, Relative output factor measurements of a
5 mm diameter radiosurgical photon beam using polymer gel dosimetry,
Med. Phys. 32, 15131520 2005.
90
E. Pappas, T. G. Maris, A. Papadakis, F. Zacharopoulou, J. Damilakis, N.
Papanikolaou, and N. Gourtsoyiannis, Experimental determination of the
effect of detector size on profile measurements in narrow photon beams,
Med. Phys. 33, 37003710 2006.
91
H. Whitney, G. X. Ding, C. W. Coffey and J. C. Gore, Gel dosimetry for
measuring a complex 3D dose distribution in stereotactic radiosurgery,
215
Med. Phys. 34, 2393 abstract 2007.

A. S. Beddar, A new scintillator detector system for the quality assurance of Co-60 and high-energy therapy machines, Phys. Med. Biol. 39,
253263 1994.
93
D. Letourneau, J. Pouliot, and R. Roy, Miniature scintillating detector
for small field radiation therapy, Med. Phys. 26, 25552561 1999.
94
M. Westermark, J. Arndt, B. Nilsson, and A. Brahme, Comparative dosimetry in narrow high-energy photon beams, Phys. Med. Biol. 45, 685
702 2000.
95
F. Chen, C. F. O. Graeff, and O. Baff, K-band EPR dosimetry: Smallfield beam profile determination with miniature alanine dosimeter, Appl.
Radiat. Isot. 62, 267271 2005.
96
S. Duane, D. Nicholas, H. Palmans, B. Schaeken, J. Sephton, P. Sharpe,
R. Thomas, M. Tomsej, K. Tournel, D. Verellen, and S. Vynckier, Alanine and ion chamber dosimetry in helical tomotherapy, Radiother. Oncol. 81, S45 2006.
97
R. K. Rice, J. L. Hansen, L. M. Chin, B. J. Mijnheer, and B. E. Bjrngard,
The influence of ionization chamber and phantom design on the measurement of lung dose in photon beam, Med. Phys. 15, 884890 1988.
98
C. H. Sibata, H. C. Mota, A. S. Beddar, P. D. Higgins, and K. H. Shin,
Influence of detector size in photon beam profile measurements, Phys.
Med. Biol. 36, 621631 1991.
99
G. Bednarz, S. Huq, and U. F. Rosenow, Deconvolution of detector size
effect for output factor measurement for narrow Gamma Knife radiosurgery beams, Phys. Med. Biol. 47, 36433649 2002.
100
P. D. Higgins, C. H. Sibata, L. Siskind, and J. W. Sohn, Deconvolution
of detector size effect for small field measurement, Med. Phys. 22,
16631666 1995.
101
F. Garcia-Vicente, J. M. Delgado, and C. Peraza, Experimental determination of the convolution kernel for the study of the spatial response of a
detector, Med. Phys. 25, 202207 1998.
102
W. Ulmer and W. Kaissl, The inverse problem of a Gaussian convolution
and its application to the finite size of the measurement chambers/
detectors in photon and proton dosimetry, Phys. Med. Biol. 48, 707727
2003.
92
103
215
P. Nizin, G. X. Qian, and H. Rashid, Zero-field dose data for 60C and
other high-energy photon beams in water, Med. Phys. 20, 13531360
1993.
104
P. S. Nizin and R. B. Mooij, An approximation of central-axis absorbed
dose in narrow photon beams, Med. Phys. 24, 17751780 1997.
105
P. S. Nizin and R. B. Mooij, Tissue-air ratios for narrow 60C gamma-ray
beams, Med. Phys. 25, 14581463 1998.
106
P. K. Kijewski, B. E. Bjarngard, and P. L. Petti, Monte Carlo calculations
of scatter dose for small field sizes in a 60C beam, Med. Phys. 13, 7477
1986.
107
G. X. Ding, D. M. Duggan, and C. W. Coffey, Commissioning stereotactic radiosurgery beams using both experimental and theoretical methods, Phys. Med. Biol. 51, 25492566 2006.
108
G. X. Ding, Dose discrepancies between Monte Carlo calculations and
measurements in the buildup region for a high-energy photon beam,
Med. Phys. 29, 24592463 2002.
109
G. X. Ding, Using Monte Carlo simulations to commission photon beam
output factors-A feasibility study, Phys. Med. Biol. 48, 38653874
2003.
110
A. Ahnesj, Collimator scatter in photon therapy beams, Med. Phys.
22, 267278 1995.
111
A. Mack, G. Mack, D. Weltz, S. G. Scheib, H. Bottcher, and V. Seifert,
High precision film dosimetry with GAFCHROMIC films for quality
assurance especially when using small fields, Med. Phys. 30, 23992409
2003.
112
M. Stasi, B. Baiotto, G. Barboni, and G. Scielzo, The behavior of several
microionization chambers in small intensity modulated radiotherapy
fields, Med. Phys. 31, 27922795 2004.
113
C. McKerracher and D. I. Thwaites, Verification of the dose to the isocentre in stereotactic plans, Radiother. Oncol. 64, 97107 2002.
114
F. Sanchez-Doblado, G. H. Hartmann, J. Pena, R. Capote, M. Paiusco, B.
Rhein, A. Leal, and J. I. Lagares, Uncertainty estimation in intensitymodulated radiotherapy absolute dosimetry verification, Int. J. Radiat.
Oncol. Biol. Phys. 68, 301310 2007.
Endovascular image-guided interventions EIGIs

Stephen Rudin,a Daniel R. Bednarek,b and Kenneth R. Hoffmannc
Toshiba Stroke Research Center, University at Buffalo, State University of New York, Biomedical Research
Building, Room 445, 3435 Main Street, Buffalo, New York 14214
Received 24 September 2007; accepted for publication 13 November 2007;

Minimally invasive interventions are rapidly replacing invasive surgical procedures for the most
prevalent human disease conditions. X-ray image-guided interventions carried out using the insertion and navigation of catheters through the vasculature are increasing in number and sophistication. In this article, we offer our vision for the future of this dynamic field of endovascular imageguided interventions in the form of predictions about 1 improvements in high-resolution detectors
for more accurate guidance, 2 the implementation of high-resolution region of interest computed
tomography for evaluation and planning, 3 the implementation of dose tracking systems to control
patient radiation risk, 4 the development of increasingly sophisticated interventional devices, 5
the use of quantitative treatment planning with patient-specific computer fluid dynamic simulations,
and 6 the new expanding role of the medical physicist. We discuss how we envision our predictions will come to fruition and result in the universal goal of improved patient care. 2008
Key words: endovascular, image-guided intervention, detectors, cone beam CT, computer fluid
dynamics CFD
I. INTRODUCTION
The three greatest killers in the developed world are heart
disease, cancer, and stroke or neurovascular disease, with the
last being the greatest cause of disability. After drug treatment, the primary interventional therapy for vascular diseases and an important treatment modality for cancer are
endovascular image-guided interventions EIGIs. EIGIs
generally involve the insertion of a catheter into the femoral
artery, which is then threaded under fluoroscopic guidance
through the vasculature to the site of the pathology to be
treated. The pathology may be a portion of a vessel that is
stenosed, or totally occluded, either by chronic plaque formation or by acute clot localization. The pathology may also
be a portion of a vessel that is weakened and bulges to form
an aneurysm, which may hemorrhage. Finally, it may be a
hypervascular region such as a tumor bed or an arteriovenous malformation. The EIGI may involve delivery of a
drug or embolic material through the catheter, or may be the
delivery of a device such as a stent that may be used to keep
a stenosed vessel open. During the course of the intervention,
rapid-sequence fluoroscopic x-ray imaging is usually relied
on for guidance.
Increasingly, EIGIs are replacing invasive surgical procedures because of the reduced mortality, morbidity, and discomfort to the patient.1 For example, surgical clipping of
aneurysms, which just a few years ago was the standard of
care, is now being replaced by the deployment of embolic
material, such as detachable coils delivered through a catheter under fluoroscopic guidance into the aneurysm, thus replacing an invasive external approach most likely through
the skull with a minimally invasive endovascular
intervention.24 In this work, we will discuss the basic features and current status of EIGI in terms of the imaging
301
system, radiation dose, interventional equipment and devices, procedure characteristics and treatment planning, and
personnel involved. We will first give an overview of the
current state of EIGI and then make predictions about each
feature category, providing the rationale and evidence for
each prediction and the strategy we suggest for making each
forecast become a reality in a 710 year time frame.
II. CURRENT STATUS
Before an EIGI procedure is scheduled, diagnostic procedures to find and analyze the pathological region are done
preferably using noninvasive imaging modalities, such as
multislice computed tomography MSCT or magnetic resonance imaging MRI, and followed up by higher-resolution
minimally invasive angiographic procedures done in dedicated angiographic or special procedure x-ray suites.5,6 Typically, EIGI procedures, if indicated, are then done under realtime x-ray image guidance in such a special procedure suite.7
Thus, such a suite may be used for the initial diagnostic
study and subsequently for the intervention. Rotating anode
x-ray tubes, that have become more robust but have changed
little in decades in overall design, are used as the source of x
rays; however, dynamic x-ray detector designs are in the process of changing from the decades old x-ray image intensifier
XII to the new flat panel detector FPD.8 The indirect FPD
and the XII designs both use an x-ray absorbing structured
CsI phosphor, but differ greatly thereafter. The XII converts
the phosphors light to electrons that are accelerated. This
electron distribution is focused on an output phosphor to
produce an intensified image of light photons and thus create
a gain in brightness.9 In contrast, an indirect FPD uses an
array of solid state photodiodes with thin film transistor
TFT switches and, currently, no additional gain to generate
0094-2405/2008/351/301/9/$23.00
301
302
Rudin, Bednarek, and Hoffmann: Endovascular image-guided interventions
electronic signals that are directly digitized by circuitry off to

the side of the x-ray sensitive region. Direct FPDs employ a
photoconductor, such as amorphous selenium, to absorb the
x rays and directly produce a charge, which is read out by an
array of TFTs. While the XIIs have a number of
disadvantagessuch as glare, pincushion distortion due to
the curved x-ray sensitive phosphor, S distortion due to the
earths magnetic field, and large physical volumethey have
the advantages of inherently large and variable gain, adjustable magnification hence pixel size, and flexible optical output that is usable with a variety of optical image receptors or
cameras.9 However, the transition to FPDs for dynamic x-ray
imaging has been somewhat disappointing, due to the substantial FPD additive instrumentation noise that is similar in
magnitude to the signal obtained in the lower range of standard fluoroscopic exposures. This noise quantity when expressed in units of exposure may be referred to as the instrumentation noise equivalent exposure INEE.10 Other
disadvantages of FPDs have been the image lag or
persistence,11,12 ghosting due to charge retention in the photodiode TFT array,12,13 and the substantial artifacts that require corrections prior to image display. The advantages of
FPDs, however, are their smaller physical size and lack of
distortion, which enable improved cone-beam computed tomography CBCT to be accomplished with the same angiographic gantry used for the diagnostic or EIGI procedures.14
For both detector systems that are currently in use, image
quality may not be optimal especially in fine detail EIGIs.
The details of interventional devices, such as stent struts, and
small vessels, such as perforators in the brain, typically cannot be visualized because of the limited spatial resolution
capabilities of current conventional detectors, although the
development of some specialized CCD-based microangiographic detectors has begun.15
As EIGI procedures become more widespread and more
complex, there is a continuing concern regarding lengthy
procedure times and the associated radiation doses, which
can be some of the highest of medical radiological
studies.16,17 Currently, there are dose reduction techniques
available, such as spectral shaping performed automatically
using selected metallic filters and the option for reducing
frame rates during fluoroscopic and angiographic acquisitions. However, other techniques, such as spatially modulating the beam for optimal dose utilization1821 and recording
and displaying the details of patient dose deposition,22,23 are
not commonly used.
Even though EIGI procedures can be lengthy and complex, the actual EIGI devices that are used remain only
somewhat more complex in design than those that were used
when the field began.24 Guidewires may vary in characteristics, such as torsion rigidity, flexibility, and kink resistance,
due to the characteristics of the materials and windings used,
and catheters may similarly vary depending upon the material and construction of the wall.2530 However, they are basically still passive cylinders and tubes that are actuated by
pulling, pushing, rotating, or, in the case of balloons, inflating, all simple mechanical actions initiated at the proximal
end outside the patient. Also, while devices, such as stents,
302
and inserted material, such as embolizing coils or glues, have

advanced somewhat over the years,31 they are still rather
simple in design and mostly passive in nature having few if
any electronic, computerized, or motorized components.
Even specialized active devices, such as the rotational
atherectomy Rotablator Boston Scientific Corp, Natick,
MA, are basically simple rapidly rotating wires with diamond burrs at the distal tip for grinding down plaque.32 Nevertheless, these mostly passive or simple devices deployed
under EIGI procedures have had a major impact already in
substituting for surgically invasive procedures, such as aneurysm clipping or open-heart coronary bypass procedures.
Currently, unlike in radiation therapy, quantitative treatment planning is lacking in preparation for EIGI procedures.
Also, although CBCT is now available, it is not used to
quantitatively plan a procedure, such as to determine in advance what guidewires, catheters, and endovascular devices
can be passed through vessels using computer analysis of the
mechanical properties of the vessels as well as a library of
such properties for the devices combined with the derived
tortuosity and other characteristics of the vessel. It is generally believed that blood flow has a major impact on the initiation of vessel pathologies,33,34 as well as the outcomes of
subsequent EIGIs,35,36 yet there is very little done clinically
to quantitatively determine the detailed flow or to predict the
effect on blood flow of a proposed intervention from information gained with computer fluid dynamic CFD calculations in time to have any clinical impact.
Because of the present rather improvised state of clinical
EIGI procedures, the personnel active during such procedures consists of a mixture of clinicians and technologists,
but rarely includes medical physicists or engineers. This contrasts with radiation therapy procedures where the treatment
team generally includes a medical physicist.
III. PREDICTION, EVIDENCE/RATIONALE, AND
STRATEGY FOR IMPLEMENTATION
We predict that in 710 years there will be dramatic
changes in the field of EIGI procedures in all areas: Imaging
equipment and doses, devices, methods of planning, as well
as the role of the medical physicist. In what follows we list
our predictions in each of these areas of EIGI, provide our
rationale supported by evidence from current research trends,
and suggest strategies for how these forecasted advances
may come to fruition.
III.A. Imaging systems and dose
III.A.1. Prediction 1
Higher resolution, lower noise, and real-time image
receptorsat first for the region of interest (ROI) near the
intervention and then across the full field of view (FOV)_
will become the standard of care for many EIGI procedures.
The rationale for this prediction is that, increasingly,
MSCT and MRI are taking over the task of vascular diagnosis with angiographic special procedure rooms being devoted
increasingly to EIGI. The fact that MSCT and magnetic resonance angiography MRA procedures require only a venous
303
303
FIG. 2. Fluoroscopic road-mapped frames done at 74 kVp and 50 mA of 3

mm diameter and 15 mm long coronary stent systems Multi-link Vision,
Guidant Corp, Santa Clara, CA. End markers indicate extent of expanded
balloon on delivery catheter. Stents are deployed in carotid arteries of the
same rabbit. A Using 2 ms pulsed XII imaging with temporal filtering
standard on commercial systems, stent deployed in right carotid. B Using
8 ms pulsed HSMAF imaging with no temporal filtering, stent deployed in
left carotid showing improved visualization of details of stent struts and
vessel edges.
FIG. 1. Improved resolution with new SSXII right compared with XII
left. A and B Bar pattern images taken at 70 kVp with 2 in. PMMA
filtration added with the same exposure for each detector. C and D Images of a neurovascular phantom containing thin wires left to right:
100 m Au, 25 m Cu, 50 m Pt, 100 m iodine contrast media-filled
tube and a new asymmetric vascular stent having a blood flow diverting
low-porosity polyurethane patch indicated by four Pt markers designed for
closing off flow into an aneurysm were taken at 74 kVp with 21.4 mm Al
added filtration RQA 5 Spectrum with the same exposure. A Bar pattern
with only 3.1 Lp/mm visible for XII in 5 in. magnification mode. B Bar
pattern with 10 Lp/mm clearly visible for SSXII. C With the XII, wires are
less clear: Cu wire and Pt wire bends are unseen and stent struts and patch
markers are not clearly visible. D With the SSXII, Cu wire is barely
visualized but Pt wire bends and stent struts and patch markers are clearly
visible.
puncture for injection of the contrast media and that image

quality is improving to the point where definitive diagnoses
are possible in many cases make this trend plausible. Once a
diagnosis has localized the site of pathology, there may not
be a need for superior image quality except over the site of
the intervention. Superior resolution is needed in the ROI for
both real-time fluoroscopy to actively guide the ongoing intervention and also for superior angiography to evaluate the
progress of the intervention throughout the procedure and to
help determine when a satisfactory endpoint has been
reached.
To achieve this improvement especially over the ROI of
the interventional site, either a vast improvement in current
FPDs or new alternative detectors with the desired characteristics will be required. For example, large area CCDs have
been considered.37 Methods for reducing the effect of instrumentation noise on FPDs have been proposed; one such proposal being to provide inherently increased signal using a
direct x-ray converter material having avalanche gain.38 For
indirect FPDs, methods include using a-Se avalanche photoconductors in place of a-Si photodiodes39,40 and placing an
amplifier at each pixel of the FPD.41,42 These methods enable
signal amplification prior to sending the charge packet repMedical Physics, Vol. 35, No. 1, January 2008
resenting the signal through the TFT lines, where the noise is
added and then onto the sides of the sensor where it can be
amplified and digitized. Were this achievable in practice, the
problem for FPDs of reducing the size of the pixels would
still remain. Alternate detector designs actively being pursued by the authors consist of a high-sensitivity microangiographic fluoroscope HSMAF10 and a detector called the
solid state x-ray image intensifier SSXII. The HSMAF employs a light image intensifier between the CsI phosphor and
a digital videocamera with a fiber optic taper FOT input.
This detector has a small FOV sufficient to be used for EIGI
and is capable of real-time fluoroscopy as well as very-highresolution angiography. The SSXII consists of an array of
electron multiplying charge-coupled devices EMCCDs,43,44
each viewing a portion of the CsI phosphor through its own
FOT coupling.45 EMCCDs have all the advantages of standard CCDs: high speed, low noise, high resolution, and standard solid state production techniques. However, in addition,
they have a row of hundreds of multiplying elements each
with a small gain of a few percent. When cascaded together
in this row, they can result in a total gain from 1 to 2000
times, depending upon the variable gain voltage setting of
only around 20 V. The mosaic array of the SSXII also has the
advantage of being extensible in FOV by adding modules.
The INEE achievable for both such detectors is a fraction of
the exposure typical of low-dose fluoroscopy, hence they
have the desirable feature of being quantum limited throughout their range of operation. Additionally, unlike FPDs they
have no lag or ghosting and are available with small pixel
sizes. The potential advantages of SSXII and HSMAF detectors compared with conventional detectors are illustrated by
some examples in Figs. 1 and 2. A quantitative evaluation
using linear cascade analysis46 can be made of these new
high-resolution detectors45,47; however, recent additions to
these analytical methods will include the effects of other important factors that affect total system performance, such as
focal spot blurring and scatter in the patient.48,49
304
High-resolution CBCT will become available for the ROI
of the interventional site with 3D roadmapping and automated indication of the catheter or guidewire tip superimposed.
The rationale for this prediction is that once a vascular
pathology has been diagnosed, superior image quality is
needed only over the site of the intervention and not over the
full volume FOV. Currently, there is some effort in both academic and commercial organizations to develop 3D roadmapping capability. The syngo iPilot product50 Siemens
Medical Solutions USA, Inc., Malvern, PA or the Dynamic
3D Roadmap product51 Philips Medical Systems, N.A,
Bothell, WA are such implementations with current detector
technology. Clearly, the higher the resolution available at the
interventional site, the greater the potential for a successful
intervention.
To achieve this goal, higher-resolution ROI detectors,
such as those in Prediction 1 above, mounted on stable detector holders added to angiographic CBCT-capable gantries
will be required. The associated software will also be
needed. Our group has shown that most of the artifacts due
to FOV truncation that one expects when the CT FOV is
smaller than the object being imaged can be eliminated by
combining the ROI data with a lower-.resolution full FOV
CT image sequence. This provides sufficient information
about those parts of the object that move in and out of the
ROI of the projection views but which are not actually located within the ROI or volume of interest of the 3D
object.52,53 These high-resolution ROI-CBCT images can
then be used as the basis for enhanced treatment planning by
calculating the simulation of the transport of a device
through the vasculature53,54 and for improved flow information, as will be discussed next.
There will be greater accounting of patient dose distribution during EIGI with techniques to minimize integral dose
as well as deterministic effects.
Although patient dose for EIGI procedures can be some
of the largest of all medical procedures, it is highly nonuniform; hence, any determination of potential biological effects
will depend upon a detailed knowledge of the dose distribution. Although the European Union EU requires the measurement of dose-area product and the International Electrotechnical Commission Standard IEC 606012-43, 2000 and
the U.S. Food and Drug Administration FDA require that
fluoroscopic equipment include, during the procedure, the
display of air kerma rate and cumulative air kerma at an
interventional reference point.55,56 This is only a first step in
an accurate accounting of patient dose in EIGI. A past attempt to provide more detailed patient entrance surface dose
distributions in the form of a commercial product Caregraph, Siemens Medical Systems Corp., Malvern, PA was
apparently discontinued; however, as EIGI procedures become more complex, potentially longer, and with more nonMedical Physics, Vol. 35, No. 1, January 2008
304
uniform dose application due to ROI techniques, recording

the dose distribution will become increasingly important.
As discussed in the section on ROI-CBCT, future patient
dose may be made highly nonuniform using collimation and
semi-radio-opaque metallic filters.57 To implement a more
comprehensive dose tracking system, our group uses the
digital signals available from new angiographic c-arm gantries, together with software that calculates and conveniently
displays the cumulative entrance skin dose distribution as
well as dose rate in real time.58,59 This enables a clinician to
obtain feedback on the consequences of their actions and to
modify the procedure to reduce the risk of deterministic effects, such as erythema, epilation, and dermal necrosis.17 In
the future, such tracking of the beam should also be able to
allow real-time accounting of effective dose and the increased risk of stochastic effects, such as cancer, which may
be especially important in procedures that include exposure
of the breast or thyroid.
One way to reduce the risk of entrance skin effects is to
change the entrance skin exposure location and thus to
spread the surface dose. Through determination of the 3D
vasculature, future systems may guide the interventionist in
selecting an optional projection to spread the dose when a
deterministic threshold dose is being reached, while maintaining the needed interventional site image view to continue
the procedure.
Since serious deterministic effects may not be evident until weeks after the procedure, it is important to be able to
correlate the cause with the effect for proper treatment. The
patient record should contain information about the distribution of dose on the skin as recommended by the FDA.60,61
Using the results of a dose tracking system, the electronic
medical record will be able to log this information for future
reference with the risk level indicated. The dose from subsequent procedures will be able to be added to that already
recorded to provide a measure of cumulative risk. It is postulated that advances in radiobiological knowledge will allow modeling of cellular repair occurring between procedures so a more accurate estimate of cumulative risk from
repeated exams can be obtained. This will further allow individual risk factors that increase radiation sensitivity such
as collagen vascular disease, diabetes mellitus, and certain
pharmaceuticals and chemicals62 to be taken into account.
III.B. Endovascular devices
III.B.1. Prediction 4
Endovascular devices will become finer, more patientspecific, more biocompatible, and more complex with the beginning of remotely actuated active components using shape
memory alloys (SMA), micromachines, microfluidics, and microelectronics; higher-resolution imaging will be relied upon
to assist in accurate device deployment.
Since minimally invasive EIGI procedures, even with the
current crop of simple and somewhat passive devices, are
rapidly replacing many invasive surgical procedures, it
would appear to make sense that this trend would continue
and be combined with advances in miniaturization. With
305
more accurate models of a specific patient vascular pathology obtained from CBCT, more patient-specific devices can
be created and deployed; however, care should be taken that
new endovascular devices are designed with constituent materials that are compatible with future imaging procedures
that might be required for the patient presently being treated,
so there is no negative effect such as artifact generation on
the images obtained.
As new more biocompatible materials are developed, devices such as bioabsorbable stents under investigation6366
should come into common usage. The initial development of
remote actuation is also happening. For example, external
magnetic fields are actively being used in the cardiovascular
area for accurately guiding new ablation probes for electrophysiology EP treatment,6769 and are being proposed in
the neurovascular arena for guiding the accurate deployment
of new flow diversion stents Fig. 1D for treatment of
aneurysms.70 The use of the existing magnetic field in MRI
systems is being reported for steering catheters.71 There is
the beginning commercial availability of remote catheter
steering systems, such as the one by Hansen Medical
Hansen Medical Inc., Mountain View, CA72,73 designed for
EP. However, these appear to be implemented with wires tied
to external motors and hence are hard to project to the full
range of EIGI, which may require navigation through a tortuous vessel. Other robotic systems have also been
reported.74 Remote actuation using inductive heating of
shape memory polymers is also being considered.75 There
are also a number of new more complex devices under development for the treatment of acute ischemic stroke, such as
ultrasonic clot busters, and complex catheter heads with high
speed saline streams to implement suction and hence reduce
clots.76 New technologies in radio-frequency ablation including software for image registration and fusion, thermal
modeling, electromagnetic tracking, semiautomated robotic
needle guidance, and multimodality imaging77 may also be
adaptable to EIGI.
To achieve the full potential for new patient-specific, biocompatible, active EIGI devices, basic components of the
process for creating these devices will have to be worked
out. First, improved morphological models of the patientspecific pathology will be achieved using the high-resolution
ROI-CBCT imaging methods described previously and using
computer models parameterizing the desired interventional
effect such as flow modification. See, for example, the CFD
calculations described next. An actual physical model of the
pathology might then be constructed using rapid-prototyping
techniques,78,79 and if the clinical circumstances permit, a
patient-specific device could be tested on the physical model
before actual clinical use. Next, if possible and if appropriate, a new biocompatible material will be selected so that the
device can be absorbed once the desired intervention has
been achieved, although reaching this goal may take longer
than the envisioned time period. Finally, the basic manufacturing process for generating the required functional elements of a new EIGI device will have to be further developed. For example, although it is doubtful that self-propelled
yet tethered devices will be commercialized in 710 years,
305
electrically activated shaped memory alloys SMAs80 could

be developed to supply the remote mechanical force needed
for remote actuation. Functions, such as bending a guidewire
or catheter to change its direction and guiding it into a desired vessel branch, could then be accomplished with more
complex accommodations to the vessel geometry occurring
when sealed catheter microjoints and robust electrical leads
are developed. Alternatively for rotary motion, basic micromachines that could be electrically activated would also have
to be developed. Thus, rotational motion now used in cardiovascular applications for shaving plaque or for forming an
intravascular ultrasound IVUS image could then be replaced with electrically activated rotating distal devices,
which could be smaller and able to travel deeper into smaller
vessels. Remote actuation would also enable IVUS for the
neurovasculature, which is presently not possible because of
the size and rigidity of current probes. Finally, although
probably somewhat further into the future, there is the possibility for achieving very high spatial resolution using optical coherence tomography OCT imaging81 if it can be applied intravascularly for viewing close to the surfaces of
vessel lumens and endovascular devices. There is also the
possibility of using nanotechnology for creating ultrasmall
endovascular sensors e.g., nanotube transistors for ionizing
radiation measurement82 and nanoavalanche photodiodes for
photon detectors39 that might be used for guiding interventions.
III.C. Procedure and treatment planning
III.C.1. Prediction 5
Treatment planning for EIGI procedures will be more extensive and involve simulations to verify the optimal selection of devices and equipment for delivery as well as flow
calculations before, during, and after the intervention to
check the intervention as it proceeds.
The rationale for this prediction is that as more detailed
3D morphological information about a vascular pathology
becomes available from improved imaging studies and as
computer power continues to increase in availability at reduced cost, the use of simulations has begun83 and should
naturally continue to blossom. Evidence for this happening
in other areas, such as in radiation treatment planning, is
apparent. Patient-specific dose delivery regimens and resulting distributions are computer simulated with clinical CT
sequences as a basic part of the input data. For EIGI, however, there are two parts to a treatment plan that will have to
be simulated: The device delivery based upon vessel morphology and mechanical device properties, and the blood
flow results as determined using CFD calculations.
The implementation of increasingly sophisticated EIGI
treatment plans will probably occur gradually. Current research into simulations for device transport through tortuous
vasculature has begun; for example, with the tracking of a
rigid stent of given dimensions and determining tolerances
for passage along the parent vessel.53 Increasingly sophisticated models84 of the human vasculature, which include not
only the morphology as derived from CT sequences for pa-
306
306
the course of the EIGI to assist the interventionalist in making better decisions while the procedure is occurring. To further increase computer capability, it may be necessary to
have a fast digital link to a super-computer facility to enable
these complex calculations to be done so that actionable information is available during the EIGI.
III.D. Personnel involved
III.D.1. Prediction 6
FIG. 3. CFD result showing streamlines and local shear stress distributions
for a patient-specific aneurysm untreated and treated with an asymmetric
vascular stent Ref. 87. Notice effect of flow diversion on streamlines and
wall shear stress WSS distribution. A Streamlines, untreated; B streamlines, treated; C wall shear stress, untreated; D wall shear stress, treated.
tients to be treated but also that of vessel flexibility and in

the more distant future the inclusion of biological components to the model85such as vessel wall cellular and molecular properties and constituentswill make treatment
planning more accurate. There will be improved selection
criteria for devices and delivery equipment as well as indications of potential vessel wall damage if nonoptimal deliveries must be made. The second important part of the EIGI
treatment plan consisting of CFD simulation of the blood
flow will depend upon continued progress being made in
adapting the processes of solving the NavierStokes equation
to the conditions of EIGIs and validating these results.8690
The verification link between actual blood flow situations,
angiogramswhere the flow of contrast media rather than
blood is recorded, and CFD calculation results, must be accomplished. At present, this validation process is being actively pursued using phantom experiments with both blood
equivalent optical dye and denser more viscous radiographic
contrast media.91,53 Once there is confidence that the CFD
results are valid, initial flow conditions established by MRA
or ultrasound studies will be used together with the morphology determined by CBCT sequences as inputs to the CFD
calculations. The treatment plan might then consist of simulating the deployment of a flow modifying device, such as a
stent, for treatment of stenoses or the deployment of a specialized flow diverting stent92,87,93 for treatment of aneurysms. Not only could velocity distributions and flow streamlines be simulated, but also wall shear stressa possible
predictor of pathologycould be estimated as well Fig. 3.
As models get more sophisticated and computer power
increases, it may be advantageous to do both the delivery and
the CFD parts of the simulation almost in real time during
The medical physicist will take a more active role in individual clinical EIGI procedures as well as in the training
of interventionalists using simulation systems.
If the future of EIGI procedures is as predicted herein,
with more patient intensive treatment planning, the analogy
of the medical physicists role to that in radiation therapy
treatment planning will be more plausible. As EIGI interventions become more technologically sophisticated, there will
be a need for a professional clinical physicist to help in the
selection of devices and delivery systems through simulations and quantitative analysis, as well as in overseeing the
appropriate interpretation and use of CFD results for individual patient interventions. These responsibilities, of course,
will be in addition to the primary continuing responsibility
for the specification and quality assurance of all imaging
systems.
Just as the therapy physicist is trained and certified in that
branch of medical physics, the vascular medical physicist
will have to be certified in EIGI physics as well. This implies
the necessity for new or augmented curricula in medical
physics training programs. While much of the basic hemodynamics and the CFD calculations may be presently in the
domain of biomechanical engineers, we expect that future
EIGI teams will consist of many medical professionals to
enable this rapidly advancing field to reach its vast potential.
However, it is our expectation that the vascular clinical
medical physicist will be uniquely qualified to integrate the
physics of imaging with the relevant aspects of biomechanical engineering involving interventional devices to be able to
oversee the technical aspects of EIGI facilities as well as to
head their scientific and engineering training programs for
medical personnel. Not only will there be a new syllabus of
EIGI physics related concepts to teach, but we expect that
simulation-based training of the physicians9499 who perform
EIGI procedures will become the standard. The interventionist will be able to practice and develop the necessary skills
for catheter navigation and device deployment in a programmable simulated environment with tactile and visual feedback. The EIGI medical physicist with the collaboration of
computer scientists will play a key role in both the use and
the development of such systems.
IV. CONCLUSIONS
In the 1966 movie Fantastic Voyage, there was an account
of what future endovascular brain surgery could be like. It
involved the shrinking of a submarine with a team of people
that was injected into a patient and instructed to take a route
through the vasculature to the brain to remove a blood clot in
307
a vessel. While we do not quite predict such a capability, we

believe there will be considerable movement in the direction
of miniaturization and remote actuation and guidance. We
have made some specific forecasts regarding this rapidly advancing and exciting field of EIGI. We have predicted that
imaging detectors will be markedly improved to better guide
the use of newer and more complex interventional devices
and delivery equipment using radiation more appropriately.
We have described how we believe treatment planning for
EIGI will be enhanced from the present rather informal
qualitative planning to an increasingly sophisticated quantitative procedure, involving a new labor-intensive role for
medical physicists somewhat akin to the role that medical
physicists play in radiation treatment planning. Ultimately,
we expect that advances in EIGI will lead to greatly improved patient procedures in which the medical physicist
will play a crucial role.
ACKNOWLEDGMENTS
We thank the members of the University at BuffaloToshiba Stroke Research Center including Dr. Ciprian N.
Ionita for supplying the angiographic images, Andrew
Kuhls-Gilcrist for the SSXII image comparison, and Dr. Hui
Mengs group for supplying the CFD images. The work was
supported in part by the National Institutes of Health under
Grant Nos. EB002873, NS43024, NS47242, HL52567, and
EB02916, the National Science Foundation under Grant No.
BES-0302389, the UB Interdisciplinary Research Development Fund IRDF, and an equipment grant from Toshiba
Medical Systems Corp.
a

[email protected]
b
Electronic mail:[email protected]
c
1
R. T. Higashida, B. J. Lahue, M. T. Torbey, L. N. Hopkins, E. Leip, and
D. F. Hanley, Treatment of unruptured intracranial aneurysms: A nationwide assessment of effectiveness, AJNR Am. J. Neuroradiol. 28, 146
151 2007.
2
C. J. Koebbe, E. Veznedaroglu, P. Jabbour, and R. H. Rosenwasser, Endovascular management of intracranial aneurysms: current experience
and future advances, Neurosurgery 59, S3-93S3-102 2006.
3
C. J. Koebbe, A. Pandey, E. Veznedaroglu, and R. H. Rosenwasser, The
evolution and future directions of endovascular therapy, Clin. Neurosurg. 53, 191195 2006 review.
4
D. Fiorella and H. H. Woo, Emerging endovascular therapies for symptomatic intracranial atherosclerotic disease, Stroke 38, 23912396
2007.
5
O. Saborowski and M. Saeed, An overview on the advances in cardiovascular interventional MR imaging, Magn. Reson. Mater. Phys. Biol.
Med. 20, 117127 2007.
6
S. D. Flamm, Cross-sectional imaging studies: What can we learn and
what do we need to know, Semin. Vasc. Surg. 20, 108114 2007.
7
P. L. Rauch, A tour of the interventional vascular imaging suite, in
Intravascular Brachytherapy Fluoroscopically Guided Interventions,
AAPM Monograph No. 28 AAPM, College Park, Maryland, 2002, pp.
547616.
8
J. A. Rowlands and J. Yorkston, Flat panel detectors for digital radiography, in Handbook of Medical Imaging, Vol. 1, Physics and Psychophysics, edited by J. Beutel, H. L. Kundel, and R. L. Van Metter SPIE,
Bellingham, Washington, 2000, Chap. 4.
9
J. T. Bushberg, J. A. Siebert, E. M. Leidholdt, and J. M. Boone, Fluoroscopy, in The Essential Physics of Medical Imaging, 2nd ed. Lippincott Williams and Wilkins, Philadelphia, Pennsylvania, 2002, Chap. 9,
307
pp. 231254.
S. Rudin, G. Yadava, G. Josan, A. Kuhls, H. Rangwala, Y. Wu, C. Ionita,
and D. R. Bednarek, New light-amplifier-based detector designs for high
spatial resolution and high sensitivity CBCT mammography, Proc. SPIE
6142, R1R11 2006.
11
P. R. Granfors, R. Aufrichtig, G. E. Possin, B. W. Giambattista, Z. S.
Huang, J. Liu, and B. Ma, Performance of a 41 41 cm2 amorphous
silicon flat panel x-ray detector designed for angiographic and R&F imaging applications, Med. Phys. 30, 27152726 2003.
12
A. K. Bloomquist, M. J. Yaffe, G. E. Mawdsley, D. M. Hunter, and D. J.
Beideck, Lag and ghosting in a clinical flat-panel selenium digital mammography system, Med. Phys. 33, 29983005 2006.
13
J. H. Siewerdsen and D. A. Jaffray, A ghost story: Spatio-temporal response characteristics of an indirect-detection flat-panel imager, Med.
Phys. 26, 16241641 1999.
14
C. Schmidgunst, D. Ritter, and E. Lang, Calibration model of a dual gain
flat panel detector for 2D and 3D x-ray imaging, Med. Phys. 34, 3649
3666 2007.
15
A. Ganguly, S. Rudin, D. R. Bednarek, K. R. Hoffmann, and I. Kyprianou, Micro-angiography for neuro vascular imaging, Part 1: Experimental measurements and feasibility, Med. Phys. 30, 30183028 2003.
16
Food and Drug Administration, FDA Public Health Advisory, Avoidance
of serious x-ray induced skin injuries to patients during fluoroscopicallyguided procedures, September 30, 1994, Center for Devices and Radiological Health, Rockville, Maryland available at http://www.fda.gov/
cdrh/fluor.html.
17
L. K. Wagner, M. D. McNeese, M. V. Marx, and E. L. Siegel, Severe
skin reactions from interventional fluoroscopy: Case report and review of
the literature, Radiology 213, 773776 1999.
18
S. Rudin and D. R. Bednarek, Region of interest fluoroscopy, Med.
Phys. 19, 11831189 1992.
19
S. Rudin, D. R. Bednarek, L. R. Guterman, A. Wakhloo, L. N. Hopkins,
L. Fletcher, and P. Massoumzadeh, Implementation of region of interest
fluoroscopy using the road mapping mode of a real-time digital radiographic unit, Radiographics 15, 14651470 1995.
20
S. Rudin, L. R. Guterman, W. Granger, D. R. Bednarek, and L. N. Hopkins, Neuro-interventional radiologic application of region of interest
ROI imaging techniques, Radiology 199, 870873 1996.
21
S. Rudin, D. R. Bednarek, M. Kezerashvili, W. Granger, J. Serghany, L.
R. Guterman, L. N. Hopkins, P. Massoumzadeh, B. Szymanski, and R.
Loftus, Clinical application of region of interest ROI radiologic imaging techniques, Radiographics 16, 895902 1996.
22
N. A. Gkanatsios, W. Huda, K. R. Peters, and J. A. Freeman, Evaluation
of an on-line patient exposure meter in neuroradiology, Radiology 203,
837842 1997.
23
S. Balter, D. W. Fletcher, H. M. Kuan, D. Miller, D. Richter, H. Seissl,
and T. B. Shope, Techniques to estimate radiation dose to skin during
fluoroscopically guided procedures, in Intravascular Brachytherapy/
Fluoroscopically-Guided Interventions, edited by S. Balter, R. Chan, and
T. Shope AAPM Summer School Proceedings, Madison, Wisconsin,
2002, pp. 731764.
24
C. T. Dotter and M. P. Judkins, Transluminal treatment of arteriosclerotic
obstruction. Description of a new technic and a preliminary report of its
application, 1964, Radiology 172, 904920 1989 classical article.
25
H. Ceschinski, H. Henkes, H. C. Weinert, W. Weber, D. Kuhne, and H.
Monstadt, Torquability of microcatheter guidewires: the resulting torsional moment, Biomed. Mater. Eng. 10, 3142 2000.
26
J. Carey, A. Fahim, and M. Munro, Design of braided composite and
cardiovascular catheters based on required axial, flexural and torsional
rigidities, J. Biomed. Mater. Res., Part B: Appl. Biomater. 70B, 7381
2004.
27
J. Carey, D. Emery, and P. McCracken, Buckling test as a new approach
to testing flexural rigidities of angiographic catheters, J. Biomed. Mater.
Res., Part B: Appl. Biomater. 76B, 211218 2006.
28
P. Bloss, W. Rothe, P. Wunsche, C. Werner, A. Rothe, G. D. Kneissl, W.
Burger, and E. Rehberg, Investigations of the pushability behavior of
cardiovascular angiographic catheters, Biomed. Mater. Eng. 13, 327343
2003.
29
G. H. Zoarski, J. M. Mathis, and J. R. Hebel, Performance characteristics
of microcatheter systems in a standard tortuous pathway, AJNR Am. J.
Neuroradiol. 19, 15711576 1998.
30
J. M. Anderson, The future of biomedical materials, J. Mater. Sci.
Mater. Med. 17, 10258 2006 review.
10
308
31
S. H. Kim, B. A. Campanella, A. I. Qureshi, A. J. Ringer, A. S. Boulos,

D. K. Lopes, L. R. Guterman, and L. N. Hopkins, in Management of
Cerebral Aneurysms, edited by P. D. Le Roux, H. R. Winn, and D. W.
Newell Saunders, Philadelphia, 2004, pp. 883902.
32
W. L. Ginete, Rotational Atherectomy: Current Perspectives,
www.med.umn.edu/img/assets/7172/Rotational_Atherectomy_073004
.swf
33
Flow-Dependent Regulation of Vascular Function, edited by J. A. Bevan,
G. Kaley, and G. M. Rubanyi Oxford University Press, New York,
1995.
34
P. F. Davies, J. A. Spaan, and R. Krams, Shear stress biology of the
endothelium, Ann. Biomed. Eng., Special Issue: International Biofluid
Mechanics Symposium 3312, 17141718 2005.
35
S. G. Imbesi and C. W. Kerber, Analysis of slipstream flow in a widenecked basilar artery aneurysm: Evaluation of potential treatment regimens, AJNR Am. J. Neuroradiol. 22, 721724 2001.
36
A. K. Wakhloo, B. B. Lieber, J. Seong, C. Sadasivan, M. J. Gounis, L.
Miskolczi, and J. S. Sandhu, Hemodynamics of carotid artery atherosclerotic occlusive disease, J. Vasc. Interv. Radiol. 15, S11121 2004 review.
37
S. Vedantham, A. Karellas, and S. Suryanarayanan, Solid-state fluoroscopic imager for high resolution angiography. Physical characteristic of
an 8 cm 8 cm experimental prototype, Med. Phys. 31, 14621472
2004.
38
D. C. Hunt, K. Tanioka, and J. A. Rowlands, X-ray imaging using avalanche multiplication in amorphous selenium: investigation of depth dependent avalanche noise, Med. Phys. 34, 976986 2007.
39
O. Hayden, R. Agarwal, and C. M. Lieber, Nanoscale avalanche photodiodes for highly sensitive and spatially resolved photon detection, Nat.
Mater. 5, 352356 2006.
40
W. Zhao, D. Li, A. Reznik, B. Lui, D. C. Hunt, K. Tanioka, and J. A.
Rowlands, Indirect flat-panel detector with avalanche gain: design and
operation of the avalanche photoconductor, Proc. SPIE 5745, 352360
2005.
41
M. H. Izadi, K. S. Karim, A. Nathan, and J. A. Rowlands, Low-noise
architecture for advanced diagnostic medical x-ray imaging applications,
Proc. SPIE 6142, T1T11 2006.
42
F. Taghibakhsh and K. S. Karim, Amplified pixel sensor architectures for
low dose computed tomography using silicon thin film technology, Proc.
SPIE 6510, W1W8 2007.
43
J. Hynecek, ImpactronA new solid state image intensifier, IEEE Trans.
Electron Devices 48, 22382241 2001.
44
M. S. Robbins and B. J. Hadwen, The noise performance of electron
multiplying charge-coupled devices, IEEE Trans. Electron Devices 50,
12271232 2003.
45
A. T. Kuhls, G. Yadava, V. Patel, D. R. Bednarek, and S. Rudin, Progress
in electron-multiplying CCD EMCCD based, high-resolution, highsensitivity x-ray detector for fluoroscopy and radiography, Proc. SPIE
6510, C1C11 2007.
46
I. A. Cunningham, Applied linear systems theory, in Handbook of
Medical Imaging, Vol. 1. Physics and Psychophysics, edited by J. Beutel,
H. L. Kundel, and R. L. Van Metter SPIE, Bellingham, Washington,
2000, Chap. 2.
47
A. Ganguly, S. Rudin, D. R. Bednarek, and K. R. Hoffmann, Micro
angiography for Neuro vascular imaging, part 2: Cascade model analysis, Med. Phys. 30, 30293039 2003.
48
I. Kyprianou, S. Rudin, D. R. Bednarek, and K. R. Hoffmann, Generalizing the MTF and DQE to include x-ray scatter and focal spot unsharpness: Application to a new micro-angiographic system for clinical use,
Med. Phys. 32, 613626 2005.
49
G. K. Yadava, I. S. Kyprianou, S. Rudin, D. R. Bednarek, and K. R.
Hoffmann, Generalized performance evaluation of x-ray image intensifier compared with a microangiographic system, Proc. SPIE 5745, 419
429 2005.
50
Siemens iPilot, Axiom No. 2, March, 2006, pp. 1017, pamphlet from
Siemens Medical Corp., http://www.medical.siemens.com/siemens/
sv_SE/rg_marcom_FBAs/files/brochures/magazin_archive_ax/AXIOM_
Innovation_in_Intervention_March_2006.pdf
51
Philips
Xpervision
Newsletter,
November,
2006,
http://
www.medical.philips.com/us/products/cardiovascular/assets/docs/
XperVision_Newsletter_int.pdf
52
R. Chityala, K. R. Hoffmann, D. R. Bednarek, and S. Rudin, Region of
interest ROI computed tomography CT: comparison with full field of
308
view FFOV and truncated CT for a human head phantom, Proc. SPIE
5745, 583590 2005.
53
K. R. Hoffmann, S. Rudin, H. Meng, L. N. Hopkins, L. Guterman, and E.
Levy, in Multidimensional Image Processing, Analysis, and Display, edited by S. G. Armato and M. S. Brown RSNA, Oak Brook, Illinois,
2005, pp. 173184.
54
S. Schafer, V. Singh, K. R. Hoffmann, P. B. Nol, and J. Xu, Planning
image-guided endovascular interventions: guidewire simulation using
shortest path algorithms, Proc. SPIE 6509, C1-C10 2007.
55
International Electrotechnical Commission, Report 606012-43, Medical
electrical equipment Part 2-43: Particular requirements for the safety of
x-ray equipment for interventional procedures, Geneva, Switzerland,
2000.
56
U. S. Food and Drug Administration FDA, HHS: Electronic Products
Federal performance standard for diagnostic x-ray systems and their major components, Federal Register, June 10, 2005 Vol. 70, Number 111,
pp. 3399734042.
57
P. Massoumzadeh, S. Rudin, and D. R. Bednarek, Filter material selection for region of interest radiologic imaging, Med. Phys. 25, 161171
1998.
58
P. M. Dinu, D. R. Bednarek, D. Wobschall, R. Peterson, C. N. Ionita, S.
Rudin, M. Zeng, K. Chugh, and K. R. Hoffmann, X-ray beam tracking
system for a fluoroscopic c-arm unit, Med. Phys. 30, 1424 2003.
59
K. Chugh, P. M. Dinu, D. R. Bednarek, D. Wobschall, S. Rudin, K. R.
Hoffmann, R. Peterson, and M. Zeng, A computer-graphic display for
real-time operator feedback during interventional x-ray procedures,
Proc. SPIE 5367, 464473 2004.
60
Food and Drug Administration, Important Information for Physicians and
other Health Care Professionals, Recording information in the patients
medical record that identifies the potential for serious x-ray-induced skin
injuries following fluoroscopically-guided procedures, September 15,
1995, Center for Devices and Radiological Health, Rockville Maryland
available at http://www.fda.gov/cdrh/xrayinj.html.
61
D. L. Miller, S. Balter, L. K. Wagner, J. Cardella, T. W. Clark, C. D.
Neithamer, Jr., M. S. Schwartzberg, T. L. Swan, R. B. Tobwin, K. S.
Rholl, and D. Sacks, SIR standards of Practice Committee, Quality
improvement guidelines for recording patient radiation dose in the medical record, J. Vasc. Interv. Radiol. 15, 423429 2004.
62
L. K. Wagner, in Intravascular Brachytherapy / Fluoroscopically-Guided
Interventions, edited by S. Balter, R. Chan, and T. Shope AAPM Summer School Proceedings, Madison, Wisconsin, 2002, pp. 195218.
63
J. A. Ormiston, M. W. I. Webster, and G. Armstrong, First-in-human
implantation of a fully bioabsorbable drug-eluting stent: the BVS poly-Llactic acid everolimus eluting coronary stent, Catheter. Cardiovasc. Interv. 69, 128131 2007.
64
R. Erbel et al., Temporary scaffolding of coronary arteries with bioabsorbable magnesium stents: a prospective, non-randomised multicentre
trial, Lancet 369, 18691875 2007.
65
J. Ormiston and M. Webster, Absorbable coronary stents. comment,
Lancet 369, 18391840 2007.
66
S. Saito, New horizon of bioabsorbable stent. comment, Catheter.
Cardiovasc. Interv. 66, 595596 2005 editorial.
67
http://www.stereotaxis.com/
68
M. P. Armacost, J. Adair, T. Munger, R. Viswanathan, F. M. Creighton,
D. T. Curd, and R. Sehra, Accurate and reproducible target navigation
with the stereotaxis Niobe Magnetic Navigation System, J. Cardiovasc.
Electrophysiol. 18, S26S31 2007.
69
C. Pappone et al., Robotic magnetic navigation for atrial fibrillation
ablation, J. Am. Coll. Cardiol. 47, 13901400 2006.
70
E. Florin, H. S. Rangwala, and S. Rudin, Method to rotate an endovascular device around the axis of a vessel using an external magnetic field,
Med. Phys. 34, 328333 2007.
71
F. Settecase, M. S. Sussman, M. W. Wilson, S. Hetts, R. L. Arenson, V.
Malba, A. F. Bernhardt, W. Kucharczyk, and T. P. L. Roberts,
Magnetically-assisted remote control MARC steering of endovascular
catheters for interventional MRI: A model for deflection and design implications, Med. Phys. 34, 31353142 2007.
72
http://www.hansenmedical.com/products/sensei.aspx
73
A. D. Al-Ahmad, J. Grossman, and P. J. Wang, Early experience with a
computerized robotically controlled catheter system, J. Interv. Card.
Electrophysiol. 12, 199202 2005.
74
L. Cercenelli, E. Marcelli, and G. Plicchi, Initial experience with a
telerobotic system to remotely navigate and automatically reposition stan-
309
dard steerable EP catheters, ASAIO J. 53, 523529 2007.

P. R. Buckley, G. H. McKinley, T. S. Wilson, W. Small IV, W. J. Benett,
J. P. Bearinger, M. W. McElfresh, and D. J. Maitland, Inductively
heated shape memory polymer for the magnetic actuation of medical
devices, IEEE Trans. Biomed. Eng. 53, 20752083 2006.
76
G. M. Nesbit, G. Luh, R. Tien, and S. L. Barnwell, New and future
endovascular treatment strategies for acute ischemic stroke, J. Vasc. Interv. Radiol. 15, S103S110 2004.
77
B. J. Wood, J. K. Locklin, A. Viswanathan, J. Kruecker, D. Haemmerich,
J. Cebral, A. Sofer, R. Cheng, E. McCreedy, K. Cleary, M. J. McAuliffe,
N. Glossop, and J. Yanof, Technologies for guidance of radiofrequency
ablation in the multimodality interventional suite of the future, J. Vasc.
Interv. Radiol. 18, 924 2007 review.
78
J. Seong, C. Sadasivan, M. Onizuka, M. J. Gounis, F. Christian, L. Miskolczi, A. K. Wakhloo, and B. B. Lieber, Morphology of elastaseinduced cerebral aneurysm model in rabbit and rapid prototyping of elastomeric transparent replicas, Biorheology 42, 345361 2005.
79
J. R. Sherman, H. S. Rangwala, C. N. Ionita, A. C. Dohatcu, D. R.
Bednarek, K. R. Hoffmann, and S. Rudin, Investigation of new flow
modifying endovascular image-guided interventional EIGI techniques in
patient specific aneurysm phantoms PSAPs using optical imaging,
Proc. SPIE in press.
80
M. Kohl, Shape Memory Microactuators Springer, New York, 2004.
81
M. I. T. Research, Laboratory of Electronics RLE, J. G. Fujimoto, PI,
http://www.rle.mit.edu/rleonline/research/
LaserMedicineandMedicalImagingGroup.html
82
X.-W. Tang, Y. Yang, W. Kim, Q. Wang, P. Qi, H. Dai, and L. Xing,
Measurement of ionizing radiation using carbon nanotube field effect
transistor, Phys. Med. Biol. 50, N23N31 2005.
83
C. Karmonik, C. M. Strother, X. Chen, F. Deinzer, R. Klucznik, and M.
E. Mawad, Stent-assisted coiling of intracranial aneurysms aided by virtual parent artery reconstruction, AJNR Am. J. Neuroradiol. 26, 2368
2370 2005.
84
K. Barth, F. Cassot, D. A. Rfenacht, and J. H. D. Fasel, Anatomically
shaped internal carotid artery aneurysm in vitro model for flow analysis to
evaluate stent effect, AJNR Am. J. Neuroradiol. 25, 17501759 2004.
85
S. Baek, K. R. Rajagopla, and J. D. Humphrey, A theoretical model of
enlarging intracranial fusiform aneurysms, J. Biomech. Eng. 128, 142
149 2006.
86
M. D. Ford, G. R. Stuhne, H. N. Nikolov, D. F. Habets, S. P. Lownie, D.
W. Holdsworth, and D. A. Steinman, Virtual angiography for visualization and validation of computational models of aneurysm hemodynamics, IEEE Trans. Med. Imaging 24, 15861592 2005.
87
M. Kim, C. N. Ionita, R. V. Tranquebar, K. R. Hoffmann, D. B. Taulbee,
H. Meng, and S. Rudin, Evaluation of an asymmetric stent patch design
for a patient specific intracranial aneurysm using computational fluid dy75
309
namic CFD calculations in the computed tomography CT derived lumen, Proc. SPIE 6143, G1G8 2006.
88
J. R. Cebral, R. S. Pergolizzi, Jr., and C. M. Putnam, Computational fluid
dynamics model of intracranial aneurysms: qualitative comparison with
cerebral angiography, Acad. Radiol. 14, 804813 2007.
89
J. Cebral, A. Radaelli, A. Frangi, and C. Putman, Hemodynamics before
and after bleb formation in cerebral aneurysms, Proc. SPIE 6511,
65112C165112C9 2007.
90
S. Appanaboyina, F. Mut, R. Lhner, C. Putman, and J. Cebral, Patient
specific modeling of intracranial aneurysmal stenting, Proc. SPIE 6511,
65112D165112D11 2007.
91
Z. Wang, C. N. Ionita, S. Rudin, K. R. Hoffmann, A. B. Paxton, and D. R.
Bednarek, Angiographic analysis of blood flow modification in cerebral
aneurysm models with a new asymmetric stent, Proc. SPIE 5369, 307
318 2004.
92
S. Rudin, Z. Wang, I. Kyprianou, K. R. Hoffmann, Y. Wu, H. Meng, L. R.
Guterman, B. Nemes, D. R. Bednarek, J. Dmochowski, and L. N. Hopkins, Measurement of flow modification in phantom aneurysm model:
comparison of coils and a longitudinally and axially asymmetric stent
initial findings, Radiology 231, 272276 2004.
93
B. B. Lieber, C. Sadasivan, L. Miskolczi, L. Cesar, J. Seong, and A. K.
Wakhloo, Flow divertors to treat cerebral aneurysms: preliminary results
in the rabbit elastase-induced aneurysm model, in Proceedings BIO2006,
ASME 2006 Summer Bioengineering Conference, June 2125, 2006,
Amelia Island Plantation, Amelia Island, Florida.
94
T. Alderliesten, M. K. Konings, and W. J. Niessen, Simulation of minimally invasive vascular interventions for training purposes, Comput.
Aided Surg. 9, 315 2004.
95
T. Alderliesten, M. K. Konings, and W. J. Niessen, Modeling friction,
intrinsic curvature, and rotation of guidewires for simulation of minimally
invasive vascular interventions, IEEE Trans. Biomed. Eng. 54, 2939
2007.
96
C. Duriez, S. Cotin, J. Lenoir, and P. Neumann, New approaches to
catheter navigation for interventional radiology simulation, Comput.
Aided Surg. 11, 300308 2006.
97
D. L. Dawson, Training in carotid artery stenting: Do carotid simulation
systems really help?, Vascular 14, 256263 2006 review.
98
D. A. Gould, D. O. Kessel, A. E. Healey, S. J. Johnson, and W. E.
Lewandowski, Simulators in catheter-based interventional radiology:
training or computer games?, Clin Radiol. 61, 556561 2006 review,
D. A. Gould, D. O. Kessel, A. E. Healey, S. J. Johnson, and W. E.
Lewandowski, Clin. Radiol. 61, 903 2006 erratum.
99
A. Sengupta, T. Kesavadas, K. R. Hoffmann, R. E. Baier, and S. Schafer,
Evaluating tool-artery interaction force during endovascular neurosurgery for developing haptic engine, Stud. Health Technol. Inform. 125,
418420 2007.
An outlook on x-ray CT research and development

Ge Wanga and Hengyong Yub
Biomedical Imaging Division, VT-WFU School of Biomedical Engineering and Science, Virginia Polytechnic
Institute and State University, Blacksburg, Virginia 240601
Bruno De Manc
CT and X-ray Laboratory, GE Global Research, Niskayuna, New York 12309
Received 5 October 2007; revised 20 December 2007; accepted for publication 20 December 2007;
published 25 February 2008
Over the past decade, computed tomography CT theory, techniques and applications have undergone a rapid development. Since CT is so practical and useful, undoubtedly CT technology will
continue advancing biomedical and non-biomedical applications. In this outlook article, we share
our opinions on the research and development in this field, emphasizing 12 topics we expect to be
critical in the next decade: analytic reconstruction, iterative reconstruction, local/interior reconstruction, flat-panel based CT, dual-source CT, multi-source CT, novel scanning modes, energy-sensitive
CT, nano-CT, artifact reduction, modality fusion, and phase-contrast CT. We also sketch several
representative biomedical applications. 2008 American Association of Physicists in Medicine.
DOI: 10.1118/1.2836950
Key words: analytic reconstruction, iterative reconstruction, local/interior reconstruction, flat-panel
based CT, dual-source CT, multi-source CT, scanning mode, energy-sensitive CT, nano-CT, artifact
reduction, modality fusion, phase-contrast CT
I. INTRODUCTION
The evolution of civilization has been largely driven by the
need for extending humans capabilities. Arguably, the most
important way for us to sense the world is by visual perception. However, the human vision is severely limited by the
opaqueness of many natural and artificial objects. Hence,
how to achieve an inner vision was often pondered in various scenarios through history. Mainly credited to the pioneering works of Cormack1 and Hounsfield,2 in the last century, x-ray computed tomography CT is the first imaging
modality that allows accurate non-destructive interior image
reconstruction of an object from a sufficient number of x-ray
projections. Hounsfields x-ray CT prototype immediately
generated a tremendous excitement in the medical community and inspired a rapid technical development with an ever
strong momentum. Also, x-ray CT as the first trans-axial tomography model promoted the development of other tomographic modalities for biomedical applications and beyond,
such as magnetic resonance imaging, ultrasound tomography,
nuclear tomography, optical tomography, molecular tomography, and so on.
Since its introduction in 1973,2 x-ray CT has revolutionized radiographic imaging and become a cornerstone of every modern radiology department. Closely correlated to the
development of x-ray CT, the research for higher performance system architectures and more advanced image reconstruction algorithms has been intensively pursued for important biomedical applications. A famous CT scanner is the
dynamic spatial reconstructor DSR built at the Mayo Clinic
in 1979.3,4 With the ability to acquire 240 contiguous 1 mm
slices in a time window of 0.01 s, the heart, lung, and blood
flow can be vividly observed. This DSR system is considered
as the precursor of the electron-beam CT scanner, the dual1051
Med. Phys. 35 3, March 2008
source CT scanner and other similar systems. Early 1990s,

single-slice helical/spiral CT became the standard scanning
model,5 and helical/spiral cone-beam CT was proposed.6,7
Elscint developed a two-slice CT scanner in the mid 90s and
GE came out with the first four-slice CT scanner a few years
later, quickly followed by all other major CT
manufacturers.8,9 In 2004, Toshiba first successfully developed 256-slice CT systems.10 With the fast evolution of the
technology, true volumetric cone-beam CT scanners in helical and other scanning modes are emerging as the next generation biomedical CT.11,12 On the algorithm side, over the
past several years, triggered by an exact and efficient solution to the long-standing helical cone-beam CT problem,13,14
there has been a remarkable surge in studies on image reconstruction algorithms.15 Now, there are already a number of
exact reconstruction schemes and algorithms that deal with
general scanning trajectories and appropriately truncated
data.
Will the progress of this magnitude in the CT field continue in the next decade? Based on the impressive track
record in this field, we firmly believe it will. Then, what are
the most important research topics/directions for the x-ray
CT research and development? The answer to this big question is difficult to formulate because any prediction of that
nature must be limited by our partial knowledge and inadequate ability to look into the future. Having acknowledged
that, in this article we would like to share our opinions
for the purpose of information and inspiration, which
would hopefully guide our efforts and facilitate the future
advancement.
This outlook article is organized as follows. In Sec. II, we
will quantitatively overview the CT literature to give a sense
of the dynamics of this field. In Sec. III as the main part of
0094-2405/2008/353/1051/14/$23.00
1051
1052
Wang, Yu, and De Man: CT outlook
this article, we will present 12 topics to cover the main

trends in the CT field. In Sec. IV, we will discuss biomedical
implications of the proposed research and development. Finally, in Sec. V we will make concluding remarks.
1052
TABLE I. Representative search data on CT studies and application, where
the percentage means the number of hits for the past five years over that for
the past ten years searched on 9/4/07.
Rule
II. QUANTITATIVE LITERATURE ANALYSIS

Our methodology for literature analysis is to search the
ISI Web of Knowledge http://portal.isiknowledge.com: Science Citation Index Expanded SCI-EXPANDED from
1975 until now search performed on September 4, 2007.
Each topic search is conveniently performed with one or
more terms within article titles, keywords, and abstracts. Our
analysis was intended to give a quantitative impression but is
necessarily selective and by no means exhaustive. A reader
can easily verify our data, modify our searches and reach
his/her own conclusions.
Does the CT research have an increasing momentum?
The answer is definitely yes, despite the great developments in all competing modalities. With the search rule xray and CT, there are 5857 hits for 19752007, of which
we have only 1210 hits for 19871996 and 4553 hits for the
past ten years 19972007. With the search rule Cone-beam
and CT, we have 896 hits for 19752007, of which we
have only 117 hits for 19922001, and 777 hits for the past
five years 20022007. More detailed yearly numbers of the
retrieved articles are shown in Fig. 1. Some further representative search results over the period 19972007 are summarized in Table I. Evidently, the above data are quite informative, indicating steady and increasing efforts for the past ten
years over a wide spectrum of CT research, development and
applications. However, there should be some noise and artifacts in the results since the keywords and their combinations
could be misleading in some cases. Therefore, these data
should not be interpreted without caution.
We then retrieved all 363 x-ray and CT papers for the
year 2007, excluded irrelevant ones, and categorized them,
as shown in Table II. Clearly, the biomedical applications
remain the mainstream for x-ray CT, and a major portion of
efforts is being devoted to research and development of
methods and systems. Also, there are some non-biomedical
applications, which are about one third of the medical applications.
CT
CT
CT
CT
CT
CT
CT
CT
CT
CT
CT
and
and
and
and
and
and
and
and
and
and
and
algorithm
cancer
cardiac
dose
dynamic
dual-energy
screen
therap
diagno
fusion
mouse
No. of Hits
for 97-07
No. of Hits
for 02-07
2785
11 468
2252
5595
2572
235
2771
12 099
27 025
1586
787
1829
7642
1517
3609
1456
142
1812
7296
15604
1134
416
65.7
66.6
67.4
64.5
56.7
60.4
65.4
60.3
57.7
71.5
52.9
III. TWELVE TOPICS FOR THE NEXT DECADE

III.A. Analytic reconstruction
Since Katsevichs 2002 paper on exact and efficient helical cone-beam CT reconstruction,14 intensified research efforts have been made in this area.15 Various sophisticated
formulas have been proposed for exact reconstruction from
projection data that can be longitudinally truncated and even
transversely truncated, and for not only a standard helical
trajectory but also a quite general class of scanning loci Fig.
2.1624 However, in the general case such as saddle-like
scanning paths24 and nonstandard helical trajectories, these
algorithms are far less computationally efficient as compared
to the popular filtered backprojection FBP with spatially
invariant filtering. In the 9th International Meeting on Fully
3D Image Reconstruction in Radiology and Nuclear Medicine Lindau, Germany, July 913, 2007, Katsevich presented an important progress towards exact yet efficient general cone-beam reconstruction algorithms for two classes of
scanning loci.25 The first class curves are smooth and of positive curvature and torsion. The second class consists of
circle-plus curves, with the segment for the plus part starting
below the circle-like trajectory and ending above it.26 However, there are other important classes of trajectories for
which exact and efficient algorithms are desired and highly
nontrivial. Therefore, it remains an open challenge to formulate such exact and efficient algorithms for many other types
of scanning curves. Theoretical unification of these new exact algorithms is also worthwhile.2729
TABLE II. Categorization of 262 relevant papers selected from the 363 articles in English retrieved under the rule x-ray and CT for 2007
searched on 9/4/07.
Category
FIG. 1. Numbers of retrieved papers during 19972007 with the searching

rules a x-ray and CT and b cone-beam and CT, respectively.
Exponential fitting shows that there is a 5% increment yearly for the former,
while a 39% increment for the latter. Note that the numbers for 2007 are
incomplete.
Medical Physics, Vol. 35, No. 3, March 2008
Methods & systems

Medical applications
Pre-clinical applications
Non-biomedical applications
No. of Papers
72
127
14
49
27.5
48.5
5.3
18.7
1053
FIG. 2. Representative general scanning trajectories. a A non-standard helix, b a saddle curve and c a circle-plus-line combination.
Furthermore, it has been well recognized that the way to

formulate exact and efficient reconstruction is not
unique.17,27,30 The key is to select an appropriate weight
function. Different choices of the weights would have distinct impacts on the image noise distribution. Depending on
each specific clinical application, the preferred image noise
distribution may be uniform or least noisy in a region of
interest ROI while being more tolerable outside the ROI. In
other words, an ideal exact cone-beam reconstruction formula needs to be not only efficient but also produce the best
possible image noise distribution, which is application dependent. The governing theory guiding this procedure has
been studied from different aspects but it has not been thoroughly developed yet.
In contrast to impressive progress in solving the so-called
long objection problem reconstruction of a long object from
longitudinally truncated cone-beam data, cardiac cone-beam
CT is what we call the quasi-short object problem reconstruction of a short portion of a long object from longitudinally truncated cone-beam data involving the short object
and deserves more research efforts. To solve the quasi-short
object problem, the circular cone-beam scan only permits
approximate reconstruction and the helical cone-beam scan
has inefficient photon utilization. The saddle-like cone-beam
scan can combine exact reconstruction and good photon utilization, representing a promising direction.24,31,32 Again, we
will need exact, efficient and optimized algorithms in the
context of solving the quasi-short object problem.
In the CT applications, approximate cone-beam reconstruction algorithms have been dominating despite the rapid
development of exact cone-beam reconstruction algorithms.6
1053
There are two reasons. First, in a good number of applications, the data completeness required for exact reconstruction
cannot be satisfied due to physical constraints. Hence, the
only choice is to perform an approximate reconstruction.
Second, even if an exact reconstruction is practically feasible, oftentimes the approximate algorithms can deliver
similar or even better performance as compared to the exact
reconstruction counterparts, which are not necessarily optimal in terms of the noise characteristics. Figure 3 shows two
approximate reconstructions of excellent quality.33
It is our opinion that the popularity of approximate conebeam reconstruction algorithms will definitely sustain in the
near future. With the development of exact cone-beam reconstruction algorithms, the approximate cone-beam reconstruction algorithms will be surely improved as well. In addition
to various approximations to be improved or made on an
individual basis, a promising research direction for approximate cone-beam reconstruction would be to adapt new exact
cone-beam reconstruction algorithms.3436 By doing so, the
merits of exact cone-beam reconstruction algorithms can be
largely inherited at a much lower computational cost and
modified to have certain practically desirable features. This
approach should be particularly attractive in the cases of incomplete and/or inconsistent data such as cardiac CT, contrast studies, artifact reduction, and so on.
III.B. Iterative reconstruction
While the very first CT scanners used algebraic iterative

reconstruction ART,37,38 FBP39 soon became the gold standard for CT reconstruction. A few years later, statistical iterative reconstruction was successfully introduced for emission tomography,40 because FBP reconstruction from a low
SNR signal-noise-ratio emission data set would produce
quite poor image quality. A simultaneous update variation of
ART, called simultaneous ART, was presented in 1984.38 In
the past decade, thanks to the increasing computational
power, statistical iterative reconstruction has become a hot
research topic for CT,4146 with a focus on noise suppression,
artifact reduction and dual energy/energy-sensitive
imaging.46,47,92 Recent advances in statistical iterative
reconstruction48 promise to achieve dramatic improvements
in image quality, as illustrated in Fig. 4. Figure 4 shows the
FIG. 3. Images reconstructed from clinical data using

an approximate helical cone-beam FBP algorithm see
Ref. 33. a A reformatted abdominal image display
window width/length: 381/83HU and b a reformatted cardiac image display window setting: 800/
216HU duplicated from Ref. 33 with the permission
of the Institute of Physics Publishing.
1054
1054
FIG. 4. Images analytically and iteratively reconstructed from clinical data. a A FBP reconstruction from a 300 mAs dataset, b a FBP reconstruction from
a 75 mAs dataset, and c an iterative reconstruction from the same 75 mAs dataset Ref. 48 Courtesy of J.-B. Thibault.
impact of a 4x reduction in x-ray tube current mA. With

FBP, the noise level increases dramatically the standard deviation of the noise increases by a factor of 2 theoretically.
With iterative reconstruction, the noise level is not increased
or even descreased relative to the full-dose FBP counterpart,
suggesting a 4x dose reduction in this particular case. New
iterative reconstruction schemes will surely emerge, as will
be further discussed for Topics J and K. We believe that it is
only a matter of time for statistical iterative reconstruction to
become available on commercial CT scanners.
Several factors have prevented statistical iterative reconstruction from being deployed on CT products so far. The
first is the huge computational cost, about two to three orders
of magnitude larger than that of FBP. Despite recent advances in high-performance image reconstruction based on
field-programmable gate arrays, multi-core processors such
as IBMs Playstation 3 Cell Broadband Engine, and graphics
processing units, we expect that first product realizations will
be slow tens of seconds per slice. The second is the robustness of iterative algorithms across different protocols and
applications. Currently, some significant parameter tuning
needs to be done for practical use of an iterative method in a
specific application. The best imaging performance is usually
very sensitive to the particular choice of parameters. The
third is the potentially new look of statistically reconstructed
images. Since images are reconstructed in a statistically optimal fashion, their noise and artifact characteristics can be
rather different from those of FBP. Radiologists are used to
the FBP image appearance and associated various image
quality trade-offs. Consequently, statistical reconstruction
may give an impression of somewhat reduced diagnostic
value in certain cases. In fact, its spatial resolution characteristics are indeed very different from the FBP counterpart
and depend on many factors. A careful quantitative analysis
of image noise and spatial resolution as a function of location, contrast, and measurement statistics shows indeed that
statistical iterative reconstruction has the potential to improve the information content of reconstructed images.49
In perspective, iterative reconstruction offers distinct advantages relative to analytic reconstruction in important
cases when data are incomplete, inconsistent and rather
noisy. Even in the cases where analytic reconstruction performs well, there is no fundamental reason why iterative reconstruction would perform any worse. Hence, we predict
that the fast development in computing methods and hardware will lead to a shift from analytic to statistical reconstruction or at least a fusion of analytic and iterative approaches. This evolution will happen gradually, because of
the high computational cost as well as the time needed for
the radiologists to adopt this new technology.
III.C. Local/interior CT
There is a long track record for the exact reconstruction of

an ROI from a minimum dataset, starting from the work on
half-scan fan-beam reconstruction.50 The benefits include
shorter data acquisition time, less radiation dose and more
imaging flexibility. The most remarkable recent finding is the
work by Katsevich in 2002 that demonstrates the feasibility
of the exact regional reconstruction within a long object from
longitudinally truncated data collected along a PI -turn of
a helical scanning trajectory.14 The subsequent backprojection filtration variant51 and generalization into the quite arbitrary scanning case16,18 significantly enriched the local CT
theory. As illustrated in Fig. 5, the latest results along this
direction are the one-sided inverse Hilbert transform based
local reconstruction algorithm52 and the truly truncated Hilbert transform based interior reconstruction algorithm
coupled with some a priori knowledge in an ROI to be
reconstructed,53,54 which is somehow challenging the conventional wisdom that an interior problem reconstruction of
an interior region from projection data along lines through
the region does not have a unique solution.55 Clearly, exact
solutions to the interior problem have tremendous implications for numerous biomedical and other applications.56
It is highly desirable to establish a unified theory for exact
local/interior reconstruction that covers the exact reconstruction schemes and methods from a minimum dataset in the
two-dimensional 2D and three-dimensional 3D cases
from parallel- and divergent-beam geometries. It should be
valuable to refine the reported stability analysis and reflect
the sampling geometry and data noise in an optimal fashion.
1055
FIG. 5. Theoretically exact solution to the interior problem. The exactly

recoverable regions are not the same according to the different sufficient
conditions-based on the truncated Hilbert Transform: a The exact reconstruction region by Defrise et al. Ref. 52, b that by Ye et al. Ref. 53
where fx is the object image along a given PI line within a compact
support and gx is the corresponding Hilbert Transform, c a slice of the
thorax phantom with a T-shaped local ROI, and d the image reconstructed
from a purely local dataset plus prior knowledge on a sub-region in the ROI
using the POCS method Ref. 53.
Also, it is critical to design numerically stable, robust and

efficient algorithms for this purpose. The state-of-the-art
framework for local/interior reconstruction is the Hilbert
transform analysis. Perhaps other possibilities exist for us to
gain a thorough understanding of this amazing problem.
As an alternative technology for local/interior reconstruction, theoretically exact fan-beam lambda tomography LT
was also available for a general scanning locus based on the
2D Calderon operator.57,58 However, exact cone-beam LT is
generally impossible based on the 3D Calderon operator for
practical scanning loci such as a helix.58 It is promising to
develop exact LT reconstruction methods in terms of the 2D
Calderon operator in some planes for data collected along a
general scanning locus. As a related area, partial derivatives
of a dynamically changing image volume can be directly
reconstructed from original cone-beam data. Thus, velocity
tomography may be improved from these partial derivatives
based on the general velocity field constraint equation.59
III.D. Flat-panel based CT
While clinical x-ray systems are conventionally limited to

projection images, digital, solid-state flat-panel x-ray detectors FPDs have enabled 3D reconstruction in these modalities. In 1999, GE Healthcare first introduced FPD-based
x-ray systems, in which FPDs replaced film in radiography
and mammography. Then, GE introduced an FPD-based cardiology system, replacing the conventional image intensifier
II tube. Today, all of the major diagnostic imaging system
1055
manufacturers GE, Hologic, Philips, Siemens, and Toshiba

offer FPD-based x-ray systems. GE, Hologic, Canon and
Shimadzu make FPDs independently. FPDs in use today employ pixel sizes ranging from 70 70 m2 to 200
200 m2, in arrays of up to 14 million pixels, and active
areas of up to 43 43 cm2. Some FPDs are of the indirect
conversion type, in which a scintillator converts the x rays to
visible light, and a 2D array of photodiodes converts the light
to electronic charges to be digitized. In an alternative direct
conversion technology, x-rays are converted directly to electronic charges and then digitized.
Even prior to the development of FPDs, 3D reconstruction of multi-angle digital projection images from an IIbased C-arm cardiology system was demonstrated.60 Once
FPDs were incorporated into such systems, the non-ideal
characteristics of the II tube were circumvented, and CT reconstruction of data from C-arm systems became practical.
Today most clinical FPD-based C-arm systems have the ability to acquire and reconstruct 3D CT images. Digital Breast
Tomosynthesis61 is a new modality using key elements of
FPD-based mammography systems. The x-ray source is
moved along an arc and a relatively small number of projections are acquired, enabling reconstruction of approximately
1.0-mm-thick image planes. This technology is currently in
advanced development. Researchers have also incorporated
FPDs into a conventional CT gantry to achieve high spatial
resolution in medium to large field-of-view scans.62 This architecture was primarily intended for small animal
imaging.63,64 FPDs are also under investigation in a few
other application-specific areas, such as breast CT
scanners,65,66 infra-operative imaging systems,67 and compact or mobile CT systems for head or dental imaging.68
FPD-based CT is still in its infancy, and substantial developments are expected in the future.
III.E. Dual-source CT
Most commercial clinical scanners until today are based

on the so-called third-generation architecture. A single x-ray
tube and a single detector assembly are positioned face to
face and rotated jointly around the patient. Berninger and
Redington69 presented the idea of replicating the source and
detector, as illustrated in Fig. 6a. The dynamic spatial reconstructor may have been the first real multi-source
prototype.4 About 25 years later, Siemens developed the first
commercial two-tube-two-detector scanner, also known as
dual-source CT.70 The main advantage of this architecture is
its improved temporal resolution. In todays state-of-the-art
CT scanners, the gantry rotation time is reduced to about
0.35 s, but it is mechanically challenging to reduce that time
even further, which justifies the renewed interest in multisource architectures. For the dual-source CT system, the
minimum rotation interval is 90 + , compared to 180 +
for a third-generation scanner. For a cardiac fan angle of
25, this gives about 44% reduction in acquisition time.
The dual-source CT has been claimed to have a dose benefit because the second detector is smaller and hence the dose
is reduced at the edge of the field of view. However, the
1056
1056
FIG. 6. CT architecture with three sources and three

detectors. a Berninger and Redington first proposed to
replicate the x-ray tube and detector assembly Ref. 69
and b the triple-source system in the helical conebeam geometry allows a perfect data mosaic pattern for
exact reconstruction. Figure 6a is duplicated from
Ref. 69 with the permission of GE Company.
exact same dose profile is achieved with a third-generation

architecture and a more aggressive bowtie. A more fair comparison would compare patient dose at an identical noise
level throughout the field of view. In fact, the cross scatter
from the second source into the first detector is reported to
result in a scatter-to-primary ratio as high as 100% for obese
patients,71 corresponding to a severe dose penalty. Therefore,
more research will be needed on both hardware scatter rejection and software scatter correction methods for dualsource CT.
A next step in this area is to use three sources and three
detectors.7274 For a cardiac fan angle of 25, a triplesource system with symmetric spacing would result in 59%
reduction in acquisition time. In terms of exact cone-beam
reconstruction, unlike the dual-source cone-beam geometry
the triple-source cone-beam scanner allows a perfect mosaic
pattern of truncated cone-beam data to satisfy the Orlov
condition,75 as shown in Fig. 6b. In the near future, manufacturers may or may not adopt the multi-source-multidetector strategy, depending on their ability to conquer temporal resolution with other means, such as faster rotation,
virtual rotation see next sub-section, or motion compensation techniques.76 Triple-source CT has unique theoretical
merits relative to dual-source CT. However, the brute-force
hardware approach of dual-source and triple-source CT
comes with important technical and cost trade-offs, and
might limit them to be niche cardiac scanners, just like
electron-beam CT EBCT.
III.F. Multi-source CT
Almost all modern CT architectures are based on one or

more single-spot x-ray tubes possibly with focal spot
wobble. The best known counter example is the EBCT
scanner,77 which uses a huge x-ray source surrounding the
patient with an electron beam that is deflected to produce
x-rays along four half-circle target rings. The distributed nature of the x-ray source makes it possible to rotate the focal
spot freely, thereby eliminating any mechanical bottlenecks.
This architecture is fairly complex, limited in source intensity, and not so compatible with true volumetric scanning,
but it is extremely fast and therefore it is very useful for
cardiac scanning. A miniature version of the e-beam CT

scanner was also conceptualized for micro-CT cardiac
studies.78 Most of the drawbacks of EBCT are eliminated by
going to a distributed source with discrete electron emitters
and appropriate source-detector topologies.79 As shown in
Fig. 7a, another class of x-ray sources has multiple spots
distributed along the longitudinal axis in the z direction.80
It has been shown that the combined information from the
multiple spots in the z direction effectively eliminates conebeam artifacts.81
Another example of a distributed x-ray source with a deflected e beam is the transmission x-ray source developed by
NovaRay formerly known as Cardiac Mariners and Nexray,
Palo Alto, California, USA, with thousands of focal spots.
This area source was first used to demonstrate the concept of
inverse-geometry CT.82 In addition to eliminating cone-beam
artifacts, this architecture also has the benefit of a small
photon-counting detector and very good detection efficiency.
A related architecture is based on discrete electron emitters,83
resulting in a 2D array source with tens of focal spots Fig.
7b. This source architecture is more compact than the
above and perhaps more compatible with the concept of a
virtual bowtie, where the operation of each spot is modulated
in real time to optimize image quality and minimize dose,
depending on the patient anatomy. In the past decades, advances in detector technologies defined the so-called slice
wars. We expect that in the next decade dramatic advances
in distributed x-ray sources may define a new revolution in
CT and give birth to a wide class of new multi-source CT
architectures, including line sources, inverse-geometry CT,
and ultimately a rebirth of stationary CT, which means that
neither the source nor the detector is in motion during the
data acquisition process.
III.G. Novel scanning modes
A very important trend in the CT field is the evolution in

scanning modes for improved imaging performance at reduced patient dose. The introduction of single-/multidetector-row/cone-beam helical CT was only the first step
towards fully volumetric CT. In practice, a number of more
flexible scanning schemes are needed. An earliest example of
1057
1057
FIG. 7. Novel source configurations. a A distributed source with multiple spots along the z direction to eliminate cone-beam artifacts Ref. 80, and b an
inverse-geometry architecture with tens of spots combining the benefits of a line source in z, a small photon-counting detector, and a virtual bowtie Ref. 83.
Figure 7a is duplicated from Ref. 80 with the permission of GE Company.
a dynamic helical pitch trajectory is bolus-chasing CT

angiography.84,85 In this application the table speed is adjusted on the fly to synchronize the CT imaging aperture with
the contrast bolus peak. Also, the so-called shuttle mode is a
scanning scheme where the patient table travels back and
forth, enabling better dynamic applications than with conventional scanning, including liver perfusion and brain perfusion. These applications are based on cone-beam reconstruction algorithms specifically designed for dynamic
helical pitch source trajectories.86 Another example is a combined cardiac and thorax scan, where the table speed is reduced in the cardiac region for phase-coherent imaging, and
increased as soon as the scan aperture is past the cardiac
region. Finally, the latest efforts on saddle-like cone-beam
scanning seem also promising for exact solution of a quasishort object problem reconstruction of a short portion of a
long object from longitudinally truncated cone-beam data involving the short object.24,31 Especially, the compositecircling scanning mode Fig. 8 has some advantages over
the existing scanning modes from perspectives of both engineering implementation and clinical applications because of
its symmetry in mechanical rotation and the compatibility
with the physiological conditions.24,31
To minimize patient dose in CT, it is critical to detect
every transmitted photon and to treat it in a statistically optimal fashion during reconstruction. It is equally important to
optimize the flux profile through the patient based on the
patient attenuation profile and the organ sensitivities. A first
way to achieve this is to pre-attenuate the x-rays with a socalled bowtie filter. Bowtie filters are becoming more advanced in targeting regions of interest, sometimes even using
one or more moving parts.87 A second way is to modulate the
tube current and possibly other scan parameters such as kVp
or filtration as a function of rotation angle, longitudinal position, and cardiac phase.88 A major advantage of an inverseMedical Physics, Vol. 35, No. 3, March 2008
geometry architecture with an array source is the ultimate

flexibility to customize the flux profile with the so-called
virtual bowtie.83
Key to all the above methods is knowledge about the
patient anatomy. This knowledge may be based on scout
scans, patient data or atlases.89 One can take huge advantage
of this knowledge not only to perform simple tasks such as
FIG. 8. Compositing-circling scanning mode. In such a CT system, the scanning trajectory is a composition of two circular motions: While an x-ray
focal spot is rotated on a plane facing an object to be reconstructed, the
x-ray source is also rotated around the object on the gantry plane. Once a
projection dataset is acquired, exact or approximate reconstruction can be
done Copyright by G. Wang and H. Y. Yu, US Provisional Patent Application, 2007.
1058
FIG. 9. Energy dependence of the linear attenuation coefficient varies across

different elements and tissues. The plot shows the linear attenuation coefficients for bone continuous curve and iodine curve with a discontinuity at
the K edge. While for the shown iodine concentration the average attenuation is very similar to that of bone; their energy dependence is very
different.
bowtie selection and patient centering but also ultimately to

compute optimal pulse sequences, avoid sensitive organs,
and perform true ROI imaging.
III.H. Energy-sensitive CT
The idea of decomposing the linear attenuation coefficient

into two or more basis functions in order to eliminate beam
hardening artifacts for material discrimination was proposed
soon after the invention of CT.90 The physical foundation is
that the linear attenuation coefficients are a function of energy and that this energy dependence is different for different
tissues and elements. Figure 9 also illustrates that various
elements have different attenuation properties as functions of
photon energy.91 Previously, two measurements were acquired at different tube kVp settings, also called dual-energy
CT. It was then realized that iterative reconstruction can
eliminate the need for a second measurement by incorporating prior information.43,46,92 Technological advances are now
making it possible to acquire two or more measurements
simultaneously at different effective x-ray energies. These
methods can be primarily evaluated by signal-to-noise ratio
and measurement registration. The process of material decomposition results in a noise amplification. Noise can be
minimized if the effective energies of the measurements are
well separated, and better algorithms are designed.
Dual-layer detectors are fairly mature93,94 and the corresponding measurements are perfectly registered but of all
solutions have the poorest energy separation. Dual-source
CT is an existing technology,70 its energy separation is better
than dual-layer detectors but there is a significant time misregistration about 100 ms between corresponding measurements, not suitable for dynamic applications. Fast kVp
switching is an existing technology,95 with better energy
separation than the dual-layer detector, and very recently it
1058
has been shown that its time misregistration can be as good

as one or a few views 1 ms.96 Advances in compact
monochromatic x-ray sources may result in improved energy
selectivity, but this solution does not address the time misregistration. The ultimate architecture for energy-sensitive
CT may use an advanced polychromatic source to cover a
spectrum of interest and a photon-counting-based energydiscriminating detector to record all these polychromatic
photons and sort them into respective spectral bins.97 While
this solution would give perfect energy separation and registration, the technology is still immature and suffers from
severe count rate limitations. Along with dedicated hardware
development, algorithm development, especially statistical
methods, will definitely facilitate this type of application.
The fact that all manufacturers are pursuing different approaches makes the area of energy-sensitive CT an extremely
interesting area. None can change the laws of physics but
enormous efforts in the coming years will show how dramatic advances in CT source, detector, and reconstruction
technologies will help us all to approach the fundamental
performance limits of energy-sensitive CT.
III.I. Nano-CT
The worlds first and only sub-100 nm nano-CT scanner

nanoXCT98 was recently developed by the Xradia company
Concord, CA. This system is a revolutionary microscope
for non-invasive investigations involving semiconductor
analysis, drug discovery, molecular imaging, stem-cell research and materials development. It allows 50 nm resolution using proprietary condenser and objective optics. For
most samples in nanotechnology, the x-ray attenuation length
for low Z materials is very long, resulting in poor image
contrast. The Xradia system can significantly increase image
contrast in the Zernike phase contrast mode.
Since this century, nanotechnology has gained tremendous
momentum through both governmental and private investments. Clearly, nano-CT may be a strategic enabling component for the immediate future research and education. In a
broad range of nano-CT applications, interior tomography is
not only valuable but also necessary. For example, in the
case of in-situ imaging of cells or tissue specimens at the
cellular/molecular level, we require little morphological
changes and minimum radiation exposure, have the water or
air component as reference and a volume of interest VOI
much smaller than the specimen. Since the recent theoretical
and numerical results53,54,56,99 demonstrated that the interior
problem can be solved in a theoretically exact and stable
fashion assuming that a small subregion within the interior
region is known, it becomes now feasible to meet the abovementioned interior reconstruction need for nano-CT.
In the next decade, we believe that the existing nano-CT
scanners will be further advanced, unique reconstruction algorithms will be developed with multi-scale and interior imaging capabilities, and more nano-CT applications will be
identified in the fields including but not limited to life science, preclinical and clinical imaging studies in vitro, pharmaceutical research, and so on.
1059
1059
FIG. 10. Representative CT image artifacts. a A shoulder phantom image with streaks caused by photon starvation, b a patient with spine implants
generating metal artifacts, and c a moving head leading to motion artifacts Duplicated from Ref. 100 with the permission of the Radiological Society of
North America.
III.J. Artifact reduction
III.K. Modality fusion
Reduction of image artifacts has been a central topic in

the CT field.100 The paradigm shift towards volumetric CT,
novel architectures and dynamic and quantitative imaging
demands a more effective reduction of various artifacts. The
well-known scattering artifacts become more and more serious with the increasing cone angle and dual-source CT configurations. Beam hardening artifacts must be suppressed to
extract energy-dependent information.101 Motion artifacts remain a main challenge for cardiac CT and contrast-enhanced
studies.102,103 More than a dozen types of artifacts are well
known to the field Fig. 10, most of which assume new
forms and present new problems associated with the development of new CT technologies. Therefore, the fight against
these artifacts remains active and requires new tools.
While many traditional artifact reduction algorithms are
ad hoc and approximate, the future efforts may rely on more
solid physical models and more rigorous inherent data
integrity.101,104,105 In this context, data consistency conditions
were proposed to suppress motion artifacts, minimize beam
hardening and so on. The invariance of the integral was suggested as a possible mechanism for this purpose. Nevertheless, much more work is required to advance this frontier. To
a large degree, artifact reduction is very similar to image
reconstruction. In both cases, the goal is to find an optimum
subject to a set of constraints. Given the complexities imposed by the artifact-related constraints, the iterative approach will play a more important role. Several iterative
schemes have been well studied so far.106108 New iterative
schemes deserve major attention and refinement, such as the
alternating iteration scheme for metal artifact reduction.109
Only with optimized reduction of various artifacts, the future
CT technology will deliver its ultimate performance that
should be spatially, dynamically, spectrally and quantitatively correct.
A distinguished trend in modern biomedical imaging is

the area of multi-modal imaging, in which two or more imaging systems are synergistically integrated for much better
performance, improving or enabling biomedical applications.
A primary example is the positive emission tomography
PET/CT systems.110 Another example is the hybrid optical
tomography systems such as those proposed for magnetic
resonance
imaging
MRI-based
diffuse
optical
tomography,111 CT/MRI integration,112 and CT-based bioluminescence tomography.113 Recently, the Siemens micro-CTPET-single photon photo emission computed tomography
SPECT system Inveon Fig. 11 became commercially
available as an integrated preclinical imaging platform. From
the perspective of x-ray CT research and development, an
unprecedented potential would be unlocked by identifying
new combinations of complementary imaging modes and improving the existing multi-modal systems, such as hybrid
CT-angio and CT-cardiac cath systems. There are good possibilities for one-stop imaging centers or suites to emerge
where all the imaging tasks can be streamlined in a taskspecific fashion, which is in some sense an extension of the
currently already available trail based multi-modal small animal imager.
In addition to the architectural issues, we emphasize that
there are excellent opportunities for algorithm development
in this area. Traditionally, each component modality of a
fusion-based system can be independently considered for image reconstruction. Then, all reconstructed images are retrospectively combined via post-processing for further analysis.
However, there is generally some or strong correlation
among the datasets acquired by multiple imaging tools applied to study the same individual object. Ideally, we should
not solve the imaging problems for these modalities separately but couple these imaging problems with implicit or
1060
1060
FIG. 12. Ordinary-x-ray-tube-based phase-contrast imaging system with

three 1D gratings G0, G1 and G2. The key idea is to utilize the so-called
Talbot effect, which is a periodic self-imaging phenomenon, to extract phase
shift information Duplicated from Ref. 121 with the permission of Nature
Publishing Group.
FIG. 11. Imaging platform Inveon for fusion of CT, PET and SPECT in
preclinical applications Courtesy of Siemens Company.
explicit relationships describing dependence among the involved datasets. Such an integrated inverse problem may require an iterative solution containing several loops each of
which assumes other image reconstructions known and refines an intermediate image, or have more sophisticated
forms like a truly simultaneous iterative solution.114 Theoretical studies are needed to establish the solution existence,
uniqueness and stability with new iterative reconstruction
schemes, as already mentioned for Topics B and J. In the
cases of no unique solutions, regularization issues must be
addressed with the aid of a priori knowledge in the form of
constraints, penalty terms and so on.
X-ray phase-contrast imaging has been a hot topic over

the past decade.117120 It is traditionally implemented via interferometry, diffractometry and in-line holography, respectively. Interferometry and diffractometry are restricted by the
availability of a synchrotron radiation facility. The in-line
holography approach was originally proposed by Gabor in
1948, for which he was awarded the Nobel Prize. Wilkins et
al. developed in-line holography techniques with a microfocus polychromatic x-ray source.120 However, a microfocus 100 m x-ray tube offers limited x-ray flux. In
2006, Pfeiffer et al. made a major improvement so that such
a phase-contrast imaging system can be built on a hospitalgrade x-ray tube.121,122 As shown in Fig. 12, the key idea is
to utilize the Talbot effect generated by x-ray gratings for
sensing waveform deformation. It is anticipated that more
research efforts will be devoted to x-ray phase-contrast imaging including x-ray phase-contrast tomography using the
micro-focus or the grating-based framework.123 The forward
model will be a theoretically and computationally challenging problem, particularly for relatively large specimen. Accordingly, the development of phase-contrast reconstruction
algorithms will become an area of intense research.
III.L. Phase-contrast CT
In all mainstream x-ray CT imaging modalities, the contrast mechanism has been attenuation based for over a century. As a result, weak-absorbing tissues are not imaged well,
and radiation dose has been a general concern. On the other
hand, x-ray phase-contrast imaging relies on the diffraction
properties of structures, promising to have much better contrast at much lower dose. Specifically, for x-rays the refraction index of biological soft tissues is approximately 1.0,
namely, n = 1 + i, where and represent attenuation
and diffraction, respectively. In the range of 15 150 KeV,
is about three orders of magnitude greater than , making
phase-contrast imaging about three orders of magnitude
more sensitive than attenuation-based methods.115 Typical
values of refraction indexes of breast tissue are plotted in
Ref. 116.
IV. BIOMEDICAL APPLICATIONS

Image quality can essentially be summarized by four main
performance characteristics. First, high-contrast image resolution spatial resolution is the ability to distinguish adjacent
high-contrast features. Second, low-contrast image resolution
contrast resolution measures the ability to differentiate a
low-contrast feature from its background. Image noise imposes a grainy appearance due to random fluctuations of the
x-ray photon flux, and is a key factor in limiting low-contrast
resolution. Third, temporal resolution determines the ability
to capture structures in motion. Fourth, quantitative accuracy
is desired to relate the image pixel values to physically
meaningful quantities in the absolute sense, which is particularly important in dual/multi-energy CT. Image artifacts are
structured interferences of any type and clearly need to be
1061
minimized at all times, but are generally hard to quantify.

The aforementioned 12 topics of x-ray CT research and development are all aimed at major improvements in image
quality and are well motivated by the imaging needs of important biomedical applications.
Just like for clinical CT, there is a strong need for improved technologies for micro-CT of small animals, especially genetically engineered mice.124126 Most CT research
results can be similarly applied in either clinical or preclinical scenarios. CT of large animals will remain important for
veterinary medicine and research.
In terms of spatial resolution, CT has come a long way.
Only a few years ago routine studies were done at 5 7 mm
slice thickness, while todays scanners offer half-mm isotropic spatial resolution. Still, there is an opportunity to dramatically improve upon the state-of-the-art CT resolution
clinically available. As an example, we are working to establish the feasibility of a clinical micro-CT CMCT prototype
targeting 60 100 m resolution for temporal bone imaging.
There has been an explosive growth in the development of
micro-CT scanners with an emphasis on high spatial
resolution126 but none of them allow patient studies due to
the narrow imaging aperture, long acquisition time, and dose
limitations. Our proposed CMCT system consists of a special
micro-CT scanner, a clinical CT scanner, and a crossmodality registration mechanism involving both hardware
and software.127 This system will integrate the strengths of
micro-CT and clinical CT techniques to achieve the spatial
resolution that is several times or even an order of magnitude
finer than any commercially available scanner for clinical
temporal bone imaging. Our preliminary system design and
simulation results indicate that such a system is possible at a
clinically acceptable dose, provided motion artifacts can be
effectively eliminated. An enabling technology for CMCT at
a minimum patient dose is to perform local CT reconstruction and/or local tomosynthesis of a volume of interest
VOI, by physically monitoring the patient motion and compensating for it during the reconstruction, subject to various
constraints. This technique is particularly useful for imaging
of the inner ear, because of its small dimensions, highcontrast structures, fine features, and stationary nature, and
may be extended for other applications such as bone cancer
studies. Another potentially useful solution would be to apply exact local reconstruction,53,99 as discussed earlier, which
can be immediately applied to dental CT as well. In such a
configuration, x-rays would only go through a tooth of interest and its neighborhood. Based on the known CT numbers
in the neighborhood air, the tooth can be reconstructed accurately, allowing characterization of bony loss and so on.
Improvements in low-contrast resolution will lead to reduced radiation dose and superior diagnostic performance.
Worldwide there are active discussions and public concerns
on radiation induced genetic, cancerous and other
diseases.128,129 CT is considered a radiation-intensive procedure, yet becoming more and more widespread. In the mid1990s, CT scans only accounted for 4% of the total x-ray
procedures but they contributed to 40% of the collective
dose.130 The recent introduction of helical, multi-slice and
1061
cone-beam technologies has increased and continues increasing the usage of CT. There is a substantial room for dose
reduction using algorithmic means for attenuation-based CT.
Other possible directions include more dose-efficient scanning protocols, better a priori knowledge, more effective reconstruction methods and smarter post-processing strategies.
Phase-contrast based CT, if successfully developed and clinically utilized, will revolutionize its x-ray imaging applications by virtually eliminating the dose problems and greatly
boosting contrast resolution, which may open doors to novel
functional and molecular imaging studies, and overtake or
complement some MRI applications. X-ray phase-contrast
CT/micro-CT is a technically very challenging and relatively
immature area, but it has the potential to have huge impact in
certain pre-clinical and clinical applications.
As far as temporal resolution is concerned, cardiac CT is
by far the most prominent application.131 Most common CT
targets are relatively motionless, such as bony structures,
head, neck, and abdomen. However, the closer we get to the
heart, the poorer the image quality becomes, because of the
cardiac motion. The coronary arteries are too elusive to be
clearly captured in challenging cases such as when the cardiac rate is too high or irregular, since the relative rest phase
of the coronary arteries is only about 60 ms. Most demanding cardiac imaging used to be done by EBCT, which is
expensive and rarely available. In addition to its remarkable
applications for dynamic anatomical imaging of cardiac
structures, EBCT is also a powerful tool for physiological
imaging. Following the introduction of the Siemens dualsource scanner, it seems that more efforts are warranted to
design better multi-source or distributed source systems and
methods for cardiac CT. The ideas and schemes already exist
on triple-source and multi-source CT systems, and hopefully
will be refined and realized in the not too distant future.
Finally, CT images with a high quantitative accuracy will
be more informative in general and extremely useful with
energy-sensitive CT in particular. It is anticipated that more
advanced x-ray sources such as portable synchrotron radiation devices will be much more accessible one or two decades from now. In the meantime, innovative sources and
detectors will be developed to generate spectrally resolved
projection data directly or indirectly. Statistical reconstruction can be applied to design and refine iterative algorithms
that fully utilize all detected photons, and add a spectral dimension to CT images. Presumably, this type of energysensitive CT will dramatically improve the characterization
of heart and lung diseases as well as various cancers. Along
this direction, scattering-based CT, x-ray induced fluorescence CT, and CT in other hybrid modes may also become
useful.
V. CONCLUDING REMARKS
When the first author of this paper graduated with a CT
dissertation in 1992, he considered to switch to a different
imaging field. His former mentor, Dr. Vannier, who is an
authority in x-ray imaging and a certified radiologist, advised
him not to do so because CT was just too practical and
1062
useful. After over a decade, Dr. Vanniers advice seems still

valid. Even if just a few of the aforementioned possibilities
are realized, CT will become much more powerful than it is
now. The ideal CT system should waste no information related to absorption, scattering, diffraction, energy, time of
flight, prior information, and so on. Correspondingly, the imaging model will become more complicated: going from a
simple line integral model and the radiation transport equations for describing attenuation and scattering to Maxwells
equations for accurately modeling the wave nature of phasecontrast imaging. Although the computational obstacles appear overwhelming at this time, we remain optimistic about
our collective creativity and capabilities. The dream of whole
body CT imaging could come through in the future to provide thorough, detailed and individualized image data whenever needed for healthcare.
As indicated earlier, this outlook paper is meant to be
suggestive and not exhaustive. As a result, the citations are
not comprehensive relative to the huge related literature
base. Nevertheless, to the best of our knowledge it reflects
the current trends in CT and, while our insights are unavoidably biased, the reality should not be too far from the targets
we hope for and believe in. It is our intension to keep refining our predictions in this framework as time goes by and to
update this outlook in a few years. Thus, we highly welcome
comments and critiques from colleagues and peers.
ACKNOWLEDGMENTS
This work is partially supported by the NIH Grant Nos.
EB002667, EB004287, EB007288, and EB006837. The authors also express their gratitude to Dr. Michael Cable, Dr.
Paul Fitzgerald, Dr. Ming Jiang, Dr. Xuanqin Mou, Dr. Bob
Senzig, Dr. Xiangyang Tang, Dr. Xizeng Wu, Dr. Wenbing
Yun, and Dr. Jun Zhao for their valuable comments and suggestions.
a

c
1
A. M. Cormack, Representation of a function by its line integrals, with
some radiological applications, J. Appl. Phys. 349, 27222727 1963.
2
G. N. Hounsfield, Computerized transverse axial scanning tomography: Part I. Description of system, Br. J. Radiol. 46, 10161022 1973.
3
R. A. Robb et al., High-sped three-dimensional x-ray computed tomography: The dynamic spatial reconstructor, Proceedings of the IEEE,
713, 308319 1983.
4
E. L. Ritman, R. A. Robb, and L. D. Harris, Imaging Physiological Functions: Experience with the DSR, Praeger, Philadelphia, 1985.
5
W. A. Kalender, Thin-section three-dimensional spiral CT: Is isotropic
imaging possible?, Radiology 1973, 578580 1995.
6
G. Wang, T. H. Lin, P. C. Cheng, and D. M. Shinozaki, A general cone
beam reconstruction algorithm, IEEE Trans. Med. Imaging 123, 486
496 1993.
7
G. Wang et al., Scanning cone-beam reconstruction algorithms for x-ray
microtomography, Proceedings of SPIE, Vol. 1556, pp. 99112 1991.
8
K. Taguchi and H. Aradate, Algorithm for image reconstruction in multislice helical CT, Med. Phys. 254, 550561 1998.
9
M. Kachelriess, S. Schaller, and W. A. Kalender, Advanced single-slice
rebinning in cone-beam spiral CT, Med. Phys. 274, 754772 2000.
10
S. Mori et al., Physical performance evaluation of a 256-slice CTscanner for four-dimensional imaging, Med. Phys. 316, 13481356
2004.
11
G. Wang, C. R. Crawford, and W. A. Kalender, Multirow detector and
b
1062
cone-beam spiral/helical CT, IEEE Trans. Med. Imaging 199, 817821
2000.
12
L. A. Feldkamp, L. C. Davis, and J. W. Kress, Practical cone-beam
algorithm, J. Opt. Soc. Am. A 1A, 612619 1984.
13
A. Katsevich, An improved exact filtered backprojection algorithm for
spiral computed tomography, Adv. Appl. Math. 324, 681697 2004.
14
A. Katsevich, Theoretically exact filtered backprojection-type inversion
algorithm for spiral CT, SIAM J. Appl. Math. 626, 20122026 2002.
15
G. Wang, Y. B. Ye, and H. Y. Yu, Appropriate and exact cone-beam
reconstruction with standard and nonstandard spiral scanning, Phys.
Med. Biol. 526, R1R13 2007.
16
Y. B. Ye, S. Y. Zhao, H. Y. Yu, and G. Wang, Exact reconstruction for
cone-beam scanning along nonstandard spirals and other curves, Proceedings of SPIE, Vol. 5535, pp. 293300, Denver, CO 2004.
17
Y. B. Ye and G. Wang, Filtered backprojection formula for exact image
reconstruction from cone-beam data along a general scanning curve,
Med. Phys. 321, 4248 2005.
18
Y. B. Ye, S. Y. Zhao, H. Y. Yu, and G. Wang, A general exact reconstruction for cone-beam CT via backprojection-filtration, IEEE Trans.
Med. Imaging 249, 11901198 2005.
19
J. Pack and F. Noo, Cone-beam reconstruction using 1D filtering along
the projection of M-lines, Inverse Probl. 214, 11051120 2005.
20
J. D Pack, F. Noo, and R. Clackdoyle, Cone-beam reconstruction using
the backprojection of locally filtered projections, IEEE Trans. Med. Imaging 241, 7085 2005.
21
T. L. Zhuang, S. Leng, B. E. Nett, and G. H. Chen, Fan-beam and
cone-beam image reconstruction via filtering the backprojection image of
differentiated projection data, Phys. Med. Biol. 4924, 54895503
2004.
22
Y. Zou, X. Pan, and E. Y. Sidky, Theory and algorithms for image
reconstruction on chords and within regions of interest, J. Opt. Soc. Am.
A 2211, 23722384 2005.
23
G. Wang and Y. B. Ye, Nonstandard spiral cone-beam scanning methods,
apparatus, and applications, US Provisional Patent Application 60/
588,682, Filing date: July 16, 2004 2004.
24
G. Wang and H. Y. Yu, Scanning method and system for cardiac imaging, Patent disclosure submitted to Virginia Tech. Intellectual Properties
on July 17 2007.
25
A. Katsevich and M. Kapralov, Theoretically exact FBP reconstruction
algorithms for two general classes of curves, 9th International Meeting
on Fully Three-Dimensional Image Reconstruction in Radiology and
Nuclear Medicine, pp. 8083, Lindau, Germany 2007.
26
A. Katsevich, Image reconstruction for a general circle-plus trajectory,
Inverse Probl. 235, 22232230 2007.
27
A. Katsevich, A general scheme for constructing inversion algorithms
for cone-beam CT, Int. J. Math. Math. Sci. 21, 13051321 2003.
28
Y. C. Wei, H. Y. Yu, J. Hsieh, W. X. Cong, and G. Wang, Scheme of
computed tomography, J. X-Ray Sci. Technol. 154, 235270 2007.
29
S. Y. Zhao, H. Y. Yu, and G. Wang, A unified framework for exact
cone-beam reconstruction formulas, Med. Phys. 326, 17121721
2005.
30
G. H. Chen, An alternative derivation of Katsevichs cone-beam reconstruction formula, Med. Phys. 3012, 32173226 2003.
31
H. Y. Yu and G. Wang, Cone-beam composite-circling scan and exact
image reconstruction for a quasi-short object, Int. J. Biomed. Imaging
2007, Article ID 87319, 10 pages 2007.
32
J. D. Pack, F. Noo, and H. Kudo, Investigation of saddle trajectories for
cardiac CT imaging in cone-beam geometry, Phys. Med. Biol. 4911,
23172336 2004.
33
X. Y. Tang, J. Hsieh, R. A. Nilsen, S. Dutta, D. Samsonov, and A. Hagiwara, A three-dimensional-weighted cone beam filtered backprojection
CB-FBP algorithm for image reconstruction in volumetric CT - helical
scanning, Phys. Med. Biol. 514, 855874 2006.
34
X. Y. Tang and J. Hsieh, A filtered backprojection algorithm for cone
beam reconstruction using rotational filtering under helical source trajectory, Med. Phys. 3111, 29492960 2004.
35
X. Y. Tang and J. Hsieh, Handling data redundancy in helical cone beam
reconstruction with a cone-angle-based window function and its
asymptotic approximation, Med. Phys. 346, 19891998 2007.
36
X. Y. Tang, J. Hsieh, R. A. Nilsen, and S. M. McOlash, Extending
three-dimensional weighted cone beam filtered backprojection CB-FBP
algorithm for image reconstruction in volumetric CT at low helical
pitches, Int. J. Biomed. Imaging 2006, Article ID 45942, 8 pages 2006.
1063
37
R. Gordon, R. Bender, and G. T. Herman, Algebraic reconstruction techniques ART for three-dimensional electron microscopy and x-ray photography, J. Theor. Biol. 29, 471482 1970.
38
A. H. Andersen and A. C. Kak, Simultaneous algebraic reconstruction
technique SART: a superior implementation of the ART algorithm,
Ultrason. Imaging 6, 8194 1984.
39
L. A. Shepp and B. F. Logan, The Fourier reconstruction of a head
section, IEEE Trans. Nucl. Sci. NS213, 2134 1974.
40
A. J. Rockmore and A. Macovski, A maximum likelihood approach to
emission image reconstruction from projections, IEEE Trans. Nucl. Sci.
234, 14281432 1976.
41
J. Nuyts et al., Iterative reconstruction for helical CT: A simulation
study, Phys. Med. Biol. 434, 729737 1998.
42
B. De Man et al., Reduction of metal streak artifacts in x-ray computed
tomography using a transmission maximum a posteriori algorithm,
IEEE Trans. Nucl. Sci. 473, 977981 2000.
43
I. A. Elbakri and J. A. Fessler, Statistical image reconstruction for polyenergetic x-ray computed tomography, IEEE Trans. Med. Imaging 212,
8999 2002.
44
J. A. Fessler, Statistical image reconstruction methods for transmission
tomography, in Medical Image Processing and Analysis, edited by M.
Sonka and J. M. Fitzpatric SPIE Press, Bellingham, Washington, 2000,
Vol. 3, pp. 170.
45
J. A. Fessler and A. O. Hero, Space-alternating generalized expectationmaximization algorithm, IEEE Trans. Signal Process. 4210, 2664
2677 1994.
46
I. A. Elbakri and J. A. Fessler, Segmentation-free statistical image reconstruction for polyenergetic x-ray computed tomography with experimental validation, Phys. Med. Biol. 4815, 24532477 2003.
47
P. E. Kinahan, A. M. Alessio, and J. A. Fessler, Dual energy CT attenuation correction methods for quantitative assessment of response to cancer therapy with PET/CT imaging, Technol. Cancer Res. Treat. 54,
319327 2006.
48
J. B. Thibault, A. K. Sauer, A. C. Bouman, and J. Hsieh, Threedimensional statistical modeling for image quality improvements in
multi-slice helical CT, 8th Meeting on Fully ThreeDimensional Image
Reconstruction in Radiology and Nuclear Medicine, pp. 271274, Salt
Lake City, UT 2005.
49
M. Iatrou, B. De Man, and S. Basu, A comparison between filtered
backprojection, post-smoothed weighted least squares, and penalized
weighted least squares for CT reconstruction, IEEE Nuclear Science
Symposium Conference Record, 2006, pp. 28452850, San Diego, CA
2006.
50
D. L. Parker, Optimal short scan convolution reconstruction for fanbeam
CT, Med. Phys. 92, 254257 1982.
51
Y. Zou and X. C. Pan, Exact image reconstruction on PI-lines from
minimum data in helical cone-beam CT, Phys. Med. Biol. 496, 941
959 2004.
52
M. Defrise, F. Noo, R. Clackdoyle, and H. Kudo, Truncated Hilbert
transform and image reconstruction from limited tomographic data, Inverse Probl. 223, 10371053 2006.
53
Y. B. Ye, H. Y. Yu, Y. C. Wei, and G. Wang, A general local reconstruction approach based on a truncated Hilbert transform, Int. J. Biomed.
Imaging 2007, Article ID 63634, 8 pages 2007.
54
Y. B. Ye, H. Y. Yu, and G. Wang, Exact interior reconstruction with
cone-beam CT, Int. J. Biomed. Imaging 2007, Article ID 10693, 5 pages
2007.
55
F. Natterer, The Mathematics of Computerized Tomography Society for
Industrial and Applied Mathematics, Philadelphia, 2001.
56
G. Wang, Y. B. Ye, and H. Y. Yu, General VOI/ROI reconstruction
methods and systems using a truncated Hilbert transform, Patent disclosure submitted to Virginia Tech. Intellectual Properties on May 15 2007.
57
H. Y. Yu and G. Wang, A general formula for fan-beam lambda tomography, Int. J. Biomed. Imaging 2006, Article ID 10427, 9 pages 2006.
58
H. Y. Yu, Y. C. Wei, Y. B. Ye, and G. Wang, Lambda tomography with
discontinuous scanning trajectories, Phys. Med. Biol. 5214, 43314344
2007.
59
H. Y. Yu and G. Wang, A general scheme for velocity tomography,
Signal Process. 885, 11651175 2008.
60
D. Saintfelix et al., In-vivo evaluation of a new system for 3D computerized angiography, Phys. Med. Biol. 393, 583595 1994.
61
L. T. Niklason et al., Digital tomosynthesis in breast imaging, Radiology 2052, 399406 1997.
1063
62
W. Ross, D. D. Cody, and J. D. Hazle, Design and performance characteristics of a digital flat-panel computed tomography system, Med. Phys.
336, 18881901 2006.
63
S. Greschus et al., Potential applications off lat-panel volumetric CT in
morphologic and functional small animal imaging, Neoplasia 78, 730
740 2005.
64
F. Kiessling et al., Volumetric computed tomography VCT: A new
technology for noninvasive high-resolution monitoring of tumor angiogenesis, Nat. Med. 1010, 11331138 2004.
65
J. M. Boone et al., Computed tomography for imaging the breast, J.
Mammary Gland Biol. Neoplasia 112, 103111 2006.
66
B. Chen and R. Ning, Cone-beam volume CT breast imaging: Feasibility
study, Med. Phys. 295, 755770 2002.
67
L. T. Holly and K. T. Foley, Image guidance in spine surgery, Orthop.
Clin. North Am. 383, 451461 2007.
68
N. Clinthorne, Intra-operative imaging, J. Laparoendosc Adv. Surg.
Tech. A 155, 549555 2005.
69
W. Berninger and R. Redington, Multiple purpose high speed tomographic x-ray scanner, US Patent No. 4,196,352A 1980.
70
T. G. Flohr et al., First performance evaluation of a dual-source CT
DSCT system, Eur. Radiol. 162, 256268 2006.
71
Y. Kyriakou and W. A. Kalender, Intensity distribution and impact of
scatter for dual-source CT, Phys. Med. Biol. 5223, 69696989 2007.
72
J. Zhao, M. Jiang, T. Zhuang, and G. Wang, An exact reconstruction
algorithm for triple-source helical cone-beam CT, J. X-Ray Sci. Technol.
143, 191206 2006.
73
J. Zhao, M. Jiang, T. G. Zhuang, and G. Wang, Minimum detection
window and inter-helix PI-line with triple-source helical cone-beam scanning, J. X-Ray Sci. Technol. 142, 95107 2006.
74
J. Zhao, Y. N. Jin, Y. Lu, and G. Wang, A reconstruction algorithm for
triple-source helical cone-beam CT via filtered backprojection, 9th International Meeting on Fully Three-Dimensional Image Reconstruction in
Radiology and Nuclear Medicine, pp. 205208, Lindau, Germany 2007.
75
S. S. Orlov, Theory of three dimensional reconstruction I: Conditions for
a complete set of projections, Sov. Phys. Crystallogr. 20, 312314
1975.
76
B. De Man, P. Edic, and S. Basu, An iterative algorithm for timeresolved reconstruction of a CT scan of a beating heart, 8th Meeting on
Fully ThreeDimensional Image Reconstruction in Radiation and
Nuclear Medicine, Salt Lake City, UT, 356359 2005.
77
D. Boyd et al., High-speed, multi-slice, x-ray computed tomography,
Proceedings of SPIE, Vol. 372, pp. 139150, Pacific Grove, CA 1982.
78
G. Wang et al., Top-level design and preliminary physical analysis for
the first electron-beam micro-CT scanner, J. X-Ray Sci. Technol. 124,
251260 2004.
79
B. De Man et al., Stationary computed tomography system and method,
US Patent No. 7,280,631 B22005.
80
J. S. Price, W. Block, and M. Vermilyea, Extended multi-spot computed
tomography x-ray source, US Patent No. 6,983,035 B22006.
81
Z. Yin, B. De Man, and J. Pack, Analytical cone-beam reconstruction
using a multi-source inverse geometry CT system, Proceedings of SPIE,
Vol. 6510, San Diego, CA 2007.
82
T. G. Schmidt, R. Fahrig, N. J. Pelc, and E. G. Solomon, An inversegeometry volumetric CT system with a large-area scanned source: A feasibility study, Med. Phys. 319, 26232627 2004.
83
B. De Man et al., Multi-source inverse geometry CT: A new system
concept for X-ray computed tomography, Proceedings of SPIE, Vol.
6510, San Diego, CA 2007.
84
G. Wang and M. W. Vannier, Bolus-chasing angiography with adaptive
real-time computed tomography, U.S. Patent No. 6,535,821, allowed on
11/26/2002, issued on 3/18/2003 2003.
85
E. W. Bai et al., Study of an adaptive bolus chasing CT angiography, J.
X-Ray Sci. Technol. 141, 2738 2006.
86
A. Katsevich, S. Basu, and J. Hsieh, Exact filtered backprojection reconstruction for dynamic pitch helical cone beam computed tomography,
Phys. Med. Biol. 4914, 30893103 2004.
87
E. Tkaczyk et al., Simulation of CT dose and contrast-to-noise as function of bowtie shape, Proceedings of SPIE, Vol. 5368, pp. 403410, San
Diego, CA 2004.
88
M. Kalra, N. Naz, S. Rizzo, and M. Blake, Computed tomography radiation dose optimization: Scanning protocols and clinical applications of
automatic exposure control, Curr. Probl Diagn. Radiol. 345, 171181
2005.
1064
89
B. De Man and B. Senzig, Method and system for radiographic imaging

with organ-based radiation profile prescription, US Patent No.
20070147579 A12007.
90
R. E. Alvarez and A. Macovski, Energy-selective reconstructions in
x-ray computerized tomography, Phys. Med. Biol. 215, 733744
1976.
91
J. H. Hubbell and S. M. Seltzer, Tables of X-ray mass attenuation coefficients 1 keV to 20 MeV, for elements Z1 / 4 1 to 92 and 48 additional
substances of dosimetric interest. National Institute of Standards and
Technology, Gaithersburg, MD, 1995.
92
B. De Man et al., An iterative maximum-likelihood polychromatic algorithm for CT, IEEE Trans. Med. Imaging 2010, 9991008 2001.
93
R. Carmi, G. Naveh, and A. Altman, Material separation with dual-layer
CT, IEEE Nuclear Science Symposium Conference Record, pp. 1876
1878, Wyndham El Conquistador Resort, Puerto Rico 2005.
94
J. E. Tkaczyk et al., Atomic number resolution for three spectral CT
imaging systems, Proceedings of SPIE, 6510, paper No. 651009, San
Diego, CA 2007.
95
S. Kuribayashi, Dual energy CT of peripheral arterial disease with
single-sorurce 64slice MDCT, presented at 9th Annual International
Symposium on Multidetector-Row CT, San Francisco, CA 2007.
96
X. Wu et al., Dual kVp computed tomography via view-to-view kV
switching: Temporal resolution and energy separation submitted.
97
P. M. Shikhaliev, Tilted angle CZT detector for photon counting/energy
weighting x-ray and CT imaging, Phys. Med. Biol. 5117, 42674287
2006.
98
nanoXCT, available from: http://www.xradia.com/Products/nanoxct.html.
99
H. Y. Yu, Y. B. Ye, and G. Wang, Local reconstruction using the truncated Hilbert transform via singular value decomposition, Phys. Med.
Biol. under review.
100
J. F. Barrett and N. Keat, Artifacts in CT: Recognition and avoidance,
Radiographics 246, 16791691 2004.
101
X. Q. Mou, S. J. Tang, and H. Y. Yu, A beam hardening correction
method based on HL consistency, Proc. of SPIE, Vol. 6318, San Diego,
CA 2006.
102
R. P. Zeng, J. A. Fessler, and J. M. Balter, Estimating 3-D respiratory
motion from orbiting views by tomographic image registration, IEEE
103
R. P. Zeng, J. A. Fessler, and J. M. Balter, Respiratory motion estimation
from slowly rotating X-ray projections: Theory and simulation, Med.
Phys. 324, 984991 2005.
104
Y. C. Wei, H. Y. Yu, and G. Wang, Integral invariants in fan-beam and
cone-beam computed tomography, IEEE Signal Process. Lett. 139,
549552 2006.
105
H. Y. Yu and G. Wang, Data consistency based general motion artifact
reduction in fan-beam CT, IEEE Trans. Med. Imaging 262, 249260
2007.
106
G. Wang, M. W. Vannier, and P. C. Cheng, Iterative X-ray cone-beam
tomography for metal artifact reduction and local region reconstruction,
Microsc. Microanal. 5, 5865 1999.
107
J. A. Fessler, Penalized weighted least-squares image-reconstruction for
positron emission tomography, IEEE Trans. Med. Imaging 132, 290
300 1994.
108
Bruno De Man, Iterative reconstruction for reduction of metal artifacts
in CT, Ph.D. Dissertation, University of Leuven, Belgium 2001.
109
J. A. OSullivan and J. Benac, Alternating minimization algorithms for
transmission tomography, IEEE Trans. Med. Imaging 263, 283297
1064
2007.
T. Beyer et al., A combined PET/CT scanner for clinical oncology, J.
Nucl. Med. 418, 13691379 2000.
111
D. A. Boas and A. M. Dale, Simulation study of magnetic resonance
imaging-guided cortically constrained diffuse optical tomography of human brain function, Appl. Opt. 4410, 19571968 2005.
112
K. Kagawa et al., Initial clinical assessment of CT-MRI image fusion
software in localization of the prostate for 3D conformal radiation
therapy, Int. J. Radiat. Oncol., Biol., Phys. 382, 319325 1997.
113
G. Wang, Y. Li, and M. Jiang, Uniqueness theorems in bioluminescence
tomography, Med. Phys. 318, 22892299 2004.
114
J. Nuyts et al., Simultaneous maximum a posteriori reconstruction of
attenuation and activity distributions from emission sinograms, IEEE
115
A. Momose and J. Fukuda, Phase-contrast radiographs of nonstained rat
cerebellar specimen, Med. Phys. 224, 375379 1995.
116
R. A. Lewis, Medical phase contrast X-ray imaging: Current status and
future prospects, Phys. Med. Biol. 4916, 35733583 2004.
117
A. Momose, Recent advances in x-ray phase imaging, Jpn. J. Appl.
Phys., Part 1 449A, 63556367 2005.
118
X. Z. Wu and H. Liu, Clarification of aspects in in-line phase-sensitive
X-ray imaging, Med. Phys. 342, 737743 2007.
119
X. Z. Wu, H. Liu, and A. M. Yan, x-ray phase-attenuation duality and
phase retrieval, Opt. Lett. 304, 379381 2005.
120
S. W. Wilkins et al., Phase-contrast imaging using polychromatic hard
X-rays, Nature London 384, 335338 1996.
121
F. Pfeiffer, T. Weitkamp, O. Bunk, and C. David, Phase retrieval and
differential phase-contrast imaging with low-brilliance x-ray sources,
Nat. Phys. 24, 258261 2006.
122
C. Kottler, C. David, F. Pfeiffer, and O. Bunk, A two-directional approach for grating based differential phase contrast imaging using hard x
rays, Opt. Express 153, 11751181 2007.
123
M. Jiang, C. L. Wyatt, and G. Wang, X-ray phase-contrast imaging with
three 2D gratings, Int. J. Biomed. Imaging 2007, Article ID 27152
2007.
124
D. W. Holdsworth, Micro-CT in small animal and specimen imaging,
Trends Biotechnol. 208, S34S39 2002.
125
M. J. Paulus, A review of high-resolution x-ray computed tomography
and other imaging modalities for small animal research, Lab Anim. 30,
3645 2001.
126
G. Wang, Micro-CT scanners for biomedical applications: An overview, Adv. Imaging 16, 1827 2001.
127
G. Wang et al., Design, analysis and simulation for development of the
first clinical micro-CT scanner, Acad. Radiol. 124, 511525 2005.
128
D. J. Brenner, C. D. Elliston, E. J. Hall, and W. E. Berdon, Estimated
risks of radiation-induced fatal cancer from pediatric CT, AJR, Am. J.
Roentgenol. 1762, 289296 2001.
129
A. Berrington de Gonzalez and S. Darby, Risk of cancer from diagnostic
x rays: Estimates for the UK and 14 other countries, Lancet 363, 345
351 2004.
130
O. W. Linto and F. A. Mattler, Jr. , National conference on dose reduction in computed tomography, emphasis on pediatrics, AJR, Am. J.
Roentgenol. 181, 321329 2003.
131
G. Wang, S. Y. Zhao, and D. Heuscher, A knowledge-based cone-beam
x-ray CT algorithm for dynamic volumetric cardiac imaging, Med. Phys.
298, 18071822 2002.
110
Assessing the future of diffuse optical imaging technologies for breast

cancer management
Bruce J. Tromberg
Beckman Laser Institute and Medical Clinic, University of California Irvine, Irvine, California 92612
and NCI Network for Translational Research in Optical Imaging, Beckman Laser Institute,
Irvine, California 92612
Brian W. Poguea and Keith D. Paulsen

Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire 03755
Arjun G. Yodh
Department of Physics & Astronomy, University of Pennsylvania, Philadelphia, Pennsylvania 191046396
David A. Boas
NMR Center, Department of Radiology, Massachusetts General Hospital, Harvard Medical School,
Boston, Massachusetts 02129 and NCI Network for Translational Research in Optical Imaging,
Beckman Laser Institute, Irvine, California 92612
Albert E. Cerussi
Beckman Laser Institute and Medical Clinic, University of California Irvine, Irvine, California 92612
Received 25 January 2008; revised 2 April 2008; accepted for publication 3 April 2008;
published 22 May 2008
Diffuse optical imaging DOI is a noninvasive optical technique that employs near-infrared NIR
light to quantitatively characterize the optical properties of thick tissues. Although NIR methods
were first applied to breast transillumination also called diaphanography nearly 80 years ago,
quantitative DOI methods employing time- or frequency-domain photon migration technologies
have only recently been used for breast imaging i.e., since the mid-1990s. In this review, the state
of the art in DOI for breast cancer is outlined and a multi-institutional Network for Translational
Research in Optical Imaging NTROI is described, which has been formed by the National Cancer
Institute to advance diffuse optical spectroscopy and imaging DOSI for the purpose of improving
breast cancer detection and clinical management. DOSI employs broadband technology both in
near-infrared spectral and temporal signal domains in order to separate absorption from scattering
and quantify uptake of multiple molecular probes based on absorption or fluorescence contrast.
Additional dimensionality in the data is provided by integrating and co-registering the functional
information of DOSI with x-ray mammography and magnetic resonance imaging MRI, which
provide structural information or vascular flow information, respectively. Factors affecting DOSI
performance, such as intrinsic and extrinsic contrast mechanisms, quantitation of biochemical components, image formation/visualization, and multimodality co-registration are under investigation in
the ongoing research NTROI sites. One of the goals is to develop standardized DOSI platforms that
can be used as stand-alone devices or in conjunction with MRI, mammography, or ultrasound. This
broad-based, multidisciplinary effort is expected to provide new insight regarding the origins of
breast disease and practical approaches for addressing several key challenges in breast cancer,
including: Detecting disease in mammographically dense tissue, distinguishing between malignant
and benign lesions, and understanding the impact of neoadjuvant chemotherapies. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.2919078
I. INTRODUCTION
The field of optical imaging in medicine has seen continuous
growth for decades, and there is a wide array of technologies
and applications in development and clinical use. The range
of different optical hardware choices available leads to a
complicated process of convergence upon the optimal system
2443
Med. Phys. 35 6, June 2008
for each specific clinical need. For example, even within the
categories of light sources, detectors, and light delivery, there
are many dozens of options to choose from, and the choices
made can drastically alter the data that is possible to be obtained. Two central benefits from optical imaging are the
ability to gain molecular-level information and submicro-
0094-2405/2008/356/2443/9/$23.00
2443
2444
Tromberg et al.: Diffuse optical imaging technologies for breast cancer management
Role in Breast Imaging
Population
benefiting
Size of
Target cost
Clinical trial per scan
Screening
All
society
10,000s
Screening
high risk groups
Select
groups
1000s
Diagnosis following
abnormal mammogram
Integration into
clinical imaging systems
Monitoring
Therapy
Those
previously
screened
Cancer
patients
$100
500+
$1000
100s
$2,000+
FIG. 1. A schematic illustration of the issues involved in determining a new

technologies role in breast imaging is shown, with the main factors being the
type of application left column and which population this then impacts,
which will then determine the size of the clinical trial required to prove the
technology is effective. Then ultimately there is a target cost per scan which
will need to be considered for the technology to be economically feasibly in
the current healthcare market.
scopic structural information, about the tissue function. Fundamental studies indicate that the molecular and cellular
specificity are possible, in both endogenous tissue markers
and with exogenous contrast agents. However, rapid medical
translation can be slowed by the lack of convergence to a
single technology, and yet strategic industry and academic
collaborations can be used to help advance translation. The
National Cancer Institute NCI has funded specific Networks for Translational Research in Optical Imaging
NTROI through a competitive review process, whose mission has been to define technologies which have the highest
potential promise for impact in cancer, and find ways to test
and validate them in multicenter trials. The Network preparing this report was funded to advance optical imaging technologies that have high potential to translate into successful
multicenter trials in a setting related to breast cancer management. The focus of this paper is to outline these promising technological innovations and look towards what development time scales they should be expected on.
In any imaging technology development, there is a key set
of trade-offs which are dictated largely by the application
needs. In Fig. 1, an illustration of the key needs in breast
cancer detection, diagnosis, and management are listed on
the left, along with estimates of the subject population who
would benefit from the system, and an order of magnitude
estimate of the scope of a clinical trial which would be
needed to demonstrate the system utility. Additionally, the
application area of the system will ultimately have a cost
which makes the system financially viable for the application. For example, population wide screening programs
would have widespread societal impact, yet would require
clinical trials with potentially tens of thousands of people to
prove the system efficacy, and ultimately could only be cost
effective if the cost per person was as low as mammography,
Medical Physics, Vol. 35, No. 6, June 2008
2444
in the range of 100s of dollars per exam. In contrast, if an

optical technology is used as part of a magnetic resonance
imaging MRI scanner, then the population imaged would
be considerably smaller, but the efficacy could be demonstrated with only hundreds of patients, and ultimately the
cost per scan could be similar to MRI, which is in the range
of a few thousand dollars per exam. While Fig. 1 is just a
rough illustration, it is a useful guide to think about how the
area of application helps determine the clinical trial needed
to prove utility, and ultimately the cost effectiveness targets
for the commercial scale system.
In the following sections the key opportunities for having
an impact upon breast cancer management are outlined. The
structure of the following sections is that each of the most
compelling applications are outlined, together with a discussion of the most useful technologies using Spectroscopy, Imaging and Tomography. In some cases, single institutional
trials are sufficient to study the problem and advance the
technology, and in some cases multicenter trials are required
to achieve a meaningful impact. This is in the context of both
the trials ongoing presently, and those which will likely be
carried out in the next decade. One goal of this work is also
to help define areas where government and industry collaboration might be partnered, to advance the field of optical
imaging in order to have useful impact in breast cancer.
II. MONITORING OF NEOADJUVANT
CHEMOTHERAPY RESPONSE
One of the most compelling applications to be studied in
the past few years has been the ability to track neoadjuvant
chemotherapy response in breast cancer tumors. The ability
to take measurements of light transmission through bulk tissue and recover absorption and scattering values, at multiple
wavelengths, has provided the critical ability to quantify tissue chromophore concentrations in vivo. Convergence on
one technology in this area is critical, since the numbers of
subjects who would be imaged is relatively small and so
unifying the technical implementation is critical to achieve
meaningful summary data which can demonstrate efficacy.
The cost of the instrument, the system accuracy, and impact
upon patient management are all key factors, and choosing
an application where no good tools exist today is also critical. As can be seen in Fig. 1, the population benefiting from
this system would be those with locally advanced breast cancer, and because the need is acute, the ability to complete a
clinical trial demonstrating efficacy is a little easier than in
screening applications. Interestingly also, since the application is not widespread but targeted to those who need the
system to assess their treatment efficacy, the cost per scan
ultimately could be substantially higher, as treatments can
typically be in the thousands to tens of thousands if they are
truly effective. The NCI sponsored NTROI has been advancing the development of Diffuse Optical Spectroscopy to be
used to track response to neoadjuvant chemotherapy in locally advanced breast cancer cases.
Neoadjuvant chemotherapy treatment is in widespread
use, and is growing in its use in treatment of other types of
2445
2445
FIG. 2. The Laser Breast Scanner LBS system developed at the University of California Irvine, Beckman Laser Institute, which has been produced for
multicenter trials of monitoring neoadjuvant chemotherapy response in breast cancer. The system electronics and console are shown in A, and the tissue
probe with light source fibers and light detectors in B, and the procedure to take measurements across the tumor region right breast and the controlateral
breast left breast are shown, with the probe being moved in 1 cm increments, to allow measurement of the on-tumor and off-tumor values of the tissue optical
index TOI.
cancer beyond breast. One rationale for this therapy is to

allow tumor reduction prior to lumpectomy surgery of the
tumor mass, thereby allowing smaller surgeries with better
normal tissue sparing. The other major rationale is to reduce
the probability of distant metastases prior to mastectomy or
lumpectomy. However, about 30% of women with locally
advanced breast cancer will not respond at all to neoadjuvant
therapy,1,2 so it is critical to determine which subjects are
responding and which are not. While most of the decisions in
this treatment scheme are done at the end of the chemotherapy that was prescribed, it is possible that earlier detection of nonresponse would be beneficial so that alternative
therapies could be attempted earlier in the disease management process. The concept of neoadjuvant chemotherapy and
monitoring the response has also been proposed as the optimal way to test if the patient will respond to chemotherapy, and this testing is ideally done prior to surgery. The
optical imaging system being proposed for multicenter studies should provide both molecular absorber information and
ultrastructural information from scattering. Early pilot studies have indicated that changes in tumor absorption can be
observed within the first days after treatment,37 and that the
specificity of these changes can be as high as 95%100%.
The clinical goal in using the system would be to accurately
screen treated patients and determine those who were not

responding, such that they could be shifted to another more
promising therapeutic approach.
A multicenter trial is under way with the system constructed at the University of California Irvines Beckman Laser Institute, called the Laser Breast Scanner LBS Fig.
2.8,9 The LBS is a bedside-capable system that combines
frequency-domain measurements with steady-state broadband tissue spectroscopy to measure complete broadband
quantitative near-infrared NIR absorption and reduced scattering spectra of breast tissue in vivo. This technology solution is ideal for the task of scanning women for determining
response to neoadjuvant chemotherapy for several reasons.
The key factors involved in this decision are:
Low patient numbers per medical center, requiring multicenter trial collaboration
Requirement of low cost for initial multicenter clinical
trials.
Requirement of established system hardware and standard operating procedures for multicenter trials.
Most tumors undergoing neoadjuvant therapy are large,
and often near the breast surface, making the surface
scanning approach a good option.
2446
Mulitiple track measurements over the tumor will avoid

the confusion associated with interpreting imaging data
in multicenter setting.
The LBS instrumentation and theory have been described in
several previous publications.10 The approach is highly stabilized and rich in information because of using multiple
fiber-coupled laser diodes that are intensity modulated, with
solid state detection. The diffusely transmitted spectra measured from the tissue are fit to a model of diffuse light transport, and are used to recover average tissue concentrations of
oxygenated hemoglobin ctO2Hb, deoxygenated hemoglobin ctHHb, total hemoglobin concentration ctTHb
= ctO2Hb+ ctHHb tissue hemoglobin oxygenation saturation
stO2 = ctO2Hb/ ctTHb 100% water ctH2O, and lipid
concentration. A contrast function is used to combine DOS
measurements to locate the location of maximum lesion optical contrast: Tissue Optical Index, TOI= ctHHb ctH2O/%
lipid. The parameters of this contrast function were determined from an evaluation of diffuse optical spectroscopy
DOS measurements in a larger population of 58 malignant
breast lesions. Tissue scattering is reported by the results of a
power law fit of the form transport scattering coefficient
ASP, where is the optical wavelength. The probe is
moved along a linear position on the surface of the tissue
Fig. 2C, and future versions of the system use will focus
on 2D scanning in both x and y directions along the breast
surface to map the tumor from above. The measurements
must be made by placing the probe on the breast with light
pressure to ensure contact. For comparison, a surface 2D
scan is also performed at an identical location on the contralateral breast.
The goals for the multicenter studies ongoing with this
system are:
To assess therapeutic efficacy early in treatment with
the LBS to allow future studies of tailoring treatment
regimens for maximizing response, minimizing toxicity,
and increasing overall survival;
To provide quantitative, noninvasive monitoring of tumor physiology and treatment response, correlated with
histopathology and MRI, will provide new insights regarding mechanisms of chemotherapy;
To examine how dissemination of the LBS technology
into general practice works, and increase the availability
of potentially valuable technologies to a broader
population.
III. SCREENING FOR SUBPOPULATIONS OF
WOMEN
Screening women for cancer has become one of the largest success stories in preventative medicine, with annual
mammography being the standard of care for women over
the age of 50 in most of the developed world. But still there
are populations where this screening program does not work
well, and this is in women with:
Highly radio-dense breast tissue.
Known higher risk due to genetic factors.
2446
Complex breast tissue due to previous surgery.

Younger women who are within the age range where
annual screening is recommended.
In each of these groups, there are research studies ongoing to
determine the best way to manage or indicate regular screening for cancer, yet the ideal situation of nonionizing radiation
or more sensitive tools is attractive for these applications.
The problems with screening in these groups largely fall into
the logistics of:
Contrast to noise problems with detecting abnormalities
through dense or complex tissues
Difficulty in imaging or scanning the entire breast volume
Difficulty in justifying the cost of screening programs,
prior to knowing the sensitivity and specificity of the
system proposed.
This latter issue is really a key factor, since in many cases the
first two problems may be fixed for given systems. The major factor is that once a system is chosen, there needs to be a
large clinical trial done to assess the sensitivity and specificity of detection of tumors in each subpopulation as indicated
in Fig. 1, and these values can then help determine if the
system is going to have clinical utility. There are major biological and clinical factors that influence the imaging, and so
pilot trials have already been completed to assess the factors
associated with age, body mass index, breast density, menstrual cycle effects, menopausal status, and hormone replacement therapy status.9,1113 Now that these are reasonably well
understood, it is possible to design screening studies which
will have matched controls to assess the system performance
in the relevant subpopulation. As indicated in Fig. 1, screening has maximal benefit, but because the application is applied widespread the cost to society is enormous, and drives
down the target cost per scan to very low values. The optimal
societal benefit in screening is likely in those not well served
by existing mammography programs, such as young women
and radio-dense women. As the numbers of subjects in these
groups is still very large, the medical impact of a successful
technology would be enormous, albeit potentially costly.
For the problem of screening these specific subpopulations of women, there are two technological approaches
which will work. The LBS mentioned in the previous section
is a logical choice to screen women with a lower cost solution. This system would require scanning the entire breast
surface and is maximally sensitive to the outer few centimeters of tissue under the probe. Alternatively, a breast spectral
tomography system could be used for transmission tomography imaging of the breast. A couple of such systems have
been developed to an advanced clinical trial stage Fig. 3A
with approval for sale in several countries, and several laboratory research systems have been developed and used in
large clinical trials Figs. 3B and 3C. The advantage of
these systems is that the tumor characteristics can be imaged
while the patient is lying in a prone position, and there is
reasonably good coverage of most of the breast. There are
still problems with imaging very small or very large breasts,
as there are in mammography, yet for a majority of women,
2447
2447
FIG. 3. Tomographic imaging systems developed for breast tumor characterization and imaging during therapy have been developed and large clinical trial
results have been published. The commercial system from ART Inc. see Ref. 15 is shown in a photograph A, and research at Dartmouth resulted in the
system shown in B Refs. 14, 28, and 29, and a schematic of the system at the University of Pennsylvania is shown in C Refs. 6 and 30. Each system
uses multiple sets of measurements at multiple wavelengths of NIR light, transmitted through the breast to reconstruct images of hemoglobin, oxygen
saturation, water, and scattering values.
this imaging can be done readily with a visit taking less than
30 min. The utility of these systems in a truly widespread
screening application is still unproven, but clinical trials are
ongoing to assess sensitivity and specificity in subgroups.
IV. OPTICAL AS AN ADJUNCT IMAGING
TOOL
Optical imaging as an adjunct device to mammography or
ultrasound imaging is a natural choice, because the information gained with the optical signal is distinctly different from
that of the clinical imaging information. Near-infrared scanning of suspected abnormalities could provide the added information required to better determine the malignancy status,
and either help determine whether biopsy is needed, or ultimately track suspected abnormalities over time to detect malignant types of changes. There have been several commercial ventures aimed at this particular technology space in the
health care market, and two devices have gained approvals in
some countries and in an evaluation phase in the U.S., including ComfortScan ART Inc., Montreal Canada and
Computed Tomography Laser Mammography system
CTLM Imaging Diagnostic Systems, Inc., Plantation FL.
Research in this area has also been carried out by a European
consortium and by Philips Medical Systems. The key to success in this area has been to find an optimal way to maximize
the clinical use of information of the optical image alongside
2448
2448
(A)
(B)
(C)
FIG. 4. Hybrid Modality systems have been developed and used in pilot clinical trials to characterize tumors and test the synergy in having mutual information
and the ability to combine information data sets to create a new class of hybrid images. The University of Pennsylvania system was the first such device A
Refs. 26 and 27, and similar systems were built into a 3Tesla MRI at Dartmouth B Refs. 18 and 20, and into an x-ray tomosynthesis system at the
Massachusetts General Hospital C Ref. 31.
standard mammography. This is more of a clinical workflow

question and significant focus has been on making a workstation model which could be truly integrated into a radiology department. So interestingly in this application, the actual success may depend more upon the business model, and
the ability to integrate the information into a standard Radiology workflow, rather than the system, yet system performance must also be at a high level.
Academic research studies have examined the combination of optical tomography as an adjunct to mammography,
and publications by Poplack et al., report on the first blinded,
age-matched control trial using NIR spectral tomography as
additional information to mammography, and the data indicate that for tumors with sufficiently high hemoglobin signatures, that there is definitely added information from the NIR
images.14 The analysis of data from the ART system indicates that when the mammography and NIR images are used
side by side with co-localized analysis, that the ROC values
can be in the range which should be highly useful

clinically.15 The clinical trials to assess these systems in
larger cohorts are still ongoing, and useful conclusions
should be reached within a few years about the feasibility of
using NIR as an adjunct to mammography.
V. OPTICAL SPECTROSCOPY INTEGRATED INTO
CLINICAL IMAGING TOOLS
One key way for optical imaging to readily be integrated
into medicine is as an add-on to larger and currently accepted clinical imaging systems, such as mammography, ultrasound US, x-ray computed tomography or tomosynthesis, and Magnetic Resonance Imaging MRI. Research in
system integration has taken place in all of these areas, with
a primary focus on breast imaging, to follow up on some
earlier screening. Illustrative systems are shown in Fig. 4,
with the first MR coupled system shown A, and a Philips/
2449
NIH sponsored 3T MR-coupled system being tested currently B, and the NIR system built into an x-ray breast
tomosynthesis system C. While the primary screening modality for the breast is x-ray mammography, there has been
comparatively little study of how to introduce optics into this
modality, indicating that integration with the other modalities
would lead to applications which are not in the screening
area. Thus the potential for adding optics into a secondary
imaging modality then lies in the area of providing information which can differentiate the diagnosis of the lesion, or
perhaps track therapy response.
If the key role for optics integrated into clinical systems is
to provide functional information for a targeted region within
the breast, then the system integration needs to be carried out
to maximally utilize the information from the two systems in
a complementary or ideally synergistic manner. A key
method of integration is then to maximize the spectral content of the optical hardware and utilize the spatial prior relations from the clinical imaging system.1620 This approach to
integrating spectral constraints and spatial prior information
as a constraint has been demonstrated in tumor imaging,20
and can now be implemented with optimized software algorithms becoming available through the network. As illustrated in Fig. 1, DOS imaging within other imaging devices
would have a cost that was likely matched to the system it
was coupled to, such that DOS coupled to ultrasound would
likely have to be quite inexpensive to make an impact,
whereas DOS coupled MR could be considerably more expensive since the latter cost is limited by the MR mainly. The
economic feasibility of these hybrid systems is a little uncertain at present, however performance will likely have a major
influence on the cost which is appropriate.
Optical tomography has been integrated into a breast
x-ray tomosynthesis system Fig. 4C at the Massachusetts
General Hospital, and is undergoing clinical trials.21,22 Because of the widespread use of x-ray imaging of the breast,
this combination makes fundamental sense, and may provide
additional information where mammography has limited efficacy, such as in the dense breast or for distributed or complex lesions.
Additionally, soft tissue imaging with US is the natural
follow on procedure for certain diagnoses, and integration of
optical spectroscopy into this has been productive and led the
way towards region-based estimation of tissue
properties.2325 This approach makes sense for US, as the
region to be characterized is usually known when US is being used, and the goal is then to provide as much information
about the tumor as is feasible. The ability to image hemoglobin, oxygen saturation, and scatter has all been shown, with
strong potential for added diagnostic information.
The integration of diffuse tomography into MRI was a
key development which was initially pioneered at the University of Pennsylvania26,27 and this work is being carried on
to explore ways to synergize the mutual information through
the use of MR-derived spatial constraints in the diffuse optical imaging DOI spectroscopy,18,20 and the exponential
growth of breast MRI in recent years makes these developments of heightened importance Figs. 4A and 4B. With
2449
MR, there is increased ability to constrain the optical property recovery both spatially as well as in chromophore concentrations, as the MR can yield water and fat maps which
can be used to constrain the spectral recovery of hemoglobin,
oxygen saturation, scatter, and exogenous agents.
The integration of NIR into clinical imaging is a natural
step in the progression of integration of clinical information
into a streamlined series of information sources. One illustration of this with widespread acceptance is the newly used
hybrid PET/CT scanners, where the PET information has
gained much wider appreciation through the ability to superimpose the scan on the high resolution CT image, and further
to allow enhancement of the image quality through postprocessing based upon the CT. Similar to this, overlay of NIR
spectral data on the MR scans and enhancement of the NIR
spectral data quantitatively should allow easier clinical acceptance of this new information stream.
VI. CONCLUSIONS
The four applications listed here are key areas where the
healthcare need overlaps very well with the emerging technology which is available in the next decade. Each area requires refinement of the NIR technology or completion of
multicenter clinical trials to verify that the technology will
truly be a useful solution to the problem. NIR spectral tomography of tissue has had a rich history of development in
NIH funded work, and some of the designs discussed in this
overview are a result of decades of development. Advanced
stage clinical designs that will be advanced towards multicenter trials are desperately needed to carry out these validation trials. Coordination of these trials with the established
medical clinical trials bodies such as the National Cancer
Institutes Imaging Clinical Trials program or through the
American College of Radiology Imaging Network ACRIN
would be the ideal way to ensure that these technologies are
advanced appropriately with maximal overlap with other information sources.
In some cases the business model involved in the system
development can lead to success or failure of the technology,
however designs which have been vetted through both peerreviewed development as well as business plan evaluation
would be ideal to ensure that the right technology is advanced with the greatest promise for success. Ensuring that
academic-industry partnerships have a pathway for this cooperation is critical, in a manner which is not too encumbered by limited funding issues, but also provides room for
competition and intellectual property preservation. This is a
complex landscape in which to develop advanced technologies, but the National Cancer Institute has made the first
solid step in this process through sponsorship of the Network
for Translational Research in Optical Imaging. Development
of optical contrast agents which will provide true molecular
or compartmental specificity for cancer still remain to be
developed to the point of FDA approval and commercial
distribution, however large amounts of research are still under way in this area. It is likely that these exogenous contrast
agents will be used in breast cancer imaging or therapy, but
2450
the timeline for FDA approval for human use is still quite
uncertain at this time. The technologies outlined here for
intrinsic contrast imaging are capable and ready to be used in
molecular contrast imaging as well, so it is likely that after
the first round of multicenter trials begin in endogenous imaging, that this may facilitate the means for studying contrast
agents in a multicenter setting as well.
This paper is a summary report from one of these sponsored networks for translational research in optical imaging,
where the outcome of the network has been exactly focused
on providing a convergence in the space of advanced NIR
technology, breast cancer health care need, and business development capacity. The clinical trials proposed here will be
carried out over the next decade or so, and will likely have a
direct impact on both the technologies, businesses, and
medical practice involved in breast cancer detection and
management.
ACKNOWLEDGMENT
This work has been funded by the Network for Translational Research in Optical Imaging, within the National Cancer Institute U54CA105480.
a

[email protected]
H. M. Kuerer, K. K. Hunt, L. A. Newman, M. I. Ross, F. C. Ames, and S.
E. Singletary, Neoadjuvant chemotherapy in women with invasive breast
carcinoma: Conceptual basis and fundamental surgical issues, J. Am.
Coll. Surg. 1903, 350363 2000.
2
L. Esserman, Neoadjuvant chemotherapy for primary breast cancer: Lessons learned and opportunities to optimize therapy, Ann. Surg. Oncol.
111 Suppl., 3S8S 2004.
3
A. Cerussi, D. Hsiang, N. Shah, R. Mehta, A. Durkin, J. Butler, and B. J.
Tromberg, Predicting response to breast cancer neoadjuvant chemotherapy using diffuse optical spectroscopy, Proc. Natl. Acad. Sci. U.S.A.
10410, 40144019 2007.
4
B. J. Tromberg, A. Cerussi, N. Shah, M. Compton, A. Durkin, D. Hsiang,
J. Butler, and R. Mehta, Imaging in breast cancer: Diffuse optics in
breast cancer: Detecting tumors in pre-menopausal women and monitoring neoadjuvant chemotherapy, Breast Cancer Res., 76, 279285
2005.
5
N. Shah, J. Gibbs, D. Wolverton, A. Cerussi, N. Hylton, and B. J. Tromberg, Combined diffuse optical spectroscopy and contrast-enhanced
magnetic resonance imaging for monitoring breast cancer neoadjuvant
chemotherapy: A case study, J. Biomed. Opt. 105, 051503 2005.
6
R. Choe, A. Corlu, K. Lee, T. Durduran, S. D. Konecky, M. GrosickaKoptyra, S. R. Arridge, B. J. Czerniecki, D. L. Fraker, A. DeMichele, B.
Chance, M. A. Rosen, and A. G. Yodh, Diffuse optical tomography of
breast cancer during neoadjuvant chemotherapy: A case study with comparison to MRI, Med. Phys. 324, 11281139 2005.
7
D. B. Jakubowski, A. E. Cerussi, F. Bevilacqua, N. Shah, D. Hsiang, J.
Butler, and B. J. Tromberg, Monitoring neoadjuvant chemotherapy in
breast cancer using quantitative diffuse optical spectroscopy: A case
study, J. Biomed. Opt. 91, 230238 2004.
8
N. Shah, A. Cerussi, C. Eker, J. Espinoza, J. Butler, J. Fishkin, R. Hornung, and B. Tromberg, Noninvasive functional optical spectroscopy of
human breast tissue, Proc. Natl. Acad. Sci. U.S.A. 988, 44204425
2001.
9
N. Shah, A. E. Cerussi, D. Jakubowski, D. Hsiang, J. Butler, and B. J.
Tromberg, The role of diffuse optical spectroscopy in the clinical management of breast cancer, Dis. Markers 192-3, 95105 2003.
10
F. Bevilacqua, A. J. Berger, A. E. Cerussi, D. Jakubowski, and B. J.
Tromberg, Broadband absorption spectroscopy in turbid media by combined frequency-domain and steady-state methods, Appl. Opt. 3934,
64986510 2000.
11
R. Cubeddu, C. DAndrea, A. Pifferi, P. Taroni, A. Torricelli, and G.
1
2450
Valentini, Effects of the menstrual cycle on the red and near-infrared

optical properties of the human breast, Photochem. Photobiol. 723,
383391 2000.
12
B. W. Pogue, S. Jiang, H. Dehghani, C. Kogel, S. Soho, S. Srinivasan, X.
Song, T. D. Tosteson, S. P. Poplack, and K. D. Paulsen, Characterization
of hemoglobin, water, and NIR scattering in breast tissue: Analysis of
intersubject variability and menstrual cycle changes, J. Biomed. Opt.
93, 541552 2004.
13
A. Cerussi, N. Shah, D. Hsiang, A. Durkin, J. Butler, and B. J. Tromberg,
In vivo absorption, scattering, and physiologic properties of 58 malignant breast tumors determined by broadband diffuse optical spectroscopy, J. Biomed. Opt. 114, 044005 2006.
14
S. P. Poplack, K. D. Paulsen, A. Hartov, P. M. Meaney, B. W. Pogue, T.
D. Tosteson, S. K. Soho, and W. A. Wells, Electromagnetic breast imaging - Pilot results in women with abnormal mammography, Radiology
2432, 350359 2007.
15
X. Intes, Time-domain optical mammography SoftScan: Initial results,
Acad. Radiol. 128, 934947 2005; see erratum also in Acad. Radiol.
1210, 1355 2005.
16
A. Li, G. Boverman, Y. Zhang, D. Brooks, E. L. Miller, M. E. Kilmer, Q.
Zhang, E. M. C. Hillman, and D. A. Boas, Optimal linear inverse solution with multiple priors in diffuse optical tomography, Appl. Opt.
4410, 19481956 2005.
17
B. Brooksby, S. Jiang, H. Dehghani, B. W. Pogue, K. D. Paulsen, J. B.
Weaver, C. Kogel, and S. P. Poplack, Combining near infrared tomography and magnetic resonance imaging to study in vivo breast tissue:
Implementation of a Laplacian-type regularization to incorporate MR
structure, J. Biomed. Opt. 105, 050504-110 2005.
18
B. Brooksby, B. W. Pogue, S. Jiang, H. Dehghani, S. Srinivasan, C.
Kogel, T. Tosteson, J. B. Weaver, S. P. Poplack, and K. D. Paulsen,
Imaging breast adipose and fibroglandular tissue molecular signatures
using hybrid MRI-guided near-infrared spectral tomography, Proc. Natl.
Acad. Sci. U.S.A. 10323, 88288833 2006.
19
P. K. Yalavarthy, B. W. Pogue, H. Dehghani, and K. D. Paulsen, Weightmatrix structured regularization provides optimal generalized leastsquares estimate in diffuse optical tomography, Med. Phys., 346,
20852098 2007.
20
C. M. Carpenter, B. W. Pogue, S. Jiang, H. Dehghani, X. Wang, K. D.
Paulsen, W. A. Wells, J. Forero, C. Kogel, J. B. Weaver, S. P. Poplack,
and P. A. Kaufman, Image-guided optical spectroscopy provides
molecular-specific information in vivo: MRI-guided spectroscopy of
breast cancer hemoglobin, water, and scatterer size, Opt. Lett. 328,
933935 2007.
21
G. Boverman, Q. Fang, S. A. Carp, E. L. Miller, D. H. Brooks, J. Selb, R.
H. Moore, D. B. Kopans, and D. A. Boas, Spatio-temporal imaging of
the hemoglobin in the compressed breast with diffuse optical tomography, Phys. Med. Biol. 5212, 36193641 2007.
22
Q. Zhang, T. J. Brukilacchio, A. Li, J. J. Stott, T. Chaves, E. Hillman, T.
Wu, M. Chorlton, E. Rafferty, R. H. Moore, D. B. Kopans, and D. A.
Boas, Coregistered tomographic x-ray and optical breast imaging: Initial
results, J. Biomed. Opt. 102, 024033 2005.
23
Q. Zhu, S. H. Kurtzma, P. Hegde, S. Tannenbaum, M. Kane, M. Huang,
N. G. Chen, B. Jagjivan, and K. Zarfos, Utilizing optical tomography
with ultrasound localization to image heterogeneous hemoglobin distribution in large breast cancers, Neoplasia 73, 263270 2005.
24
Q. Zhu, Optical tomography with ultrasound localization: Initial clinical
results and technical challenges, Technol. Cancer Res. Treat. 43, 235
244 2005.
25
Q. Zhu, M. Huang, N. Chen, K. Zarfos, B. Jagjivan, M. Kane, P. Hedge,
and S. H. Kurtzman, Ultrasound-guided optical tomographic imaging of
malignant and benign breast lesions: Initial clinical results of 19 cases,
Neoplasia 55, 379388 2003.
26
V. Ntziachristos, A. G. Yodh, M. Schnall, and B. Chance, Concurrent
MRI and diffuse optical tomography of breast after indocyanine green
enhancement, Proc. Natl. Acad. Sci. U.S.A. 976, 27672772 2000.
27
V. Ntziachristos, A. G. Yodh, M. D. Schnall, and B. Chance, MRIguided diffuse optical spectroscopy of malignant and benign breast lesions, Neoplasia 44, 347354 2002.
28
S. P. Poplack, K. D. Paulsen, A. Hartov, P. M. Meaney, B. W. Pogue, T.
D. Tosteson, M. R. Grove, S. K. Soho, and W. A. Wells, Electromagnetic
breast imaging: Average tissue property values in women with negative
clinical findings, Radiology 2312, 571580 2004.
29
B. W. Pogue, S. P. Poplack, T. O. McBride, W. A. Wells, K. S. Osterman,
2451
U. L. Osterberg, and K. D. Paulsen, Quantitative hemoglobin tomography with diffuse near-infrared spectroscopy: Pilot results in the breast,
Radiology 2181, 261266 2001.
30
J. P. Culver, R. Choe, M. J. Holboke, L. Zubkov, T. Durduran, A. Slemp,
V. Ntziachristos, B. Chance, and A. G. Yodh, Three-dimensional diffuse
optical tomography in the parallel plane transmission geometry: Evalua-
2451
tion of a hybrid frequency domain/continuous wave clinical system for

breast imaging, Med. Phys. 302, 235247 2003.
31
S. A. Carp, T. Kauffman, Q. Fang, E. Rafferty, R. Moore, D. Kopans, and
D. Boas, Compression-induced changes in the physiological state of the
breast as observed through frequency domain photon migration measurements, J. Biomed. Opt. 116, 064016 2006.
Vision 20/20: Increased image resolution versus reduced

radiation exposure
Erik L. Ritmana
Department of Physiology & Biomedical Engineering, Mayo Clinic College of Medicine,
200 First Street SW, Rochester, Minnesota 55905
Received 6 December 2007; revised 8 April 2008; accepted for publication 8 April 2008;
This is a review of methods, currently and potentially, available for significantly reducing x-ray
exposure in medical x-ray imaging. It is stimulated by the radiation exposure implications of the
growing use of helical scanning, multislice, x-ray computed tomography for screening, such as for
coronary artery atherosclerosis and cancer of the colon and lungs. Screening requires highthroughput imaging with high spatial and contrast resolution to meet the need for high sensitivity
and specificity of detection and classification of specific imaged features. To achieve this goal
beyond what is currently available with x-ray imaging methods requires increased x-ray exposure,
which increases the risk of tissue damage and ultimately cancer development. These consequences
limit the utility of current x-ray imaging in screening of at-risk subjects who have not yet developed
the clinical symptoms of disease. Current methods for reducing x-ray exposure in x-ray imaging,
mostly achieved by increasing sensitivity and specificity of the x-ray detection process, may still
have potential for an up-to-tenfold decrease. This could be sufficient for doubling the spatial
resolution of x-ray CT while maintaining the current x-ray exposure levels. However, a spatial
resolution four times what is currently available might be needed to adequately meet the needs for
screening. Consequently, for the proposed need to increase spatial resolution, an additional order of
magnitude of reduction of x-ray exposure would be needed just to keep the radiation exposure at
current levels. This is conceivably achievable if refraction, rather than the currently used attenuation, of x rays is used to generate the images. Existing methods that have potential for imaging the
consequences of refracted x ray in a clinical setting are 1 by imaging the edge enhancement that
occurs at the interfaces between adjacent tissues of different refractive indices, or 2 by imaging
the changes in interference patterns resulting from moving grids which alter the refraction of x rays,
that have passed through the body, in a predictable fashion, and 3 theoretically, by an image
generated from the change in time-of-flight of x-ray photons passing through the body. Imaging
phase shift or change in time-of-flight, rather than attenuation, of x-ray photons through tissues
presents formidable technological problems for whole-body 3D imaging. However, if achievable in
a routine clinical setting, these approaches have the potential for greatly expanding the use of x-ray
imaging for screening. This overview examines the increased contrast resolution and reduced radiation exposure that might be achievable by the above-mentioned methods. 2008 American
Association of Physicists in Medicine. DOI: 10.1118/1.2919112
Key words: screening, x-ray refractive index, x-ray attenuation
I. INTRODUCTION
Shortly after its introduction to clinical practice in 1973,
computed tomography was rapidly shown to greatly increase
the specificity and sensitivity of image information content.1
Since the mid 1990s fast, multislice, helical scanning multidetector, or multislice, computed tomography MDCT or
MSCT,2,3 which generates a 3D array of tiny cubic picture
elements voxels, has advanced to include screening applications. Examples include CT colonography,4 lung imaging
for early detection of cancer,5 and coronary artery
calcification.6
Because of the clinical appeal of these developments,
there is considerable pressure for further technological improvement to increase the spectrum of clinical applications,
especially in screening for early stages of disease. This, however, is likely to increase the radiation exposure involved,
2502
and this is raising some concern. The following discussion

explores both the probable image resolution requirements for
effective screening and its implications for associated radiation exposure, and how this may be minimized by modifications of attenuation-based x-ray scanning as well as by use of
non-attenuation-based x-ray imaging.
II. BACKGROUND
II.A. Increasing concern about x-ray exposure
In general, the duration during which the patient profits

from any treatment resulting from the image-based diagnosis should not be curtailed by the onset of undesirable deleterious consequences of the radiation exposure. Some estimate of that time interval in which deleterious consequences
develop can be deduced from previous studies of the tempo-
0094-2405/2008/356/2502/11/$23.00
2502
2503
Erik L. Ritman: Reduced x-ray exposure for screening
ral relationship between radiation exposure and consequences. For instance, Imaizumi et al.7 recently showed that
a significant linear radiation dose response for thyroid nodules, including malignant tumors and benign tumors, exists
in the survivors of atom bomb exposure.
With the introduction of 3D imaging CT scanners, radiation exposures are increasingly being expressed as volume
computed tomography dose index CTDIvol in mGy and
dose-length product DLP in mGycm. Although rarely experienced in clinical diagnostic or interventional practices,
after a 2 Gy single dose, skin reddening occurs.8 However,
because not all tissues are equally radiosensitive, gonads, for
instance, are more sensitive than bone marrow colon
lung stomach and these are more sensitive than bladder,
liver, and thyroid, which in turn are more sensitive than skin.
A reasonable approximation of the more biologically meaningful effective dose E can be obtained by the equation
1
E = K DLP,
where K = 0.017 mSv/ mGy cm for a thorax scan9 and DLP

is the dose length product the length typically being 30 cm
for a thorax.
A typical multislice heart/calcification CT scan involves
approximately 30 mSv and a CT coronary angiogram scan
up to 60 mSv.10 The National Radiological Protection Board
of the U.K. Ref. 11 showed that effective radiation doses in
patients 01-year-old were typically higher than corresponding values in adults. This is particularly worrisome in the
cases such as congenital heart diseases, which are often eminently treatable provided an accurate diagnosis is available
and x-ray imaging is often the method of choice. A complex
diagnostic study of congenital heart disease or an electrophysiological interventional study could involve as much as
a 600 mSv exposure.12 To put this in perspective, for a population in which each person is exposed to 100 mSv it is likely
that the chance of developing cancer increases from 20 in a
nonexposed population to 20.01%.13
The ability to discriminate between different tissues in a
CT image depends on the difference in attenuation coefficient of tissue in two adjacent voxels being greater than the
noise in the detected x-ray signal at best this is the quantum
noise. Thus, to distinguish a 1 mm diameter cubic voxel of
muscle tissue with attenuation coefficient = 0.212/ cm at 60
keV Ref. 14 from a 1 mm x diameter voxel of blood
= 0.222/ cm within a 20 cm diameter, water-equivalent,
torso = 0.210/ cm, the signal is given by the BeerLambert law, which relates the transmitted flux I to the
illuminating flux Io by
I = I oe x
2
0.1 0.21020
I thru blood voxel = Ioe0.222

I thru muscle voxel = Ioe
I = Ioe
0.21020 0.2220.1
0.2120.1 0.21020
20 0.2120.1
Ioe0.210
= 1.5 10 Io ,
5
where Io is the x-ray exposure and I is the difference in

transmitted x ray through a voxel of blood and of muscle.
2503
FIG. 1. K is the kinetic energy released in matter and is the area exposed.
Extrapolation of these data to 511 keV indicates that use of the positron
annihilation method would involve almost ten times higher KERMA than
that at 60 keV. With permission from Ref. 51.
As one standard deviation of the quantum noise= I0.5, then

the signal I has to be greater than 1.5 102Io0.5. This
condition is met if Io 6 7 109 photons/ exposure of a
1 mm2 resolution unit. According to Motz et al.,15 this
would involve an exposure of 0.3 Gy, more than an order of
magnitude higher than the exposure with a current CT scan.
In order to achieve higher spatial resolution at constant
contrast resolution, the x-ray exposure needs to increase by
at least the cube of the resolution, i.e., by decreasing voxel
size from 1 to 0.5 mm3, the exposure should increase at least
eightfold.16 Low-energy x-ray photons i.e., 50 keV provide more contrast because each photon is more likely to
interact with matter by virtue of the photoelectric effect.
However, at this photon energy fewer primary photons are
transmitted through the body, so that the signal-to-noise ratio
in the x-ray image is poorer for a given x-ray photon flux
exposure. On the other hand, when an x-ray photons energy
increases beyond 50 keV, it is less likely to interact with
matter but when it does interact the damage potential per
photon increases as illustrated in Fig. 1. This trade-off between exposure and x-ray photon energy results in an optimal x-ray photon energy, in terms of minimum number of
photons per unit x-ray interaction with tissue, at about 60
keV in clinically relevant conditions.
II.B. Proposed image resolution requirements for
medical screening
Because the radiation exposure required to achieve a desired image quality depends so strongly on the desired image
resolution, the pathophysiological rationale for the level of
image resolution likely to be needed for screening purposes
needs to be established.
Screening for a disease requires both low false negative
and false positive detections as well as high true positive and
high true negative detections, i.e., high specificity and sensitivity. For imaging this requires spatial, contrast, and/or temporal resolutions depending on the type of pathological pro-
2504
cess to be detected, the stage in the evolution of that

pathological process, as well as where in the body that process is occurring. As spatial, contrast, and temporal resolutions are generally interdependent, it is important to know
the upper limits of these resolutions that are necessary for a
desired sensitivity and specificity.
II.B.1. Spatial resolution
The basic functional unit BFU of organs seems like a
reasonable goal for the highest spatial resolution of clinical
whole-body CT imagers used for screening. BFUs are the
smallest accumulations of diverse cells that function like the
organ they are in, e.g., the hepatic lobule in the liver or
nephron in the kidney. Generally, they occupy about
0.01 mm3, i.e., about 200 m in diameter. The volume of a
BFU is essentially the upper limit of a spherical assembly of
cells, immersed in a suitable nutrient medium, which can
survive without its own blood supply. However, each BFU
has its own capillary blood supply to support the extra energy needed for the physiological function e.g., secretion or
filtration of the BFU. Similarly, early solid cancers could
grow to this size in perfused tissues before a blood supply is
needed to sustain further growth.17 However, once a blood
supply to the small aggregation of cancerous cells is formed,
then growth into larger tumors, as well as dissemination of
cancer cells throughout the body, is facilitated. Thus, an isotropic voxel resolution of 100 m seems like a reasonable
goal because it would allow us to detect, measure the number, size, and perhaps functional status especially as reflected in its perfusion, but this would generally require use
of a contrast agent of individual BFUs or of early cancers.
Unfortunately, due to x-ray photon statistics alone, this spatial resolution would require almost two decades more in
x-ray exposure than is delivered by current MSCT scanners.
Additional x-ray exposure may be needed depending on the
contrast differential between a BFU and its surrounding
stroma. Additional radiation may also be needed to overcome
the noise generated by x-ray beam hardening and x-ray
scatter contamination of the primary x-ray image data.
II.B.2. Contrast resolution
BFUs differ little in x-ray contrast from surrounding tissues because they are elementally not very different from the
connective tissue in which they are embedded. Similarly, discrimination of some macroscopic tissue components that are
of particular clinical interest, such as the cartilage in joints,
and tissues that are starting to be infiltrated by pathological
proteins e.g., amyloid in heart muscle or elements e.g.,
iron in the liver, also is difficult for attenuation-based x-ray
CT imaging. With current attenuation-based x-ray imaging,
exploitation of any attenuation-based contrast difference
would require either very high radiation exposures to provide
the signal-to-noise ratio needed for discriminating these tiny
BFUs from their surrounding tissue matrix, or would require
administering a contrast-enhancing agent generally involving
a high-Z element such as iodine that selectively accumulates
in or avoids BFUs.
2504
II.B.3. Volume that needs to be imaged

The volume that needs to be imaged is determined by the
need to image an entire organ. This is because the absence of
a detected abnormality in an organ will only then have meaning in the case of a cancer or some other disease process that
starts as focal pathology. Thus, the gut would require that the
entire abdomen be imaged and the lung requires that the
entire thorax be imagedboth being approximately 3040
cm in cephalocaudal height and in transverse diameter. This
is currently achieved with multislice helical scanning CT at
approximately 0.6 mm3 resolution. Note that spatial resolution cannot be better than twice the size of a voxel.
II.B.4. Temporal resolution
In addition to 3D anatomy, the rate, amplitude, synchrony,
and the spatial distribution of the movement attributes of
some organs and of the blood flow within organs are important indicators of the organs functions. Even if the dynamics is not of direct clinical interest, the minimizing of blurring of anatomic features due to motion is of direct interest.
Moreover, temporal and spatial resolutions are directly interrelated in that increasing one often involves reducing the
other if x-ray exposure is kept constant.18 There are two
types of organ motion; one can be predictively cyclic, such
as the normal heart beat or breathing, and the other is unpredictable and therefore, generally not reproducible, such as
the peristaltic movement of the bowel or the passage of a
bolus of injected contrast agent through the blood stream.
These physiological factors raise three issues: one is the need
to image fast enough to stop the motion blurring i.e., the 3D
imaging shutter time, which would need to be fast enough
to stop-action the fastest moving phase of the motion. In the
heart this requires that CT scan data be acquired within 100
ms even in the slow diastolic filling phase of the heart
cycle at heart rates less than 70 beats/min. Unless a priori
knowledge of the time at which the quiescent or most dynamic phase of the cyclic activity occurs, continuous observation is needed at a rate required to ensure adequate imaging of the most dynamic phase. This approach generally
results in retrospective selection of the desired images and
discarding the remaining imageswhich often represent the
majority of those images. This approach results in considerable x-ray exposure that ultimately does not contribute toward information extracted from the available x-ray image
data.
II.C. Enhancement of exposure-to-information content
of x-ray images
A goal is to have the modulation of the transmitted x-ray

signal be greater than the noise in that signal for unequivocal
detection of the signal.19 However, this needs to be achieved
without increasing radiation exposure. Contrast resolution20
depends on signal noise especially in x-ray quantum-limited
images and contamination by scattered x rays which add
noise21 and reduce the dynamic range of the electronic image
gray scale available for detecting small changes of contrast
2505
in the x-ray image. Another contributor to altered gray-scale

values is beam hardening due to a polychromatic bremsstrahlung x-ray beam becoming hardened i.e., less attenuating as it traverses tissue, especially bone, which is
highly attenuating.22
Radiation exposure has been, and is continuing to be, decreased by increasing the sensitivity of x-ray detection by
improved detector technology i.e., make every transmitted
x-ray photon count and by increasing specificity by selectively rejecting scatter without use of antiscatter grids and by
use of beam filtration to generate a narrow spectrum or ideally the use of monochromatic radiation. The following summarizes some present and ongoing developments.
However, as will be shown, these approaches are unlikely
to adequately reduce radiation exposure, and some of them
are likely to be technically impractical to implement in, or
with, clinical CT scanners used in a screening mode.
2505
II.C.2. Contrast media

The use of contrast media,32 primarily iodine-based, reduces the x-ray exposure needed to differentiate an opacified physiological space e.g., blood vessel lumen from its
surrounding tissues e.g., brain tissue. The impact of the
selective opacification may be limited in certain organs
when intravenous injections are used, as this also opacifies
the microcirculation in overlying tissues or cardiac chambers
and thereby reduces the desired signal. Zeman et al.33
showed that, for a tissue thickness equivalent to 20 cm of
water, the element providing optimal image contrast should
have a Z value ranging from 60 to 70 iodine Z = 53. However, use of contrast media in screening applications is generally not indicated because of logistic implications and because of occasional side effects such as hematoma at the
injection site, embolism downstream to the injection site,
aggravation of kidney disease or allergic reaction.34
II.C.3. Prospective gated image acquisition
II.C.1. Reduce x-ray scatter

The scatter component in an x-ray image has been shown
to double when the x-ray photon energy is reduced from 100
to 20 keV, and also to double if body diameter is increased
from 5 to 20 cm.23 These two opposing factors are traded off
by selecting higher x-ray voltages for greater body diameters. Partial removal of scattered x rays from the x-ray imager has been achieved by two methods. The most commonly used method involves increasing the distance between
the subject and the detector.24 Use of a large object-todetector distance can greatly reduce the scatter that is detected, but with finite focal spot dimensions this is at the cost
of reduced spatial resolution.25 Use of a mechanical Venetian blind-like collimator26 preferentially reduces the x-ray
scatter contamination of the image and thereby increases the
contrast resolution, but at the cost of loss of signal due to the
collimators vanes stopping some of the transmitted primary
x-ray photons.27 This latter aspect can be partially overcome
by using software to model the scatter and subtracting that
computed scatter signal from the detected signal28 instead of
using a mechanical collimator. However, this approach usually degrades the spatial resolution of the scatter-corrected
image.
Other methods that might possibly be used are technically
and/or economically more challenging. A method used in
nuclear medicine imaging involves detecting the x-ray photons with an energy-sensitive detector. If a monochromatic
x-ray source can be used, the contribution of the lower energy scattered photons can be eliminated from the x-ray
transmission image data. Another, potential, method borrows
from positron emission tomography PET technology29 by
using time-of-flight information to eliminate scattered photons by virtue of them generally being delayed in their arrival at the detector relative to the ballistic, nonscattered,
photons. This approach requires knowledge of the time of
launch of a very brief pulse of x-ray photons, such as it
might be possible to generate with field-emission x-ray
sources30 or with a pulsed laser/electron-beam interaction.31
Prospective gated image acquisition, used in the case of

the predictable cyclic motion such as occurs in the heart, can
also reduce radiation exposure. This method involves acquiring incremental sets of partial-scan data only during a fixed
phase of that cyclic motion progressively over many sequential heart cycles during one breath-hold. Total radiation
exposure/scan sequence can thereby be reduced fourfold
relative to retrospectively gated tomographic reconstruction,
in which the x-ray exposure is continuous during the entire
breath-hold duration of the scan.35
II.C.4. Hardware and software tricks
Hardware and software tricks can also be used to reduce patient exposure. The simplest method is to merely reduce the x-ray source current. This so-called low-dose CT
can reduce exposure to 1/5,36 but at the cost of reduced image quality. This approach is probably suitable for monitoring an abnormality known to be present, rather than for detecting new ones. Another method involves varying the
radiation exposure at different angles of view around the
subject so as to keep the transmitted x-ray signal constant at
the level needed for the desired signal-to-noise ratio.37
Harpen38 showed the use of a bow tie-type of beam shaping filter to better match the dynamic range in the transmitted
x-ray image gray scale to the image detector dynamic range.
This can decrease noise power in a CT image by up to 25%,
especially in larger diameter fields-of-view. Another approach is to restrict the x-ray exposure to only the organ of
interest e.g., the heart. This would reduce the amount of
tissue exposed but requires use of special limited data set
tomographic reconstruction algorithms to provide usable
images.39,40
II.C.5. Monochromatic x rays
Use of monochromatic x rays instead of polychromatic
bremsstrahlung x rays41 would limit the radiation exposure
to just the useful x-ray photons, and also make the image
2506
2506
contrast information more quantitative by eliminating beam

hardening. In favorable circumstances this may reduce x-ray
exposure. Simulations by Baldelli et al.42 indicated the reduction of radiation in the chest and abdomen to be by
10% 40%, but actually an increase in the head and lumbar
spine by up to 45%. This latter effect may well have been
due to a suboptimal match between the x-ray photon energy
and the average x-ray path length through the body.
III. MINIMIZING X-RAY EXPOSURE BY USE OF
NONATTENUATION ASPECTS OF X-RAY/
MATTER INTERACTION TO GENERATE IMAGES
Instead of the gray scale within an image corresponding
to the traditional attenuation coefficient of tissue within a
pixel or voxel, it can also be made to correspond to a parameter related to the refractive index of that piece of tissue,
which can be expressed as the difference from the refractive
index of water.
III.A. Complex refractive index
The complex refractive index n of x rays passing

through matter can be expressed43 as
n = 1 i ,
where the imaginary part represents the attenuation ,

and the real part represents refraction . The refractive index of x ray in matter is very close to unity. If is the
dielectric function of matter, then
n = = 1 1 /2,
where is the susceptibility, which is very small because

= e2 / m2, e is the electron charge, is electron density
in the material, m is electron mass, and is the frequency of
the x ray.
Refraction should not be confused with diffraction,44
which typically involves change of direction of the x-ray
photon by many degrees depending on the spacing between
atoms encountered by the x-ray photons, rather than very
small fractions of a degree as occurs with refraction.45 Nonetheless, the two are related.46 The refractive index of x ray is
directly related to the phase velocity phase of x ray,
whereas the group velocity group of radiation is the velocity
with which energy propagates. Group velocity relates to
phase velocity as47
group = c2/phase ,
cially at lower energy x-ray photon energy, are currently

used to convey this small difference in refractive index.
Figure 2 illustrates that the refractive index for clinically
relevant x rays through breast tissue theoretically provides
orders of magnitude greater contrast than does attenuation,
especially of low atomic weight elements.49 Moreover, the
amount of x-ray exposure needed to obtain a usable image
should also be greatly reduced50 relative to attenuation-based
imaging.15 However, because phase is a concept which assumes an ensemble of x-ray photons, its use inevitably involves a number of photons per image pixel. The reduction
in radiation exposure, therefore, may not be as great as might
be expected from the magnitudes of versus when change
in phase is used to make an image.
Although monochromatic x rays emitted from a very
small focal spot or from a large, but very distant x-ray source
such as can be achieved with a synchrotron can generate
the coherent x-ray needed for phase-interference-based imaging, conventional polychromatic x ray generated with relatively large focal spots can be used.51
Four general approaches to utilizing x-ray refraction are
as follows:
i
so that if phase = c1 + 1.5 106 at 60 keV, then group = c1

1.5 106.
This means that as an x-ray photon traverses 1 mm of a
vacuum in approximately 3.3 1012 s, i.e., 3.3 picoseconds, the delay to be expected after a 60 keV x-ray photon
traverses 1 mm of water is 5 1018 s, and at 511 keV this
difference is 2 1021 s! Because of this minuscule difference, phase interference, as may be manifested in Moir patterns or as development of optical caustics48 much like the
bright lines at the bottom of a wavy swimming pool, espeMedical Physics, Vol. 35, No. 6, June 2008
FIG. 2. Complex refractive index n of breast tissue, where is contribution by photon refraction and is contribution from attenuating factors.
Note that at 60 keV contributes 104 times the value of . Figure modified
with permission from Ref. 51.
ii
Refraction of the x rays, especially at sites of very

rapid change in refractive index such as occurs at the
edges of tissue fibers or at surfaces of small ducts and
blood vessel walls52 results in edge enhancement
e.g., much like an optical caustic at some distance
beyond the refraction site.45 This approach, called by
some diffraction enhanced imaging DEI, is probably
limited to imaging several centimeters of tissue depths
and tissues in which adjacent tissues have marked differences in refractive index properties, such as can be
the case in cartilage in joint spaces of fingers or toes,
and in fibro-vascular microstructures of breast
tissue.53
Differences in phase shift that occur due to passage
through materials of different refractive indices can be
2507
used to cause Moir-type patterns when mixed with

a reference beam such as in the Bonse-Hart interferometer principle.54 The Moir pattern can be mathematically unraveled to provide a map of the phase
shift distribution. This works well for a small thickness of tissue in which a relatively small number e.g.,
less than ten of phase shifts occur and hence can be
accounted for but, after passage through 30 cm of
tissue the amount of phase shift, combined with the
shorter photon wavelength needed to provide adequate penetration of the body, would make unraveling of the phase shift quite challenging.
A variant of this method involves use of Talbot grating
interferometers placed on the detector side, but adjacent
to, the person i.e., the phase grating and the other just
in front of the detector i.e., the absorption grating.55
The phase interference may be finer than the pixel resolution of an area detector, so that an additional absorption grating with a suitable pitch put in front of the detection plane would allow the detection of intensity
variations that correspond to the derivative of the wave
front phase taken along the direction perpendicular to
the grating lines. This technique can be combined with a
phase-stepping method, which involves several exposures which differ in phase such that any nonuniform
intensities and in-line phase contrast causing edgeenhancement are eliminated.
iii
A third, conceivable, approach is to cyclically modulate the illuminating x-ray flux with a cycle duration
i.e., wavelength very much longer e.g., 1010 than
an x-ray photon wavelength. The Moir pattern resulting from the refraction of these waves would be more
manageable in terms of fewer phase wraparounds.
The downside of this approach is that a sufficient
number of photons N would be needed, per modulation wavelength, to provide the needed sensitivity to
detect the small phase shift . To measure the
phase difference between the two trains of n x-ray
pulses for the purposes of illustration, each with a
Gaussian distribution of N photons, the number of
photons needed to detect a phase difference between the two identical trains of pulses is
= FWHM/2.355N n,
where FWHM is the duration of the pulse of x ray at the

half-maximum intensity of each pulse. Thus, for a
phase-shift through 1 mm water to be detected 5
1018 s, this can be achieved with a train of ten
pulses, each with a 1015 s FWHM x-ray pulse duration containing N = 1000 photons. This involves detection of 10,000 photons/detector pixelwhich is the
same as regular attenuation-based imaging. The only
advantage with this method would be that there would
be increased contrast between low-Z elements.
2507
III.B. Direct measurement of spatial distribution of

transit times
The fourth, and perhaps the ultimate, goal might be to

directly measure the spatial distribution of the transit times
of individual x-ray photons. As the velocity of x ray through
water differs little from that through air, then imaging of just
the presence or absence of tissue at 1 mm resolution would
require time discriminations in the attosecond ranges, and
that is not technologically practical at the clinical x-ray level
at present. Nonetheless, this approach seems to have potential for so dramatically expanding use of clinical x-ray imaging that it is worth exploring in some detail, especially as
technology continues to develop rapidly at the rate described
by Moores law, i.e., doubling in speed or capacity every
year and a half.56 State-of-the-art streak cameras have temporal resolution of 200300 femtoseconds,57 so development
of a technique to decrease this a thousand-fold would require
15 years to achieve if Moores law is also applicable to this
technology and holds over that period of time.
The overwhelming advantage of velocity measurement
would be that only one transmitted photon would provide
this measurement although, as will be shown later, the delay
of arrival of several transmitted photons would need to be
measured to provide an acceptable degree of certainty,
whereas attenuation estimation needs more than 104 photons
to be transmitted to provide a useful signal-to-noise measurement per picture element pixel in the detector.15 Hence,
the exposure for time-of-flight-delay imaging could potentially require only one ten-thousandth the radiation needed
for conventional, attenuation-based x-ray imaging.
Two ways in which time-of-flight measurements might be
made are illustrated in Fig. 3. One is to launch a very brief,
well-characterized x-ray pulse at a precisely known time and
then measure the precise arrival time at a detector. The other
is to use a positron annihilation source of gamma rays which
launches two photons at exactly the same time in essentially
opposite directions, one to a reference detector and the other
through the subject and then to the imaging detector. In both
cases, the difference in time between the launch and detection of the x-ray photon then provides the time-of-flight delay of the photon. However, the positron/electron annihilation approach using a mirror image reference detector
array is almost certainly not practical for two reasons. One is
that at that photon energy there is only a very small difference in the photon velocity as it passes through matter and
second, the fact that the gamma ray photon pairs do not
travel in exactly opposite directions. This deviation from
180 can be as much as 30 m radians which at 60 cm if a
point source is used to illuminate a 30 cm diameter torso
translates to 18 mm although statistically this might be half
thatresulting in an unacceptable blurring of the image, let
alone the added uncertainty in time delay resulting from the
up to 0.27 mm longer path length. However, if the reference
detector is positioned close to the gamma ray source, then it
could be used to provide a timing signal for individual annihilations, resulting in a gamma ray pair. In this approach, a
source generating 6 1010 photons/ s would ensure illumi-
2508
2508
FIG. 3. Two approaches to time-of-flight x-ray imaging. Left panel: a very brief x-ray pulse is launched and the change in time of arrival of the x ray at the
detector used to generate the image. Right panel: a positron annihilation results in two gamma ray photons which are launched simultaneously in opposite
directions. The difference in time of detection between the reference and imaging detector pair is the signal used to generate the image.
nation of each pixel in the detector array with at least five

photons within 100 ms see the next sections for details of
several issues related to characteristics of the x-ray detection
process which require x-ray exposure to be greater than
needed to just overcome the attenuation by the body.
There are three major reasons why more than a single
photon needs to enter a detector pixel, thereby somewhat
reducing the exposure advantage of the time-of-flight
method.
III.B.1. Spatial heterogeneity of illumination of the
x-ray detector areas
If the x-ray or positron-emitting source is effectively a
single point source of x ray generating a random distribution
of photons in 4 steradians, then, as illustrated schematically
in Fig. 4, some detector pixels in the detector array are likely
to have several hits before others have even one hit. If the
number of detected hits in a detection pixel follows a Poisson distribution with a mean dependent on the number of
photons that survive and are detected, then
Px = eNdNdx/x ! ,
where Nd is the mean number of photons detected in each

pixel, and x is the thickness of the transradiated object.
The number detected follows a binomial distribution, dependent on the number that survive Nd so that the probability of detection P is
PNd = pdNd1 pdNNd,
where pd= detection probability or detector efficiency, and

N is the number of photons that survive with the number that
survive passing through the chest following a binomial distribution, then, depending on No, the absorption parameter ,
and the thickness x, then
PN = exN1 exNoN .
FIG. 4. If each pixel in this 9-pixel detector array has an equal chance of
being hit by a random x-ray photon emitted from a point x-ray source, and
detection efficiency is 100%, then approximately double the number of photons are needed to be reasonably sure that each pixel receives at least one
photon. The panels show the distribution of the number of detected photons
that might occur after the array is exposed by the indicated number of
photons.
Table I shows how the combination of attenuation by body

thickness, the efficiency of the detector, and launched number of photons combine to influence the percentage of pixels
that detect fewer than 5 photons. If we tolerate that 0.01% of
the detector pixels detect 5 photons/ pixel, this means that
a 10002 array of detector elements would have 100 pixels
each receiving less than 5 photons. For this array, an x-ray
detection efficiency of 70% and imaging through a 10 cm
tissue path length, this would mean that if we accept 100
pixels with less than five detections would require launching,
2509
2509
TABLE I. Each detector pixel should detect at least one photon to prevent dead spots in the image. The random spatial distribution of photons see Fig. 4,
attenuation of photon flux by the body, and the inefficiency of the detector combine to decrease the chance of a detector pixel detecting an x-ray photon. This
decrease is mitigated by increasing the x-ray flux destined for each typical detector pixel. The numbers in bold indicate those conditions for which less than
0.01% of detector pixels each detect less than five photons.
X-ray flux mean number of photons aimed
at 1 detector pixel
Efficiency of detector fraction of
transmitted photons detected by detector
Attenuation by body thickness
5 cm
10 cm
15 cm
20 cm
25 cm
60
80
0.5
0.7
0.9
0.95
0.5
0.7
0.09
0.95
0.4
9.7
41
73
97
0.01
1.4
18
52
87
0.00
0.14
6.4
32
62
Percent of detector pixels with 5 detections

0.00
0.02
0.00
0.00
0.00
0.00
0.08
2.2
0.12
0.00
0.00
0.42
4.9
21.0
5.6
1.17
0.77
9.7
30
57
33
17
15
42
60
76
66
48
33
73
0.00
0.00
1.53
19
52
0.00
0.00
0.18
8
36
0.00
0.00
0.10
6
32
on average, 100 photons at each detector pixel, i.e., 1% of

the number of photons needed for attenuation-based imaging. This random pixel exposure effect can be largely eliminated if a scanning x-ray source illuminates one detection
pixel pair at the time and it moves to the next pixel immediately after an x-ray photon is detected. However, if a mechanical scanning process is involved, this would almost certainly incur a clinically unacceptable increase in the duration
required to acquire such an image.
III.B.2. X-ray detector time-of-detection uncertainty
A second issue is that the detectors have an inherent delay
between x ray entering into the detector and its detection.
This delay overwhelms the temporal resolution of a detected
event. Figure 5 schematically illustrates the problem in terms
of the probability of detection in a detector as a function of
depth into the detector Fig. 5. Hence, we have uncertainty in
the time of detection due to this trade-off between depth of
penetration and probability of detection with depth. If, however, the detection duration is kept fixed from the time of
x-ray launch, then the situation depicted schematically in
Fig. 6 results if each photons launch or in the positron
FIG. 5. As detector thickness increases the probability of a photon being

detected increases in a 1 ex relationship, where is the attenuation
coefficient of the detector material. However, the thicker the detector, the
accuracy of the estimated arrival time of the photon falls off hyperbolically.
0.5
0.7
100
0.9
0.95
annihilation case, the detection in the reference detector

triggers the stop-clock of the detection device. If the launch
time of each x-ray photon also turns off the detection process
after a preset time interval, then for a number of photons
with a time distribution of the chance Pt that a photon is
detected at time t is
Pt = k1 egt,
10
where k is a constant related to the detection efficiency of the

detector mechanism, is attenuation coefficient of the detector crystal, g is the group velocity of x ray i.e., photon
velocity in the detecting mechanism, and t is the time after
entry into the crystal.
The change in probability of detection, resulting from a
delayed arrival by t, is P = kegtt. If the detector crystal
is 1 cm in length, = 10 cm1, the velocity of the x ray
through the crystal is g = 3 1010 cm/ s, and a time differential of 1015 s is to be determined, then the difference in
number of counts due to the delay in the difference signal
is approximately 3 104. Hence, at least 104 detections
would be needed, which requires at least as many photons as
would attenuation-based x-ray imaging. Moreover, this is an
ambiguous situation because this approach cannot distinguish between a decrease in detected photons due to increased attenuation by the subject or whether there is an
increased delay of the photon arrival. However, as illustrated
in Fig. 7, if we use a detector system that can detect the
depth of penetration of the photon, such as would be functionally equivalent to a CCD coupled to a microchannel plate
intensifier which is exposed to x ray side-on, we can have
an effectively infinitely thick detector that captures all photons entering it, but without loss of timing information because each depth of interaction can be accounted for in terms
of added delay due to passage through that distance within
the detector. The time uncertainty is now largely determined
by the thickness of each microchannel and the size of the
CCD detection pixelwhich can be of the order of 10 m,
and by the delay between the interaction of the x-ray photon
and the microchannel and subsequent electron transfer into
the CCD pixel. This could therefore result in 3 1014 s
time-resolution steps. This time discrimination alone may be
2510
2510
FIG. 6. If the x-ray detection process of a detector is cut off after a short
time interval from a selected time zero, then the total number of photons
area under curve detected decreases with delay of the x-ray pulse left
panel. Hence, the difference in the area under the curves is related both to
delay of the pulse due to any difference in attenuation right panel within
the subject.
FIG. 8. If a brief bunch of x-ray photons is allocated a time of occurrence

with a clock cycle time that is greater than the time between photons, then
a binning transformation of the bunchs frequency/time distribution occurs
which affects the computed mean time of arrival of the photons. However, if
the x-ray photon bunchs time distribution is well known, then the transformation of the time distribution allows computation of the arrival time with
subcycle time precision.
enough to distinguish different x-ray transit times through

the torso, but not quite enough to detect a 100 m difference in tissue path length.
III.B.3. Precision of time-delay detection
A third issue is the need for increased number of x-ray
photon detections if the precision of the time-of-detection
has to be better than one clock cycle interval e.g., 3
1014 s. Super-resolution has been achieved with analogto-digital converter systems by dithering the input signal
with a well-described signal perturbation, e.g., by the addition of a Gaussian distribution of noise.58 The duration and
time-intensity profile of the x-ray pulse, whether due to a
field-emission burst29 or due to the time-intensity characteristics of a combined laser pulse and electron beam system, or
due to the finite thickness of the positron emitter source,
transforms by virtue of the temporal discretization into a
postdigitized time-intensity profile as illustrated schematically in Fig. 8. For situations in which the time difference of
interest is less than the clock precision, we can use the
known time distribution of sequential photons to provide
super-resolution time difference detection as follows. If the
stochasticity in the time-of-flight for the multiple pairs of
particles is approximated by a Gaussian distributionwith a
standard deviation equal to a constant times the clock precision, and we assume a correlation of 0.80 between the reference and imaging detectors in the case of the positron anni-
hilation approach, then one time-of-flight explains 64% of

the variability in the other. Table II shows that a tenfold
increase of temporal resolution can be achieved with an exposure sufficient to result in an average of 128 detected
photons/detector element. Thus, even if extra photons are
needed to increase the clock-cycle precision by one order of
magnitude, this two decade inefficiency, combined with a
25-fold 128/5 increase needed to ensure at least 128 photons for each detector pixel, would still result in a 40-fold
105 / 250 lower x-ray exposure than achievable with current
attenuation-based systems. If a sufficiently fast clock cycle is
available thereby eliminating the need to use the dithering
principle, then a 200-fold lower x-ray exposure would be
plausiblewhich would allow reducing the voxel size to
100 m3 without an increase in x-ray exposure currently
experienced with MDCT scanners.
III.B.4. Temporal resolution
Temporal resolution requirements are the same as those
for attenuation-based x-ray imaging. However, as the radiation exposure needed is reduced several fold, the exposure
time needed to generate an image should potentially also be
reduced, although this and the repetition rate of image generation would depend very much on the technology involved
in implementing the time-of-flight-based x-ray imaging
method.
IV. CONCLUSION
FIG. 7. If the depth x of x-ray interaction into a detector can be detected,

then the time of entry into the detector can be calculated as x / g, where g
is the speed of x ray through the detector. The precision of this distance
measurement would be less than 10 m i.e., 33 fs if a microchannel
platelike arrangement were used.
Current, attenuation-based x-ray imaging methods cannot

be expected to increase much more in spatial and contrast
resolution due to the undesirable x-ray exposure consequences. If time-of-flight x-ray imaging can be developed for
clinical applications, the power of x-ray imaging could be
greatly enhanced by virtue of its potential for a two decade
reduction of radiation exposure at current image resolutions
2511
2511
TABLE II. For the positron annihilation approach a pair of photons is emitted in opposite directions from a
source of finite thickness such that the arrival times of the photons at the detector pixel has a Gaussian
distribution. As the number of detected event pairs is increased, the discrimination of time differences t
between the reference detector pixel and imaging detector pixel can be a fraction of the time resolution t
of the clock used to measure the times of the event occurrences.
# Observed events pairs
Discrimination t / t
2
0.75
4
0.55
or its potential for resolution down to 100 m without increasing the radiation exposure beyond that currently experienced at 600 m voxel resolution.
ACKNOWLEDGMENTS
The author thanks Professor Dr. Ulrich Bonse, Technical
University Dortmund, Dortmund, Germany; Professor Dr.
Winfried Schuelke, Institute of Physics, University of Dortmund, Dortmund, Germany; Professor James L. Ritman,
University of Bochum and Research Center Jlich, Institute
for Nuclear Physics, Jlich, Germany, for their input on some
of the x-ray/matter interaction issues; Dr. Vernon S. Pankratz
and Mr. Jeffrey Slezak of the Mayo Clinic Biostatistics Section for performing the statistical analysis underlying the
analysis illustrated in Tables I and II, and Ms. Mara Lukenda
for typing and formatting this manuscript.
a
Telephone: 507-255-1939; Fax: 507-255-1935. Electronic mail:

[email protected]
1
H. L. Baker, Jr., The impact of computed tomography on neurological
practice, Radiology 116, 637640 1975.
2
H. Nishimura and O. Miyazaki, CT system for spirally imaging subject
on a moveable bed synchronized to x-ray tube revolution, U.S. Patent
4,789,929 1988.
3
C. Kondo, S. Mori, M. Endo, K. Kusakabe, K. Suzuki, N. Hattori, and M.
Kusakabe, Real-time volumetric imaging of human heart without electrocardiographic gating by 256-detector row computed tomography: Initial experience, J. Comput. Assist. Tomogr. 29, 694698 2005.
4
P. J. Pickhardt, J. R. Choi, I. Hwang, J. A. Butler, M. L. Puckett, H. A.
Hildebrandt, R. K. Wong, P. A. Nugent, and P. A. Mysliwiec, Computed
tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults, N. Engl. J. Med. 349, 21912200 2003.
5
D. F. Yankelevitz, R. Gupta, B. Zhao, and C. I. Henschke, Small pulmonary nodules: Evaluation with repeat CTpreliminary experience, Radiology 212, 561566 1999.
6
B. Ohnesorge, T. Flohr, R. Fischbach, A. F. Kopp, A. Knez, S. Schroder,
U. J. Schopf, A. Crispin, E. Klotz, M. F. Reiser, and C. R. Becker, Reproducibility of coronary calcium quantification in repeat examinations
with retrospectively ECG-gated multisection spiral CT, Eur. Radiol. 12,
15321540 2002.
7
M. Imaizumi et al., Radiation dose-response relationships for thyroid
nodules and autoimmune diseases in Hiroshima and Nagasaki Atomic
Bomb survivors 5558 years after radiation exposure, JAMA, J. Am.
Med. Assoc. 295, 10111022 2006.
8
L. K. Wagner, P. J. Eifel, and R. A. Geise, Potential biological effects
following high x-ray dose interventional procedures, J. Vasc. Interv. Radiol. 5, 7184 1994.
9
R. L. Morin, T. C. Gerber, and C. H. McCollough, Physics and dosimetry in computed tomography, Cardiol. Clin. 21, 515520 2003.
10
A. J. Einstein, M. J. Henzlova, and S. Rajagopalan, Estimating risk of
cancer associated with radiation exposure from 64-slice computed tomography coronary angiography, JAMA 298, 317323 2007.
11
P.C. Shrimpton, M.C. Hillier, M.A. Lewis, M. Dunn, Doses from computed tomography CT examination in the UK2003 Review, National
Radiological Protection Board, Chilton, Dideo, Oxon OX110RQ, 2003.
12
N. L. Keat, Real time CT and CT fluoroscopy. Report MDA/00/10,
8
0.4
16
0.3
32
0.2
64
0.15
128
0.10
Medical Devices Agency, London, 2000.

Statistical research and applications. Probability of developing or dying
of cancer, National Cancer Institute: http://srab.cancer.gov/devcan, National Cancer Institution Vol. 206, pp. 541545 1983.
14
D. R. White, L. H. Peaple, and T. J. Crosby, Measured attenuation coefficients at low photon energies 9.8859.32 keV for 44 materials and
tissues, Radiat. Res. 84, 239252 1980.
15
J. W. Motz and M. Danos, Image information content and patient exposure, Med. Phys. 5, 822 1978.
16
L. Grodzins, Optimal energies for x-ray transverse tomography of small
samples, Nucl. Instrum. Methods 206, 541545 1983.
17
M. J. Friedrich, Studying cancer in 3 dimensions: 3-D models foster new
insights into tumorigenesis, JAMA 290, 19771979 2003.
18
T. Behrenbeck, J. H. Kinsey, L. D. Harris, R. A. Robb, and E. L. Ritman,
Three-dimensional spatial, density and temporal resolution of the dynamic spatial reconstructor, J. Comput. Assist. Tomogr. 6, 11381147
1982.
19
A. Rose, Sensitivity performance of the human eye on an absolute
scale, J. Opt. Soc. Am. 38, 196 1948.
20
D. A. Chesler, S. J. Riederer, and N. J. Pelc, Noise due to photon counting statistics in computed tomography, J. Comput. Assist. Tomogr. 1,
6474 1977.
21
P. D. Johns and M. Yaffe, Scattered radiation in fan beam imaging systems, Med. Phys. 9, 231239 1982.
22
P. S. Rao and R. J. Alfidi, The environmental density artifact: A beamhardening effect in computed tomography, Radiology 141, 223227
1981.
23
A. L. C. Kwan, J. M. Boone, and N. Shah, Evaluation of x-ray scatter
properties in a dedicated cone-beam CT breast CT scanner, Med. Phys.
32, 29672975 2005.
24
J. Persliden and G. A. Carlsson, in Developments in Radiography and
Fluoroscopy Imaging, edited by H. Cole and C. P. Lawinski British
Institute of Radiology, London, 2006, Vol. 18, pp. 7581.
25
E. N. C. Milne, The role and performance of minute focal spots in
roentgenology with special references to magnification, CRC Crit. Rev.
Radiol. Sci. 2, 269310 1971.
26
G. Bucky, ber die Ausschaltung der Sekunderstrahlen Verhandlung der
Deutsch Rtg Gesellsch, Leipzig: GeorgThieme Verlag, 1913, Vol. 9, pp.
3032.
27
R. F. Wagner, G. T. Barnes, and B. S. Askins, Effect of reduced scatter
content on radiographic information, contrast and patient exposure: A
qualitative demonstration, Med. Phys. 7, 1318 1980.
28
R. Ning, X. Tang, and D. Conover, X-ray scatter correction algorithm for
cone beam CT imaging, Med. Phys. 31, 11951202 2004.
29
S. Surti, J. S. Karp, L. M. Popescu, M. E. Duabe-Witherspoon, and M.
Werner, Investigation of time-of-flight benefit for fully 3-D PET, IEEE
30
W. P. Dyke, Advances in field emission, Sci. Am. 210, 108116 1964.
31
W. Leemans, R. Schoenlein, A. Chin, E. Glover, R. Govil, P. Volfbeyn, S.
Chattopadhyay, K.-J. Kim, and C.V. Shank. Femtosecond x-rays from
90 Thomson scattering, Proc. IEEE Particle Accelerator Conf. 1, 174
176 1995.
32
Z. Parvez, R. Moncada, and M. Sovak, Contrast Media: Biologic Effect
and Clinical Application CRC Press Inc., Baton Rouge, 1987, Vol. IIII.
33
H. D. Zeman and D. P. Siddons, Contrast agent choice for intravenous
coronary angiography, Nucl. Instrum. Methods Phys. Res. A 291, 6773
1990.
34
W. H. Shehadi and G. Toniolo, Adverse reactions to contrast media,
Radiology 137, 299302 1980.
35
H. T. Abada, C. Larchez, B. Daoud, A. Sigal-Cinqualbre, and J. F. Paul,
MDCT of the coronary arteries: Feasibility of low-dose CT with ECG13
2512
pulsed tube current modulation to reduce radiation dose, AJR, Am. J.

Roentgenol. 186, S387S390 2006.
36
W. Huda, E. M. Scalzetti, and G. Levin, Technique factors and image
quality as functions of patient weight at abdominal CT, Radiology 217,
430435 2000.
37
W. A. Kalender, H. Wolf, C. Suess, M. Gies, H. Greess, and W. A. Bautz,
Dose reduction in CT by on-line tube current control: Principles and
validation on phantoms and cadavers, Eur. Radiol. 9, 323328 1999.
38
M. D. Harpen, A single theorem relating noise and patient dose in computed tomography, Med. Phys. 26, 22312234 1999.
39
R. M. Lewitt, Processing of incomplete measurement data in computed
40
A. Faridani, F. Keinert, F. Naterrer, E. L. Ritman, and K. T. Smith, in
Signal Processing Part II (IMA Control Theory and Applications), edited
by F. A. Grunbaum, J. W. Helton, and P. Khargonekar Springer-Verlag,
New York, 1990, Vol. 23, pp. 241255.
41
F. A. Dilmanian, Computed tomography with monochromatic x-rays,
Am. J. Physiol. Imaging 7, 175193 1992.
42
P. Baldelli, A. Taibi, A. Tuffanelli, and M. Gambaccini, Dose comparison between conventional and quasi-monochromatic systems for diagnostic radiology, Phys. Med. Biol. 49, 41354146 2004.
43
B. L. Henke, E. M. Gullikson, and J. C. Davis, X-ray interactions: photoabsorption, scattering, transmission, and reflection at E
= 50 30 000 eV , Z = 1 92, At. Data Nucl. Data Tables 54, 181342
1993.
44
J. Kosanetzky, B. Knoerr, G. Harding, and U. Neitzel, X-ray diffraction
measurements of some plastic materials and body tissues, Med. Phys.
14, 526553 1987.
45
M. N. Wernick, O. Wirjadi, D. Chapman, Z. Zhong, N. P. Galatsanos, Y.
Yang, J. G. Brankov, O. Oltulu, M. A. Anastasio, and C. Muehleman,
Multiple-image radiography, Phys. Med. Biol. 48, 38753895 2003.
46
A. M. Anastasio and D. Shi, On the relationship between intensity diffraction tomography and phase-contrast tomography, Proc. SPIE 5535,
2512
13611368 2004.
G. Woan, The Cambridge Handbook of Physics Formulas Cambridge
University Press, Cambridge, 2000, p. 219.
48
Van Nostrands Scientific Encyclopedia, 5th ed., edited by D. M. Considine Van Nostrand Reinhold Co., New York, 1976.
49
A. Momose and J. Fukuda, Phase-contrast radiographs of nonstained rat
cerebellar specimen, Med. Phys. 22, 375379 1995.
50
T. J. Davis, D. Gao, T. E. Gureyev, A. W. Stevenson, and S. W. Wilkins,
Phase-contrast imaging of weakly absorbing materials using hard
x-rays, Nature London 373, 595598 1995.
51
R. A. Lewis, Medical phase contrast x-ray imaging: Current status and
future prospects, Phys. Med. Biol. 49, 35733583 2004.
52
T. Weitkamp, C. David, C. Kottler, O. Bunk, and F. Pfeiffer, Tomography with grating interferometers at low-brilliance sources, Proc. SPIE
6318, 63180S 2006.
53
T. Takeda, A. Momose, J. Wu, Q. Yu, T. Zeniya, T.-T. Lwin, A.
Yoneyama, and Y. Itai, Vessel imaging by interferometric phase-contrast
x-ray technique, Circulation 105, 17081712 2002.
54
F. Beckmann, U. Bonse, F. Busch, and O. Gnnewig, X-ray microtomography microCT using phase contrast for the investigation of organic
matter, J. Comput. Assist. Tomogr. 21, 539553 1997.
55
T. Weitkamp, A. Diaz, B. Nhammer, F. Pfeiffer, T. Rohbeck, P. Cloetens,
M. Stampanoni, and C. David, Hard x-ray phase imaging and tomography with a grating interferometer, Proc. SPIE 5535, 137142 2004.
56
G. E. Moore, Cramming more components onto integrated circuits,
Electronics 38, 114117 1965.
57
C. Martel, S. Fourmaux, L. Lecherbourg, H. Bandulet, and J. C. Kieffer,
New ultra fast x-ray streak camera for the advanced laser light source
facility, Proc. SPIE 6703, 67030K 2007.
58
H. Shimizu, Y. Iikura, Y. Sasano, and N. Takeuchi, Resolution improvement in an analog-to-digital converter by the superposed dither signal,
Electron. Commun. Jpn. 64, 18 1981.
47
Targeted radionuclide therapy

Lawrence E. Williamsa
Radiology Division, City of Hope National Medical Center, Duarte, California 91010
Gerald L. DeNardob
Internal Medicine, University of California Davis Medical Center, 1508 Alhambra Boulevard, Suite 3100,
Sacramento, California 95816
Ruby F. Meredithc
Department of Radiation Oncology, Wallace Tumor Institute WTI No. 117, University of Alabama
at Birmingham, Birmingham, Alabama 35294
Received 4 March 2008; revised 22 April 2008; accepted for publication 10 May 2008;
Targeted radionuclide therapy TRT seeks molecular and functional targets within patient tumor
sites. A number of agents have been constructed and labeled with beta, alpha, and Auger emitters.
Radionuclide carriers spanning a broad range of sizes; e.g., antibodies, liposomes, and constructs
such as nanoparticles have been used in these studies. Uptake, in percent-injected dose per gram of
malignant tissue, is used to evaluate the specificity of the targeting vehicle. Lymphoma B-cell has
been the primary clinical application. Extension to solid tumors will require raising the macroscopic
absorbed dose by several-fold over values found in present technology. Methods that may effect
such changes include multistep targeting, simultaneous chemotherapy, and external sequestration of
the agent. Toxicity has primarily involved red marrow so that marrow replacement can also be used
to enhance future TRT treatments. Correlation of toxicities and treatment efficiency has been limited by relatively poor absorbed dose estimates partly because of using standard phantom organ
sizes. These associations will be improved in the future by obtaining patient-specific organ size and
activity data with hybrid SPECT/CT and PET/CT scanners. 2008 American Association of
Physicists in Medicine. DOI: 10.1118/1.2938520
Key words: radionuclide therapy, antibodies, absorbed dose
I. PRESENT CLINICAL SITUATION
A possible imaging result with a cancer patient is shown in
Fig. 1, whereby multiple lesions appear using nuclear or
other technology. It is important to find a method to target
both seen and unseen tumor locations anywhere in the body.
It would be desirable that this strategy be as specific as possible to the tumor with minimal collateral toxicity. Targeting
of this sort is measured in units of percent-injected dose per
gram of tissue % ID/ g. The mouse would generally be the
test species for preclinical development of radiopharmaceuticals for targeted radionuclide therapy TRT. Following IV
injection into a 20 g animal, nonspecific uptake should be
100% ID/20 g = 5% ID/g
for any of the tissues. Extrapolating the calculation to a

70-kg man, the comparable nontargeting value is
1.4% ID/ kg. Tumor uptakes u exceeding these respective
tissue sample numbers can then be taken as evidence of specific targeting of the agent of interest. Tumor absorbed dose
rate for particulate emissions is also proportional to the tumor u value so that applicability of uptake to radiotherapy is
direct.
While intravenous injection is probably the only way to
target lesions at any site, direct injections may be used to
enhance tumor accumulation. Typically, this intervention oc3062
curs postsurgery. For example, intraperitoneal applications of

antibodies to ovarian cancer1 and directly into brain tumor
sites2 are currently being evaluated.
Many agents have been investigated as possible tumortargeting vehicles. Some, such as antibodies, may also have
intrinsic antitumor effects. While the following will describe
radionuclide-effected therapy, antitumor chemical species
may also be loaded into some tumor seekers such as liposomes. A partial list is given in Table I that includes the
traditional use of the 131I ion as postsurgery treatment of
choice for some types of thyroid cancer. This compilation is
capable of expansion using either further engineering of the
agents listed, direct invention of novel entities, or combinations of multiple technologies; e.g., liposomes with antibodies on their outer surface.
Optimization of tumor targeting is complicated due to the
number of possible agents. Using an antibody as an example,
a single amino acid exchange can cause order-of-magnitude
variation in blood circulation times of the mutated protein.13
Since there are several thousand amino acids in the original
antibody sequence and essentially 20 possible amino acid
inserts at each location, the number of variations is astronomical. Thus, only a very limited subset of all possible
agents will ever be investigated experimentally.
Only four agents have progressed to FDA-approved clinical therapies as shown in Table II. Note that two of these
therapies involve direct catheter injection into tumor sites in
0094-2405/2008/357/3062/7/$23.00
3062
3063
Williams, DeNardo, and Meredith: Targeted radionuclide therapy
3063
Increasing injected total activity A0 cannot be used to

arbitrarily raise tumor absorbed doses because of limits set
by normal organ toxicities.17 Bone marrow is the most probable absorbed dose-limiting tissue in TRT when the activity
is injected IV. Presently, red marrow RM absorbed dose is
difficult to estimate due to uncertainty in patient-specific information. Recall that internal emitter absorbed dose D is
usually estimated by the simple vector equation
D = S A ,
FIG. 1. Anterior gamma camera image of a medullary thyroid cancer patient. Image obtained at 48 h post-IV injection of 5 mCi 185 MBq of
111
In-cT84.66 anti-CEA antibody. Multiple bone lesions are seen in the pelvis and both femurs. Part of the right lobe of liver appears superiorly see
Ref. 37.
the liver. Intravenous injections have been limited to treatment of lymphoma using anti-B cell anti-CD20 monoclonal
antibodies.
II. CURRENT LIMITATIONS
Because of the small number of FDA-approved agents, it
is clear that significant constraints have hindered development of TRT. Primarily, the limitation has been relatively
poor tumor uptake. This has led to correspondingly low absorbed doses and lack of tumor response. Table III contains
estimated absorbed doses for a standard phase III lymphoma
trial14 and an experimental colon cancer therapy.16 These
protocols were at previously determined maximal values of
activity per body mass14 or per body surface area.16 Notice
that both lymphoma and colon cancer regimens have
achieved similar numerical tumor absorbed dose values.
Clinical application of TRT to the former disease state has
followed from the relatively high sensitivity of B-cells to
ionizing radiation.
where S is a rectangular matrix relating source organs to

targets and A is a vector containing the integrals of source
organ activities A over time. S contains information on the
radionuclides emission energies and probabilities as well as
the geometry of the source and target organs. For particulate
emitters that stop in the source, we concern ourselves with
diagonal elements of S where source and target are the same
tissue. In this case, S matrix elements are inversely proportional to the total organ mass. For red marrow, this parameter
is unknown since the individual patient has genetic, agerelated, and therapy-associated variability. One would expect
that the RM mass would be significantly different fromand
probably lower thanvalues found in phantom-based absorbed dose estimation algorithms. There is no present-day
routinely used method for estimating this unknown tissue
mass value. Thus, phase I TRT trial escalation has generally
not been based on absorbed dose toxicity but usually on the
injected activity per meter squared or activity per total body
mass. Neither of these parameters, which are based on traditional chemotherapy treatment planning, has correlated well
with patient marrow toxicity.18
Given the above constraints, an approximate formula to
estimate the circulating blood contribution to marrow absorbed dose has been developed by the AAPM task group19
TABLE I. A partial list of tumor-targeting agents.

Agent
Molecular weight Da or physical size
NaI
MIBGa
Octreotideb
SHALsc
Nucleotidesd
Antibodiese
Liposomesf
Nanoparticlesg
Morpholinosh
Spheresi
154 Da
130 Da
100 Da
2 kDa
10 kDa
25150 kDa
100 nm
10 nm
2 kDa
30 m
Application
Representative radiolabels
Thyroid cancer follicular and papillary

Neuroendocrine tumors
As above
Lymphoma
Solid tumors
Lymphomas, solid tumors
Solid tumors
Solid tumors
Solid tumors
Hepatic lesions
131
See Ref. 3.
See Ref. 4.
c
See Ref. 5.
d
See Ref. 6.
e
See Ref. 7.
See Ref. 8.
See Ref. 9.
h
See Ref. 10.
i
See Refs. 11 and 12.
I
I
90
Y, 111In
131
I
99m
Tc, 111In
90
Y and 131I
111
In, 99mTc
131
188
Re
Y
90
3064
3064
TABLE II. FDA-Approved TRT protocols.

Agent
SIR spheres plastic
Theraspheresb glass
Zevalinc Ibritumomab
Bexxard Tositumomab
a
Label
Disease
Method
Y 99mTc MAA for images

Y 99mTc MAA for images
90
Y 111In for images
131
I
Metastatic liver cancer

Primary liver cancer
B-cell lymphoma
B-cell lymphoma
Catheter via hepatic artery

Catheter via hepatic artery
IV injection
IV injection
90
90
See Ref. 11.

See Ref. 12.
c
See Ref. 14.
d
See Ref. 15.
b
ARM = f Ablood 1500/5000,

*
*
where f is a fraction on the order of 0.3 and the ratio represents the correction for marrow mass 1500 g in a particular
standard phantom20 in a patient with an assumed 5000 g
total blood mass. In the ideal case of no direct marrow targeting of the radioactive agent or its label, the result of Eq.
3 is used as an input into Eq. 2 to estimate total RM
absorbed dose.
Issues of accurate absorbed dose estimation have been at
the heart of TRT since its inception in the 1950s; e.g., there is
little knowledge of the absorbed dose to remnant thyroid
tissue during 131I treatment. While MIRD-type calculations
are often used to justify clinical trials to the FDA, use of the
associated phantom mass values contained in programs like
MIRDOSE3 Ref. 20 and OLINDA Ref. 21 cannot be directly
applied to a given patient. It has been shown, for example,
that hepatic and splenic masses may differ by factors of twoand threefold in individuals undergoing TRT of colon
cancer.22 Patient-specific treatment planning23 requires actual
knowledge of patient anatomy organ separations and organ
sizes masses. This problem is thus one of image fusion
whereby the activity found using quantitative nuclear imaging can be assigned correctly to a particular anatomic
structure.
III. FUTURE DEVELOPMENTS
Medical physicists can be expected to contribute directly
to the growth of TRT in the next 710 years. Advances
would be of two distinctly different but complementary
types. Initially there will have to be a strategy for develop-
ing, testing, and generally enhancing the tumor-targeting capability of radiopharmaceuticals. Computer modeling of biodistributions and comparisons of figures of merit are aspects
of this work relevant to the physicist. This is particularly
important for solid tumors. Beyond the engineering and inventive phase will be the necessity of establishing improved
methods of patient-specific absorbed dose calculations.
These estimates will become more useful when they relate to
voxels rather than whole organ values as indicated above.
Among target tissue calculations, red marrow absorbed dose
estimates will continue to be the most important.
III.A. Pharmaceutical development and testing
Engineering is the operative word for the future. TRT

agents of the present time are usually constructed by crude
analogy to naturally occurring entities in the bloodstream of
common mammals. Protein configurations appear to be one
of the most likely for future manipulation. After in vitro testing, the putative tumor seekers would be evaluated via sequential animal three-dimensional imaging rather than by
biodistribution studies as are done today. One particularly
fruitful area of this research would be the use of mouse-sized
PET or SPECT scanners to obviate the need for large numbers of animals being sacrificed at each time point in the
biodistribution evaluation. Statistically, this method is also
preferred since the same animals are being followed serially.
Flexibility of usage may favor SPECT over PET in that more
radionuclide labels are available for the former technique.
Use of PET scanners requires availability of suitable positron emitters as labels. Two types of labeling may be anticipated: Radioiodine and radiometal. In the case of proteins,
TABLE III. Comparison of two TRT clinical antibody protocols involving 90Y. Absorbed doses were computed using the MIRD formalism Refs. 20 and 21
and refer to whole organ mean values. Ranges are shown in parentheses. The clinical protocols were at maximum activity per body mass Ref. 14 or body
surface area Ref. 16 determined from previously established hematological toxicity levels.
Agent
Zevalina
Ant-CEAb
Disease
NHL
Colon cancer
See Ref. 14.

See Ref. 16.
Tumor absorbed dose

cGy
RM absorbed dose
cGy
Liver absorbed dose

cGy
1484
61 24 300
1320
46 6400
71
18 221
64
19 198
532
234 1856
912
534 1719
3065
there is a developmental advantage in using radioiodines.

Iodine can be directly attached to tyrosine moieties in the
amino acid sequence with protein amounts as small as
100 g. This is in contrast to the radiometal case where
milligram amounts are required since a chelator must be first
attached to the protein. For iodines, the physical half-life of
124
I is 100 h and appears to be a good theoretical choice for
PET. Limitations of availability and accurate quantitation
have stymied extensive use. Recent work24 has shown that
measurement of the amount of 124I is possible with a commercial PET imager and human-scale phantoms.
Iodination of tumor-tracking agents has one disadvantage.
Because of thyroid needs, multiple enzymes have evolved to
extract iodine atoms from sites on proteins and other agents.
Using an iodinated species for extended periods in vivo may
lead to unwanted targeting of freed activity to stomach, gut,
and thyroid. Because of this limitation, use of radiometal
labels of proteins and other entities is expected to increase.
This will require production facilities for 64Cu and 86Y, for
example. With their relatively simple gamma spectra, these
labels will probably become the marker of choice in many
preclinical and clinical trials that study entities with relatively rapid biological targeting i.e., biological half-life
12 h. Species that remain in the blood for periods exceeding one day, however, will remain difficult to label with these
radiometals and may require use of 124I.
Several ideas have shown promise in the enhancement of
tumor therapy using IV-injected agents. One of these is the
use of a sequence of steps to optimize the placement of activity on or within the tumor.25 In this strategy, a large unlabeled molecule, which is at least partially an antibody to the
tumor-associated antigen, is injected initially. After some
time, and perhaps a clearance step, a low-mass radiolabeled
entity is injected. The labeled agent would be an antibody
fragment or other small ligand that reacts to the complex of
original antigen and the first large molecule. Since the labeled molecule would be of relatively low molecular weight,
it would diffuse more rapidly to the lesion sites. Such quick
movement would reduce red marrow absorbed dose induced
by the circulating activity in the blood.26 Limitations to the
method involve the possible increase in the renal absorbed
dose due to the anticipated urinary excretion of the radiolabeled entity.
A second targeting variation that manipulates the blood
curve of the radiotracer shows significant promise for reducing bone marrow toxicities as estimated via Eq. 3. Here the
clinicians would use an antigen-bearing protein column located outside the patient.27 Blood would be shunted from the
patient to the column at a calculated time when the estimated
marrow absorbed dose approached toxic limits. At that point,
the activity level in the bloodstream would essentially be set
to zero such that no further bone marrow toxicity would be
anticipated. Note that this method requires that the patient be
placed into an external circulatory system.
Other invasive strategies have been proposed. If the marrow is expected to be the absorbed dose-limiting tissue, the
patient may have a marrow sample sequestered prior to the
therapy with subsequent transplantation if the RM is suffiMedical Physics, Vol. 35, No. 7, July 2008
3065
ciently depleted. It has become conventional to replace platelets and white cells in patients undergoing TRT when these
cells are below certain limits. In the future, such procedures
may become economical enough such that the marrow is no
longer the activity-limiting tissue. In that case, other solid
organs such as the kidneys or liver may be considered limiting. This strategy may lead to greater total activities being
injected and larger tumor absorbed doses being achieved.
As in the case of external beam, one anticipates that considerable effort will be expended to enhance TRT-induced
absorbed dose by means of circulating chemotherapeutic and
radiation-enhancing agents.28 At present, it is unclear what
chemical dose levels and what timing relative to the IV injection of the radiotherapeutic agent are optimal. Again, preclinical and phase I trials will be needed to evaluate the
effectiveness of this approach. Because of the multiple possible combinations of enhancing and TRT agents, extensive
animal testing will become more important in the near term.
Following from the above multiple agent therapies, it is
probable that future collaborations would be expanded between medical oncology and radiation oncology. Historically,
chemotherapeutic drugs have been studied using only blood
and excreta chemical measurements. In the next 710 years,
it would be anticipated that present-day and future chemotherapeutic drugs would be radiolabeled and then followed,
probably with PET scanning, in patients. Because of concern
that the radioactive tag may affect the biodistribution, labeling will preferably be done by substituting radioactive isotopes into the chemical structure. While organic isotopes 11C,
15
O, and 13N are of primary importance, there will be other
possible labels such as 18F in the case of 5FU. Availability of
local cyclotrons is implicit in this approach due to the short
half-lives of the positron labels required.
III.B. Improved absorbed dose estimation
Improvement in patient-specific absorbed dose estimates

will be one outcome achievable by medical physicists in the
immediate time frame. This result may be termed a type-II
absorbed dose estimate whereas the MIRD phantom result
is a type-I calculation.29 Phantom results will still be needed
for FDA approval of the trial; patient-specific calculations
will evolve to be used to assess clinical outcomes for an
individual. In order to achieve such specificity, knowledge of
the patients organ masses and their spatial separations will
be essential.
Instrumentation enhancements, now common in nuclear
medicine, will make it possible to directly improve patient
absorbed dose estimates. This follows from the two uncertainties implicit in Eq. 2: Evaluating A and S for an individual. Considering the present difficulty of estimating the
activity A at depth, it is important to realize that there are at
least six different techniques30 that have been applied to this
long-standing physics problem. With hybrid scanners such as
SPECT/CT and PET/CT, gamma attenuation factors will be
evaluated directly without the need for external transmission
sources or complicated fusions of different modality images.
Additionally, the activity being imaged will be assignable to
3066
a specific anatomic structure in the body; a structure that will

also have a definite geometric volume and location relative
to other tissues. As a result, uncertainties using the phantom
calculations in OLINDA Ref. 21 will be reduced and the true
organ or tumor size obtained. Koral and co-workers have
demonstrated31 that quantitative SPECT permits tumor absorbed dose estimates that do correlate with lymphoma outcomes. With hybrid scanners, we expect that such improved
correlations will occur for other tumor types.
Variation of tumor mass during therapy is a feature of
TRT that has been observed in clinical lymphoma trials.32
This result cannot be anticipated from prior imaging sessions
since the effect arises out of the relatively high magnitude of
the therapeutic absorbed dose. In the future, simple relationships such as Eq. 2 will not generally be valid for tumors.
Instead, dosimetrists will use the differential form of the
equality whereby absorbed dose rate is evaluated over a set
of time intervals during the therapy regimen. A mathematical
model of the target mass variation over time can be used in
the determination of S in this case.
Equation 2 has other important limitations. As defined
above, it refers to average total organ or tumor absorbed
doses. While of gross interest, such single value representations of the ionizing radiation would not be as useful as
absorbed dose-volume histograms that have been popularized in external beam work. In this case, a voxel-based set of
sources would be introduced and a voxel-by-voxel result calculated by convolution. This result, when available, will lead
to greater absorbed dose heterogeneity and make treatment
planning a more complicated task. A technical problem in
this context would be the tracking of individual regions from
one imaging session to the next during the imaging or subsequent therapy. For beta emitters, brake radiation effects can
be included in this formulation.
It has been shown in both murine and some human studies
that tumor uptake is an inverse function of tumor massat
least for radiometal-labeled entities.33,34 We would anticipate
that this result would become more important in future clinical trials implying adjuvant treatment of patients postsurgery.
Applications of this idea to occult lesions would also be of
interest since these sites would probably have relatively
small sizes. One colon cancer trial involving unlabeled antibodies has been reported.35
Like politics, absorbed dose is essentially a local phenomenon. Given surgical specimens and information on cellular
localization, an optimal radionuclide label for a given tumor
type can also become possible. Proof of localization within
tumor cell nuclei or at least at their surfaces would imply use
of alpha emitters such as 213Bi or Auger emitters as 111In.
Heterogeneous distributions of activity throughout the tumor,
on the other hand, would indicate longer-range beta producers such as 90Y to enhance cross-fire effects. Similar arguments can be applied to normal tissue toxicity. Thus, future
improvement in therapy will depend upon greater cooperation between surgeons and pathologists, radiation oncologists, and therapy physicists.
3066
IV. CONCLUSIONS
Continued growth in lymphoma treatment using TRT is
expected with more patients being entered into RIT trials as
part of a first-line therapy. Helping to expedite this expansion
is the unexplained, growing incidence of B-cell lymphomas.
We would anticipate that T-cell antibodies would become
available within the next 5 years. While 90Y facilitates outpatient treatments, the utilization of 131I as a label is expected to continue due to lower radionuclide cost and the
possibility of directly imaging therapeutic administrations to
verify treatment planning.
With better dosimetry, repeat TRT procedures would become possible. In the next 57 years, repeated lymphoma
therapies would become a topic of importance to patients
who have recurrence or were originally undertreated. Patientspecific normal organ absorbed doses would be tabulated to
make sure that the relevant total is within the levels determined in external beam therapy.17 It is likely that the toxicity
levels due to TRT-derived absorbed doses will be found to be
different from those seen in external beam therapies.
We can also expect extensive use of engineering and invention to contribute a number of new targeting
pharmaceuticals.36 Improvement in chelates will enable
greater control of radiometals so that direct targeting of the
freed label to the marrow becomes improbable.
Improved patient-specific absorbed dose estimates will
lead to better control of toxicities and improved clinical results for lymphomas and, eventually, solid tumors. The
present tenuous connection between estimated marrow absorbed dose and its clinical toxicity will become more statistically significant as the marrow mass becomes available via
marrow tracers used as adjuncts prior to the TRT protocols.
As a result, escalation in phase I and other FDA-approved
trials would occur with normal organ absorbed doses as the
control variables and not simplistic parameters such as total
activity or total activity per body surface area or body mass.
Labeling and tracking of chemotherapeutics will evolve out
of this work and lead to greater understanding and safety of
their administration to individual patients. Combinations of
radiation-enhancing compounds with TRT will expand
within these two converging threads of knowledge. Patientspecific treatment is one logical outcome of this work.
ACKNOWLEDGMENTS
This work was partially supported by NIH Grant Nos.
PO1-43904 and CA 33572 L.W..
a
Author to whom correspondence should be addressed. Telephone: 626359-8111

X61488;
Fax:
626-930-5451;
Electronic
mail:
[email protected]
b
Telephone: 916-734-3787; Fax: 916-451-2857; Electronic mail:
[email protected]
c
Telephone: 205-975-0222; Fax: 205-975-0784; Electronic mail:
[email protected]
1
A. J. Jacobs, M. Fer, F. M. Su, H. Breitz, J. Thompson, H. Goodgold, J.
Cani, J. Heap, and P. Weiden, A phase I trial of rhenium-186-labeled
monoclonal antibody administered intraperitoneally in ovarian carcinoma:
Toxicity and clinical response, Obstet. Gynecol. N.Y., NY, U. S. 82,
586593 1993.
3067
P. Riva, A. Arista, V. Tison, C. Sturiale, G. Francheschi, A. Spinelli, N.

Riva, M. Casi, G. Moscatelli, and M. Frattarelli, Intralesional radioimmunotherapy of malignant gliomas: An effective treatment in recurrent
tumors, Cancer 73, 10761082 1994.
3
B. Shapiro, J. C. Sisson, D. M. Wieland, T. J. Mangner, S. M. Zempel, E.
Mudgett, M. D. Gross, J. E. Carey, K. R. Zasadny, and W. H. Beierwaltes, Radiopharmaceutical therapy of malignant pheochromocytoma
with 131I metaiodobenzylguanidine: Results from ten years of experience, J. Nucl. Biol. Med. 35, 269276 1991.
4
A. Frilling, F. Weber, V. Cicinnati, and C. Broelsch, Role of radiolabeled
octreotide therapy in patients with metastatic neuroendocrine neoplasms,
Exp. Rev. Endocrinology Metabolism 2, 517527 2007.
5
G. L. DeNardo, A. Natarajan, S. Hok, J. Perkins, M. Cosman, S. J. DeNardo, F. C. Lightstone, G. L. Mirick, L. A. Miers, and R. L. Balhorn,
Pharmacokinetic characterization in xenografted mice of a series of firstgeneration mimics for HLA-DR antibody, Lym-1, as carrier molecules to
image and treat lymphoma, J. Nucl. Med. 48, 13381347 2007.
6
B. J. Hicke, A. W. Stephens, T. Gould, Y.-F. Chang, C. K. Lynott, J. Heil,
S. Borkowski, C.-S. Hilger, G. Cook, S. Warren, and P. G. Schmidt,
Tumor targeting by an aptamer, J. Nucl. Med. 47, 668678 2006.
7
T. A. Waldmann, Monoclonal antibodies in diagnosis and therapy, Science 252, 16571662 1991.
8
J. W. Park, Liposome-based drug delivery in breast cancer treatment,
Breast Cancer Res. 4, 9599 2002.
9
J. F. Kukowska-Latallo, K. A. Candido, Z. Cao, S. S. Nigavekar, I. J.
Majoros, T. P. Thomas, L. P. Balogh, M. K. Khan, and J. R. Baker, Jr.,
Nanoparticle targeting of anticancer drug improves therapeutic response
in animal model of human epithelial cancer, Cancer Res. 65, 53175324
2005.
10
G. Liu, S. Dou, G. Mardirossian, J. He, S. Zhang, X. Liu, M. Rusckowski,
and D. J. Hnatowich, Successful radiotherapy of tumor in pretargeted
mice by 188Re-radiolabeled phosphorodiamidate morpholino oligomer,
as synthetic DNA analogue, Clin. Cancer Res. 12, 49584964 2006.
11
R. J. Lewandowski, K. G. Thurston, J. E. Goin, C. Y. Wong, V. L. Gates,
M. Van Buskirk, J. F. Geschwind, and R. Salem, 90Y microsphere
TheraSphere treatment for unresectable colorectal cancer metastases of
the liver: Response to treatment at targeted absorbed doses of 135
150 Gy as measured by 18F fluorodeoxyglucose positron emission tomography and computer tomographic imaging, J. Vasc. Interv. Radiol.
16, 16411651 2005.
12
A. S. Kennedy, D. Coldwell, C. Nutting, R. Murthy, D. E. Wertman, Jr.,
S. P. Loehr, C. Overton, S. Meranze, J. Niedzwecki, and S. Sailer, Resin
90
Y-microsphere brachytherapy for unresectable colorectal liver metastases: Modern USA experience, Int. J. Radiat. Oncol., Biol., Phys. 65,
412425 2006.
13
V. Kenanova, T. Olafsen, L. E. Williams, N. H. Ruel, J. Longmate, P. J.
Yazaki, J. E. Shively, D. Colcher, A. A. Raubitschek, and A. M. Wu,
Radioiodinated versus radiometal-labeled anti-carcinoembryonic antigen
single-chain Fv-Fc antibody fragments: Optimal pharmacokinetics for
therapy, Cancer Res. 67, 718726 2007.
14
G. A. Wiseman, C. A. White, R. B. Sparks, W. D. Erwin, D. A. Podoloff,
D. Lamonica, N. L. Bartlett, J. A. Parker, W. L. Dunn, S. M. Spies, R.
Belanger, T. L. Witzig, and B. R. Leigh, Biodistribution and dosimetry
results from a phase III prospectively randomized controlled trial of Zevalin radioimmunotherapy for low-grade, follicular, or transformed B-cell
non-Hodgkins lymphoma, Crit. Rev. Oncol. Hematol. 39, 181194
2001.
15
J. M. Vose, R. L. Wahl, M. Saleh, A. Z. Rohatiner, S. J. Knox, J. A.
Radford, A. D. Zelenetz, G. F. Tidmarsh, R. J. Stagg, and M. S. Kaminski, Multicenter phase II study of iodine I-131 tositumomab for
chemotherapy-relapsed/refractory low-grade and transformed low-grade
B-cell non-Hodgkins lymphoma, J. Clin. Oncol. 18, 13161323 2000.
16
J. Y. C. Wong, S. Shibata, L. E. Williams, C. S. Kwok, A. Liu, D. Z. Chu,
D. M. Yamauchi, S. Wilcznski, D. N. Ikle, A. M. Wu, P. J. Yazaki, J. E.
Shively, J. H. Doroshow, and A. A. Raubitschek, A phase I trial of
90Y-anti-carcinoembryonic antigen chimeric T84.66 radioimmunotherapy
with 5-fluorouracil in patients with metastatic colorectal cancer, Clin.
Cancer Res. 9, 58425852 2003.
17
B. Emami, J. Lyman, A. Brown, L. Coia, M. Goitein, J. E. Munzenrider,
B. Shank, L. J. Solin, and M. Wesson, Tolerance of normal tissue to
therapeutic irradiation, Int. J. Radiat. Oncol., Biol., Phys. 21, 109122
1991.
18
G. L. DeNardo, S. J. DeNardo, S. Shen, D. A. DeNardo, G. R. Mirick, D.
3067
J. Macey, and K. R. Lamborn, Factors affecting 131-I-Lym-1 pharmacokinetics and radiation dosimetry in patients with non-Hodgkins lymphoma and chronic lymphocytic leukemia, J. Nucl. Med. 40, 13171326
1999.
19
J. A. Siegel, B. W. Wessels, E. E. Watson, M. G. Stabin, H. M. Vriesendorp, E. W. Bradley, C. C. Badger, A. B. Brill, C. S. Kwok, D. R. Stickney, K. F. Eckermann, D. R. Fisher, D. J. Buchsbaum, and S. E. Order,
Bone marrow dosimetry and toxicity in radioimmunotherapy, Antibody, Immunoconjugates, Radiopharm. 3, 213233 1990.
20
M. G. Stabin, MIRDOSE: Personal computer software for internal absorbed dose assessment in nuclear medicine, J. Nucl. Med. 37, 538546
1996.
21
M. G. Stabin, R. B. Sparks, and E. Crowe, OLINDA/EXM: The secondgeneration personal computer software for internal absorbed dose assessment in nuclear medicine, J. Nucl. Med. 46, 10231027 2005.
22
L. E. Williams, A. Liu, A. A. Raubitschek, and J. Y. C. Wong, A method
for patient-specific absorbed dose estimation for internal beta emitters,
Clin. Cancer Res. 5, 3015s3019s 1999.
23
G. Akabani, W. G. Hawkins, M. B. Eckblade, and P. K. Leichner,
Patient-specific dosimetry using quantitative SPECT imaging and threedimensional discrete Fourier transform convolution, J. Nucl. Med. 38,
308314 1997.
24
C. H. Holdsworth, M. Dahlbom, A. Liu, L. Williams, C. S. Levin, M.
Janecek, and E. J. Hoffman, Expanding the versatility of a more accurate
accelerated Monte Carlo simulation for 3D PET: Data correction of PET
emission scans using 124-I, Nuclear Science Symposium Conference
Record 2001 IEEE 2002, Vol. 4, pp. 21052109.
25
D. A. Goodwin, Tumor pretargeting: Almost the bottom line, J. Nucl.
Med. 36, 876879 1995.
26
D. M. Goldenberg and R. M. Sharkey, Novel radiolabeled antibody conjugates, Oncogene 26, 37343744 2007.
27
O. Linden, J. Kurkus, M. Garkavij, E. Cavallin-Stahl, M. Ljungberg, R.
Nilsson, T. Ohlsson, B. Sandberg, S.-E. Strand, and J. Tennvall, A novel
platform for radioimmunotherapy: Extracorporeal depletion of biotinylated and 90Y-labeled rituximab in patients with refractory B-cell lymphoma, Cancer Biother. Radiopharm. 20, 457466 2005.
28
S. Welt, G. Ritter, C. Williams, Jr., L. S. Cohen, A. Jungbluth, E. A.
Richards, L. J. Old, and N. E. Kemeny, Preliminary report of a phase I
study of combination chemotherapy and humanized A33 antibody immunotherapy in patients with advanced colorectal cancer, Clin. Cancer Res.
9, 13471353 2003.
29
L. E. Williams, A. Liu, D. M. Yamauchi, G. Lopatin, A. A. Raubitschek,
and J. Y. C. Wong, The two types of correction of absorbed dose estimates for internal emitters, Cancer 94, 12311234 2002.
30
D. J. Macey, L. E. Williams, H. B. Breitz, A. Liu, T. K. Johnson, and P.
B. Zanzonico, A Primer for Radioimmunotherapy and Radionuclide
Therapy AAPM Report No. 71 Medical Physics, Madison, WI, 2001.
31
K. F. Koral, Y. Dewaraja, L. A. Clarke, J. Li, R. Zasadny, S. G. Rommelfanger, I. R. Francis, M. S. Kaminski, and R. L. Wahl, Tumor-absorbed
dose estimates versus response in tositumomab therapy of previously untreated patients with follicular non-Hodgkins Lymphoma: Preliminary
report, Cancer Biother. Radiopharm. 15, 347355 2000.
32
C. Hindorf, O. Linden, L. Stenberg, J. Tennvall, and S.-E. Strand,
Change in tumor-absorbed dose due to decrease in mass during fractionated radioimmunotherapy in lymphoma patients, Clin. Cancer Res. 1,
4003S4006S 2003.
33
L. E. Williams, R. B. Duda, R. T. Proffitt, B. G. Beatty, J. D. Beatty, J. Y.
C. Wong, J. E. Shively, and R. J. Paxton, Tumor uptake as a function of
tumor mass: A mathematic model, J. Nucl. Med. 29, 103109 1988.
34
D. J. Macey, S. J. DeNardo, G. DeNardo, J. L. Goodnight, and M. W.
Unger, Uptake of In-111 labeled monoclonal antibody ZME018 as a
function of tumor size in a patient with melanoma, Am. J. Physiol.
Imaging 3, 16 1988.
35
G. Riethmuller, E. Schneider-Gadicke, G. Schlimok, W. Schmiegel, R.
Raab, K. Hoffken, R. Gruber, H. Pichlmaier, H. Hirche, R. Pichlmayr, P.
Buggisch, and J. Witte, Randomised trial of monoclonal antibody for
adjuvant therapy of resected DukesC colorectal carcinoma, Lancet 343,
11771183 1994.
36
J. Carlsson, E. Aronsson Forsell, S. O. Hietala, T. Stigbrand, and T.
3068
Tennvall, Tumor therapy with radionuclides: Assessment of progress and

problems, Radiother. Oncol. 66, 107117 2003.
37
Reprinted by permission of the Society of Nuclear Medicine from J. Y. C.
Wong, G. E. Thomas, D. Yamauchi, L. E. Williams, T. L. Odom-Maryon,
3068
A. Liu, J. M. Esteban, M. Neumaier, S. Dresse, A. M. Wu, F. J. Primus,

J. E. Shively, and A. A. Raubitschek, Clinical evaluation of indium-111labeled chimeric anti-CEA monoclonal antibody, J. Nucl. Med. 38,
19511959 1997.
The role of photodynamic therapy PDT physics

Timothy C. Zhua and Jarod C. Finlay
Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania 19104
Received 14 April 2008; revised 7 May 2008; accepted for publication 8 May 2008;
Photodynamic therapy PDT is an emerging treatment modality that employs the photochemical
interaction of three components: light, photosensitizer, and oxygen. Tremendous progress has been
made in the last 2 decades in new technical development of all components as well as understanding of the biophysical mechanism of PDT. The authors will review the current state of art in PDT
research, with an emphasis in PDT physics. They foresee a merge of current separate areas of
research in light production and delivery, PDT dosimetry, multimodality imaging, new photosensitizer development, and PDT biology into interdisciplinary combination of two to three areas. Ultimately, they strongly believe that all these categories of research will be linked to develop an
integrated model for real-time dosimetry and treatment planning based on biological
response. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.2937440
Key words: PDT, spectroscopy, implicit dosimetry, explicit dosimetry, dynamic process
I. INTRODUCTION
II. THE PROBLEMS IN CLINICAL PDT
Photodynamic therapy PDT is an emerging cancer treatment modality based on the interaction of light, a photosensitizing drug, and oxygen.1 PDT has been approved by the
U.S. Food and Drug Administration for the treatment of microinvasive lung cancer, obstructing lung cancer, and obstructing esophageal cancer, as well as for premalignant actinic keratosis and age-related macular degeneration. Studies
have shown some efficacy in the treatment of a variety of
malignant and premalignant conditions including head and
neck cancer,2,3 lung cancer,46 mesothelioma,7 Barretts
esophagus,8,9 prostate,1012 and brain tumors.9,1315 Unlike radiation therapy, PDT uses nonionizing radiation and can be
administered repeatedly without cumulative long-term complications since it does not appear to target DNA.
There has been tremendous progress made in the last 2
decades in new technologies and in understanding of the
basic biophysical mechanisms of PDT. The most important
question to be answered is: What determines PDT efficacy
for a particular patient, photosensitizer, and treatment protocol? Answering this question will require a unified understanding of the interactions of the three basic components:
light, photosensitizer, and tissue oxygenation. We have categorized the current basic research in PDT into five areas: 1
light sources, light transport, and light delivery in tissue; 2
PDT dosimetry; 3 optical and anatomic imaging; 4 new
photosensitizers; and 5 PDT biology. Among these, the development of new photosensitizers and PDT biology is traditionally considered outside of the realm of PDT physics
and will only be briefly described for completeness. All five
areas are linked by a quantitative understanding of the dynamic processes involved in the photochemical interaction
that drives PDT. In the next section, we will describe these
areas separately.
II.A. Modeling the dynamic process of PDT
3127
Photodynamic therapy is inherently a dynamic process.

There are three principal components: photosensitizer, light,
and oxygen, all of which interact on time scales relevant to a
single treatment. The distribution of light is determined by
the light source characteristics and the tissue optical properties. The tissue optical properties, in turn, are influenced by
the concentration of photosensitizer and the concentration
and oxygenation of the blood. The distribution of oxygen is
altered by the photodynamic process, which consumes oxygen. Finally, the distribution of photosensitizer may change
as a result of photobleaching, the photodynamic destruction
of the photosensitizer itself. To account for these interactions, a dynamic model of the photodynamic process is required.
At the most fundamental level, the photodynamic process
depends on the photosensitizer molecule itself. Figure 1
shows the energy level, or Jablonski, diagram for a typical
type-II photosensitizer. The photochemical reaction is initiated by the absorption of a photon of light by a photosensitizer molecule in its ground state S0, promoting it to an
excited state S1. Both this state and the ground state are
spectroscopic singlet states. One essential property of a good
photosensitizer is a high intersystem crossing ISC yield,
i.e., a high probability of transition from S1 to an excited
triplet state T1. In the T1 state, the photosensitizer can
transfer energy to molecular oxygen 3O2, exciting it to its
highly reactive singlet state 1O2. The details of this energy
transfer process are beyond the scope of this article but have
been an area of active study.16
Two approaches have been used to study PDT dynamics.
First, a microscopic model takes into account diffusion of
oxygen and photosensitizer from blood vessels and can determine the singlet oxygen concentration in cells microscopically. Foster et al. were the first to propose such a quantita-
0094-2405/2008/357/3127/10/$23.00
3127
3128
T. C. Zhu and J. C. Finlay: PDT physics

Energy
transfer to
oxygen
ISC
S1
3128
T1
1
Photon
absorption
S0
O2
Sensitizer
O2
Oxygen
FIG. 1. Energy level diagram for a typical type II photosensitizer and oxygen. The sensitizer in its ground state S0 absorbs a photon of light and is
excited to its first singlet state S1. It spontaneously decays to its excited
triplet state T1 via ISC. From T1, energy is transferred to ground state
molecular oxygen 3O2, creating reactive singlet oxygen 1O2.
tive model and to verify its results in multicell tumor

spheroid models.17,18 These models demonstrate that the
dominant effect of the fluence rate in photodynamic therapy
arises because the photochemical process itself consumes
oxygen. If the rate of photochemical oxygen consumption is
greater than the rate at which oxygen can be resupplied by
the vasculature or ambient medium, an induction of transient
hypoxia by PDT can result. This effect has been modeled
theoretically and has been demonstrated in cell spheroids,17
animals,19 and human tissues,20 and continues to be an area
of active research.21 A second, macroscopic model has also
been developed.22,23 This is an empirical model that does not
take into account the actual oxygen and photosensitizer diffusion processes microscopically but instead approximates
them with simpler functions. It does, however, explicitly account for the larger-scale spatial variation in fluence rate
based on the diffusion approximation. This model provides a
quantity reacted singlet oxygen that can be used directly for
clinical PDT dosimetry, and that relates directly to the threedimensional distribution of photosensitizers, light fluence
rate, and a mean tissue oxygenation distribution.22,23
II.B. Current areas of basic PDT research
II.B.1. Light source, transport, and delivery

PDT became popular after the invention of the laser,
which allowed the production of monochromatic light that
could be easily coupled into optical fibers. The wavelength
of light used for PDT is typically in the wavelength range
between 600800 nm, the so called therapeutic window.24
In this wavelength range, the energy of each photon h is
high enough 1.5 eV to excite the photosensitizer and yet
is low enough so that the light has sufficient penetration into
the tissue. Early lasers were based on either argon gas lasers
488 and 514.5 nm or frequency doubled Nd:yttriumaluminun-garnet solid state lasers 532 nm, which were
then used to pump a dye laser to produce light in the desired
wavelength range. With the development of the diode laser,
the laser source has become portable and a turn-key operaMedical Physics, Vol. 35, No. 7, July 2008
tion. Development of more powerful and cheaper laser

sources, e.g., fiber lasers, which couple more efficiently into
optical fiber, is an active area of research.
The invention of optical fiber allowed light to be directed
easily to deliver irradiation to desired regions without the
requirement of a straight light path and is another enabler of
PDT. Currently, most PDT procedures are performed with
optical fibers. By attaching diffuse scattering tips of various
geometrical shapes at the exit end of the fiber, point, linear,
and planar light sources can be produced.25 Further development in light delivery devices that produce a light field with
various geometrical shapes and in power distribution modulation that covers a larger area and has higher power is still
an active area of research.
Light distributions can be modeled using the diffusion
approximation to radiative transport.26 The finite-element
method is commonly used to solve the diffusion equation27
in an optically heterogeneous medium with arbitrary geometry. Various boundary conditions are used to describe various tissue-tissue, tissue-air, and tissue-water interfaces.26,28
However, the solution is often inaccurate in regions without
a sufficient number of multiple scattering near the light
source or nonscattering medium. In these regions Monte
Carlo simulation provides more accurate results but is much
slower.29 Active research is on going to solve the Boltzmann
equation of light transport directly to provide accurate result
while improve the speed of calculation.26,30
II.B.2. PDT dosimetry
Three general strategies have been developed for dosimetry based on the cumulative dose of singlet oxygen, which
is presumed to be predictive of tissue damage. Explicit dosimetry refers to the prediction of singlet oxygen dose on the
basis of measurable quantities that contribute to the photodynamic effect.31 The quantities of interest are typically the
distributions of light, photosensitizer, and oxygen. The distributions of photosensitizer and oxygenation can be measured via optical spectroscopy, as described above. In current
clinical practice, however, the quantity most straightforward
to measure is the light dose. Flat photodiode detectors have
been used to measure the incident irradiance at the tissue
surface in intraoperative PDT.32 These measurements, however, may not accurately reflect the fluence rate in the tissue
itself because they neglect contribution of backscattered
light. Detectors based on optical fibers overcome this problem by collecting light isotropically.33 The effect on the measured fluence rate is significant.34 Because these probes collect light via multiple scattering, the interface between the
scattering tip and the surrounding medium can change the
sensitivity of the detector by as much as a factor of 2, requiring careful calibration.35
Because complete explicit dosimetry requires measurement of three different parameters, it is inherently challenging. Two alternatives have been suggested that require measurement of only a single parameter. Direct dosimetry relies
on the detection of singlet oxygen itself, either through its
own phosphorescence emission or via singlet-oxygen-
3129
sensitive chromophores. Implicit dosimetry31 uses a quantity such as fluorescence photobleaching of the photosensitizer, which is indirectly predictive of the production of
singlet oxygen. Strategies for direct and implicit dosimetry
are under development and will be discussed in later sections.
II.B.3. Anatomic and optical imaging
The most commonly used medical imaging modalities include ultrasound, computer tomography CT, magnetic
resonance, magnetic resonance spectroscopy, single photon
emission computer tomography, and positron emission tomography PET. The first three modalities produce excellent
anatomical images, while the latter three provide functional
information e.g., oxygen perfusion or tissue metabolism at
the expense of image resolution. Diffuse optical tomography
DOT is a viable new biomedical imaging modality.36 This
technique images the absorption and scattering properties of
biological tissues and has been explored as a diagnosis tool
in breast,3742 brain,43,44 and bones and joints.45 Some preliminary attempts have been made to perform DOT for
prostate.4649 DOT at the treatment wavelength can be used
as input to calculate light fluence rate distribution for PDT.
In addition, it provides access to a variety of physiological
parameters that cannot otherwise be measured easily. Another modality, optical coherence tomography OCT uses
coherent light to obtain high resolution images. However,
OCT is rarely used in PDT because of the limitation of penetration depth 1 mm.50
Most image reconstruction of DOT is based on solving
the inverse problem for the diffusion equation.27 This is an
ill-posed problem because of the strong scattering in the turbid medium. As a result, the image resolution of DOT is
limited when compared with other imaging modalities, such
as magnetic resonance imaging MRI or CT. The use of
spectroscopic information and/or a priori anatomic information is common to provide additional constraint to produce
reliable DOT reconstruction. Interested readers can find more
information from some excellent review articles.51,52
The concept of absorption and fluorescence spectroscopy
as a modality for the diagnosis of disease dates back several
decades.53 In recent years, the potential for spectroscopy to
diagnose cancer54 and to monitor the progress of cancer
treatment55 has been increasingly appreciated. In addition to
its role in diagnosis, spectroscopy is particularly applicable
to PDT in determining the local drug and oxygen
concentrations.56
Photodynamic therapy can cause changes in the concentration and oxygenation of blood in tissue both directly,
through photochemical oxygen consumption, and indirectly,
through effects on the vasculature and general physiological
responses. The monitoring of these responses may, therefore,
be predictive of treatment outcome. For oxygen monitoring,
the difference in absorption spectra between oxy- and deoxyhemoglobin is used to determine the hemoglobin oxygen
saturation, i.e., the fraction of total hemoglobin that is in its
oxygenated state. This quantity can be related to the oxygen
3129
concentration in the blood using the Hill curve.57 This measurement does not directly measure the concentration of oxygen in the tissue itself, however, it is possible to model the
relationship between the vascular and tissue oxygenation.
This concept has been investigated in animal models in
which the blood flow and/or oxygenation were monitored
and changes correlated with outcome.58,59
II.B.4. New photosensitizers
Various photosensitizing drugs have been developed. The
first-generation photosensitizer, hematoporphyrin derivative,
is a mixture of porphyrin monomers and oligomers that is
partially purified to produce the commercially available
product, porfimer sodium, marketed under the tradename
Photofrin. Photofrin was approved for treatment of early
stage lung cancer in 1998 and for Barretts esophagus in
2003. The clinical applicability of Photofrin has been limited by two factors. First, its absorption peak occurs at too
short a wavelength 630 nm to allow deep penetration in
tissue. Second, administration of Photofrin results in cutaneous photosensitivity lasting up to 6 weeks. These limitations have inspired the development of a second generation
of photosensitizers with longer-wavelength absorption peaks
and more rapid clearance from skin. Among these was benzoporphyrin derivative monoacid A BPD-MA, or verteporfin. In preclinical trials, it was observed that verteporfin preferentially targeted neovasculature. This selectivity has been
exploited for the treatment of choroidal neovascularization
CNV, an abnormal growth of vessels in the retina associated with age-related macular degeneration AMD, the leading cause of blindness in the developed world. Verteporfin
was approved in the U.S. under the tradename Visudyne for
CNV treatment in 2000. Tetra m-hydroxyphenyl chlorin
mTHPC, Foscan is another second generation photosensitizer, a pure synthetic chlorin compound, which is activated
by 652 nm light.60 The major advantages of mTHPC are a
short duration of skin photosensitivity 15 days, a high
quantum yield for singlet oxygen, and depth of tumor necrosis of up to 10 mm in preclinical models.61 mTHPC has been
used for treatment of pleural mesothelioma,7 head-and-neck
cancers,62,63 esophagus,64,65 prostate,10,66 pancreas,67 arthritic
joints,68 and skin cancers69 and was approved in Europe for
PDT of head-and-neck and varieties of other tumors in 2001.
Another development of note is the prodrug
-aminolevulinic acid ALA. Unlike other PDT drugs, ALA
itself is not a photosensitizer. When taken up by cells, however, it is converted by a naturally occurring biosynthetic
process into the photosensitizer protoporphyrin IX PpIX.
ALA can be applied topically, and was approved by the FDA
in 1999 for the treatment of actinic keratosis AK. Table I
summaries several of the more widely used photosensitizers
currently available.
II.B.5. PDT biology
From the point of view of biological response, PDT is
fundamentally different from other cancer therapies. Unlike
ionizing radiation, PDT achieves its cytotoxic effects prima-
3130
3130
TABLE I. An incomplete list of photosensitizers currently undergoing human clinical trials.
Photosensitizer
Porfimer sodium
ALA-PpIX
Methyl aminolevulate-PpIX
Hexyl aminolevulate-PpIX
BPD-MA
mTHPC
Trade name
Approval
Excitation Drug-light
nm
interval
Clearance
time
Photofrin
Levulan Keratastick
Metvix
Hexvix
Verteporfin, Visudyne
Foscan
1998, 2003 FDA

1999 FDA
2004 FDA
2005 EU
2000 FDA
Phase I trials, 2001 EU
630
405, 635
405, 635
405
689
652
48150 h
1418 h
3h
13 h
15 min
48110 h
46 weeks
2 days
2 days
2 days
5 days
15 days
Phase I trials
Phase I trials
Phase I and II trials
732
762
664
3h
30 min
1h
2 h
Phase I trials
672
2436 h
Motexafin Lutetium
MLu, Lutex, Lutrin
Pd-bacteriopheophorbide
Tookad
Taloporfin sodium
LS11
mono-L-aspartyl chlorin e6
Silicon pthalocyanine 4
PC-4
rily though damage to targets other than DNA. The specific

subcellular targets damaged by PDT depend on the photosensitizers localization within the cell, which varies among
photosensitizers and cell lines. Different types of damage can
lead to different mechanisms of cell death. Damage to mitochondria in particular can lead to apoptosis even at relatively
low light doses.70 Recently, the role of autophagic cell death
in PDT has been increasingly recognized.71 In addition to the
variation among photosensitizers in their subcellular targets,
there is considerable variability in the macroscopic targeting
of photosensitizers. Photosensitizers that are retained in the
vasculature can destroy tumors via vascular damage rather
than direct cell killing. Some photosensitizers may act as
vascular agents at short times after injection and at high fluence rates, where only the vasculature is sufficiently oxygenated, and produce direct cell kill at low fluence rates and
Light Delivery
Dosimetry
Light Dosimetry
with Feedback
Complex
Dosimetry with
Feedback,
Real-time TPS
Lung, Barretts esophagus

AK
AK
Detection of bladder tumors
CNV
Head and neck, prostate, pancreas,
esophagus, mesothelioma
Prostate, atherosclerosis
Prostate
CNV, liver and colorectal metastasis
Skin
long times after injection.72 An additional level of complexity arises from the fact that the response to PDT is not confined to the cells where the singlet oxygen is deposited but
can involve physiological73 and immunological74,75 responses as well.
III. CURRENT DEVELOPMENTS
Much of the research in the early decades of PDT and its
related fields proceeded in five almost independent areas, as
illustrated in the top row of Fig. 2. The problems associated
with light source and light delivery system development, dosimetry, and optical imaging were treated as physics problems, while photosensitizer development and PDT biology
were treated as problems of chemistry and biology, respectively. In recent years, however, the most promising ad-
Photosensitizer
Development
Imaging
Explicit and,
Implicit
Dosimetry,
SOLD
Sites
Biomarkers
Molecular
Beacons
PMB, doseresponsive
imaging
PDT
Biology
Targeted
Photosensitizer,
Photoimmunotherapy
Biological
Response
Imaging
Real-time dosimetry
and treatment planning
based on biological
response
FIG. 2. Diagram illustrating the progress of PDT development from a set of disparate fields top row to a collaborative effort unifying the contributions of
biologists, chemists, physicists, and engineers. The second row illustrates the current state of art of research and represents integration of two separate fields.
The third row illustrates the future research direction and the fourth row is the ultimate integration of all disparate fields. See text for a complete description.
3131
vances have come out of interdisciplinary collaborations

among these areas. The systems and strategies currently in
preclinical and clinical trials are examples of such
collaborations.
III.A. Light dosimetry with real-time feedback
It is very difficult to assess the efficacy and toxicity of any

treatment in a clinical trial without a rigorous quantitative
measurement of the treatment given. One reason that radiotherapy has become an accepted locoregional therapy is that
the dose to tissues has been quantified accurately and curative doses can be delivered safely. In our current clinical
trials, we deliver a dose of light to the treatment sites based
upon measurements made from implanted isotropic light
detectors.34,76 These detectors allow us to prescribe the light
dose based upon actual measurements in the prostate rather
than the output of the laser. This dosimetric approach also
ensures that the same amount of light energy as measured
by the isotropic detectors is deposited uniformly in the prostate within each cohort of patients as a specific dose level.
The measurement of light fluence in vivo is necessary but
not sufficient to optimize the light fluence distribution. Optimization of the light fluence depends upon accurate calculation of the light fluence rate in the entire prostate volume,
which in turn requires accurate characterization of the in vivo
prostate optical properties as input. Many studies have been
conducted to determine the in vivo optical properties of human prostate.7779 These light fluence measurements show
that the optical properties of the prostate vary substantially
within a prostate gland as well as among separate prostate
glands. The optical properties of prostate tissue may also
vary over time during a PDT procedure.
The state of the art for prostate PDT uses only in vivo
light fluence rate monitoring to determine the light fluence
using equal weight of linear light sources. This is not sufficient since light fluence at only selected points are known.
Integrating ultrasound imaging and the PDT dose calculation
engine into the PDT delivery system, with input of tissue
optical properties and drug uptake, will allow the clinician to
optimize the weight of each light-delivery fiber, and thus
improve the light fluence distribution in the prostate gland.
Adequate treatment planning often uses various optimization
engines to optimize the distribution of light source
intensities.
III.B. Implicit dosimetry and dosimetric imaging
The most commonly cited example of implicit dosimetry

is fluorescence photobleaching. Wilson31 suggested that the
photobleaching of the photosensitizer, if moderated by the
same mechanism as the photodynamic effect, could indicate
the extent of damage. The theoretical basis for quantitative
implicit dosimetry using fluorescence photobleaching was
formalized by Georgakoudi et al.18 for the case of multicell
tumor spheroids. Fluorescence photobleaching has been used
in animal models to investigate the fluence-rate dependence
of photodynamic therapy8082 and has been shown to correlate with visible skin damage.83 A device for skin PDT doMedical Physics, Vol. 35, No. 7, July 2008
3131
simetry using photobleaching has been developed for use in

clinical trials84 and work to extend the spheroid models to
realistic tissues is ongoing.85
Fluorescence photobleaching is relatively inexpensive and
straightforward to implement, however, its quantitative interpretation involves several challenges. First, fluorescence
measurement in vivo must always account for the confounding effects of light scattering and absorption. Second, the
validity of the relationship between fluorescence photobleaching and reacted singlet oxygen dose assumes that the
photobleaching is mediated primarily by singlet oxygen reaction with the photosensitizer. There is experimental evidence to suggest that other bleaching mechanisms may be
important for the commonly used photosensitizers
Photofrin82 and ALA-induced PpIX,86 and the in vivo photobleaching of mTHPC exhibits features that are not readily
interpretable.81 Care should therefore be exercised in using
photobleaching as a dose metric.
Two methods have been developed for the detection of
singlet oxygen itself. Qin et al. have demonstrated a
chemoluminescent compound that emits light in the presence
of singlet oxygen.87 The translation of this technology into
patients will require the development of a nontoxic, singletoxygen-specific chemoluminescent probe that distributes
uniformly in tissue.
Direct measurement of singlet oxygen can be accomplished via detection of its 1270 nm phosphorescence emission, a technique known as singlet oxygen luminescence dosimetry SOLD.88 Detection of this signal presents
significant technical challenges due to the long wavelength
of the emission and the weak signal. Jarvi et al.88 have developed a system based on a photomultiplier tube capable of
detecting the long-wavelength emission of singlet oxygen.
The emission is stimulated by a pulsed laser at the treatment
wavelength, and a time-resolved measurement of the emitted
signal allows determination of the singlet oxygen lifetime in
the system being measured. An additional challenge is the
interpretation of the singlet oxygen phosphorescence signal.
The phosphorescence emission arises from a transition of
oxygen from its singlet state to its ground state. This process
therefore competes with the reaction of singlet oxygen with
cellular and extracellular substrates. This leads to an increase
in phosphorescence emission in conditions where substrates
for singlet oxygen are not plentiful. To differentiate between
the cellular signal corresponding to PDT effect and the extracellular signal, the time-resolved emission signal can be fit
as a sum of signals with different lifetimes.
III.C. Biomarkers and molecular beacons
Because the goal of PDT is ultimately to deliver the optimal treatment to the tumor, the idea of using noninvasive
detection and imaging to determine the status of tumor cells
at the molecular level is very attractive. This concept has
driven the development of molecular beacons, molecules that
target specific molecular pathways that can be imaged using,
for instance, near-infrared fluorescence. Steflova et al. have
developed a series of molecular beacons whose fluorescence
3132
is quenched in their latent state and restored under the action

of specific cellular processes of interest.89 Because these beacons target specific molecules, it is likely that they will be
tumor-type specific, however, the variety of cellular processes that can potentially be imaged using this approach
makes it very promising. Molecular beacons are a subset of
the general category of cancer biomarkers, molecules which
allow imaging of specific biological processes via various
modalities including MRI, PET,90 and optical imaging.
III.D. Enhancement and targeting of photosensitizers
Several researchers have developed methods for enhancing the ability of a photosensitizer to target tumors. One
strategy to accomplish this is to optimize the timing the administration of light to coincide with the desired distribution
of the photosensitizer.72 This approach has been used to target vasculature by applying light while the photosensitizer is
in circulation, for the treatment of AMD using Vertepofin and
the prostate using Tookad.12 In cases where the rate of production of singlet oxygen is limited by the vascular resupply
of oxygen, changes in the light fluence rate can change the
distribution of deposited singlet oxygen, with higher fluence
rates leading to preferential targeting of vascular-adjacent tissue. Thus, the combination of drug-light interval and fluence
rate can be adjusted to enhance the treatment of the desired
tissue type.
A more direct approach to photosensitizer targeting is the
conjugation of the photosensitizer to a tumor-selective molecule or particle.91 The last 2 decades have seen tremendous
progress in the development of tumor-targeted therapeutic
and imaging agents for cancer in general.92,93 These innovations will continue to inform the development of new photosensitizers targeted to specific tumor types.
Recent research has shown PDT induced increases of
VEGF, MMPs, and/or COX-2 in tumor microenvironment
that cause resistance to standard treatment. As a result, combination of therapy that suppresses these growth factors has
been proposed to enhance the therapeutic efficacy of PDT.
C225, a monoclonal antibody that inhibits the receptor tyrosine kinase activity of EGFR, has been shown to enhance
the PDT treatment in ovarian cancer.94,95 Other combined
modalities seek to target treatment-induced angiogenesis
and/or inflammation to enhance the effectiveness of PDT.96
IV. FUTURE DEVELOPMENTS AND
PREDICTIONS
We predict that future developments in PDT will continue
the trend toward interdisciplinary work and the inclusion of
more technologies and subfields, such as imaging, novel
drug design, and biological modeling into the treatment planning and dosimetry processes. Below are a few predictions of
future PDT physics research directions:
IV.A. Light sources: Faster, cheaper, larger, smaller
A significant driver in the development of new light

sources has been the advent of inexpensive, reliable diode
3132
lasers and light-emitting diodes. Diode light sources have

already been made small enough to be implantable.97 We
predict that this trend will continue, with future light sources
becoming ever smaller, cheaper, and more efficient. On the
other hand, the increases in efficiency of these light sources
will also lead to lasers with greater total power output, allowing the treatment of larger volumes of tissue using a
single source.
Simple increases in power are not the only advances we
can expect from future light sources. New delivery systems
will incorporate mechanisms to deliver a light distribution
customizable to each patients treatment plan. In our group,
we have developed a computer-controlled attenuator system
for prostate PDT designed to adjust the intensities of up to 16
implanted interstitial diffusing fibers independently and in
real time.98 Lilge et al. have investigated optical diffusing
fibers whose emission profiles along the fiber axis can be
customized to deliver much more complex light distributions
than conventional diffusers.99,100 We expect further developments in this direction to allow more and more precise control of the in vivo light distribution during PDT.
IV.B. Real-time oxygen monitoring
While it has long been recognized that oxygenation of the

target tissue is essential to PDT, measurements of oxygenation are now starting to be implemented in preclinical and
clinical trials. We anticipate that future clinical protocols involving real-time dosimetry will also include real-time oxygen monitoring, both to prevent PDT-induced hypoxia and to
take advantage of the predictive value of PDT-induced
changes in blood flow and hemoglobin oxygen saturation.58
The monitoring of hemoglobin saturation and blood flow has
been demonstrated in clinical trials.101 The translation of
measurements of hemoglobin oxygenation into a determination of the cellular oxygen concentration requires a model of
oxygen diffusion and consumption. The development of such
models is ongoing and will inform future dosimetry
developments.85
An alternative to hemoglobin monitoring is the use of
time-resolved SOLD measurements. A sufficiently sophisticated analysis of the time dependence of the SOLD signal
allows determination of the photosensitizer triplet lifetime,
which is related to the local oxygen concentration. Hence,
SOLD has the potential to provide both a measure of reacted
singlet oxygen dose and a measure of the molecular oxygen
concentration at the point of singlet oxygen generation.88
This approach has the additional advantage from the point of
view of explicit dosimetry of providing an intermediate verification of the triplet lifetime predicted by the dosimetric
model.
IV.C. Multifunctional, multimodality photosensitizers
Another area of active research in photodynamic therapy

is the development of photosensitizers and photosensitizer
delivery systems that enhance the functionality of the
photosensitizer.102 For instance, a photosensitizer may be
linked to a fluorescent molecule to enhance its delectability
3133
under fluorescence imaging.103 Pandey et al.104 have proposed combining photosensitizers with contrast agents for
optical, MRI and PET imaging, allowing image-guided PDT
treatment. Zheng et al. have combined the molecular beacon concept with a photosensitizer to produce photodynamic molecular beacons.105 These molecules exhibit
quenching that suppresses their production of 1O2 in their
latent state. They can be activated by interaction with a
tumor-specific molecular marker, in this case the matrix metalloproteinase MMP-7. The clinical implementation of strategies such as these has the potential to dramatically improve
the targeting of the photosensitizer to the tumor. An essential
component of any targeting strategy will be the ability to
verify the targeting of the drug to the desired target. The
distribution of the photosensitizer can be determined optically using fluorescence or absorption imaging. In addition,
detailed models of the distributions and kinetics of light,
photosensitizer, and oxygen will provide a quantitative relationship between the microscopic distribution and photochemical properties of the drug and the macroscopically observed treatment response.106 In cases where the drug is
activated by an endogenous agent, the results will be predictable only using such sophisticated models.
IV.D. Treatment planning integrated with dosimetry,
imaging, and light delivery devices to allow
adaptive treatment
The integration of dosimetry systems with multimodality

imaging anatomic and DOT and light delivery devices into
an integrated system has drawn great interest recently. Several groups are developing integrated systems to control light
delivery, real-time light monitoring, and volumetric light fluence calculation. The group in Lund, Sweden is developing a
multiple channel system of point sources that can be used as
either light sources or detectors.107 A personal computer controls the position and duration of the point sources to optimize light fluence distribution. Another group in Toronto,
Canada has commercialized a four-channel light source with
computerized power control in each channel. Our group has
developed an integrated system that incorporates PDT dosimetry that include light fluence rate at multiple points,108 a
three-dimensional map of optical properties,46 drug concentration distribution,56 and tissue oxygenation.98 A kernelbased algorithm was developed to calculate light fluence rate
distribution in the optically heterogeneous prostate gland,109
which, coupled with an optimization engine allows optimization of light source powers to achieve a uniform light fluence distribution.110,98 This area of research will evolve into
a totally computerized delivery, monitoring, and dosimetry
system for real-time feedback control.
IV.E. New dosimetry quantities based on modeling of
the dynamic system
The current state of art in PDT dosimetry is to explicitly

determine the quantity PDT dose, defined as the number of
photons absorbed by photosensitizing drug per gram of
tissue.111 While this quantity takes into account the consumpMedical Physics, Vol. 35, No. 7, July 2008
3133
tion of photosensitizer, it does not consider the effect of tissue oxygenation on the quantum yield of oxidative radicals.
Thus, it is only applicable in cases where ample oxygen supplies exist. It is anticipated that new dosimetry quantities
based on reacted singlet oxygen 1O2rx, e.g., a product of
singlet oxygen quantum yield and PDT dose, can be used to
account for kinetics of the oxygen consumption during PDT
process.23 Foster et al. have determined most of fundamental
parameters necessary for microscopic model of the oxygen
consumption during PDT.18,82,106,112 However, when applying models developed in the microscopic scale to a macroscopic environment, the values of many parameters may
change, and many parameters may be observable only in the
volume average, so extensive study will be required to determine the values for each specific photosensitizer.
IV.F. Physiological effects of PDT will be exploited to
develop systemic therapy
The ability of PDT to elicit an immune response has been

recognized for some time.75 The past decade has seen dramatic progress in our understanding of the mechanisms of
PDT-induced immune response.74,113 In addition it has been
demonstrated that PDT can be used to generate vaccines
against specific tumor types.114,115 This strategy is particularly attractive for the treatment of cancer, where it is likely
that PDT will be used in combination with radiation and
chemotherapy, both of which can have immunosuppressive
effects.
As the details of the mechanisms of the immune response
are better understood, it is likely that new photosensitizers
and delivery mechanisms will be developed for the express
purpose of enhancing the immunological response and/or
moderate treatment induced angiogenesis and inflammation.
Furthermore, the parameters which optimize immune response to PDT will be different from those that maximize
singlet oxygen dose or direct cell killing.116 Therefore, there
will be an increasing need for the incorporation of these effects into PDT treatment planning and dosimetry. We predict
that PDT vaccines will find a use as an adjuvant therapy,
even in cases where the tumor location or geometry precludes conventional PDT as a primary treatment. In these
cases, it may be possible to perform PDT on ex vivo tissue or
cell cultures, in which case the problems of light propagation
and tumor physiology are greatly simplified, allowing a level
of treatment optimization not feasible in vivo.
V. CONCLUSIONS
We have predicted that PDT research will increasingly
rely on combinations of widely varied fields, leading to more
and more sophisticated treatment protocols. It is tempting to
assume that such therapies will become so complex that they
can only be modeled empirically, and that the role of physics
will diminish. However, the history of PDT research demonstrates that many advances in the field are made not by
avoiding the complexity of the problem but by embracing it.
As we exploit our growing knowledge of the underlying photochemistry, biology, and physiology of PDT to develop new
3134
and better treatments, it will be more important than ever to

understand and optimize those treatment parameters over
which we have direct control. In the future, as in the past, it
will be the understanding of the physics of PDT that will
allow us to take advantage of our scientific understanding
and translate it into improved clinical treatments.
ACKNOWLEDGMENTS
This work is supported by the National Institute of Health
NIH Grant Nos. P01 CA87971 and R01 CA109456.
a

T. J. Dougherty, C. J. Gomer, B. W. Henderson, G. Jori, D. Kessel, M.
Korbelik, J. Moan, and Q. Peng, Photodynamic therapy, J. Natl. Cancer
Inst. 90, 889905 1998.
2
M. A. Biel, Photodynamic therapy and the treatment of head and neck
cancers, J. Clin. Laser Med. Surg. 14, 239244 1996.
3
W. E. Grant, P. M. Speight, C. Hopper, and S. G. Bown, Photodynamic
therapy: an effective, but non-selective treatment for superficial cancers
of the oral cavity, Int. J. Cancer 71, 937942 1997.
4
J. S. Friedberg, R. Mick, J. P. Steveson, T. Zhu, T. M. Busch, D. Shin, D.
Smith, M. Culligan, A. Dimofte, E. Glatstein, and S. M. Hahn, A phase
II trial of pleural photodynamic therapy PDT and surgery for patients
with non-small cell lung cancer NSCLC with pleural spread, J. Clin.
Oncol. 22, 21922201 2004.
5
T. L. Moskal, T. J. Dougherty, J. D. Urschel, J. G. Antkowiak, A. M.
Regal, D. L. Driscoll, and H. Takita, Operation and photodynamic
therapy for pleural mesothelioma: 6-year follow-up, Ann. Thorac. Surg.
66, 11281133 1998.
6
H. I. Pass et al., Intrapleural photodynamic therapy: Results of a phase I
trial, Ann. Surg. Oncol. 1, 2837 1994.
7
J. S. Friedberg, R. Mick, J. Stevenson, J. Metz, T. Zhu, J. Buyske, D. H.
Sterman, H. I. Pass, E. Glatstein, and S. M. Hahn, A phase I study of
Foscan-mediated photodynamic therapy and surgery in patients with mesothelioma, Ann. Thorac. Surg. 75, 952959 2003.
8
R. Ackroyd, N. J. Brown, M. F. Davis, T. J. Stephenson, S. L. Marcus, C.
J. Stoddard, A. G. Johnson, and M. W. Reed, Photodynamic therapy for
dysplastic Barretts esophagus: A prospective, double blind, randomised,
placebo controlled trial, Gut 47, 612617 2000.
9
M. Panjehpour, B. F. Overholt, J. M. Haydek, and S. G. Lee, Results of
photodynamic therapy for ablation of dysplasia and early cancer in Barretts esophagus and effect of oral steroids on stricture formation, Am. J.
Gastroenterol. 95, 21772184 2000.
10
T. R. Nathan, D. E. Whitelaw, S. C. Chang, W. R. Lees, P. M. Ripley, H.
Payne, L. Jones, M. C. Parkinson, M. Emberton, A. R. Gillams, A. R.
Mundy, and S. G. Browen, Photodynamic therapy for prostate cancer
recurrence after radiotherapy: A phase I study, J. Urol. 168, 14271432
2002.
11
D. Stripp, R. Mick, T. C. Zhu, R. Whittington, D. Smith, A. Dimofte, J. C.
Finlay, J. Miles, T. M. Busch, D. Shin, A. Kachur, Z. Tochner, S. B.
Malkowicz, E. Glatstein, and S. M. Hahn, Phase I trial of Motexfin
Lutetium-mediated interstitial photodynamic therapy in patients with locally recurrent prostate cancer, Proc. SPIE 5315, 8899 2004.
12
R. A. Weersink, A. Bogaards, M. Gertner, S. R. Davidson, K. Zhang, G.
Netchev, J. Trachtenberg, and B. C. Wilson, Techniques for delivery and
monitoring of TOOKAD WST09-mediated photodynamic therapy of
the prostate: clinical experience and practicalities, J. Photochem. Photobiol. B 79, 211222 2005.
13
M. A. Rosenthal, B. Kayar, J. S. Hill, D. J. Morgan, R. L. Nation, S. S.
Stylli, R. L. Basser, S. Uren, H. Geldard, M. D. Green, S. B. Kahl, and A.
H. Kaye, Phase I and pharmacokinetic study of photodynamic therapy
for high-grade gliomas using a noval boronated porphyrin, J. Clin. Oncol. 19, 519524 2001.
14
T. T. Goodell and P. J. Muller, Photodynamic therapy: A novel treatment
for primary brain malignancy, J. Neurosci. Nurs. 33, 296300 2001.
15
A. Bogaards, A. Varma, K. Zhang, D. Zach, S. K. Bisland, E. H.
Moriyama, L. Lilge, P. J. Muller, and B. C. Wilson, Fluorescence imageguided brain tumour resection with adjuvant metronomic photodynamic
therapy: Pre-clinical model and technology development, Photochem.
1
3134
Photobiol. Sci. 4, 438442 2005.
M. J. Paterson, O. Christiansen, F. Jensen, and P. R. Ogilby, Overview of
theoretical and computational methods applied to the oxygen-organic
molecule photosystem, Photochem. Photobiol. 82, 11361160 2006.
17
T. H. Foster, D. F. Hartley, M. G. Nichols, and R. Hilf, Fluence rate
effects in photodynamic therapy of multicell tumor spheroids, Cancer
Res. 53, 12491254 1993.
18
I. Georgakoudi, M. G. Nichols, and T. H. Foster, The mechanism of
Photofrin photobleaching and its consequences for photodynamic dosimetry, Photochem. Photobiol. 65, 135144 1997.
19
S. L. Gibson, K. R. VanDerMeid, R. S. Murant, R. F. Raubertas, and R.
Hilf, Effects of various photoradiation regimens on the antitumor efficacy of photodynamic therapy for R3230AC mammary carcnimoas,
Cancer Res. 50, 72367241 1990.
20
B. W. Henderson, T. M. Busch, L. A. Vaughan, N. P. Frawley, D. Babich,
T. A. Sosa, J. D. Zollo, A. S. Dee, M. T. Cooper, D. A. Bellnier, W. R.
Greco, and A. R. Oseroff, Photofrin photodynamic therapy can significantly deplete or preserve oxygenation in human basal cell carcinomas
during treatment, depending on fluence rate, Cancer Res. 60, 525529
2000.
21
B. W. Henderson, T. M. Busch, and J. W. Snyder, Fluence rate as a
modulator of PDT mechanisms, Lasers Surg. Med. 38, 489493 2006.
22
X. H. Hu, Y. Feng, J. Q. Lu, R. R. Allison, R. E. Cuenca, G. H. Downie,
and C. H. Sibata, Modeling of a type II photofrin-mediated photodynamic therapy process in a heterogeneous tissue phantom, Photochem.
Photobiol. 81, 14601468 2005.
23
T. C. Zhu, J. C. Finlay, X. Zhou, and J. Li, Macroscopic modeling of the
singlet oxygen production during PDT, Proc. SPIE 6427, 642708
6427202007.
24
B. C. Wilson and M. S. Patterson, in Photodynamic Therapy of Neoplastic Disease, edited by D. Kessel CRC Press, Boca Raton, 1990, Vol. 1,
pp. 129144.
25
P. R. Almond, Photodynamic Therapy Equipment and Dosimetry American Institute of Physics, Woodbury, NY, 1992.
26
A. Ishimaru, Wave Propagation and Scattering in Random Media IEEE,
New York, 1997.
27
W. Cong, L. V. Wang, and G. Wang, Formulation of photon diffusion
from spherical bioluminescent sources in an infinite homogeneous medium, Biomed. Eng. Online 3, 12 2004.
28
R. C. Haskell, L. O. Svaasand, T.-T. Tsay, T.-C. Feng, M. S. McAdams,
and B. J. Tromberg, Boundary conditions for the diffusion equation in
radiative transfer, J. Opt. Soc. Am. A 11, 27272741 1994.
29
L. Wang, S. L. Jacques, and L. Zheng, MCML-Monte Carlo modeling of
light transport in multi-layered tissues, Comput. Methods Programs
Biomed. 47, 131146 1995.
30
J. J. Duderstadt and W. R. Martin, Transport Theory Wiley, New York,
1979.
31
B. C. Wilson, M. S. Patterson, and L. Lilge, Implicit and explicit dosimetry in photodynamic therapy: A new paradigm, Lasers Med. Sci. 12,
182199 1997.
32
S. K. Hendren, S. M. Hahn, F. R. Spitz, T. W. Bauer, S. C. Rubin, T. Zhu,
E. Glatstein, and D. L. Fraker, Phase II trial of debulking surgery and
photodynamic therapy for disseminated intraperitoneal tumors, Ann.
Surg. Oncol. 8, 6571 2001.
33
J. P. Marijnissen and W. M. Star, Performance of isotropic light dosimetry probes based on scattering bulbs in turbid media, Phys. Med. Biol.
47, 20492058 2002.
34
T. G. Vulcan, T. C. Zhu, C. E. Rodriguez, A. Hsi, D. L. Fraker, P. Baas,
L. H. Murrer, W. M. Star, E. Glatstein, A. G. Yodh, and S. M. Hahn,
Comparison between isotropic and nonisotropic dosimetry systems during intraperitoneal photodynamic therapy, Lasers Surg. Med. 26, 292
301 2000.
35
J. P. Marijnissen and W. M. Star, Calibration of isotropic light dosimetry
probes based on scattering bulbs in clear media, Phys. Med. Biol. 41,
11911208 1996.
36
A. H. Hielscher, A. Y. Bluestone, G. S. Abdoulaey, A. D. Klose, J.
Lasker, M. Stewart, U. Netz, and J. Beuthan, Near-infrared diffuse optical tomography, Dis. Markers 18, 313337 2002.
37
A. Li, E. L. Miller, M. E. Kilmer, T. J. Brukilacchio, T. Chaves, J. Stott,
Q. Zhang, T. Wu, M. Chorlton, R. H. Moore, D. B. Kopans, and D. A.
Boas, Tomographic optical breast imaging guided by three-dimensional
mammography, Appl. Opt. 42, 51815190 2003.
38
J. P. Culver, R. Choe, M. J. Holboke, L. Zubkov, T. Durduran, A. Slemp,
16
3135
V. Ntziachristos, B. Chance, and A. G. Yodh, Three-dimensional diffuse

optical tomography in the parallel plane transmission geometry: Evaluation of a hybrid frequency domain/continuous wave clinical system for
breast imaging, Med. Phys. 30, 235247 2003.
39
M. J. Holboke, B. J. Tromberg, X. Li, N. Shah, J. Fishkin, D. Kidney, J.
Butler, B. Chance, and A. G. Yodh, Three-dimensional diffuse optical
mammography with ultrasound localization in a human subject, J.
Biomed. Opt. 5, 237247 2000.
40
S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, C. Kogel, S. Soho, J.
J. Gibson, T. D. Tosteson, S. P. Poplack, and K. D. Paulsen, In vivo
hemoglobin and water concentrations, oxygen saturation, and scattering
estimates from near-infrared breast tomography using spectral reconstruction, Acad. Radiol. 13, 195202 2006.
41
B. Brooksby, S. Jiang, H. Dehghani, B. W. Pogue, K. D. Paulsen, J.
Weaver, C. Kogel, and S. P. Poplack, Combining near-infrared tomography and magnetic resonance imaging to study in vivo breast tissue:
implementation of a Laplacian-type regularization to incorporate magnetic resonance structure, J. Biomed. Opt. 10, 0515040515142005.
42
B. W. Pogue, S. P. Poplack, T. O. McBride, W. A. Wells, K. S. Osterman,
U. L. Osterberg, and K. D. Paulsen, Quantitative hemoglobin tomography with diffuse near-infrared spectroscopy: Pilot results in the breast,
Radiology 218, 261266 2001.
43
J. P. Culver, A. M. Siegel, J. J. Stott, and D. A. Boas, Volumetric diffuse
optical tomography of brain activity, Opt. Lett. 28, 20612063 2003.
44
J. P. Culver, T. Durduran, D. Furuya, C. Cheung, J. H. Greenberg, and A.
G. Yodh, Diffuse optical tomography of cerebral blood flow, oxygenation, and metabolism in rat during focal ischemia, J. Cereb. Blood Flow
Metab. 23, 911924 2003.
45
Y. Xu, N. Iftimia, H. Jiang, L. L. Key, and M. B. Bolster, Threedimensional diffuse optical tomography of bones and joints, J. Biomed.
Opt. 7, 8892 2002.
46
T. C. Zhu and J. C. Finlay, Prostate PDT dosimetry, Photodiagnosis
Photodynamic Therapy 3, 234246 2006.
47
X. Zhou, T. C. Zhu, J. C. Finlay, J. Li, A. Dimofte, and S. Hahn, Twodimensional/three dimensional hybrid interstitial diffuse optical tomography of human prostate during photodynamic therapy: Phantom and clinical results, Proc. SPIE 6434, 64341Y64352Y 2007.
48
C. Li, R. Liengsawangwong, H. Choi, and R. Cheung, Using a priori
structural information from magnetic resonance imaging to investigate the
feasibility of prostate diffuse optical tomography and spectroscopy: A
simulation study, Med. Phys. 34, 266274 2007.
49
X. Zhou and T. C. Zhu, Interstitial diffuse optical tomography using an
adjoint model with linear sources, Proc. SPIE 6845, 68450C
68459C2008.
50
A. F. Low, G. J. Tearney, B. E. Bouma, and I. K. Jang, Technology
Insight: optical coherence tomographyCurrent status and future development, Nat. Clin. Pract. Cardiovasc. Med. 3, 154162 2006.
51
B. W. Pogue, S. C. Davis, X. Song, B. A. Brooksby, H. Dehghani, and K.
D. Paulsen, Image analysis methods for diffuse optical tomography, J.
Biomed. Opt. 11, 033001033017 2006.
52
A. P. Gibson, J. C. Hebden, and S. R. Arridge, Recent advances in
diffuse optical imaging, Phys. Med. Biol. 50, R1R43 2005.
53
R. Richards-Kortum and E. Sevick-Muraca, Quantitative optical spectroscopy for tissue diagnosis, Annu. Rev. Phys. Chem. 47, 555606
1996.
54
B. Swinson, W. Jerjes, M. El-Maaytah, P. Norris, and C. Hopper, Optical
techniques in diagnosis of head and neck malignancy, Oral Oncol. 42,
221228 2006.
55
I. J. Bigio and S. G. Bown, Spectroscopic sensing of cancer and cancer
therapy: current status of translational research, Cancer Biol. Ther. 3,
259267 2004.
56
J. C. Finlay, T. C. Zhu, A. Dimofte, D. Stripp, S. B. Malkowicz, T. M.
Busch, and S. M. Hahn, Interstitial fluorescence spectroscopy in the
human prostate during motexafin lutetium-mediated photodynamic
therapy, Photochem. Photobiol. 82, 12701278 2006.
57
A. V. Hill, The possible effects of the aggregation of the molecules of
hemoglobin on its dissociation curves, J. Physiol. Proc. 40, ivvii
1910.
58
G. Yu, T. Durduran, C. Zhou, H. W. Wang, M. E. Putt, H. M. Saunders, C.
M. Sehgal, E. Glatstein, A. G. Yodh, and T. M. Busch, Noninvasive
monitoring of murine tumor blood flow during and after photodynamic
therapy provides early assessment of therapeutic efficacy, Clin. Cancer
Res. 11, 35433552 2005.
3135
59
H. W. Wang, M. E. Putt, M. J. Emanuele, D. B. Shin, E. Glatstein, A. G.

Yodh, and T. M. Busch, Treatment-induced changes in tumor oxygenation predict photodynamic therapy outcome, Cancer Res. 64, 7553
7561 2004.
60
M. C. Berenbaum, S. L. Akande, R. Bonnett, H. Kaur, S. Ioannou, R. D.
White, and U. J. Winfield, Meso-tetrahydroxyphenylporphyrins, a new
class of potent tumour photosensitisers with favourable selectivity, Br. J.
Cancer 54, 717725 1986.
61
I. P. van Geel, H. Oppelaar, Y. G. Oussoren, M. A. van der Valk, and F. A.
Stewart, Photosensitizing efficacy of mTHPC-PDT compared to
photofrin-PDT in the RIF1 mouse tumour and normal skin, Int. J. Cancer
60, 388394 1995.
62
W. Rauschning, I. B. Tan, and G. Dolivet, Photodynamic therapy PDT
with mTHPC in the palliation of advanced head and neck cancer in patients who have failed prior therapies and are unsuitable for radiatiotherapy, surgery or systemic chemotherapy, J. Clin. Oncol. 2214, 5596
2004.
63
M. Biel, Advances in photodynamic therapy for the treatment of head
and neck cancers, Lasers Surg. Med. 38, 349355 2006.
64
L. B. Lovat, N. F. Jamieson, M. R. Novelli, C. A. Mosse, C. Selvasekar,
G. D. Mackenzie, S. M. Thorpe, and S. G. Bown, Photodynamic therapy
with m-tetrahydroxyphenyl chlorin for high-grade dysplasia and early
cancer in Barretts columnar lined esophagus, Gastrointest. Endosc. 62,
617623 2005.
65
J. Etienne, N. Dorme, G. Bourg-Heckly, P. Raimbert, and J. F. Flejou,
Photodynamic therapy with green light and m-tetrahydroxyphenyl chlorin for intramucosal adenocarcinoma and high-grade dysplasia in Barretts
esophagus, Gastrointest. Endosc. 59, 880889 2004.
66
C. M. Moore, T. R. Nathan, W. R. Lees, C. A. Mosse, A. Freeman, M.
Emberton, and S. G. Bown, Photodynamic therapy using meso tetra
hydroxy phenyl chlorin mTHPC in early prostate cancer, Lasers Surg.
Med. 38, 356363 2006.
67
S. P. Pereira, L. Ayaru, A. Rogowska, A. Moose, A. R. Hatfield, and S. G.
Bown, Photodynamic therapy of malignant biliary strictures using mesotetrahydroxyphenylchlorin, Eur. J. Gastroenterol. Hepatol. 19, 479485
2007.
68
A. Hansch, O. Frey, M. Gajda, G. Susanna, J. Boettcher, R. Brauer, and
W. A. Kaiser, Photodynamic treatment as a novel approach in the
therapy of arthritic joints, Lasers Surg. Med. 40, 265272 2008.
69
M. Triesscheijn, M. Ruevekamp, N. Antonini, H. Neering, F. A. Stewart,
and P. Baas, Optimizing meso-tetra-hydroxyphenyl-chlorin-mediated
photodynamic therapy for basal cell carcinoma, Photochem. Photobiol.
82, 16861690 2006.
70
N. L. Oleinick, R. L. Morris, and I. Belichenko, The role of apoptosis in
response to photodynamic therapy: What, where, why, and how, Photochem. Photobiol. Sci. 1, 121 2002.
71
E. Buytaert, M. Dewaele, and P. Agostinis, Molecular effectors of multiple cell death pathways initiated by photodynamic therapy, Biochim.
Biophys. Acta 1776, 86107 2007.
72
B. Chen, B. W. Pogue, P. J. Hoopes, and T. Hasan, Vascular and cellular
targeting for photodynamic therapy, Crit. Rev. Eukaryot. Gene Expr. 16,
279305 2006.
73
T. M. Busch, Local physiological changes during photodynamic
therapy, Lasers Surg. Med. 38, 494499 2006.
74
M. Korbelik, PDT-associated host response and its role in the therapy
outcome, Lasers Surg. Med. 38, 500508 2006.
75
M. Korbelik, Induction of tumor immunity by photodynamic therapy, J.
Clin. Laser Med. Surg. 14, 329334 1996.
76
P. Baas, L. Murrer, F. A. Zoetmulder, F. A. Stewart, H. B. Ris, N. van
Zandwijk, J. L. Peterse, and E. J. Rutgers, Photodynamic therapy as
adjuvant therapy in surgically treated pleural malignancies, Br. J. Cancer
76, 819826 1997.
77
L. K. Lee, C. Whiteburst, M. L. Pantelides, and J. V. Moore, In situ
comparison of 665 nm light and 633 nm wavelength light penetration in
the human prostate gland, Photochem. Photobiol. 62, 882886 1995.
78
M. L. Pantelides, C. Whitehurst, J. V. Moore, T. A. King, and N. J.
Blacklock, Photodynamic therapy for localised prostatic cancer: light
penetration in the human prostate gland, J. Urol. 143, 398401 1990.
79
T. C. Zhu, A. Dimofte, J. C. Finlay, D. Stripp, T. Busch, J. Miles, R.
Whittington, S. B. Malkowicz, Z. Tochner, E. Glatstein, and S. M. Hahn,
Optical properties of human prostate at 732 nm measured in vivo during
motexafin lutetium-mediated photodyanmic therapy, Photochem. Photobiol. 81, 96105 2005.
3136
80
J. C. Finlay, D. L. Conover, E. L. Hull, and T. H. Foster, Porphyrin

bleaching and PDT-induced spectral changes are irradiance dependent in
ALA-sensitized normal rat skin in vivo, Photochem. Photobiol. 73,
5463 2001.
81
J. C. Finlay, S. Mitra, and T. H. Foster, In vivo mTHPC photobleaching
in normal rat skin exhibits unique irradiance-dependent features, Photochem. Photobiol. 75, 282288 2002.
82
J. C. Finlay, S. Mitra, and T. H. Foster, Photobleaching kinetics of
Photofrin in vivo and in multicell tumor spheroids indicate multiple simultaneous bleaching mechanisms, Phys. Med. Biol. 49, 48374860
2004.
83
D. J. Robinson, H. S. de Bruijn, N. van der Veen, M. R. Stringer, S. B.
Brown, and W. M. Star, Fluorescence photobleaching of ALA-induced
protoporphyrin IX during photodynamic therapy of normal hairless
mouse skin: The effect of light dose and irradiance and the resulting
biological effect, Photochem. Photobiol. 67, 140149 1998.
84
W. J. Cottrell, A. Oseroff, and T. H. Foster, Portable instrument that
integrates irradiation with fluorescence and reflectance spectroscopies
during clinical photodynamic therapy of cutaneous disease, Rev. Sci.
Instrum. 77, 064302 2006.
85
K. K.-H. Wang, S. Mitra, and T. H. Foster, A comprehensive mathematical model of microscopic dose deposition in photodynamic therapy,
Med. Phys. 34, 282293 2007.
86
J. S. Dysart and M. S. Patterson, Photobleaching kinetics, photoproduct
formation, and dose estimation during ALA induced PpIX PDT of MLL
cells under well oxygenated and hypoxic conditions, Photochem. Photobiol. Sci. 5, 7381 2006.
87
Y. Qin, D. Xing, S. Luo, J. Zhou, X. Zhong, and Q. Chen, Feasibility of
using fluoresceinyl Cypridina luciferin analog in a novel chemiluminescence method for real-time photodynamic therapy dosimetry, Photochem. Photobiol. 81, 15341538 2005.
88
M. T. Jarvi, M. J. Niedre, M. S. Patterson, and B. C. Wilson, Singlet
oxygen luminescence dosimetry SOLD for photodynamic therapy: current status, challenges and future prospects, Photochem. Photobiol. 82,
11981210 2006.
89
K. Stefflova, J. Chen, and G. Zheng, Using molecular beacons for cancer
imaging and treatment, Front. Biosci. 12, 47094721 2007.
90
H. Minn, PET and SPECT in low-grade glioma, Eur. J. Radiol. 56,
171178 2005.
91
N. Solban, I. Rizvi, and T. Hasan, Targeted photodynamic therapy,
Lasers Surg. Med. 38, 522531 2006.
92
D. Schrama, R. A. Reisfeld, and J. C. Becker, Antibody targeted drugs
as cancer therapeutics, Nat. Rev. Drug Discov. 5, 147159 2006.
93
R. M. Sharkey and D. M. Goldenberg, Targeted therapy of cancer: New
prospects for antibodies and immunoconjugates, Ca-Cancer J. Clin. 56,
226243 2006.
94
M. G. del Carmen, I. Rizvi, Y. Chang, A. C. Moor, E. Oliva, M. Sherwood, B. Pogue, and T. Hasan, Synergism of epidermal growth factor
receptor-targeted immunotherapy with photodynamic treatment of ovarian cancer in vivo, J. Natl. Cancer Inst. 97, 15161524 2005.
95
K. A. Cengel, S. M. Hahn, and E. Glatstein, C225 and PDT combination
therapy for ovarian cancer: the plays the thing, J. Natl. Cancer Inst. 97,
14881489 2005.
96
C. J. Gomer, A. Ferrario, M. Luna, N. Rucker, and S. Wong, Photodynamic therapy: Combined modality approaches targeting the tumor microenvironment, Lasers Surg. Med. 38, 516521 2006.
97
S. K. Bisland, L. Lilge, A. Lin, R. Rusnov, and B. C. Wilson, Metronomic photodynamic therapy as a new paradigm for photodynamic
therapy: Rationale and preclinical evaluation of technical feasibility for
treating malignant brain tumors, Photochem. Photobiol. 80, 2230
2004.
98
J. Li, T. C. Zhu, X. Zhou, A. Dimofte, and J. C. Finlay, Integrated light
3136
dosimetry system for prostate photodynamic therapy, Proc. SPIE 6845,
68450Q68458Q 2008.
99
L. Vesselov, W. Whittington, and L. Lilge, Design and performance of
thin cylindrical diffusers created in Ge-doped multimode optical fibers,
Appl. Opt. 44, 27542758 2005.
100
A. Rendon, R. Weersink, and L. Lilge, Towards conformal light delivery
using tailored cylindrical diffusers: attainable light dose distributions,
Phys. Med. Biol. 51, 59675975 2006.
101
G. Yu, T. Durduran, C. Zhou, T. C. Zhu, J. C. Finlay, T. M. Busch, S. B.
Malkowicz, S. M. Hahn, and A. G. Yodh, Real-time in situ monitoring
of human prostate photodynamic therapy with diffuse light, Photochem.
Photobiol. 82, 12791284 2006.
102
E. Clo, J. W. Snyder, P. R. Ogilby, and K. V. Gothelf, Control and
selectivity of photosensitized singlet oxygen production: challenges in
complex biological systems, ChemBioChem 8, 475481 2007.
103
Y. Chen, A. Gryshuk, S. Achilefu, T. Ohulchansky, W. Potter, T. Zhong,
J. Morgan, B. Chance, P. N. Prasad, B. W. Henderson, A. Oseroff, and R.
K. Pandey, A novel approach to a bifunctional photosensitizer for tumor
imaging and phototherapy, Bioconjug. Chem. 16, 12641274 2005.
104
R. K. Pandey, L. N. Goswami, Y. Chen, A. Gryshuk, J. R. Missert, A.
Oseroff, and T. J. Dougherty, Nature: A rich source for developing multifunctional agents. Tumor-imaging and photodynamic therapy, Lasers
Surg. Med. 38, 445467 2006.
105
G. Zheng, J. Chen, K. Stefflova, M. Jarvi, H. Li, and B. C. Wilson,
Photodynamic molecular beacon as an activatable photosensitizer based
on protease-controlled singlet oxygen quenching and activation, Proc.
Natl. Acad. Sci. U.S.A. 104, 89898994 2007.
106
S. Mitra and T. H. Foster, Photophysical parameters, photosensitizer
retention and tissue optical properties completely account for the higher
photodynamic efficacy of meso-tetra-hydroxyphenyl-chlorin vs Photofrin, Photochem. Photobiol. 81, 849859 2005.
107
A. Johansson, J. Axelsson, S. Andersson-Engels, and J. Swartling, Realtime light dosimetry software tools for interstitial photodynamic therapy
of the human prostate, Med. Phys. 34, 43094321 2007.
108
T. C. Zhu, J. Li, J. C. Finlay, A. Dimofte, D. Stripp, B. S. Malkowicz, and
S. M. Hahn, In vivo light dosimetry of interstitial PDT of human prostate, Proc. SPIE 6139, 61390L61401L 2006.
109
J. Li and T. C. Zhu, Determination of in vivo light fluence distribution in
heterogeneous prostate during photodyanmic therapy, Phys. Med. Biol.
53, 21032114 2008.
110
M. D. Altschuler, T. C. Zhu, J. Li, and S. M. Hahn, Optimized interstitial
PDT prostate treatment planning with the Cimmino feasibility algorithm,
Med. Phys. 32, 35243536 2005.
111
M. S. Patterson, B. C. Wilson, and R. Graff, In vivo tests of the concept
of photodynamic threshold dose in norma rate liver photosensitized by
aluminum chlorosulphonated phthalocyanine, Photochem. Photobiol. 51,
343349 1990.
112
I. Georgakoudi and T. H. Foster, Singlet oxygen- versus nonsinglet
oxygen-mediated mechanisms of sensitizer photobleaching and their effects on photodynamic dosimetry, Photochem. Photobiol. 67, 612625
1998.
113
A. P. Castano, P. Mroz, and M. R. Hamblin, Photodynamic therapy and
anti-tumour immunity, Nat. Rev. Cancer 6, 535545 2006.
114
S. O. Gollnick, L. Vaughan, and B. W. Henderson, Generation of effective antitumor vaccines using photodynamic therapy, Cancer Res. 62,
16041608 2002.
115
M. Korbelik and J. Sun, Photodynamic therapy-generated vaccine for
cancer therapy, Cancer Immunol. Immunother. 55, 900909 2006.
116
B. W. Henderson, S. O. Gollnick, J. W. Snyder, T. M. Busch, P. C. Kousis, R. T. Cheney, and J. Morgan, Choice of oxygen-conserving treatment regimen determines the inflammatory response and outcome of photodynamic therapy of tumors, Cancer Res. 64, 21202126 2004.
Molecular imaging and the unification of multilevel mechanisms and data

in medical physics
George C. Nikiforidis,a George C. Sakellaropoulos,b and George C. Kagadisc
Department of Medical Physics, School of Medicine, University of Patras, GR 265 04, Rion, Greece
Received 18 February 2008; revised 29 May 2008; accepted for publication 29 May 2008;
published 8 July 2008
Molecular imaging MI constitutes a recently developed approach of imaging, where modalities
and agents have been reinvented and used in novel combinations in order to expose and measure
biologic processes occurring at molecular and cellular levels. It is an approach that bridges the gap
between modalities acquiring data from high e.g., computed tomography, magnetic resonance
imaging, and positron-emitting isotopes and low e.g., PCR, microarrays levels of a biological
organization. While data integration methodologies will lead to improved diagnostic and prognostic
performance, interdisciplinary collaboration, triggered by MI, will result in a better perception of
the underlying biological mechanisms. Toward the development of a unifying theory describing
these mechanisms, medical physicists can formulate new hypotheses, provide the physical constraints bounding them, and consequently design appropriate experiments. Their new scientific and
working environment calls for interventions in their syllabi to educate scientists with enhanced
capabilities for holistic views and synthesis. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.2948321
Key words: molecular imaging, data integration, unification
I. INTRODUCTION
Over the last 30 years we have experienced a rapid evolution
of technology which has led to a transformation of biological
and medical sciences. Biological entities and mechanisms
that the limitations of technology until recently did not allow
us to observe have now become an object of routine measurement and reasoning. A typical field of application of this
change is medical imaging. Until the early 1970s, the study
of the human body was based on imaging modalities that
produced projections, exploiting a variety of physical properties of the human tissues. Three dimensional 3D representations were rarely achieved through time-consuming and
mainly experimental procedures, producing rather vague final results. In contrast, modern routine medical diagnostic
imaging heavily relies on modalities that produce highresolution 3D representations of the human body featuring
both morphological and functional characteristics. The latter
often require four dimensional images, by combining the
three dimensional spatial information with other types of information, e.g., temporal, spectroscopic, etc.
Although computed tomography CT and magnetic resonance imaging MRI were initially considered to be inefficient, as it is vividly described in a recent article by Wagner,1
they proved to be indispensable medical diagnostic modalities. These modalities are mathematically ill-conditioned
problems. As such, much experimentation and cultivation of
ideas was necessary to make them appropriate for providing
clinically useful images. Instrumentation for functional tomographic imaging evolved along a path distinct from that of
anatomical imaging devices CT and MRI. The first human
tomographic images with positron-emitting isotopes PET
were presented in 1972,2 followed by single photon emission
3444
Med. Phys. 35 8, August 2008
tomography SPECT in 1973.3,4 Until the early 1990s, the

evolution of ideas and experiments within each modality occurred in a relative solitude to other modalities. It was an era
in which continuous improvement of instrumentation specifications was the driving force behind the progress in diagnostic capabilities and effectiveness.
While the usefulness of combining anatomical and functional planar images was evident to physicians even in the
1960s,5 preceding the invention of CT, the first prototypes
combining two imaging modalities SPECT/CT, PET/CT
capable of acquiring images within the same reference frame
appeared only in the late 1990s.6,7 Ironically,8 while the preclinical designs and early clinical prototypes were received
with enthusiasm, the introduction of commercial PET/CT
scanners into the clinic encountered a certain amount of resistance. The technology was accused of being disruptive,9
expensive,10 wasteful of resources, clinically unproven,11
oversold for treatment planning,12,13 and successful only as
a consequence of the marketing strategy of major equipment
vendors.11 It was, however, a manifestation of a new paradigm in the field of medical imaging.
As computer power increased exponentially and became
more readily available, medical physicists joined scientists
from other disciplines in their new perspective for tackling
with the primary clinical tasks; that of exploring and measuring the benefit of combining the available information
that originated from apparently disjoint modalities and was
encapsulated into data of variable levels of complexity. The
field of medical informatics attracted the spotlight since the
combination of information attained its own distinctive role
as a major factor for achieving better health care.14,15
In our time, the scientific vernacular of biology and medicine has been augmented with a multitude of new terms,
0094-2405/2008/358/3444/9/$23.00
3444
3445
Nikiforidis, Sakellaropoulos, and Kagadis: Molecular imaging and the unification of multilevel mechanisms
3445
TABLE I. The advent of MI allows the design of structured experiments aiming to reveal/find the causal relationships among variables and mechanisms
concerning various levels of biological organization. A unified biomedical theory may be built using top-down and bottom-up approaches as well as
combinations thereof for diagnosis or prognosis. This theory could causally connect the mutations of specific genes sequence analysis with a characteristic
pattern of gene coexpressions analysis of microarrays, the abnormal value of some immunochemical parameters flow cytometry, microscopy, the manifestation of certain macroscopic anatomical or functional anomalies CT, SPECT, etc. and finally with the observation of specific clinical signs and symptoms.
All modalities are in vivo unless otherwise indicated.
describing the function of genes, proteins, and cells. The

advances in mass spectrometry, cell sorting, and development of microarrays, with substantial contribution of physicists, are feeding the biomedical field with an abundance of
data, driving it toward a molecular perspective. A formidable
number of variables are being recorded, pertaining to the
diverse biological mechanisms inherent to the function of a
living organism. Based on different levels of biological organization, the data can be classified into: genetic, genomic
and proteomic, cellular, tissue, organ, and organism Table I.
Within this continuum, data originating from higher levels of
biological organization can be defined as high-level data,
whereas those originating from lower levels of biological
organization can be defined as low-level data. Each level is
studied with specific acquisition modalities and methods of
analysis, giving rise to a large variety of omics.16 Unfortunately, these data cannot be directly utilized. The ocean of
electronic data at our disposal is stored in a mosaic of heterogeneous database implementations, hindering the access
to and aggregation of data across implementations and creating a large gap between their potential and effective
value.17 The application of various measurement protocols
customized for the specific experimental conditions introduces another source of data inhomogeneity.
The advent of molecular imaging MI is further changing
our approach to research and patient care by directly targeting the molecular, cellular, or physiologic defects responsible
for disease. It offers a promise for more precise disease charMedical Physics, Vol. 35, No. 8, August 2008
acterization and assessment of therapeutic response.18 More

importantly, it may become the means for a paradigm shift to
our mode of scientific reasoning, where interdisciplinary collaboration becomes fundamental toward a holistic view of
the biological mechanisms.
In Sec. II we present how the attempt to combine the
various fragments of information gave rise to data integration. Sections III and III A contain the current status and
future steps in MI technology. In Sec. IV we present how MI
can become the vehicle for a new paradigm shift in medical
physics and our vision for the new scientific and working
environment that the medical physicist is going to be part of
in the years to come.
II. DATA INTEGRATION
The medical diagnostic and prognostic tasks have motivated researchers to integrate the available data, in an attempt to combine the various fragments of information. Biomedical data integration can be defined as the procedure for
the creation of a single, uniform interface to query the data
stored in many heterogeneous databases.
Integration means much more than obtaining the data in
digital form or even reducing diverse data to a common
form. Care must be taken so that integrated data serve the
end-users clinicians requirements and reflect in a well
specified manner the medical hypothesis that warranted the
specific examination. In general, integration can be the ag-
3446
gregation of semantically similar data from multiple heterogeneous sources vertical integration; the composition of semantically complementary data from multiple heterogeneous
sources horizontal integration; and the standardization of
access to semantically similar information at disparate
sources integration for application portability. For example,
with vertical integration morphology information can be obtained through the combination of data from both CT and
ultrasound US; horizontal integration would be beneficial
when both anatomy and physiology of a specific region are
sought for. Integration for application portability would
make comparable data originating from imaging apparatuses
of different specifications, e.g., 512 512 and 256 256 image matrices.
The attempts for data integration have comprised a strenuous track of research within the field of medical informatics
and a number of issues still need to be addressed. Some of
these issues are the partial availability of databases, the controversial privacy issues, the fact that useful information lies
in the natural language text of the scientific papers, and, perhaps most importantly, the lack of appropriate standards.19 It
was soon evident that obtaining multiple representations of
the same entity, e.g., through CT, SPECT, or MRI, would
enhance the ability of making better medical decisions. But
the efficient combination of available variables toward better
medical diagnosis and prognosis had been a high-complexity
process, even when the available data were of the high-level
type, i.e., results of clinical and laboratory examinations.
With the advent of molecular data, the combination of
multimodality information reached a higher level of complexity. It became obvious that analytical approaches are unable to compensate for such complexity. From the thousands
of variables stored in the integrated databases only the variables relevant to specific clinical problems should be considered. This calls for the development of efficient ways for
variable description and management. One such way is
building additional data structurescalled metadatathat
describe the characteristics of stored variables. Metadata arranged in schemata can represent the hierarchical structures
of these characteristics and help us build frameworks for the
integration of contextual information from the clinicians
point of view. Statistical modeling and procedures for statistical learning can also be helpful, providing methodologies
for feature extraction and data reduction.
Evidence-based medicine EBM has been recently established as the approach of choice to medical practice. EBM
uses the best current evidence and statistical methods to justify the value of therapeutic decisions in an individual patient. What it effectively does is to identify similarities or
matches between the specific patients patterns of clinical
and laboratory data and scientifically valid evidence. The
availability of integrated data will allow us to describe all the
relevant variables and their interactions in a deep and structured manner and finally to expose the underlying mechanisms that give rise to these patterns. We could then obtain
better specifications of ones health status and consequently
aim at personalized treatment or prevention.
Medical Physics, Vol. 35, No. 8, August 2008
3446
III. MOLECULAR IMAGING

Since understanding is our goal, seeking optimal methods
for production and analysis of integrated data might prove a
futile task. Drawing an example from the field of machine
learning, one can find a number of methodologies that are
capable of performing efficient multivariate discrimination
of cases into classes but provide little or no insight on how
these variables are semantically connected. Expanding our
focus from a possibly blind analysis of integrated data to the
quest for understanding the inter-relationship among the
variables and governing biological mechanisms can allow us
to obtain a holistic view of human organism.
MI is the field of research that provides an opportunity for
convergence of two seemingly disparate interests: the clinical interest in providing better health care and the basic scientists interest in explaining the biological mechanisms that
give rise to the phenotypic manifestations. Based on this
convergence, productive alliances between medical physicists and radiologists, on one hand, and molecular and cellular biochemists, on the other hand, are currently driving
MI.20
MI can be broadly defined as the in vivo characterization
and measurement of biologic processes at the cellular and
molecular level.21 Receptors, transporters, extracellular enzymes, and intracellular macromolecules are all potential targets for MI. MI encompasses a broad set of technologies that
couple imaging modalities and contrast agents with molecular specificity. MI can be divided into two general approaches for generation of image contrast. In the first approach, contrast reagents are engineered to interrogate
endogenously expressed proteins or nucleic acids and in the
second the reporter genes are transfected for expression regulated by environmental or tissue factors.22 It is expected that
chemically engineered exogenous reporters will be more useful than genetically encoded reporters.20
The contrast agents, called molecular probes or tracers,
consist of a signaling component that emits a detectable signal and a targeting component that confers localization. Modalities used in MI include nuclear imaging, i.e., SPECT and
PET, anatomical imaging, i.e., CT and MR imaging, optical
imaging and US.18 MI is possible with the existing imaging
modalities because some degree of tissue cell inhomogeneity can be registered within the inherent spatial resolution
limits of the resultant images.23
Despite the term, current MI modalities used to image
animals and patients do not visualize individual molecules or
even cells but rather populations thereof, since there are specific requirements for generating sufficient signal-to-noise
ratios to render imaging possible. The challenge for remote
imaging modalities i.e., PET, SPECT, MRI, and optical imaging is the direct imaging of DNA,24 which is associated
with problems of nonspecific and nontarget binding of imaging probes in case of polymorphisms. Direct targeting of
mRNA is also challenging, as it requires one to one correlation of mRNA molecules with the imaging probes, whereas
only 50 to thousands of mRNA molecules are present in
every cell. These problems are solved in the case of protein
3447
imaging since proteins are present in quite higher amounts

thousands to millions of copies per cell, that make their
imaging more feasible.25,26 By using carefully designed
probes in order to target proteins that act as enzymes MI can
provide an indirect measurement of characteristic signals that
arise from the expression of specific DNA parts. While direct
measurement seems infeasible, MI takes advantage of the
enzymatic amplification of protein function.
Contrary to the paradigms followed until now, the underlying biological question rather than the technology itself
should drive the choice of imaging technique. MI seeks to
identify and evaluate specific molecular signals which in turn
induce specific requirements for spatial resolution and sensitivity. PET and SPECT use contrast agents that are synthesized at sufficiently high specific activities and provide the
required levels of image contrast. MRI, on the other hand, a
preferred modality when high spatial resolution is required,
suffers from the need for 46 orders of magnitude higher
contrast agent concentration due to the indirect nature of enhancement produced by MRI contrast agents.24 This introduces clinical risks of toxicity as well as the danger of contrast agent concentration, perturbing the underlying
molecular signal that is being monitored by MRI. Similar
trade-offs are intrinsic to US contrast agents, optical imaging
techniques, and even x-ray contrast agents.
III.A. Future steps in MI technology
Merging PET with optical imaging techniques seems

technically feasible and there are already published results
toward this direction.27 Such a system OPET is capable of
detecting both high-energy gamma rays and optical wavelength photons and allows the noninvasively and repetitive
imaging of small animal models in vivo for the presence of
PET and optical signals. Thus, both technologies tracers can
be combined and provide simultaneous different information.
The expected improvements in detector technology will
make the construction of simultaneous OSPECT systems
with similar advantages feasible.
More exotic technologies, coming from the world of high
energy physics, are being investigated for potential medical
applications. It is obvious that they have to be seen not as
single imaging modalities but in combination with the already existing ones. For example, x-ray phase-contrast imaging reconstructs the projected absorption and refractive index
distributions of an object and offers dramatically reduced
noise levels with greatly enhanced imaging contrast.28 Neutron imaging can offer alternative tools, since neutrons are
highly penetrating. Therefore, it can image deeply seated
body structures that cannot be reached by other probes.29
Neutron stimulated emission computed tomography illuminates the body with fast neutrons with energies between 1
and 10 MeV and could identify malignancies by the way
they change concentrations of chemical elements in tissue,
long before cancer has begun to produce macroscopic anatomical changes. Terahertz THz spectroscopy can be used
to image fatty tissue, bone, teeth, and thin slices of nonfatty
tissue by using electromagnetic waves at THz frequencies.30
3447
Instruments based on THz spectrometers with resolution better than 400 m are expected to complement existing biomedical imaging technologies.
Nanoparticles have the potential to change the role of imaging technologies in diagnosis and therapy. They are colloidal vesicular systems that vary in size from 10 to 1000 nm
with the drug and/or imaging probe of interest either entrapped therein or attached thereon.31 Inherently, when
echogenic nanoparticles are bound to surfaces, they can be
modified for compatibility with MR, US, x ray, or nuclear
imaging methodologies.32 There are several examples that
prove the advantages of such bimodal agents. For example,
fluorine 19F MRI of cells labeled with different types of
liquid perfluorocarbon nanoparticles produces unique and
sensitive cell markers distinct from any tissue background
signal.33 Successful full body microSPECT/CT mouse imaging of Cd125mTe/ ZnS NPs linked to either a monoclonal antibody against mouse lung thrombomodulin mAb 201B, or
a control antibody mAb 33, has shown that nanoparticles
conjugated to mAb 201B principally target the lungs while
the nanoparticles coupled to mAb 33 accumulate in the liver
and spleen.34 Similarly, PET labeled nanoparticles have been
successfully imaged.35 PET and MRI studies are carried out
in parallel due to the lack of combined imaging systems.36,37
It is estimated that simultaneous imaging of bimodal nanoparticles will maximize the benefits from the combination of
nuclear medicine and MR information.38 Other magnetic
nanoparticles can be caused to oscillate under the influence
of an incident ultrasonic wave with the advantage of greater
sensitivity due to the absence of a large background signal.39
By the use of targeted gold nanoparticles, the combination of
US with contrast enhanced photoacoustic imaging seems
possible. This is proposed as a visual tool to compound molecular and structural information for early stage prostate
cancer detection.40
An additional promise of endogenous expression, where
contrast reagents are engineered to interrogate endogenously
expressed proteins or nucleic acids, is the combination of
imaging with therapeutic reagents such that they can both be
optimized in an iterative fashion. The effect of therapeutic
reagents can give rise to modified signals which in turn can
be identified with the use of different contrast reagents. The
latter can give evidence for new therapeutic pathways, etc. It
is now recognized that many of the survival attributes of
neoplastic cells are determined by proteins involved in
mechanisms of antiapoptosis, cell cycle regulation, and damage repair.41 Therapeutic radiationfollowing the definition
of the underlying molecular features of cancer cells by MI
can act as focused biology i.e., producing molecular events
in the irradiated tissue that targets molecular entities other
than DNA, offering new opportunities for therapeutic attacks
on cancer cells.42
IV. UNIFICATION
In its course through history research often passes through
stages of conformity. In our opinion, we are currently experiencing such a stage. New research hypotheses are rareon
3448
a global scaleand most research is conducted in deference

to the mandates of what has been established as a valid research trend at that particular time. Researchers often publish
results originating from repetitions of experiments already
performed elsewhere, perhaps responding to an indirect mandate of the peer-review publication procedure. A possible
validation of results is preferred to the statement and pursuit
of new hypotheses. Technology offers an alibi for this attitude, shifting our focus to its impressive advances. However,
we are now entering a new era in which technological
progress alone does not suffice.43,44 The current challenge for
medical physics is the development of a framework that will
lead us from the deepening of our knowledge in the details of
specific phenomena to the widening of our scope in understanding the underlying etiologies. From the deductive mode
of reasoning, we must march to the statement of new hypotheses and consequently invent the experiments to test them.
A new paradigm must be followed in order to reverse this
trend of conformity and provide new pathways for research.
MI can facilitate such a paradigm. Lying between the gene
and the cell level, it provides images and measurements of
populations of molecules, thus incorporating the statistical
nature of the mechanisms which give rise to the measured
quantities. New variables, which emerge through the analysis
of this new kind of data, could describe these quantities efficiently and offer a connection with modalities pertinent to
other levels of complexity of the human organism. It is the
potential for broadening the scope of our goals within biology and medicine that makes MI unique. Apart from aiming
at improved diagnostic and prognostic performance, MI can
become a vehicle for the development of conceptual frameworks that transect the entire range of information levels.
Clinicians have traditionally treated the various levels of
available data in a top-down fashion: they start with clinical
manifestations of diseases and they attempt to find relations
with specific patterns in low-level molecular data. On the
other hand, biologists take the opposite direction: they perform research on DNA sequences and gene expression, and
they try to link them with specific diseases, following a
bottom-up approach inside the multilevel data space. This
observation does not necessarily exclude sideway movements inside the same level or even opposite direction steps
within the main trend. For example, biologists, performing
gene coexpression experiments with the use of microarray
technology, might need to go along the top-down direction to
investigate through sequencing whether a polymorphism
causes the apparent behavior. Similarly, clinicians are moving bottom-up when their low-level findings are attributed to
clinical cases similar to the ones they originated from.
The case of rheumatoid arthritis RA can be used as an
example of a disorder in which various imaging modalities
are used to provide details at different levels of information.
The tissues within a joint, usually harmed in patients suffering from RA, can be examined with tomographic
techniques45,46 that depict the phenotypic manifestation of
the illness and/or inflammation. Physicists have helped in
this endeavor by working in the area of CT and MRI for
many years, initially in the concept of tomographic imaging
3448
and later on, in developing dual energy CT modalities and

functional MRI. Research studies in functional imaging modalities such as in the areas of optimizing data acquisition
time of flight PET,47 scatter modeling,48 CT based attenuation correction,49 etc., require a strong physics background.
Besides, US is used for both diagnostic evaluation of erosions or synovitis and therapeutic follow-up,50 whereas metabolizing bone can be observed with the aid of functional
imaging by using PET Ref. 51 and SPECT.52 At molecular
level, techniques, such as bioluminescence and fluorescence
reflectance imaging, allow imaging of the delivery of therapeutic agents. To this direction, physicists have worked in
many areas like tomographic optical imaging53 and combined procedures of PET and optical imaging.27 Furthermore,
microarrays allow an open-ended survey to comprehensively
identify the fraction of genes that define each samples
unique expression.54 IL-17RA and IL-17RC were found to
be overexpressed in RA peripheral whole blood,55 while oligonucleotide microarrays have already been used in order to
elucidate the molecular effects of antirheumatic drugs on human chondrocyte gene expression.56 Research studies conducted by physicists have contributed in optimizing the techniques for better signal acquisition and better microarray
image segmentation.57 RT-PCR has been used in order to
evaluate the potent effects of vasoactive intestinal peptide in
human RA,58 while the efficacy of a soluble ligand of
CD200R has been evaluated in established collagen-induced
arthritis in mice.59 Additionally, southern blot,60 northern
blot,61 and western blot62 analyses have been used in measuring the expression of different cell types in RA. Optical
imaging techniques have also emerged in the field of RA
Ref. 63 with the more recent advancement of coupled photoacoustic tomography and image reconstruction in joint inflammatory diseases with high sensitivity and accuracy
Fig. 1.64
A model that would efficiently combine these multilevel
fragments of information is still missing. Such a model
would not only result in the improvement of the diagnostic
and therapeutic potential of modern medicine but would also
render thenow hiddenunderlying mechanisms of interactions observable and explainable, bridging the gap between the phenotypic and the genotypic level. The concept of
probability as a measure of plausibility constitutes a core
notion for the modeling of biological mechanisms. The
Bayesian perspective to probability can assist researchers in
this task with its main advantage being that it allows reasoning based on subjective evidence accumulated in a stepwise
manner. Bayesian networks6568 constitute a graphical representation of knowledge,69 explicitly representing complexity,
and simultaneously an efficient tool for inference and reasoning under uncertainty. They are being extensively used in
medicine7072 and biology.7377 They are learning from data
and their modularity allows the inclusion of fragments of
knowledge into larger reasoning systems. Therefore, Bayesian networks can effectively capture the various constraints
posed by the different disciplines involved in MI. It is our
belief that Bayesian theory can provide a solid and transpar-
3449
3449
FIG. 1. The case of RA where various imaging modalities are used to provide details at different levels of information.
ent chassis for the conduct of experiments that will justify

the steps toward a unification of our theories for the mechanisms under study.
The apparent bottleneck in the exploitation of the currently available data cannot only be attributed to the relative
hysteresis in the development of information management
methodologies as compared to the technological breakthrough. What is more significant is the ignorance of the
sources of variance and its magnitude in our measurements.
Physicists have a key role in identifying these sources of

statistical noise through their comprehension of the physical
processes governing the instrumentation used. Joining a team
consisting of biologists, physicians, and bioinformaticists,
medical physicists will have distinct and important roles in
estimating the magnitude of this noise, designing improved
measurement apparatus, providing the physical constraints
bounding the proposed hypotheses, and setting-up novel experiments to test them.
3450
The introduction of nanoparticles described previously

may be considered as an example of this procedure. Their
use within MI gives rise to a number of considerations relevant to their characteristics. Paramagnetic, radioactive, and
chemical properties of nanoparticles and imaging probes
which are bound to them, influence both the way they interact at the molecular level and the characteristics of their imaging. Medical physicists are studying how these properties
affect parameters like spatial resolution, sensitivity, acquisition time, and dose and provide optimization guidelines specific to the particular task.
The emergence of MI offers the opportunity for studying
mechanisms acting on different biologic organization levels.
Medical physicists will be asked to provide support to the
pursuit of a future unification theory incorporating the entire
array of mechanisms underlying the phenotypic manifestations. Toward this end, a reorientation of medical physics is
needed. Currently, most medical physicists dedicate a large
part of their education and work in becoming specialists in a
particular diagnostic modality or therapeutic apparatus. In
this way, medical physicists acquire not only deep knowledge of the physical principles underlying the specific modality but work to optimize its applications. This leads to the
rapid development of the modality with impressive results.
Unfortunately, this overspecialization is not sufficient to address the issues that will be raised in the immediate future.
The new scientific and working environment, which the
medical physicist is going to be part of in the years to come,
calls for developing a broader view of the scientific questions
raised and the methodologies at hand within both medical
physics education and research activities.
The education of medical physicists should aim at developing scientists with enhanced capabilities for synthesis. The
syllabi of future medical physics courses will require substantial augmentation with carefully selected and interconnected topics from molecular and computational biology.
These types of courses will familiarize physicists with concepts and methodologies they can use in their collaboration
with scientists from other disciplines. MI can acquire a
dominant position in these courses not only as an exposition
of a new technology but mainly as a new means for understanding the mechanisms governing the function of the human organism at various levels of organizational complexity.
Apart from research in the traditional fields of medical physics, additional research activities will then naturally unfold
toward theoretical aspects of a unified conceptual
framework.
V. DISCUSSION/CONCLUSION
From a DNA strand anomaly to a clinical symptom or
sign, there are many different levels at which technology
offers means for inspection. But since they have emerged at
different points in our perpetual attempt to comprehend the
structure and function of the human organism, imaging technologies constitute instruments of fragmentary value. In the
continuum of interactions among the various entities that
modern medicine and biology have revealed, each imaging
3450
modality covers efficiently only a narrow region. As technology advances, each modality becomes increasingly better in
terms of precision within this continuum, shedding light to
new mechanisms and processes. We are thus given the opportunity to build new theories that link the mechanisms relevant to one level with mechanisms relevant to another.
The design of new complex experiments will be needed,
involving variables pertinent to many levels of biological
organization. Obviously, MI is an interdisciplinary approach
where end-user needs and challenges have to be well defined
and clearly understood; clinicians define the diagnostic or
therapeutic needs; biologists can define the biological phenomena that can be targeted; and physicists will have to investigate the appropriate physical processes and suggest imaging concepts. These will lead to the advancement of both
science and technology. New theories will be developed and
new conceptual frameworks will be established to include
them. We thus have the opportunity to build genuine interdisciplinary collaborations within which the ideas originating
from different disciplines will fuse and create a new core of
thinking.
This is expected to affect the field of medical physics in
its theoretical foundations. It is unlikely that its current structure can accommodate the increasing interactions with biology and medicine; new ways of thinking will be needed to
meet new scientific challenges. We now have the possibility
of performing a step into revolutionary science,78 exploring
alternatives to long-held, obvious-seeming assumptions.
From an epistemological viewpoint, there are many similarities with the situation in physics at the dawn of the 20th
century. Experimental data that could not be explained with
the physical theories available at that time were partially accommodated by the Bohr atomic model and later with refinements made by Sommerfeld. However, only when the concept of the wave function and the Schrdinger equation
appeared did quantum theory manage to describe and explain
the mechanisms giving rise to the phenomena. The growth of
the quantum theory offered a unifying framework for concepts initially bound to either the microscopic or macroscopic level and gave birth to electronics as a new discipline.
Medical physics might follow the same route. Its theoretical
expansion may develop as a result of our attempt to provide
a unifying view of the biological processes from the molecule to the entire organism. MI may become both a technology that produces experimental data not yet accounted for by
current theories and a bridge through which fragments of
knowledge will be assembled into a unified theory.
The role that medical physicists will play in this endeavor
cannot be predicted precisely. It will be the result of a number of factors ranging from the reorientation of medical
physics courses syllabi to the self-motivation and degree of
engagement of each individual medical physicist. Medical
physicists should closely collaborate with biologists, biochemists, physicians, and bioinformaticists. They should take
a strategic position due to their diverse role in both providing
new results in the imaging research as well as giving feedback to the research of other disciplines.
3451
Author to whom correspondence should be addressed. Telephone:

30 2610 969144; Fax: 30 2610 969166. Electronic mail:
[email protected]
b
c
1
R. F. Wagner, From medical images to multiple-biomarker microarrays,
Med. Phys. 34, 49444951 2007.
2
A. D. Chesler, Symposium on Tomographic Imaging in Nuclear Medicine, New York, 1972.
3
D. E. Kuhl, R. Q. Edwards, A. R. Ricci, and M. Reivich, Quantitative
section scanning using orthogonal tangent correction, J. Nucl. Med. 14,
196200 1973.
4
H. W. Strauss, B. L. Zaret, N. D. Martin, H. P. Wells, Jr., and M. D.
Flamm, Jr., Noninvasive evaluation of regional myocardial perfusion
with potassium 43. Technique in patients with exercise-induced transient
myocardial ischemia, Radiology 108, 8590 1973.
5
H. N. Wagner, A Personal History of Nuclear Medicine Springer, London, 2006.
6
T. Beyer, D. W. Townsend, T. Brun, P. E. Kinahan, M. Charron, R.
Roddy, J. Jerin, J. Young, L. Byars, and R. Nutt, A combined PET/CT
scanner for clinical oncology, J. Nucl. Med. 41, 13691379 2000.
7
S. C. Blankespoor, X. Xu, K. Kaiki, J. K. Brown, H. R. Tang, C. E. Cann,
and B. H. Hasegawa, Attenuation correction of SPECT using x-ray CT
on an emission-transmission CT system: Myocardial perfusion assessment, IEEE Trans. Nucl. Sci. 43, 22632274 1996.
8
D. W. Townsend, Multimodality imaging of structure and function,
Phys. Med. Biol. 53, R1R39 2008.
9
Disruptive technology: The conflict over PET/CT, http://
www.imagingeconomics.com/issues/articles/2005-06_02.asp
accessed
February 9, 2008.
10
A. Alavi, A. Mavi, S. Basu, and A. Fischman, Is PET-CT the only
option?, Eur. J. Nucl. Med. Mol. Imaging 34, 819821 2007.
11
H. Zaidi, The quest for the ideal anato-molecular imaging fusion tool,
Biomed. Imaging Interv. J. 2, e47 2006.
12
B. Wessels, The value of PET/CT is being over-sold as a clinical tool in
radiation oncology. Against the proposition, Med. Phys. 32, 14581459
2005.
13
L. Xing, The value of PET/CT is being over-sold as a clinical tool in
radiation oncology. For the proposition, Med. Phys. 32, 14571458
2005.
14
E. H. Shortliffe and J. J. Cimino, in Health Informatics, edited by K. J.
Hannah and M. J. Ball Springer, New York, 2006.
15
G. C. Kagadis, P. Nagy, S. Langer, M. Flynn, and G. Starkschall, Anniversary paper: Roles of medical physicists and health care applications of
informatics, Med. Phys. 35, 119127 2008.
16
Omics-Omics.org, http://omics.org/index.php/Main_Page, accessed February 9, 2008.
17
W. Sujansky, Heterogeneous database integration in biomedicine, J.
Biomed. Inform. 34, 285298 2001.
18
D. S. Wang, M. D. Dake, J. M. Park, and M. D. Kuo, Molecular imaging: A primer for interventionalists and imagers, J. Vasc. Interv. Radiol.
17, 14051423 2006.
19
B. Louie, P. Mork, F. Martin-Sanchez, A. Halevy, and P. Tarczy-Hornoch,
Data integration and genomic medicine, J. Biomed. Inform. 40, 516
2007.
20
R. J. Gillies, In vivo molecular imaging, J. Cell. Biochem. Suppl. 39,
231238 2002.
21
R. Weissleder and U. Mahmood, Molecular imaging, Radiology 219,
316333 2001.
22
G. D. Luker, Special conference of the American Association for Cancer
Research on molecular imaging in cancer: Linking biology, function, and
clinical applications in vivo, Cancer Res. 62, 21952198 2002.
23
R. Blasberg, PET imaging of gene expression, Eur. J. Cancer 38, 2137
2146 2002.
24
V. Sharma, G. D. Luker, and D. Piwnica-Worms, Molecular imaging of
gene expression and protein function in vivo with PET and SPECT, J.
Magn. Reson. Imaging 16, 336351 2002.
25
E. P. Krenning et al., Somatostatin receptor scintigraphy with 111InDTPA-D-Phe1- and 123I-Tyr3-octreotide: The Rotterdam experience
with more than 1000 patients, Eur. J. Nucl. Med. 20, 716731 1993.
26
J. Lister-James, S. Vallabhajosula, B. R. Moyer, D. A. Pearson, B. J.
McBride, M. A. De Rosch, L. R. Bush, J. Machac, and R. T. Dean,
Pre-clinical evaluation of technetium-99m platelet receptor-binding pepMedical Physics, Vol. 35, No. 8, August 2008
3451
tide, J. Nucl. Med. 38, 105111 1997.

D. L. Prout, R. W. Silverman, and A. Chatziioannou, Detector concept
for OPET-A combined PET and optical imaging system, IEEE Trans.
Nucl. Sci. 51, 752756 2004.
28
C.-Y. Chou, Y. Huang, D. Shi, and M. A. Anastasio, Image reconstruction in quantitative x-ray phase-contrast imaging employing multiple
measurements, Opt. Express 15, 10002 2007.
29
J. E. Bender, A. J. Kapadia, A. C. Sharma, G. D. Tourassi, B. P. Harrawood, and C. E. Floyd, Jr., Breast cancer detection using neutron stimulated emission computed tomography: Prominent elements and dose requirements, Med. Phys. 34, 38663871 2007.
30
M. C. Beard, G. M. Turner, and C. A. Schmuttenmaer, Progress towards
two-dimensional biomedical imaging with THz spectroscopy, Phys.
Med. Biol. 47, 38413846 2002.
31
Y. W. Cho, S. A. Park, T. H. Han, D. H. Son, J. S. Park, S. J. Oh, D. H.
Moon, K. J. Cho, C. H. Ahn, Y. Byun, I. S. Kim, I. C. Kwon, and S. Y.
Kim, In vivo tumor targeting and radionuclide imaging with selfassembled nanoparticles: Mechanisms, key factors, and their implications, Biomaterials 28, 12361247 2007.
32
T. Cyrus, D. R. Abendschein, S. D. Caruthers, T. D. Harris, V. Glattauer,
J. A. Werkmeister, J. A. Ramshaw, S. A. Wickline, and G. M. Lanza,
MR three-dimensional molecular imaging of intramural biomarkers with
targeted nanoparticles, J. Cardiovasc. Magn. Reson. 8, 535541 2006.
33
K. C. Partlow, J. Chen, J. A. Brant, A. M. Neubauer, T. E. Meyerrose, M.
H. Creer, J. A. Nolta, S. D. Caruthers, G. M. Lanza, and S. A. Wickline,
19F magnetic resonance imaging for stem/progenitor cell tracking with
multiple unique perfluorocarbon nanobeacons, FASEB J. 21, 16471654
2007.
34
J. D. Woodward, S. J. Kennel, S. Mirzadeh, S. Dai, J. S. Wall, T. Richey,
J. Avenell, and A. J. Rondinone, In vivo SPECT/CT imaging and biodistribution using radioactive Cd125mTe/ZnS nanoparticles, Nanotechnology 18, 175103175108 2007.
35
E. D. Pressly, R. Rossin, A. Hagooly, K. Fukukawa, B. W. Messmore, M.
J. Welch, K. L. Wooley, M. S. Lamm, R. A. Hule, D. J. Pochan, and C. J.
Hawker, Structural effects on the biodistribution and positron emission
tomography PET imaging of well-defined 64Cu-labeled nanoparticles
comprised of amphiphilic block graft copolymers, Biomacromolecules
8, 31263134 2007.
36
Z. Burbar, R. Grazioso, and J. Corbeil, PET performance of MR/PET
brain insert tomograph abstr, IEEE Nuclear Science Symposium and
Medical Imaging Conference, San Diego, California, 2006.
37
M. S. Judenhofer, C. Catana, B. K. Swann, S. B. Siegel, W. I. Jung, R. E.
Nutt, S. R. Cherry, C. D. Claussen, and B. J. Pichler, PET/MR images
acquired with a compact MR-compatible PET detector in a 7-T magnet,
Radiology 244, 807814 2007.
38
H. Zaidi, O. Mawlawi, and C. G. Orton, Point/counterpoint. Simultaneous PET/MR will replace PET/CT as the molecular multimodality imaging platform of choice, Med. Phys. 34, 15251528 2007.
39
S. J. Norton and T. Vo-Dinh, Imaging the distribution of magnetic nanoparticles with ultrasound, IEEE Trans. Med. Imaging 26, 660665
2007.
40
A. Uedono, K. Ito, T. Narumi, M. Sometani, K. Yamabe, Y. Miyagawa, T.
Murata, K. Honda, N. Hattori, M. Matsuura, K. Asai, T. Ohdaira, and R.
Suzuki, Annealing properties of open volumes in strained SiN films
studied by monoenergetic positron beams, J. Appl. Phys. 102, 064513
2007.
41
R. K. Schmidt-Ullrich, Molecular targets in radiation oncology, Oncogene 22, 57305733 2003.
42
C. N. Coleman, Linking radiation oncology and imaging through molecular biology or now that therapy and diagnosis have separated, its
time to get together again!, Radiology 228, 2935 2003.
43
G. D. Fullerton, The development of technologies for molecular imaging
should be driven principally by biological questions to be addressed
rather than by simply modifying existing imaging technologies. For the
proposition, Med. Phys. 32, 12311232 2005.
44
J. D. Hazle, The development of technologies for molecular imaging
should be driven principally by biological questions to be addressed
rather than by simply modifying existing imaging technologies. Against
the proposition, Med. Phys. 32, 12321233 2005.
45
U. M. Dohn, B. J. Ejbjerg, M. Hasselquist, E. Narvestad, J. Moller, H. S.
Thomsen, and M. Ostergaard, Detection of bone erosions in rheumatoid
arthritis wrist joints with magnetic resonance imaging, computed tomography and radiography, Arthritis Res. Ther. 10, R25 2008.
27
3452
46
M. Knupp, A. Skoog, H. Tornkvist, and S. Ponzer, Triple arthrodesis in

rheumatoid arthritis, Foot Ankle Int. 29, 293297 2008.
47
W. W. Moses, Time of flight in PET revisited, IEEE Trans. Nucl. Sci.
50, 13251330 2003.
48
H. Zaidi, Scatter modelling and correction strategies in fully 3-D PET,
Nucl. Med. Commun. 22, 11811184 2001.
49
J. P. Carney, D. W. Townsend, V. Rappoport, and B. Bendriem, Method
for transforming CT images for attenuation correction in PET/CT imaging, Med. Phys. 33, 976983 2006.
50
S. Ozgocmen, H. Ozdemir, A. Kiris, Z. Bozgeyik, and O. Ardicoglu,
Clinical evaluation and power Doppler sonography in rheumatoid arthritis: Evidence for ongoing synovial inflammation in clinical remission,
South. Med. J. 101, 240245 2008.
51
E. H. Elzinga, C. J. van der Laken, E. F. Comans, A. A. Lammertsma, B.
A. Dijkmans, and A. E. Voskuyl, 2-Deoxy-2-F-18fluoro-D-glucose
joint uptake on positron emission tomography images: Rheumatoid arthritis versus osteoarthritis, Mol. Imaging Biol. 9, 357360 2007.
52
F. M. McQueen and M. Ostergaard, Established rheumatoid arthritis
New imaging modalities, Best Pract. Res. Clin. Rheumatol. 21, 841856
2007.
53
G. Alexandrakis, F. R. Rannou, and A. F. Chatziioannou, Tomographic
bioluminescence imaging by use of a combined optical-PET OPET system: A computer simulation feasibility study, Phys. Med. Biol. 50,
42254241 2005.
54
T. C. van der Pouw Kraan, L. G. van Baarsen, F. Rustenburg, B. Baltus,
M. Fero, and C. L. Verweij, Gene expression profiling in rheumatology,
Methods Mol. Med. 136, 305327 2007.
55
S. Zrioual, M. L. Toh, A. Tournadre, Y. Zhou, M. A. Cazalis, A. Pachot,
V. Miossec, and P. Miossec, IL-17RA and IL-17RC receptors are essential for IL-17A-induced ELR+ CXC chemokine expression in synoviocytes and are overexpressed in rheumatoid blood, J. Immunol. 180, 655
663 2008.
56
K. Andreas, C. Lubke, T. Haupl, T. Dehne, L. Morawietz, J. Ringe, C.
Kaps, and M. Sittinger, Key regulatory molecules of cartilage destruction in rheumatoid arthritis: An in vitro study, Arthritis Res. Ther. 10, R9
2008.
57
A. Daskalakis, D. Cavouras, P. Bougioukos, S. Kostopoulos, D. Glotsos,
I. Kalatzis, G. C. Kagadis, C. Argyropoulos, and G. Nikiforidis, Improving gene quantification by adjustable spot-image restoration, Bioinformatics 23, 22652272 2007.
58
Y. Juarranz, I. Gutierrez-Canas, B. Santiago, M. Carrion, J. L. Pablos, and
R. P. Gomariz, Differential expression of vasoactive intestinal peptide
and its functional receptors in human osteoarthritic and rheumatoid synovial fibroblasts, Arthritis Rheum. 58, 10861095 2008.
59
E. Simelyte, G. Criado, D. Essex, R. A. Uger, M. Feldmann, and R. O.
Williams, CD200-FC, a novel antiarthritic biologic agent that targets
proinflammatory cytokine expression in the joints of mice with collageninduced arthritis, Arthritis Rheum. 58, 10381043 2008.
60
A. Komatsuda, H. Wakui, T. Nimura, and K. I. Sawada, Reversible
infliximab-related lymphoproliferative disorder associated with EpsteinBarr virus in a patient with rheumatoid arthritis, Mod. Rheumatol. 18,
3452
315318 2008.
J. K. Jackson, T. Higo, W. L. Hunter, and H. M. Burt, Topoisomerase
inhibitors as anti-arthritic agents, Inflamm. Res. 57, 126134 2008.
62
H. Okamoto, Y. Katagiri, A. Kiire, S. Momohara, and N. Kamatani, Serum amyloid A activates nuclear factor-kappaB in rheumatoid synovial
fibroblasts through binding to receptor of advanced glycation endproducts, J. Rheumatol. 35, 752756 2008.
63
U. J. Netz, J. Beuthan, and A. H. Hielscher, Multipixel system for gigahertz frequency-domain optical imaging of finger joints, Rev. Sci. Instrum. 79, 034301 2008.
64
X. Wang, D. L. Chamberland, P. L. Carson, J. B. Fowlkes, R. O. Bude, D.
A. Jamadar, and B. J. Roessler, Imaging of joints with laser-based photoacoustic tomography: an animal study, Med. Phys. 33, 26912697
2006.
65
R. G. Cowell, A. P. Dawid, and S. L. Lauritzen, Probabilistic Networks
and Expert Systems Springer, Berlin, 1999.
66
F. V. Jensen and T. D. Nielsen, Bayesian Networks and Decision Graphs,
2nd ed. Springer, New York, 2007.
67
R. E. Neapolitan, Learning Bayesian Networks Pearson Prentice-Hall,
Englewood Cliffs, NJ, 2004.
68
J. Pearl, Probabilistic Reasoning in Intelligent Systems Morgan Kaufmann, San Francisco, CA, 1988.
69
D. Poole, Logic, knowledge representation, and Bayesian decision
theory, in Computational LogicCL 2000 Springer, Berlin/Heidelberg,
2000, Vol. 1861/2000, pp. 7086.
70
N. Cruz-Ramirez, H. G. Acosta-Mesa, H. Carrillo-Calvet, L. A. NavaFernandez, and R. E. Barrientos-Martinez, Diagnosis of breast cancer
using Bayesian networks: A case study, Comput. Biol. Med. 37, 1553
1564 2007.
71
J. P. Sacha, L. S. Goodenday, and K. J. Cios, Bayesian learning for
cardiac SPECT image interpretation, Artif. Intell. Med. 26, 109143
2002.
72
G. C. Sakellaropoulos and G. C. Nikiforidis, Development of a Bayesian
network for the prognosis of head injuries using graphical model selection
techniques, Methods Inf. Med. 38, 3742 1999.
73
Bayesian Networks for Expression References, http://www.stat.wisc.edu/
~yandell/statgen/reference/network.html, accessed January 31, 2008.
74
R. Jansen, H. Yu, D. Greenbaum, Y. Kluger, N. J. Krogan, S. Chung, A.
Emili, M. Snyder, J. F. Greenblatt, and M. Gerstein, A Bayesian networks approach for predicting protein-protein interactions from genomic
data, Science 302, 449453 2003.
75
S. M. McMahon, Quantifying the community: Using Bayesian learning
networks to find structure and conduct inference in invasions biology,
Biol. Invasions 7, 833844 2005.
76
C. J. Needham, J. R. Bradford, A. J. Bulpitt, and D. R. Westhead, A
primer on learning in Bayesian networks for computational biology,
PLOS Comput. Biol. 3, e129 2007.
77
S.Y. Kim, in Computational Methods in Systems Biology Springer, Berlin, 2003, Vol. 2602, pp. 104113.
78
T. S. Kuhn, The Structure of Scientific Revolutions The University of
Chicago Press, Chicago, 1962.
61
Towards new functional nanostructures for medical imaging

Naomi Matsuuraa and J. A. Rowlands
Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5, Canada
Received 2 March 2008; revised 10 July 2008; accepted for publication 10 July 2008;
Nanostructures represent a promising new type of contrast agent for clinical medical imaging
modalities, including magnetic resonance imaging, x-ray computed tomography, ultrasound, and
nuclear imaging. Currently, most nanostructures are simple, single-purpose imaging agents based
on spherical constructs e.g., liposomes, micelles, nanoemulsions, macromolecules, dendrimers, and
solid nanoparticle structures. In the next decade, new clinical imaging nanostructures will be
designed as multi-functional constructs, to both amplify imaging signals at disease sites and deliver
localized therapy. Proposals for nanostructures to fulfill these new functions will be outlined. New
functional nanostructures are expected to develop in five main directions: Modular nanostructures
with additive functionality; cooperative nanostructures with synergistic functionality; nanostructures activated by their in vivo environment; nanostructures activated by sources outside the patient;
and novel, nonspherical nanostructures and components. The development and clinical translation
of next-generation nanostructures will be facilitated by a combination of improved clarity of the in
vivo imaging and biological challenges and the requirements to successfully overcome them; development of standardized characterization and validation systems tailored for the preclinical assessment of nanostructure agents; and development of streamlined commercialization strategies and
pipelines tailored for nanostructure-based agents for their efficient translation to the clinic. 2008
Key words: medical imaging, nanostructures, nanotechnology, contrast agents
I. INTRODUCTION
Imaging is a fundamental tool in the practice of modern
medicine. In parallel with the development of more sensitive
detectors and imaging systems, there is increasing interest in
designing new types of contrast agents for all modalities, to
sensitively and successfully diagnose specific disease pathologies. Exogenous contrast agents are designed to differentially scatter, absorb, or emit radiation such that upon introduction into a patient, the contrast agents location may be
distinguished from its surrounding tissue or background
noise. This may greatly increase the utility of imaging, as
normal and diseased tissues typically exhibit minimal native
contrast. Since the in vivo distribution of the contrast agent
depends on the patients physiology and the properties of the
agent, the radiologist can use contrast agents to reveal structural, functional, and/or molecular information about the patient to assist in disease diagnosis. At the forefront of new
contrast agent development are nanometer-scale structures,
or nanostructures, that can act as, or carry, contrast agents.
Nanostructures are constructs with physical dimensions
between a few nanometers to hundreds of nanometers in size.
Although simple nanostructures have been utilized for decades in clinical medicine e.g., colloidal and particulate carriers for pharmaceutical applications,1,2 and radiocolloids for
imaging and therapy3, there has been a recent revival in
developing new nanostructures of different materials, sizes,
and properties for biotechnological and medical applications
and, in particular, for medical imaging. Much of this re4474
newed academic and commercial interest in the creation of

new nanostructures for all imaging modalities has been due
to the convergence of several factors: The recent capability
to easily synthesize, image, and characterize such nanostructures using inexpensive, standard laboratory tools; improved
insights into the biology of diseases to create a demand for
specialized, targeted contrast agents; advances in imaging
technologies that allow preclinical contrast agents to be easily tested, particularly in animal models; and growing public
curiosity in nanotechnology combined with greater patient
awareness and strong market demand for new and improved
medical treatment options. All these factors have resulted in
increased funding and investment in basic nanotechnology
research.
Exogenous contrast agents have long been used in medical imaging. The first radiographic iodinated contrast media
was demonstrated in 1929 followed by its introduction into
general clinical use in the 1950s4. Since then, thousands of
new contrast agents for all imaging modalities have been
proposed in the research literature. These agents are typically
designed for a specific imaging modality and/or the diagnosis of particular disease/organ-dependent pathologies, which
has led to a natural and convenient end-user-based classification scheme for contrast agents based on modality or
disease/organ type. Imaging modality-dependent classification is based upon the specific contrast mechanism upon
which the imaging modality is based, independent of the
size, composition, or function of the particular contrast
agent. Similarly, contrast agents can be classified to image
0094-2405/2008/3510/4474/14/$23.00
4474
4475
N. Matsuura and J. A. Rowlands: Functional nanostructures
particular characteristics of specific diseases and disease

sites, for example, cancer, angiogenesis, atherothrombosis,
and musculoskeletal, neurological, and cardiovascular diseases.
Contrast agents may also be classified by size, independently of imaging modality. The general subdivisions are
small molecule contrast agents on the order of 1 nm generally 1000 Da; micron-scale contrast agents larger than
1000 nm; and nanoscale contrast agents between 1 and
1000 nm in size. This size-based subdivision is convenient
for both agent end-users and designers, because the in vivo
properties of nanoscale constructs are generally distinct from
those of small molecules typically purely extracellular
agents and micron-scale structures typically purely intravascular blood pool agents, and because the synthesis for
different sizes/types of agents tend to be similar. Since nanostructures are one to three orders of magnitude smaller than
cells, they can be used to obtain structural and functional
information and to interface with molecular targets on or
within cells, making them very promising for applications in
molecular imaging.511 By definition, nanoscale contrast
agents are always classified by size, irrespective of modality
or function for which they were designed.
The use and properties of different types of nanostructures
as contrast agents have been extensively reviewed in the
literature.1221 Thus, this review will not catalog all current
nanoscale contrast agents, but will rather provide a general
context for the existing literature for the nonspecialist, independent of imaging modality, disease, and/or nanostructure
type, as well as speculate on future developments. We will
introduce the underlying principles and challenges of nanoscale contrast agent development, with the aim of providing
researchers a condensed blueprint upon which innovative
new contrast agents can be rationally designed for clinical
practice.
We begin in Sec. II by briefly summarizing the general
features of nanostructures and factors that must be considered for their use as clinical nanoscale imaging contrast
agents. Section III will expand on our premise that future
development of new functional constructs will not only amplify imaging signals at disease sites for various imaging
modalities, but also deliver localized therapy. We propose
that new functional nanostructures will develop in five basic
directions: a Cooperative nanostructures with added functionality, b modular nanostructures with synergistic functionality, c smart nanostructures that can be activated by
the in vivo environment, d cross-modal nanostructures that
may be activated by means external to the patient, and e
novel, nonspherical, nanostructure components. Section IV
highlights current issues preventing the rapid translation of
preclinical nanostructures into the clinic, including a the
wide knowledge gap between highly specialized and disparate fields, b the lack of specialized, early-stage validation
tools, and c the lack of streamlined commercialization
pipelines for the efficient translation of promising new
agents into the clinic. In Section V are our conclusions.
4475
II. NANOSTRUCTURES FOR CLINICAL MEDICAL

IMAGING: CURRENT STATUS
Current nanostructures under investigation as contrast
agents for magnetic resonance imaging MRI, x-ray computed tomography CT, ultrasound US, and nuclear imaging i.e., positron emission tomography PET, and single
photon emission computed tomography SPECT include
assemblies of molecules e.g., liposomes, micelles, nanoemulsions, large macromolecules and dendrimers, or
solid structures e.g., metal, ceramic, or polymer nanoparticles. The contrast agent may be a major component of the
nanostructure, or it may be a high sensitivity, secondary
component attached to it. Despite significant differences between imaging modalities and nanostructures types, there are
some commonalities and interrelationships between nanostructure design and use in vivo. This section will briefly
summarize a the general features of nanoscale contrast
agents and b factors to be considered for the successful
design of nanoscale contrast agents for the clinic.
II.A. General features of nanoscale contrast agents
One reason why nanostructures are of interest in medical

imaging is because their size imparts them with physiological properties in vivo that are beneficial for imaging specific
pathologies, compared to their smaller molecular 1 nm
or larger vascular 1000 nm counterparts. Since normal
vascular endothelia are permeable to agents less than 2 nm
in size, small molecule contrast agents commonly used in
MRI, CT, and nuclear imaging circulate systemically. Following injection into the intravascular space, they are rapidly
diluted in the circulating plasma volume, and then rapidly
leak out of this space into the extra-vascular extra-cellular
space interstitium.22 This results in short circulation times
i.e., minutes due to their rapid metabolism and
excretion.14,16,22 This short half-life of small molecule contrast agents can limit delivery and accumulation of the agent
to the disease site and thus affect the ability to image diseased tissue above background levels. On the other hand,
larger, micron-scale contrast agents e.g., microbubble contrast agents 1000 to 8000 nm in size used in US
imaging23 cannot escape the vasculature and are consequently limited to blood vessels, which are typically
10 000 to 50 000 nm in diameter.
Comparatively, medical nanostructures, with their intermediate size, can passively extravasate to and accumulate
in various diseased sites that have certain characteristic
leaky pathological vasculature compared to normal healthy
tissue. This leaky pathological vasculature occurs at tumor
sites or in inflamed tissue, and can facilitate nanostructure
accumulation to these sites via the enhanced permeability
and retention EPR effect.24 This property, in combination
with the fact that nanostructures can be designed to have a
prolonged circulatory half-life necessary to reach their target,
has resulted in extensive interest in using nanostructures for
the detection and treatment of diseases with leaky pathological vasculature, such as cancer.13,16,21,2527
4476
Another key feature of medical nanostructures is their potential to carry significant payloads of small molecular or
particulate imaging agents within them or on their surface.
For example, a nanostructure 100 nm in diameter has a
volume equivalent to 100 000 small imaging molecules
and a surface area equivalent to the cross-sectional area of
10 000 small molecules. The localization of high concentrations of imaging agents within or on the surface of nanostructures combined with their preferential accumulation in
target sites can enhance the image signal-to-noise ratio, reduce dosage, and limit toxicity by decreasing distribution of
the imaging nanostructure in normal tissue.28
In addition to the EPR effect, nanostructures can potentially be designed to actively segregate to diseased tissue
or indicators of disease, through their conjugation to targeting moieties e.g., peptides, aptamers, or antibodies that specifically bind to surface epitopes or receptors preferentially
overexpressed at disease sites.2932 This type of active targeting may enhance the accumulation of nanostructures to disease sites, which is the reason behind the proposed use of
nanostructures for future molecular imaging applications.
Nanostructures may also be used to revive small molecule
agents that were previously limited by their innate nonideal
in vivo properties. For example, the incorporation of imaging
agents within nanostructures may increase the stability of
existing small molecule agents with short half-lives and
rapid clearance12,14,16,28 and allow molecular agents previously limited by their poor aqueous solubility to be used in
imaging and therapy applications.12,16,28,3335
II.B. Rational design of nanoscale contrast agents
Much of the recent focus in nanostructure development is

the rational design of new contrast agents for clinical imaging. In general, a rational design must consider three interrelated factors: The imaging modality proposed to detect the
contrast agent, the characteristics of the disease pathology to
be imaged, and the behavior of the contrast agent in vivo.
This is contrary to the design of nanostructures for pharmaceutical drug delivery, nonspecific imaging probes, or imaging agents for nonhuman use, which typically require only
two of the three interrelated factors in their design Fig. 1.
For imaging, since the nanostructure must provide sufficient signal to identify disease pathologies and address a specific biological question, the contrast agent is typically designed to maximize its imaging efficacy for a particular
imaging modality. The imaging modality and thus contrast
mechanism is selected based on the spatial and temporal
resolution desired; the location, distribution, orientation, and
concentration of the target i.e., the diseased tissue; the
specificity and sensitivity that can be achieved; the source
and locations of background and nonspecific signal; and the
requirement for target quantification.
In parallel, the nanostructure design must take into consideration the accessibility of the target and microscopic locations of the epitopes e.g., vascular, interstitial, membrane,
or intracellular, and its ability to reach the disease site by
overcoming biological delivery barriers as determined by its
4476
FIG. 1. The rational design of new nanostructures for clinical imaging gray
must holistically consider the imaging modality blue, the imaging target
yellow, and the in vivo properties pink. This differs from the design of
contrast agents for nonhuman subjects green, nonspecific agents purple,
or targeted drug delivery vehicles orange that typically only have to consider two of these three factors.
in vivo properties i.e., its pharmacokinetics, pharmacodynamics, and biodistribution.3639 There are also specific imaging modality-related accessibility issues that must be considered, e.g., the critical condition that relaxation-enhancing
MRI contrast agents have access to water.
Target accessibility must not only consider spatial accessibility, but how long the agent can circulate in the patient to
reach its target before being cleared from the system. The
nanostructures need to be stable enough to circulate in the
blood long enough to localize to disease sites while retaining
their payload/targeting ligands e.g., at least 2 to 3 h are
required for all blood to complete passage through a remote
vascular bed40. Additionally, the nanostructures must be designed to take into account possible elimination by the
mononuclear phagocytic system MPS. This can happen
within a few minutes and minimizes the quantity of nanostructures passing by the target site and, if actively labeled,
significantly decreases their interactions with target antigens.
To avoid MPS elimination the nanostructure surface is typically cloaked, frequently with a hydrophilic polymer e.g.,
polyethylene glycol PEG38,39,41,42 with low immunogenicity and antigenicity.
In opposition to the strategy of avoiding MPS elimination,
nanostructures can be designed for maximal uptake by macrophages in vivo to track their fate for direct and indirect
imaging of disease. For example, certain dextran-coated,
long-circulating, superparamagnetic iron oxide SPIO nanoparticles are preferentially taken up by macrophage cells, and
these cell-encapsulated long-lived entities can thus passively target pathological inflammatory processes for MR
imaging of atherosclerotic plaques.43 However, for effective
in vivo use, the degree to which only target macrophages are
labeled must be established for their use in this capacity.
Alternatively, cell tracking the passive ex vivo labeling
4477
of cells with nanoscale imaging nanostructures followed by

reinjection into the patient may be another way to counter
issues related to MPS elimination while achieving labeling
of specific cells without issues related to active, in vivo targeting. This concept is in the early stages of being used to
track labeled stem cells and cancer cells,44,45 and to image
retransplanted cell populations e.g., islets of Langerhans46
in animal models. Such ex vivo labeling methods will not
only require validation that the labeled cells will survive reinjection, but that they will retain their cellular functions as
compared to healthy, unlabeled cells.
Active targeting applications for nanoscale contrast agents
require the appropriate selection of a suitable biomarker, the
development of a suitable conjugation strategy for the effective binding of the targeting ligand to the nanostructure surface, and the affinity of the selected targeting ligand to the
biological target of interest. It is also critical to consider the
immunogenicity of the targeting ligands and proposed strategies for conjugation to the nanostructure in their selection.
For example, a common conjugation strategy in in vitro biodiagnostic applications is the avidin-biotinylated ligand peptide or antibody linkage system, which is rapidly cleared by
the liver47 and is thus not feasible for in vivo clinical imaging.
For translation of research-level nanostructures to clinical
imaging, the nanostructures biocompatibility i.e., the potential short-term and long-term toxicity to the patient is arguably their biggest hurdle. This aspect differentiates
nanostructures for use in humans from those used for
ex vivo/in vitro cell studies, biodiagnostics, and/or in vivo
animal studies where such probes have found significant utility and success. The nanoscale quantum dot QD particle, a
successful optical probe for nonhuman applications, is an
example of one such clinically limited imaging agent due to
its toxicity.14,27,35 In addition to not affecting cellular function, for potential clinical utility, the nanoscale imaging
nanostructure must be formulated with either completely
nontoxic components, and/or must biodegrade to clearable
components, and/or must be small enough i.e., with a hydrodynamic diameter 5.5 nm to permit complete elimination from the body.48 These toxicity and elimination studies
must be evaluated in both healthy and unhealthy patients.
Such constraints drastically limit the choice of materials and
types of nanostructures that can be used for the development
of new nanoscale imaging agents for clinical applications.
To a large extent, these three interrelated factors of imaging modality, imaging target, and in vivo properties will 1
restrict the type of construct e.g., liposomes, micelles, emulsions, polymers, dendrimers, and solid nanoparticles that
can be used, 2 determine its parameters i.e., its
composition/structure, size, size distribution, surface properties, and stability, and 3 dictate its in vivo properties. The
particular design challenges and constraints of new nanostructures for clinical medical imaging have resulted in an
individualized approach to optimizing unifunctional, imaging constructs tailored for a single modality and specific disease sites. Such an individualized approach forces an early,
preclinical stop-go decision on the new agents potential
4477
clinical utility. This motivates researchers to either synthesize new agents with only incremental modifications to
clinically accepted technologies or to partially develop
high-risk/high-reward-type agents without rigorous testing or
specific optimization for clinical use. This current paradigm
for agent development is inefficient, expensive, leads to duplication of effort, and has a disproportionate reliance on
luck and/or an ab initio, perfectly defined imaging/biology
problem. This suggests that new, more general approaches to
the design of functional nanostructures are needed.
III. TOWARDS NEW FUNCTIONAL
NANOSTRUCTURES
In our opinion, the next decade will see the rational design of a new generation of nanostructures for medical imaging progressing towards the inclusion of added functionality, with a focus on effective treatment and improved patient
outcome resulting from more sensitive and specific detection
of disease. Already, the role of nanostructures in medical
imaging is expanding toward this goal, from simple detection
of disease to characterization/diagnosis of disease, local application of therapy, real-time assessment of response to
therapy, and as a tool in the drug discovery and development
process.11,4952 The continuing evolution of current nanostructures towards multifunctional, nanoscale medical devices has been explicitly and implicitly anticipated in the
recent literature.11,16,18,26,27,29,5358
We propose that new functional nanostructure designs
will move away from the current paradigm of relatively
simple, spherical nanostructures encapsulating imaging
agents and/or targeting diseased tissues by size-specific or
mono-ligand attachment. In this section, concepts that can
facilitate the development of novel, multifunctional nanostructures will be briefly described, along with proposals on
how new nanostructures could fulfill these functions. These
concepts are classified into five main directions, which are
outlined below: a Modular nanostructures with additive
functionality, b cooperative nanostructures with synergistic
functionality, c nanostructures activated by their in vivo
environment, d nanostructures activated by sources outside
the patient, and e novel, nonspherical nanostructures and
components.
III.A. Evolution of modular nanostructures
with additive functionality
The most obvious tactic for the evolution of multifunctional nanostructures that will permit multiple imaging
and/or therapy applications is a linear add-on approach,
where individual components i.e., imaging agents and/or
therapy agents are combined to form a single multifunctional construct.10,59,60
This add-on approach has been explored in preclinical,
multimodal imaging applications. Numerous multimodal
nanostructures have been proposed, including many dualmodal nanostructures e.g., MRI/SPECT,15,61 MRI/PET,61
MRI/optical,20,6264 and CT/MRI65, and a few tri-modal reporters e.g., for PET, MRI, and fluorescence imaging,66 and
4478
for MR, CT, and US imaging67. It is worth noting that

complementary advances in multi-modal imaging systems
have occurred, which are, in turn, expected to help drive the
rapid development and implementation of new, corresponding multi-modal imaging nanostructures into the clinic. For
example, dual-modality technologies are in various stages of
development. PET/CT and SPECT/CT imaging systems are
commonplace in the clinic as they can provide functional or
molecular information using PET/SPECT molecular contrast
agents in the context of human anatomy using CT imaging.68
Other dual-modality imaging systems include PET/MRI,
optical/PET,61 optical/SPECT,61 MRI/US, and MRI/CT systems. The introduction of new technological systems into the
clinic may pave the way for new niche nanoscale agents to
gain a foothold in the clinical environment.
Similar to this multimodal concept of combining imaging
probes into the same construct, extra nanostructure functionality may be added by combining imaging and therapy
agents. For many decades, FDA-approved nanostructures
have been used for the delivery of cancer therapeutics e.g.,
liposomes, albumin-based particles, nanoparticles, in clinical trials e.g., using polymeric micelles or polymer-based
particles, and in preclinical development e.g., with dendrimers, inorganic or other solid particles.13,25,69,70 Since the
nanostructure is already optimized to localize to its target,
the integration of imaging agents to existing therapy agents
in the same construct, and vice versa, can allow for localized
imaging as well as treatment.16,18,26,27,58,71,72 Furthermore,
the addition of imaging agents to a therapeutic nanoscale
construct may alter the acceptable toxicity profile compared
to an imaging agent alone. Since both imaging and therapeutic nanostructures have been independently used in
humans,12 combining imaging and therapy agents is expected
to be the most likely path for future functional nanostructures
to be translated to clinical use.
One of the more intriguing subconcepts to this add-on
approach is the potential to make designer agents via
modular nanoscale construct assembly. By decoupling the
nanostructure components from each other, independently
assessing and optimizing these components, and reassembling them into nanostructures using optimized components,
constructs tailored for specific imaging modalities and disease pathologies may be easily designed. These new composite nanostructures may be rapidly translated to the clinic.
Although the properties of the overall construct may be
somewhat modified using this approach, this differs greatly
from molecular conjugates, whose behaviors in vivo are so
particular to their specific molecular structures that individual characterization of even slightly different agents is a
necessity.
Tailoring the components of the nanostructure in a
pseudo-modular fashion may be accomplished because many
imaging nanostructures already are comprised of an intrinsic
bi-component, core-shell-type structure, which may facilitate
similar types of molecules or components replacing one another. Such replacements can drastically change its function
as an imaging and/or therapy agent, but minimally change its
in vivo properties. For example, liposomal and micelle-based
4478
nanostructures can carry contrast agents for virtually all modalities based on this principle.15,17 Also, the structure of
nanoscale constructs may accommodate payloads of mixed
molecular or particulate agents within their volumes or on
their surfaces.
Multifunctional nanostructures comprised of truly modular components will require significant technical advancements, particularly in the area of coatings to integrate different modular components. However, technical progress
may be fast because this approach has already proved to be a
valuable tool for the early, preclinical characterization of
new nanostructures. Not only will this lead to faster assessment and iteration of clinically useful nanostructures at an
early stage of development, new common platform technologies and strategies can thus be developed for the future effective integration of disparate modular components for
clinical imaging. For example, many new imaging nanostructures are being designed to contain an optical marker for
preclinical cross-validation.20,6264 The high sensitivity and
subcellular spatial resolution of optical agents can be used to
assess different targeting strategies in vitro, without altering
the overall properties of the nanostructure. Useful information about the agent may then be obtained prior to moving on
to in vivo models. In the same manner, other validation molecules e.g., those for histopathology60 may be integrated
into early-stage nanostructures to confirm the behavior and
pharmacokinetics of preclinical agents for faster translation
into the clinic.
In general, designing multimodal and multifunctional
nanostructures will remain challenging compared to designing unimodal or unifunctional nanostructures. The integration of two or more different agents within the nanostructure
will add a level of complexity to their synthesis, will decrease the loading capability of each individual agent, and
may change the overall nanostructures pharmacokinetics
and stability. To counter this, nanostructures that provide intrinsic contrast in more than one imaging modality can be
exploited. For example, Gd-chelates can provide contrast in
both MR and CT.73 Perfluorocarbon droplets and their derivatives have been used to provide contrast in CT,7476 US,77
and/or MR7880 imaging, and have also been demonstrated as
drug delivery vehicles81 and as radiosensitizers.82
III.B. Evolution of cooperative nanostructures systems
with synergistic functionality
The development of new functional nanostructures may

also occur through the evolution of nanostructures designed
to have an increased and predictable ability to interact with
their targets and/or other nanostructures. Instead of changing
the components of a single nanostructure to increase its functionality in an additive function, it may be possible to design
nanostructures such that their cooperative interaction between multiple constructs and/or their intended target allows
them to increase their utility as an imaging and/or therapeutic
agent.
Modifications of the constructs size or surface properties
may lead to changes in their interactions with the target rela-
4479
tive to normal tissue such that the contrast agents can accumulate disproportionately i.e., nonlinearly in the disease
site, or result in a spatial distribution of individual nanoscale
constructs that will otherwise provide signal amplification.
Such signal amplification is significant in imaging applications where increasing signal to noise can decrease the lower
limit for detection currently on the order of hundreds of
millions of cells, as well as in pharmaceutical delivery
where any advantage in therapeutic ratio may play a role in
improved patient outcome.
The simplest example of nonlinear, preferential accumulation of nanostructures is the localization that occurs via the
EPR effect at disease sites with leaky pathologies. In addition to size-dependent accumulation, further nonlinear accumulation may occur by modifications to the nanostructures
surface. This is the underlying concept behind enhancing
size-dependent nanostructure localization in diseased tissue
through active targeting, or the simple attachment of targeting molecules to the nanostructures surface for further imaging or therapeutic amplification.13,16,21,2527 In opposition
to the concept of preferential accumulation at disease sites,
the nanostructures surface may be tailored to assist its clearance and/or elimination from nontarget tissue to reduce background noise and toxicity.8385
More complex strategies for nonlinear accumulation are
possible. This concept is particularly beneficial for the preferential discrimination of disease sites from normal tissues
that express the same biomarkers of disease but at different
levels. The large surface area of the nanostructure compared
to targeting ligands which range from 1 to 20 nm in
size potentially allows the attachment of multiple targeting
ligands. This may facilitate multivalent binding to cell surface receptors such that collections of low-affinity ligands
can have very high effective affinities.8688 This not only
can greatly expand the range of useful, active targeting
ligands, but may allow nanostructures to be conjugated to
multiple weak-binding ligands for disease-specific biomarkers such that they would bind only to sites expressing abnormally high levels of the biomarker, but not at normal, lower
levels. Further enhancement of preferential targeting may be
accomplished by using several sets of nanostructures, each
labeled with weak-binding ligands for different biomarkers
expressed by a disease. A similar strategy, but more technically challenging, would be to use several different biomarkers on the same nanostructure such that disease targets that
express relative proportions of a series of agents will result in
preferential nanostructure accumulation.
The accumulation of nanostructures at a disease site may
be augmented through the use of a series of different sets of
targeted nanostructures. After saturating the disease site with
a first set of targeted nanostructures, it may be possible to
add a second, labeled nanostructure that could attach only to
multiple nanostructures of the first kind, and so on, such that
imaging and/or therapeutic nanostructures could preferentially, and controllably, accumulate at the disease site. In order to prevent nonspecific, uncontrolled accumulation, which
could lead to thrombosis and/or high background imaging
signals, each type of nanostructure must be fully cleared
4479
from the system before introduction of the next, which may

be feasible if each nanostructures surface ligands require
multivalent binding for attachment.
Other cooperative effects between nanostructures can be
used to help diagnose disease, e.g., through exploitation of
the expected spatial distribution of targeted nanostructures at
disease sites. Since agglomerated iron oxide nanostructures
result in a significantly larger MR signal than the same mass
of material dispersed into multiple, smaller units,89 targeted
iron oxide nanoparticles accumulated at a disease site may
cooperatively result in an amplified signal. Other preclinical examples of nanostructures that may result in amplified
imaging signals after accumulation compared to their dispersed forms include perfluorocarbon droplets, which give
an enhanced backscatter signal in US imaging,90 and aggregated gold nanostructures, which result in enhanced
photothermal-based imaging and therapy in live cells.54 Also,
as the relaxation properties in Gd-based MRI contrast agents
are dependent on their attached ligands and/or surrounding
environment,91 and size and surface effects have been shown
to affect the properties of iron-based nanostructures,92 their
aggregation and/or successful attachment to their target may
significantly increase their signal such that they could be
distinguished from background noise.
There are significant challenges in designing new agents
using these proposed cooperative and synergistic approaches.
These challenges lie in the effective and controlled conjugation of different targeting molecules to the surfaces of nanoscale structures, and the effective synthesis of new agents
with such nonlinear properties, which will, by definition, be
very sensitive to these factors. The determination of effective
amplification strategies for nanostructure imaging will result
from the development of highly nonlinear contrast agent
configurations, and correlations with disease parameters,
such as biomarker/target distribution in vivo. Appropriate
combinations of various low-affinity targeting ligands,
and/or their disease-dependent distribution, must also be well
understood in order to detect disease sites compared to normal tissue.
III.C. Evolution of smart multifunctional
nanostructures that may be activated by the in vivo
environment
Recently, new types of smart contrast agents, designed

to activate exclusively in the presence of their intended target, have been proposed.6,7,49,51 Due to the absence of a
background signal, these agents have a significant signal advantage over simple targeted agents. Most of the more sophisticated examples of these sensor-type, activatable contrast agents in the research literature are small molecule
agents in the early stages of development, including MR
agents that exhibit an enhancement in contrast in the presence of -galactosidase,93 and optical imaging probes activatable by enzymes.6,49,51,94,95 The development of activatable nanostructures rather than activatable molecular agents
may potentially result in even higher signal to noise ratios,
because the activation may be magnified by the larger size/
4480
volume of the nanostructure. They can also have a significantly longer half-life to accumulate and localize in greater
numbers in diseased tissue.
The most straightforward type of environmentally activatable nanostructures simply dissolve and/or aggregate in certain in vivo environments. In fact, the controlled, environmentally dependent degradation of nanostructures into their
molecular components in vivo has been the basis for pharmaceutical delivery of drugs for decades.96,97 More recently,
pH- and temperature-responsive polymeric nanostructures
for potential imaging and drug delivery applications have
been reported in the preclinical research literature.98103
Simple nanostructures can also aggregate in certain in vivo
environments resulting in a significant change in imaging
signal. For example, preclinical iron oxide-based nanostructures for MRI have been shown to undergo reversible clustering in the presence of enzymes, chemical compounds,104
or analytes e.g., glucose105 or calcium106.
One intriguing concept that has not been extensively explored in medical imaging is the controlled in vivo assembly
of nanostructures at the target site. Many nanostructures are
composed of atoms or molecules that assemble together via
intermolecular bonding using weak chemical bonds. Since
nanostructures have large surface-area to mass ratios, surface
effects e.g., intermolecular bonding and surface-dominated
reactions e.g., surface oxidation, saturation of dangling surface bonds are strongly enhanced in nanoscale constructs.
For example, the weak noncovalent forces between molecular components that dominant at the nanoscale i.e., the constructs stability are highly sensitive to slight changes in
their surrounding environmental conditions e.g., temperature, pH107109.
These environmentally sensitive surface properties can be
utilized to modify and/or assemble/disassemble structures
under appropriate conditions in situ, as a function of their
environmental surroundings in a simplified mimicry of biological self-assembly processes i.e., biomimetics.40,110115
Transient, metastable nanostructures can be constructed from
biological or exogenous molecules that will self-assemble
in vivo for short periods of time, depending on the local
environment or external application of energy. This concept
could be expanded further such that nanoscale reaction
vessels116118 for the fabrication of contrast nanostructures
could be assembled within the biological system. The use of
smaller molecular components in vivo that can assemble directly at the target site may allow larger nanostructures to be
built at the target without prior elimination by the MPS.
A simple example of in vivo assembly of a larger structure from smaller nanostructure components can be found in
the preclinical literature, where nanostructures i.e., perfluorocarbon nanodroplets and drug-loaded micelles were injected into an animal model. After injection, the nanostructures underwent temperature-induced coalescence and
conversion to form micron-scale gas-filled microbubbles.119
Further, it was reported that microbubbles created in vivo
not only acted as effective acoustic scatterers for US imaging, they could also be disrupted by US to enhance the delivery of their encapusulated therapeutic cargo.119
4480
The challenges associated with making complex, environmentally activatable contrast nanostructures are that such
nanostructures and their molecular components are currently difficult to synthesize and design, and significant barriers in the complex biological environment of the human
body must be overcome before reaching the target site. One
possible approach to shielding the nanoscale components
from the biological environment is to use a biological entity
as a delivery vehicle. Already, modified lipoproteins120 and
cells have been used as carriers for contrast nanostructures,
and the tracking of stem cells, cancer cells, and macrophages
have been demonstrated in vivo.4345
For environmentally activatable nanostructures to be designed and utilized, and for the considerable technical challenges of synthesizing the agents themselves to be overcome,
it is critical to understand how the contrast signal correlates
to the physical changes in the nanostructure i.e., its degradation and/or self-assembly, how these structural changes
are related to environmental conditions e.g., hypoxia/
oxygenation, pH, and temperature, and how these changes
correlate to disease pathologies.
III.D. Evolution of multifunctional cross-modal
nanostructures that may be activated by devices
external to the patient
New types of nanostructures that can be remotely activated by an energy source external to the patient are also
under investigation. External triggering from energy sources
using different imaging modalities may localize the agent to
its target site, excite the agent to emit a detectable signal
above background noise, and/or produce a local therapeutic
effect.
Externally applied US energy can push contrast nanostructures in the vasculature towards the target site using
radiation force, to increase local binding121 or to cause individual particles to attract each other and coalesce.122 This
same concept may allow US to be used to improve the localization of other i.e., non-US imaging and therapy nanostructures. This type of controlled localization of selected
structures external to the patient may also eventually facilitate the directed self-assembly of larger constructs in vivo at
a target site.
In another example of localization, externally applied
magnetic fields have been used to control and track magnetic
beads using magnetic field gradients to increase localization/
binding to target sites in preclinical systems.123126 The concept of enhanced localization of drugs using this method may
be envisioned, as magnetic particles have been coupled with
pharmaceutical agents124 and therapeutic radiolabeled
molecules.126
Imaging modalities have been used to activate an exogenous contrast nanostructure in vivo from outside the body.
For example, US can convert contrast nanostructures e.g.,
perfluorocarbon emulsions to gas bubbles, thereby significantly increasing their acoustic contrast127 and signal to noise
such that single bubbles may be imaged in vivo. The energy
of the imaging system can also be applied to nanostructures
4481
to directly induce a therapeutic effect and/or to trigger the

release of their encapsulated therapeutic payload in vivo. For
example, in MR, there has been considerable effort in heating iron oxide nanostructures using RF to cause local
hyperthermia.126,128,129
We predict a surge of interest in the development of crossfunctional nanostructures, in which the full suite of clinical
medical imaging technologies will be harnessed for combined imaging and therapy. For example, it is possible to use
the energy from one imaging modality to activate a contrast
agent, and another imaging modality for its detection. Some
preclinical examples include a demonstration that the absorption of light into nanostructures can result in the formation of
gas bubbles at their surfaces which may be detected by US.54
In another example, light was used to induce the release of
MR contrast agents from nanostructure micelles in vitro.130
For cross-modal imaging and/or therapy, it is feasible to use
metal-based nanostructures as both CT contrast agents131,132
and to enhance the RF-based thermal destruction of cancer
cells.133 Magnetic nanostructures have also been combined
with light-activated chemotherapy drugs i.e., photodynamic
cancer therapy PDT for enhanced cell death.53,134
US agents have been used to enhance drug delivery from
the vasculature into tissue across the blood-brain barrier by
causing localized release from a carrier vehicle or by increasing cell membrane permeability by mechanical or thermal
action.135,136 The enhanced delivery of secondary imaging
and/or
therapy
agents
e.g.,
therapeutic
radiopharmaceuticals,137 therapeutic gold compounds138 by
such external means can be easily envisaged. Expanding on
this concept, it may be possible for nanostructures to carry
targeting ligands or other components of self-assembly,
which may be externally triggered to release their contents
near their intended targeting site to form a larger construct
with higher signaling capabilities.
There are several key issues that must be considered in
the development of agents that may be controllably activated
by energy sources external to the patient. The nanostructures
must be activatable using a minimum amount of energy such
that normal tissue is undamaged; the energy used must result
in a high enough signal to detect small numbers of agents
and/or induce a significant therapeutic response; and the dynamic in vivo environment must be considered in the development, validation, and characterization of the more complex, externally activatable agents.
III.E. Evolution of novel, nonspherical, nanostructure
components
Most nanostructures designed for medical imaging are

simple, spherical, so-called zero-dimensional 0D constructs. 0D nanostructures are easy to synthesize by methods
that can be directly adopted from simple, decades-old, synthesis technologies. Although such nanostructures can have
simple functionality e.g., agent loadability, ligand attachment for targeting, and controlled dissolution they have
been largely adopted from other fields e.g., optoelectronic or
drug
delivery
without
any
major
structural
4481
modifications.101,139,140 The main focus in medicine has been

on finding unique medical applications for traditional nanostructures instead of designing new types of unique structures tailored for medicine. In reality, advanced synthesis
techniques developed for the fabrication of intricate 3D
nanoscale architectures remain largely unexplored and underutilized in medical imaging compared to other fields e.g.,
optoelectronics.
Currently, there are methods to synthesize 1D wire-type,
2D sheet-type, and complex 3D hierarchical nanostructures
e.g., carbon nanotubes and fullerenes141, as well as nanostructures of different shapes including rods, cubes, tetrapods, and coils.142145 New nanostructures developed to
maximize contrast phenomena can be fabricated, e.g., nanoscale magnetic coils may be synthesized to label cells or be
targeted to disease sites for tracking and imaging using MRI.
Recent advances in the biomolecular-assisted assembly of
nanostructures using biological components such as DNA,
antibodies, viruses, and proteins, as well as the development
of various hybrid materials based on proteins that bind specifically to inorganic materials,111115 may allow unique, fully
biocompatible, in vivo self-assembly of complex functional
devices to take place.
These nanostructures may in turn be used as building
blocks or components for new 2D/3D hierarchical superstructures by their controlled aggregation using noncovalent interactions between them. Novel nanoscale components may allow nonspherical hierarchical structures to
assemble in situ to form new devices at length scales that can
be adaptable according to their in vivo environment, and
thereby potentially result in contrast agents with tunable
properties. For example, nanoscale cantilevers26 could be assembled in vivo such that one end could be attached to the
disease site while the other end is free. The formation of such
structures could be binary, dependent on the initial deposition of a minimal number of priming ligands deposited at
the disease site. If such cantilever-type nanostructures can be
assembled at disease sites and are sensitive to US frequencies, they could be used as an adjunct to traditional spherical
gas or fluid-filled contrast agents for US imaging, but with
completely different, and tailored, material properties e.g.,
composition, size, surfaces, and stability. This means that
traditional limitations as determined by the imaging modality, biological target site, and/or in vivo properties may be
circumvented through the synthesis of, effectively, a new designer material that does not pre-exist in nature.
The forces acting on nonspherical nanostructures will be
orientationally and environmentally dependent, which could
be utilized in different hydrodynamic environments.26,146,147
For example, nanorods, with different functionalization at
the tips than on the lengths,144 may be designed to selfassemble into different structures depending on the diseasedependent hydrodynamic and biological environment e.g.,
within tumors compared to in normal vasculature. The understanding of the properties of particular disease pathologies may be exploited in the development of new shape/size/
surface-specific nanostructures to further differentiate slight
pathological differences, much in the same way that the EPR
4482
effect has been the key to developing imaging and pharmaceutical nanostructures for cancer detection and therapy.
The emergence of new imaging modalities should play a
role in propelling the development of new complementary
contrast agents and increase the functionality of existing
nanostructures. One example gaining in popularity is transcutaneous optical imaging, which has led to the development
of biocompatible molecular near-infrared probes.148 In addition to current fluorescence and luminescence methods, other
aspects of optical technology including absorption, emission, and scattering techniques will require the optimization
of new contrast agents to complement these different contrast
mechanisms.
The main challenge in developing new types of nanostructures for functional imaging is that they must truly be
rationally designed for them to be successful. A thorough
understanding of the imaging and disease parameters is necessary, and the dependence on variations in size, composition, and morphology should be established beforehand. The
design of new nanostructure must be balanced with practical
issues of cost and yield, which will depend greatly on the
synthesis method.
IV. STRATEGIES FOR MOVING FORWARD
Despite the innumerable nanostructures outlined in the
preclinical research literature, the translation of new nanostructures into humans is in the early stages. For example, in
2006, the only marketed nanoscale contrast agents were three
superparamagnetic iron oxide nanoparticle contrast agents
for MRI.69 To facilitate the rapid progression of new, functional nanostructures into the clinic, we must address three
main challenges: a Improving the clarity of both the desired
nanostructure properties and the challenges needed to be
overcome for their successful implementation, b determining better methods to test and validate the new nanostructures early in their development, and c developing accepted
and established strategies for the efficient translation of new
nanostructures from the lab bench into the clinic.
IV.A. Improved clarity of desired properties and
existing challenges
The development of new functional contrast agents requires the convergence of highly specialized and disparate
fields. In addition to overcoming the biological issues effective delivery to the target, biocompatibility, toxicity and fabrication issues synthesis, yield, quality assurance as outlined earlier in this review, it is also necessary to overcome
medical imaging issues to obtain sufficient contrast for imaging.
Through collaboration with physicians, biologists, and
chemists, medical physicists must determine the appropriate
design parameters, such as minimum required loading, and
ideal agent and imaging timepoints, in parallel with the spatial resolution required for disease diagnosis. For contrast
agent design and determination of minimum detection limits,
the expected number, location, and distribution of the target
should be known. The overall suitability of the particular
4482
imaging modality also depends on the expected target distribution within the patiente.g., whole body tomographic
imaging can be done with CT, MRI, and nuclear imaging,
but is not commonly undertaken with US or optical imaging.
For combined detection and therapy, the imaging properties
must be weighed with the desired effect of the therapeutic
agent, as determined by its degradation rate, concentration,
and effective target delivery. All these factors must be considered for the efficient development of new nanostructures
for medical imaging, and will facilitate the design of new
contrast agents and imaging techniques to allow not only the
agents detection but also their quantification, allowing specific information about the changes in expression levels of
the target to be obtained.
The wide knowledge gap between the disparate fields involved in the design of new functional nanostructures must
be addressed. Critical bottlenecks in nanostructure agent development are often not elucidated preventing progress in
creating suitable nanostructures for medical imaging. Recently, many institutes and government agencies have recognized the importance of communication within this interdisciplinary ensemble of highly specialized fields and have
implemented interdisciplinary teams to increase the efficiency of their interactions.57,149,150
Another simple way to link the scientific community is to
utilize the massive information sharing capability of the internet to develop a database of core parameters that identify
different types of imaging and therapeutic nanostructures,
their correlative compositions, and their imaging and in vivo
properties. Recently, the National Center for Biotechnology
Information NCBI at the National Institutes of Health
NIH implemented such a freely accessible online database
MICAD151 to provide the scientific community working
with in vivo molecular imaging and contrast agents with information aimed to foster research and development. A similar database focusing on nanoscale agents should be developed, supplemented by correlative databases that identify the
parameters desired from an imaging perspective e.g., ideal
type/number/distribution of molecular or atomic units required per unit, size range, amount of agent required, special
considerations/parameters, a biological perspective e.g.,
ideal disease targets, targeting ligands/biomarkers, delivery
paths, safety data, and a commercial/market perspective
e.g., the desired yield, purity and final application/market
demand and size. In addition to being freely accessible,
these databases should be linked, cross-searchable, and able
to accommodate dynamic adaptation by the users. Although
some specific fabrication protocols may be restricted due to
patenting and commercialization concerns, the protocols and
procedures employed for characterization of the agents as
well as their properties should be listed, which will assist in
establishing baseline standards for development of new
nanostructure agents. This massive collaborative effort
should help differentiate which nanostructures are in the
early, mid, and late stages of development; identify major
knowledge gaps and the likelihood of the successful translation of various agents to the clinic; and help scientists to
4483
converge toward the most optimal nanostructure design parameters to yield the greatest patient benefit.
IV.B. Development of early-stage validation and
characterization tools specific for nanostructure
imaging and therapy applications
First, the development of common standards and validation tools for functional nanostructure assessment will be
necessary to streamline and adequately evaluate the thousands of nanostructures under investigation for imaging and
therapy applications reported in the literature. Currently,
even agreement on what constitutes agent size has not been
standardized e.g., hydrodynamic radius versus dry radius, and something as simple as the adoption of common
units to quantify size would help alleviate the confusion between chemists who measure size using atomic mass Daltons and physicists who measure linear dimension nanometers. Also, it would be beneficial to establish goldstandard protocols for the assessment of toxicity/safety and
bioconjugation, and for allowing different agents and numerous and occasionally contradictory protocols in the literature to be compared, while reducing the abundance of research nanostructures to those that have met a known,
accepted standard. In addition, the development of validation
tools specifically for nanostructure imaging agents would be
useful. Currently, most researchers cannot image nanostructures on a subcellular scale without an integrated optical
marker. The development of a suite of subcellular and other
high-resolution imaging tools that mimic clinical systems
may allow early assessment of nanostructures, which could
lead to faster iteration of nanostructure design in the early
developmental stage.
Second, it would be very useful if common platform technologies could be developed such that the targeting ligands,
bioconjugation strategies, surface, interfacial, and capping
layers could be tested independently of the imaging construct. In this way, it may be possible to develop a consistent
description of the behavior of nanostructures in biological
systems as a function of nanostructure size, structure, and
material composition. The ability to tune the surfaces to be
efficiently capped by various targeting/shielding ligands
would also assist in differentiating the properties of the nanostructure from the properties of their surfaces e.g., structure
vs. charge. Current challenges such as hydrodynamic size
effects, pH stability of the composite structures, and reproducibility biological versus structural issues may be addressed using universal nanoscale constructs. This can lead
to predictive models of nanostructure systems, and further to
the engineering of specific nanostructures tailored to biological applications.
Already, the concept of using standardized, validated
nano-based imaging agents for the real-time assessment of
the efficacy of therapeutic regimens as a complement to, or a
substitute for, conventional end point analysis has been
proposed.26 This application can assist in the design and advancement of standardized, nanostructure-based validation
tools.
4483
IV.C. Development of commercialization pipelines,

scale-up strategies, and regulatory process
negotiation for efficient clinical translation
The move from laboratory-based probes to clinical agents

is long, arduous, and extremely costly. Since nanoscale
agents use macroscopic amounts of material, a full drug toxicity study must be conducted, and this realistically requires
a market/application that can support these costs. The fact
that IP protection must typically be obtained for clinical
translation of new agents is another issue that must be addressed. This is not unique to imaging probes, but is an issue
spanning the biotechnology and drug sectors. Suitable models to overcome considerable commercialization, scale-up,
and regulatory hurdles are necessary for this
translation.26,40,69,150,152154
Stringent regulatory approval issues will limit the translation of most new nanostructures to the clinic.26 In addition to
the assessment of their biocompatibility, biodistribution, and
production protocols, nanostructures could fall under the
drugs, medical devices, and biological agents branches of the
Food and Drug Adminstration FDA,26 further complicating
an already unpredictable process. The time and cost required
to translate a new agent for clinical use is considerable
although estimates vary, the translation of a single agent
from bench to bedside requires a commitment of 3561 million dollars and 2 to 4 years.153 In general, countries outside
the United States have a less stringent approval process, with
biomedical products typically marketed in Europe and/or
Canada for 2 3 years before they receive US regulatory
approval,40 which often provides an opportunity to gain extra
clinical experience with a product prior to FDA review.40
The transfer of knowledge into practice requires a streamlined approach and the identification and negotiation of
bottlenecks in the technology transfer pipeline. Piggybacking
off of prevalidated or semi-validated imaging or pharmaceutical agents to develop new products, prior working knowledge of the intricacies of the approval process, and strong
assessment of the nanostructures prior to clinical trials may
be a few ways to increase the chances of translation to the
clinic. The faster establishment of safe regulatory approval
protocols would ameliorate concerns about the length of time
it takes for agents to be assessed by the FDA.26
V. CONCLUSIONS
Researchers are on the path towards designing functional
nanostructures through the increased sensitivity of imaging
modalities, the advancement of nanoscale fabrication and
characterization technologies, and a better understanding of
the biology of disease, including the sequencing of the genome that has identified new potential disease targets.6,155
Future nanostructures will be designed to have added functionality through assembly of prevalidated, modular components, or will be designed with amplified functionality
through synergistic effects. Such nanostructures can be activated for imaging and/or therapy applications either by their
in vivo disease environment or through activation external to
the patient. Also, the development of novel nanostructures
4484
will facilitate the design of entirely new nanoscale agents for

imaging and therapy applications. The development and
clinical translation of next-generation nanostructures will be
facilitated by a combination of improved clarity of the in
vivo imaging challenges and the requirements to successfully
overcome them; development of standardized characterization and validation systems tailored for the preclinical assessment of nanostructure agents; and development of
streamlined commercialization strategies and pipelines tailored for nanostructure-based agents for their efficient translation to the clinic.
ACKNOWLEDGMENTS
The authors would like to acknowledge Dr. Elizabeth Gillies University of Western Ontario, Dr. Greg Stanisz Sunnybrook Health Science Centre, and Dr. John Valliant McMaster University for critical review of the manuscript. This
work was supported by the CIHR Excellence in Radiation
Research for the 21st century EIRR21st Research Training
Program, and the Ontario Institute for Cancer Research Network through funding provided by the Province of Ontario.
a

[email protected]
1
T. M. Allen and P. R. Cullis, Drug delivery systems: Entering the mainstream, Sciences N.Y. 3035665, 18181822 2004.
2
R. H. Muller and C. M. Keck, Challenges and solutions for the delivery
of biotech drugsa review of drug nanocrystal technology and lipid
nanoparticles, J. Biotechnol., 11313, 151170 2004.
3
R. M. Kniseley, Marrow studies with radiocolloids, Semin Nucl. Med.
21, 7185 1972.
4
P. Dawson, D. O. Cosgrove, and R. G. Grainger, eds., Textbook of
Contrast Media. Isis Medical Media, Oxford, 1999.
5
R. Weissleder, Molecular imaging: Exploring the next frontier, Radiology 2123, 609614 1999.
6
S. R. Cherry, In vivo molecular and genomic imaging: New challenges
for imaging physics, Phys. Med. Biol. 493, R13R48 2004.
7
A. S. Dzik-Jurasz, Molecular imaging in vivo: An introduction, Br. J.
Radiol. 76Spec. No. 2, S98S109 2003.
8
R. J. Gillies, In vivo molecular imaging, J. Cell Biochem. Suppl. 39,
231238 2002.
9
H. R. Herschman, Molecular imaging: Looking at problems, seeing solutions, Sciences N.Y. 302 5645, 605608 2003.
10
A. Hengerer, A. Wunder, D. J. Wagenaar, A. H. Vija, M. Shah, and J.
Grimm, From genomics to clinical molecular imaging, Proc. IEEE
934, 819828 2005.
11
K. C. Li, S. D. Pandit, S. Guccione, and M. D. Bednarski, Molecular
imaging applications in nanomedicine, Biomed. Microdevices 62, 113
116 2004.
12
O. M. Koo, I. Rubinstein, and H. Onyuksel, Role of nanotechnology in
targeted drug delivery and imaging: A concise review, Nanomedicine
13, 193212 2005.
13
J. R. Heath and M. E. Davis, Nanotechnology and cancer, Annu. Rev.
Med. 59, 251265 2008.
14
S. K. Sahoo, S. Parveen, and J. J. Panda, The present and future of
nanotechnology in human health care, Nanomedicine 31, 2031
2007.
15
V. P. Torchilin, PEG-based micelles as carriers of contrast agents for
different imaging modalities, Adv. Drug Delivery Rev. 542, 235252
2002.
16
V. P. Torchilin, Targeted pharmaceutical nanocarriers for cancer therapy
and imaging, AAPS J. 92, E128E147 2007.
17
C. Tilcock, Delivery of contrast agents for magnetic resonance imaging,
computed tomography, nuclear medicine and ultrasound, Adv. Drug Delivery Rev. 3713, 3351 1999.
18
S. A. Wickline and G. M. Lanza, Molecular imaging, targeted therapeutics, and nanoscience, J. Cell Biochem. Suppl. 39, 9097 2002.
19
4484
S. A. Wickline, A. M. Neubauer, P. M. Winter, S. D. Caruthers, and G. M.

Lanza, Molecular imaging and therapy of atherosclerosis with targeted
nanoparticles, J. Magn. Reson Imaging 254, 667680 2007.
20
W. J. Mulder, G. J. Strijkers, G. A. van Tilborg, A. W. Griffioen, and K.
Nicolay, Lipid-based nanoparticles for contrast-enhanced MRI and molecular imaging, NMR Biomed. 191, 142164 2006.
21
I. Brigger, C. Dubernet, and P. Couvreur, Nanoparticles in cancer
therapy and diagnosis, Adv. Drug Delivery Rev. 545, 631651 2002.
22
P. Dawson, Functional imaging in CT, Eur. J. Radiol. 603, 331340
2006.
23
D. Cosgrove, Ultrasound contrast agents: An overview, Eur. J. Radiol.
603, 324330 2006.
24
H. Maeda, T. Sawa, and T. Konno, Mechanism of tumor-targeted delivery of macromolecular drugs, including the EPR effect in solid tumor and
clinical overview of the prototype polymeric drug SMANCS, J. Controlled Release 7413, 4761 2001.
25
N. G. Portney and M. Ozkan, Nano-oncology: Drug delivery, imaging,
and sensing, Anal. Bioanal. Chem. 3843, 620630 2006.
26
M. Ferrari, Cancer nanotechnology: Opportunities and challenges, Nat.
Rev. Cancer 53, 161171 2005.
27
K. Y. Kim, Nanotechnology platforms and physiological challenges for
cancer therapeutics, Nanomedicine 32, 103110 2007.
28
H. Devalapally, A. Chakilam, and M. M. Amiji, Role of nanotechnology
in pharmaceutical product development, J. Pharm. Sci. 9610, 2547
2565 2007.
29
R. Sinha, G. J. Kim, S. Nie, and D. M. Shin, Nanotechnology in cancer
therapeutics: Bioconjugated nanoparticles for drug delivery, Mol. Cancer Ther. 58, 19091917 2006.
30
T. M. Allen, Ligand-targeted therapeutics in anticancer therapy, Nat.
Rev. Cancer 210, 750763 2002.
31
L. Nobs, F. Buchegger, R. Gurny, and E. Allemann, Current methods for
attaching targeting ligands to liposomes and nanoparticles, J. Pharm. Sci.
938, 19801992 2004.
32
B. Stella, S. Arpicco, M. T. Peracchia, D. Desmaele, J. Hoebeke, M.
Renoir, J. DAngelo, L. Cattel, and P. Couvreur, Design of folic acidconjugated nanoparticles for drug targeting, J. Pharm. Sci. 8911, 1452
1464 2000.
33
D. Horn and J. Rieger, Organic nanoparticles in the aqueous phasetheory, experiment, and use, Angew. Chem., Int. Ed. 4023, 43304361
2001.
34
B. E. Rabinow, Nanosuspensions in drug delivery, Nat. Rev. Drug Discovery 39, 785796 2004.
35
S. Nie, Y. Xing, G. J. Kim, and J. W. Simons, Nanotechnology applications in cancer, Annu. Rev. Biomed. Eng. 9, 257288 2007.
36
R. K. Jain, Transport of molecules in the tumor interstitium: A review,
Cancer Res. 4712, 30393051 1987.
37
R. K. Jain, Delivery of molecular and cellular medicine to solid tumors,
Adv. Drug Delivery Rev. 4613, 149168 2001.
38
S. M. Moghimi and J. Szebeni, Stealth liposomes and long circulating
nanoparticles: Critical issues in pharmacokinetics, opsonization and
protein-binding properties, Prog. Lipid Res. 426, 463478 2003.
39
S. M. Moghimi, A. C. Hunter, and J. C. Murray, Long-circulating and
target-specific nanoparticles: Theory to practice, Pharmacol. Rev. 532,
283318 2001.
40
C. S. S. R. Kumar, J. Hormes, and C. Leuschner, Nanofabrication Towards Biomedical Applications: Techniques, Tools, Applications, and Impact Wiley-VCH, Weinheim, 2005.
41
A. L. Klibanov, K. Maruyama, V. P. Torchilin, and L. Huang, Amphipathic polyethyleneglycols effectively prolong the circulation time of liposomes, FEBS Lett. 2681, 235237 1990.
42
J. M. Harris and R. B. Chess, Effect of pegylation on pharmaceuticals,
Nat. Rev. Drug Discovery 23, 214221 2003.
43
S. A. Schmitz, M. Taupitz, S. Wagner, K. J. Wolf, D. Beyersdorff, and B.
Hamm, Magnetic resonance imaging of atherosclerotic plaques using
superparamagnetic iron oxide particles, J. Magn. Reson Imaging 144,
355361 2001.
44
W. J. Rogers, C. H. Meyer, and C. M. Kramer, Technology insight: In
vivo cell tracking by use of MRI, Nat. Clin. Pract. Cardiovasc. Med.
310, 554562 2006.
45
E. Sykova and P. Jendelova, Migration, fate and in vivo imaging of adult
stem cells in the CNS, Cell Death Differ 147, 13361342 2007.
46
N. V. Evgenov, Z. Medarova, G. Dai, S. Bonner-Weir, and A. Moore, In
vivo imaging of islet transplantation, Nat. Med. 121, 144148 2006.
4485
47
N. C. Phillips, L. Gagne, C. Tsoukas, and J. Dahman, Immunoliposome

targeting to murine CD4+ leucocytes is dependent on immune status, J.
Immunol. 1526, 31683174 1994.
48
H. Soo Choi, W. Liu, P. Misra, E. Tanaka, J. P. Zimmer, B. Itty Ipe, M. G.
Bawendi, and J. V. Frangioni, Renal clearance of quantum dots, Nat.
Biotechnol. 2510, 11651170 2007.
49
R. Weissleder, Scaling down imaging: Molecular mapping of cancer in
mice, Nat. Rev. Cancer 21, 1118 2002.
50
M. Rudin and R. Weissleder, Molecular imaging in drug discovery and
development, Nat. Rev. Drug Discovery 22, 123131 2003.
51
T. F. Massoud and S. S. Gambhir, Molecular imaging in living subjects:
Seeing fundamental biological processes in a new light, Genes Dev.
175, 545580 2003.
52
S. K. Lyons, Advances in imaging mouse tumour models in vivo, J.
Pathol. 2052, 194205 2005.
53
B. Ross, A. Rehemtulla, Y.-E. L. Koo, R. Reddy, G. Kim, C. Behrend, S.
Buck, R. J. Schneider, M. A. Philbert, R. Weissleder, and R. Kopelman,
Photonic and magnetic nanoexplorers for biomedical use: from subcellular imaging to cancer diagnosis and therapy, Proc. SPIE 5331, 7683
2004.
54
V. P. Zharov, J. W. Kim, D. T. Curiel, and M. Everts, Self-assembling
nanoclusters in living systems: Application for integrated photothermal
nanodiagnostics and nanotherapy, Nanomedicine 14, 326345 2005.
55
O. Salata, Applications of nanoparticles in biology and medicine,
Nanobiotechnol. 21 2004.
56
T. W. Prow, J. H. Salazar, W. A. Rose, J. N. Smith, L. Reece, A. A.
Fontenot, N. A. Wang, R. S. Lloyd, and J. F. Leary, Nanomedicine:
Nanoparticles, molecular biosensors, and targeted gene/drug delivery for
combined single-cell diagnostics and therapeutics, Proc. SPIE 5318,
111 2004.
57
M. C. Roco, Nanotechnology: Convergence with modern biology and
medicine, Curr. Opin. Biotechnol. 143, 337346 2003.
58
T. M. Fahmy, P. M. Fong, J. Park, T. Constable, and W. M. Saltzman,
Nanosystems for simultaneous imaging and drug delivery to T cells,
AAPS J. 92, E171E180 2007.
59
T. Ishida, D. L. Iden, and T. M. Allen, A combinatorial approach to
producing sterically stabilized Stealth immunoliposomal drugs, FEBS
Lett. 4601, 129133 1999.
60
V. P. Torchilin, Recent advances with liposomes as pharmaceutical carriers, Nat. Rev. Drug Discovery 42, 145160 2005.
61
R. G. Blasberg, In vivo molecular-genetic imaging: Multi-modality
nuclear and optical combinations, Nucl. Med. Biol. 308, 879888
2003.
62
E. A. Schellenberger, D. Sosnovik, R. Weissleder, and L. Josephson,
Magneto/optical annexin V, a multimodal protein, Bioconjugate Chem.
155, 10621067 2004.
63
L. Josephson, M. F. Kircher, U. Mahmood, Y. Tang, and R. Weissleder,
Near-infrared fluorescent nanoparticles as combined MR/optical imaging
probes, Bioconjugate Chem. 133, 554560 2002.
64
V. S. Talanov, C. A. Regino, H. Kobayashi, M. Bernardo, P. L. Choyke,
and M. W. Brechbiel, Dendrimer-based nanoprobe for dual modality
magnetic resonance and fluorescence imaging, Nano Lett. 67, 1459
1463 2006.
65
J. Zheng, G. Perkins, A. Kirilova, C. Allen, and D. A. Jaffray, Multimodal contrast agent for combined computed tomography and magnetic
resonance imaging applications, Invest. Radiol. 413, 339348 2006.
66
M. Nahrendorf, H. Zhang, S. Hembrador, P. Panizzi, D. E. Sosnovik, E.
Aikawa, P. Libby, F. K. Swirski, and R. Weissleder, Nanoparticle
PET-CT imaging of macrophages in inflammatory atherosclerosis, Circulation 1173, 379387 2008.
67
M. A. McDonald and K. L. Watkin, Small particulate gadolinium oxide
and gadolinium oxide albumin microspheres as multimodal contrast and
therapeutic agents, Invest. Radiol. 386, 305310 2003.
68
S. R. Cherry, Multimodality in vivo imaging systems: Twice the power
or double the trouble? Annu. Rev. Biomed. Eng. 8, 3562 2006.
69
V. Wagner, A. Dullaart, A. K. Bock, and A. Zweck, The emerging nanomedicine landscape, Nat. Biotechnol. 2410, 12111217 2006.
70
R. Duncan, The dawning era of polymer therapeutics, Nat. Rev. Drug
Discovery 25, 347360 2003.
71
N. Oku, Delivery of contrast agents for positron emission tomography
imaging by liposomes, Adv. Drug Delivery Rev. 3713, 5361 1999.
72
A. Mitra, A. Nan, B. R. Line, and H. Ghandehari, Nanocarriers for
nuclear imaging and radiotherapy of cancer, Curr. Pharm. Des. 1236,
4485
47294749 2006.
S. B. Yu and A. D. Watson, Metal-based x-ray contrast media, Chem.
Rev. Washington, D.C. 999, 23532378 1999.
74
R. F. Mattrey, D. M. Long, F. Multer, R. Mitten, and C. B. Higgins,
Perfluoroctylbromide: A reticuloendothelial-specific and tumor-imaging
agent for computed tomography, Radiology 1453, 755758 1982.
75
D. M. Long, Jr., E. C. Lasser, C. M. Sharts, F. K. Multer, and M. Nielsen,
Experiments with radiopaque perfluorocarbon emulsions for selective
opacification of organs and total body angiography, Invest. Radiol.
153, 242247 1980.
76
D. M. Long, F. K. Multer, A. G. Greenburg, G. W. Peskin, E. C. Lasser,
W. G. Wickham, and C. M. Sharts, Tumor imaging with x-rays using
macrophage uptake of radiopaque fluorocarbon emulsions, Surgery St.
Louis 841, 104112 1978.
77
C. W. Hughes, R. W. Williams, M. Bradley, and G. H. Irvine, Ultrasound
monitoring of distraction osteogenesis, Br. J. Oral Maxillofac Surg.
414, 256258 2003.
78
A. M. Morawski, G. A. Lanza, and S. A. Wickline, Targeted contrast
agents for magnetic resonance imaging and ultrasound, Curr. Opin. Biotechnol. 161, 8992 2005.
79
G. M. Lanza, R. Lamerichs, S. Caruthers, and S. A. Wickline, Molecular
Imaging in MR with targeted paramagnetic nanoparticles, Med. Mundi
471, 3439 2003.
80
S. D. Caruthers, A. M. Neubauer, F. D. Hockett, R. Lamerichs, P. M.
Winter, M. J. Scott, P. J. Gaffney, S. A. Wickline, and G. M. Lanza, In
vitro demonstration using 19F magnetic resonance to augment molecular
imaging with paramagnetic perfluorocarbon nanoparticles at 1.5 Tesla,
Invest. Radiol. 413, 305312 2006.
81
Y. B. Yu, Fluorocarbon nanoparticles as multifunctional drug delivery
vehicles, J. Drug Target. 1410, 663669 2006.
82
S. Rockwell, M. Kelley, C. G. Irvin, C. S. Hughes, E. Porter, H. Yabuki,
and J. J. Fischer, Modulation of tumor oxygenation and radiosensitivity
by a perfluorooctylbromide emulsion, Radiother. Oncol. 222, 9298
1991.
83
C. Wilhelm, C. Billotey, J. Roger, J. N. Pons, J. C. Bacri, and F. Gazeau,
Intracellular uptake of anionic superparamagnetic nanoparticles as a
function of their surface coating, Biomaterials 246, 10011011 2003.
84
J. Kreuter, The influence of coatings with surfactants on the body distribution of nanoparticles after intravenous injection to rats, Clini. Materi.
1314, 131134 1993.
85
V. Labhasetwar, C. Song, W. Humphrey, R. Shebuski, and R. J. Levy,
Arterial uptake of biodegradable nanoparticles: Effect of surface modifications, J. Pharm. Sci. 8710, 12291234 1998.
86
C. B. Carlson, P. Mowery, R. M. Owen, E. C. Dykhuizen, and L. L.
Kiessling, Selective tumor cell targeting using low-affinity, multivalent
interactions, ACS Chem. Biol. 22, 119127 2007.
87
S. Hong, P. R. Leroueil, I. J. Majoros, B. G. Orr, J. R. Baker, Jr., and M.
M. Banaszak Holl, The binding avidity of a nanoparticle-based multivalent targeted drug delivery platform, Chem. Biol. 141, 107115 2007.
88
X. Montet, M. Funovics, K. Montet-Abou, R. Weissleder, and L. Josephson, Multivalent effects of RGD peptides obtained by nanoparticle display, J. Med. Chem. 4920, 60876093 2006.
89
A. Tanimoto, K. Oshio, M. Suematsu, D. Pouliquen, and D. D. Stark,
Relaxation effects of clustered particles, J. Magn. Reson Imaging 141,
7277 2001.
90
O. Couture, P. D. Bevan, E. Cherin, K. Cheung, P. N. Burns, and F. S.
Foster, Investigating perfluorohexane particles with high-frequency ultrasound, Ultrasound Med. Biol. 321, 7382 2006.
91
P. Caravan, Strategies for increasing the sensitivity of gadolinium based
MRI contrast agents, Chem. Soc. Rev. 356, 512523 2006.
92
U. I. Tromsdorf, N. C. Bigall, M. G. Kaul, O. T. Bruns, M. S. Nikolic, B.
Mollwitz, R. A. Sperling, R. Reimer, H. Hohenberg, W. J. Parak, S.
Forster, U. Beisiegel, G. Adam, and H. Weller, Size and surface effects
on the MRI relaxivity of manganese ferrite nanoparticle contrast agents,
Nano Lett. 78, 24222427 2007.
93
A. Y. Louie, M. M. Huber, E. T. Ahrens, U. Rothbacher, R. Moats, R. E.
Jacobs, S. E. Fraser, and T. J. Meade, In vivo visualization of gene
expression using magnetic resonance imaging, Nat. Biotechnol. 183,
321325 2000.
94
J. Rao, A. Dragulescu-Andrasi, and H. Yao, Fluorescence imaging in
vivo: Recent advances, Curr. Opin. Biotechnol. 181, 1725 2007.
95
S. Kumar and R. Richards-Kortum, Optical molecular imaging agents
for cancer diagnostics and therapeutics, Small 11, 2330 2006.
73
4486
R. H. Muller, K. Mader, and S. Gohla, Solid lipid nanoparticles SLN

for controlled drug deliverya review of the state of the art, Eur. J.
Pharm. Biopharm. 501, 161177 2000.
97
D. A. Wood, Biodegradable drug delivery systems, Int. J. Pharm. 71,
118 1980.
98
Y. Bae, S. Fukushima, A. Harada, and K. Kataoka, Design of
environment-sensitive supramolecular assemblies for intracellular drug
delivery: Polymeric micelles that are responsive to intracellular pH
change, Angew. Chem., Int. Ed. 4238, 46404643 2003.
99
R. Haag, Supramolecular drug-delivery systems based on polymeric
core-shell architectures, Angew. Chem., Int. Ed. 433, 278282 2004.
100
E. R. Gillies and J. M. Frechet, A new approach towards acid sensitive
copolymer micelles for drug delivery, Chem. Commun. Cambridge
200314, 16401641 2003.
101
R. Haag and F. Kratz, Polymer therapeutics: Concepts and applications,
Angew. Chem., Int. Ed. 458, 11981215 2006.
102
R. Duncan, The dawning era of polymer therapeutics, Nat. Rev. Drug
Discovery 25, 347360 2003.
103
G. B. Sukhorukov, A. L. Rogach, B. Zebli, T. Liedl, A. G. Skirtach, K.
Kohler, A. A. Antipov, N. Gaponik, A. S. Susha, M. Winterhalter, and W.
J. Parak, Nanoengineered polymer capsules: Tools for detection, controlled delivery, and site-specific manipulation, Small 12, 194200
2005.
104
D. E. Sosnovik and R. Weissleder, Emerging concepts in molecular
MRI, Curr. Opin. Biotechnol. 181, 410 2007.
105
E. Y. Sun, L. Josephson, K. A. Kelly, and R. Weissleder, Development
of nanoparticle libraries for biosensing, Bioconjugate Chem. 171, 109
113 2006.
106
T. Atanasijevic, M. Shusteff, P. Fam, and A. Jasanoff, Calcium-sensitive
MRI contrast agents based on superparamagnetic iron oxide nanoparticles
and calmodulin, Proc. Natl. Acad. Sci. U.S.A. 10340, 1470714712
2006.
107
S. Forster and T. Plantenberg, From self-organizing polymers to nanohybrid and biomaterials, Angew. Chem., Int. Ed. 415, 689714 2002.
108
G. M. Whitesides and B. Grzybowski, Self-assembly at all scales, Sciences N.Y. 2955564, 24182421 2002.
109
I. W. Hamley, Nanotechnology with soft materials, Angew. Chem., Int.
Ed. 4215, 16921712 2003.
110
S. E. Sakiyama-Elbert and J. A. Hubbell, Functional biomaterials: Design of novel biomaterials, Annu. Rev. Mater. Res. 31, 183201 2001.
111
M. Sarikaya, C. Tamerler, A. K. Jen, K. Schulten, and F. Baneyx, Molecular biomimetics: Nanotechnology through biology, Nat. Mater. 29,
577585 2003.
112
M. Sarikaya, C. Tamerler, D. T. Schwartz, and F. Baneyx, Materials
assembly and formation using engineered polypeptides, Annu. Rev.
Mater. Res. 34, 373408 2004.
113
S. Zhang, D. M. Marini, W. Hwang, and S. Santoso, Design of nanostructured biological materials through self-assembly of peptides and proteins, Curr. Opin. Chem. Biol. 66, 865871 2002.
114
G. F. Payne, Biopolymer-based materials: The nanoscale components
and their hierarchical assembly, Curr. Opin. Chem. Biol. 112, 214219
2007.
115
D. W. Wendell, J. Patti, and C. D. Montemagno, Using biological inspiration to engineer functional nanostructured materials, Small 211,
13241329 2006.
116
M. Karlsson, M. Davidson, R. Karlsson, A. Karlsson, J. Bergenholtz, Z.
Konkoli, A. Jesorka, T. Lobovkina, J. Hurtig, M. Voinova, and O. Orwar,
Biomimetic nanoscale reactors and networks, Annu. Rev. Phys. Chem.
55, 613649 2004.
117
A. Kawamura, A. Harada, K. Kono, and K. Kataoka, Self-assembled
nano-bioreactor from block ionomers with elevated and stabilized enzymatic function, Bioconjugate Chem. 185, 15551559 2007.
118
D. M. Vriezema, M. Comellas Aragones, J. A. Elemans, J. J. Cornelissen,
A. E. Rowan, and R. J. Nolte, Self-assembled nanoreactors, Chem. Rev.
Washington, D.C. 1054, 14451489 2005.
119
N. Rapoport, Z. Gao, and A. Kennedy, Multifunctional nanoparticles for
combining ultrasonic tumor imaging and targeted chemotherapy, J. Natl.
Cancer Inst. 9914, 10951106 2007.
120
J. C. Frias, M. J. Lipinski, S. E. Lipinski, and M. T. Albelda, Modified
lipoproteins as contrast agents for imaging of atherosclerosis, Contrast
Media Mol. Imaging 21, 1623 2007.
121
P. A. Dayton, S. Zhao, S. H. Bloch, P. Schumann, K. Penrose, T. O.
Matsunaga, R. Zutshi, A. Doinikov, and K. W. Ferrara, Application of
96
4486
ultrasound to selectively localize nanodroplets for targeted imaging and

therapy, Mol. Imaging 53, 160174 2006.
122
P. A. Dayton, K. E. Morgan, A. L. Klibanov, G. Brandenburger, K. R.
Nightingale, and K. W. Ferrara, A preliminary evaluation of the effects
of primary and secondary radiation forces on acoustic contrast agents,
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 446, 12641277
1997.
123
S. Martel, J. B. Mathieu, O. Felfoul, A. Chanu, E. Aboussouan, S. Tamaz,
P. Pouponneau, L. Yahia, G. Beaudoin, G. Soulez, and M. Mankiewicz,
Medical and technical protocol for automatic navigation of a wireless
device in the carotid artery of a living swine using a standard clinical
MRI system, Med. Image Comput. Comput. Assist. Interv. 10Pt 1,
144152 2007.
124
R. Jurgons, C. Seliger, A. Hilpert, L. Trahms, S. Odenbach, and C. Alexiou, Drug loaded magnetic particles for cancer therapy, J. Phys.:
Condens. Matter 18, S2893S2902 2006.
125
N. Gaponik, I. L. Radtchenko, G. B. Sukhorukov, and A. L. Rogach,
Luminescent polymer microcapsules addressable by a magnetic field,
Langmuir 204, 14491452 2004.
126
Q. A. Pankhurst, J. Connolly, S. K. Jones, and J. Dobson, Applications
of magnetic nanoparticles in biomedicine, J. Phys. D: Appl. Phys.
3613, R167R181 2003.
127
O. D. Kripfgans, J. B. Fowlkes, D. L. Miller, O. P. Eldevik, and P. L.
Carson, Acoustic droplet vaporization for therapeutic and diagnostic applications, Ultrasound Med. Biol. 267, 11771189 2000.
128
I. Hilger, R. Hiergeist, R. Hergt, K. Winnefeld, H. Schubert, and W. A.
Kaiser, Thermal ablation of tumors using magnetic nanoparticles: An in
vivo feasibility study, Invest. Radiol. 3710, 580586 2002.
129
R. Ivkov, S. J. DeNardo, W. Daum, A. R. Foreman, R. C. Goldstein, V. S.
Nemkov, and G. L. DeNardo, Application of high amplitude alternating
magnetic fields for heat induction of nanoparticles localized in cancer,
Clin. Cancer Res. 1119 Pt 2, 7093s7103s 2005.
130
M. Lepage, J. Jiang, J. Babin, B. Qi, L. Tremblay, and Y. Zhao, MRI
observation of the light-induced release of a contrast agent from photocontrollable polymer micelles, Phys. Med. Biol. 5210, N249N255
2007.
131
Q. Y. Cai, S. H. Kim, K. S. Choi, S. Y. Kim, S. J. Byun, K. W. Kim, S. H.
Park, S. K. Juhng, and K. H. Yoon, Colloidal gold nanoparticles as a
blood-pool contrast agent for X-ray computed tomography in mice, Invest. Radiol. 4212, 797806 2007.
132
J. F. Hainfeld, D. N. Slatkin, T. M. Focella, and H. M. Smilowitz, Gold
nanoparticles: A new X-ray contrast agent, Br. J. Radiol. 79939, 248
253 2006.
133
C. J. Gannon, C. R. Patra, R. Bhattacharya, P. Mukherjee, and S. A.
Curley, Intracellular gold nanoparticles enhance non-invasive radiofrequency thermal destruction of human gastrointestinal cancer cells, Nanobiotechnol. 62 2008.
134
J. R. McCarthy, F. A. Jaffer, and R. Weissleder, A macrophage-targeted
theranostic nanoparticle for biomedical applications, Small 289, 983
987 2006.
135
R. Deckers, C. Rome, and C. T. Moonen, The role of ultrasound and
magnetic resonance in local drug delivery, J. Magn. Reson Imaging
272, 400409 2008.
136
N. Y. Rapoport, D. A. Christensen, H. D. Fain, L. Barrows, and Z. Gao,
Ultrasound-triggered drug targeting of tumors in vitro and in vivo, Ultrasonics 4219, 943950 2004.
137
W. A. Volkert and T. J. Hoffman, Therapeutic radiopharmaceuticals,
Chem. Rev. Washington, D.C. 999, 22692292 1999.
138
C. F. Shaw III, Gold-based therapeutic agents, Chem. Rev. Washington, D.C. 999, 25892600 1999.
139
J. Kreuter, Nanoparticlesa historical perspective, Int. J. Pharm.
3311, 110 2007.
140
R. C. Oppenheim, Solid colloidal drug delivery systems: Nanoparticles,
Int. J. Pharm. 83, 217234 1981.
141
R. H. Baughman, A. A. Zakhidov, and W. A. de Heer, Carbon
nanotubesthe route toward applications, Sciences N.Y. 297 5582,
787792 2002.
142
C. Burda, X. Chen, R. Narayanan, and M. A. El-Sayed, Chemistry and
properties of nanocrystals of different shapes, Chem. Rev. Washington,
D.C. 1054, 10251102 2005.
143
Y. Chen and A. Pepin, Nanofabrication: Conventional and nonconventional methods, Electrophoresis 222, 187207 2001.
144
P. D. Cozzoli, T. Pellegrino, and L. Manna, Synthesis, properties and
4487
perspectives of hybrid nanocrystal structures, Chem. Soc. Rev. 3511,

11951208 2006.
145
S. Yang, X. Chen, S. Motojima, and M. Ichihara, Morphology and microstructure of spring-like carbon micro-coils/nano-coils prepared by
catalytic pyrolysis of acetylene using Fe-containing alloy catalysts, Carbon 434, 827834 2005.
146
P. Decuzzi, S. Lee, B. Bhushan, and M. Ferrari, A theoretical model for
the margination of particles within blood vessels, Ann. Biomed. Eng.
332, 179190 2005.
147
P. Decuzzi, S. Lee, M. Decuzzi, and M. Ferrari, Adhesion of microfabricated particles on vascular endothelium: A parametric analysis, Ann.
Biomed. Eng. 326, 793802 2004.
148
J. V. Frangioni, In vivo near-infrared fluorescence imaging, Curr. Opin.
Chem. Biol. 75, 626634 2003.
149
M. A. Horton and A. Khan, Medical nanotechnology in the UK: A
4487
perspective from the London Centre for Nanotechnology, Nanomedicine

21, 4248 2006.
150
L. Mazzola, Commercializing nanotechnology, Nat. Biotechnol.
2110, 11371143 2003.
151
Molecular Imaging and Contrast Agent Database MICAD database online, National Library of Medicine US, Bethesda MD: NCBI; 2004
2008. Available from http://micad.nih.gov.
152
R. H. Glassman and A. Y. Sun, Biotechnology: Identifying advances
from the hype, Nat. Rev. Drug Discovery 32, 177183 2004.
153
J. V. Frangioni, Translating in vivo diagnostics into clinical reality, Nat.
Biotechnol. 248, 909913 2006.
154
V. M. Runge, Changes in the approval process for contrast media, J.
Magn. Reson Imaging 103, 485488 1999.
155
D. A. Benaron, The future of cancer imaging, Cancer Metastasis Rev.
211, 4578 2002.
Magnetic resonance spectroscopy

Robert W. Prosta
Department of Radiology, Medical College of Wisconsin, 8701 Watertown Plank Road,
Milwaukee, Wisconsin 53226
Received 27 October 2007; revised 29 July 2008; accepted for publication 29 July 2008;
The nuclear magnetic resonance phenomenon has given rise to both magnetic resonance imaging,
which yields morphologic data, and magnetic resonance spectroscopy MRS, which yields chemical data. In humans these data are derived principally from the resonances of the hydrogen nucleus
in the low molecular weight compounds in the body. Hydrogen MRS has become a routinely used
clinical tool in the brain, prostate, and breast. Other nuclei also demonstrate this phenomenon but
each of these comes with additional difficulties, including low abundance, low sensitivity, and/or
low chemical concentrations. The future of MRS includes a drive to higher main magnetic field
strengths and new methods to create 45 orders of magnitude greater signal. The future of MRS is
bright, but in the United States it is endangered by overuse and misuse driven by the advent of
reimbursement. 2008 American Association of Physicists in Medicine.
DOI: 10.1118/1.2975225
I. INTRODUCTION
I.A. Comparing magnetic resonance imaging and
magnetic resonance spectroscopy
In Fig. 1, axial magnetic resonance MR images and spectra

are shown from the brain of a volunteer in a repeatability
study. On the morning of the second rightmost session, the
subject was profoundly different. After a late night of drinking, the subject was still inebriated that morning. The ethanol
ETOH was easily detected in her brain as shown from the
resonances in the MR spectra, while the T1-weighted MR
images were unchanged. To understand why the information
in the magnetic resonance imaging MRI and magnetic
resonance spectroscopy MRS differed, we must understand
how the two methods, imaging and spectroscopy, relate.
MRS and MRI arise from the same principle, nuclear magnetic resonance NMR, first observed by Bloch and Purcell
in 1946.1 In Blochs classical description of the phenomenon,
polarized nuclei precess about the direction of the main magnetic field with a frequency that is a product of the gyromagnetic ratio of the nucleus, , and the strength of the magnetic
field at the nucleus, B0. The origin of MRS dates to 1951
when Albert described small changes in resonant frequency
due to the bonding positions of two hydrogen nuclei in different locations on a molecule.2 Changes in the resonant frequency of populations of nuclei give rise to the information
content of both MRI and MRS. In MRI, the resonant frequency of a spin is modified by gradients imposed on the
main magnetic field. The frequency of the spin thus becomes
a function of the position of that spin relative to the gradient
where f = B0 + gr * r, gr is the magnetic field gradient,
and r is the distance from system isocenter along that gradient. In this way, spatial information is extracted and images
created. MR spectroscopy, however, detects the modification
of the magnetic field at the postion of the spin created by the
local chemical bonding environment. The frequency of resonance of a spin in a molecule is given by f = B01 . The
4530
shift imparted by the local bonding environment is given the

symbol and is called the chemical shift. The chemical shift
is a very small fraction of the resonance frequency of the
nucleus, ranging in magnitude from 105 for hydrogen 1H
to 103 for fluorine 19F. Unlike the case where resonance
frequency is shifted by gradients for MRI, is a linear function of the strength of the main magnetic field. For instance,
the difference in resonance frequency between the hydrogen
nuclei in water and the hydrogen nuclei in the methylene
groups of fat is 210 Hz at 1.5 T and 420 Hz at 3 T. While
the chemical shift of MRS and resonance shift of MRI arise
from different sources, and are typically of significantly different magnitudes, the chemical shift can interfere with imaging. The combination of imaging shifts and chemical shifts
causes fat containing tissues to appear displaced from their
real locations Fig. 2. This displacement in MRI worsens
with increasing field strength.
In MRS, the resonance offset is normalized to the operating frequency of the magnet and referenced to a standard
resonance to minimize confusion when comparing results
from laboratories that use different magnetic field strengths.
Resonance positions are then reported in parts per million
ppm. For 1H spectroscopy, the standard is the methyl proton resonance of tetramethyl silane which was chosen to be
0 ppm. Based on this standard, protons in water resonate at
4.8 ppm regardless of magnetic field strength. The resonance
of the methylene protons in adipose tissue is 1.3 ppm. The
shift between water and fat remains 3.5 ppm regardless of
field strength, although the frequency difference in hertz
changes.
I.B. A truly small signal: NMR
The NMR effect is exceedingly small. Of all of the hydrogen nuclei in the human body, only 1 in 105 is polarized
when immersed in a field of 1.5 T. The signal which can be
derived from any NMR experiment MRI or MRS is first
dependent on this polarization. In all but the most recent
0094-2405/2008/3510/4530/15/$23.00
4530
4531
Robert William Prost: Magnetic resonance spectroscopy
4531
FIG. 1. Localizing MRI and accompanying 1H MRS spectrum from a 0.56 cm3 voxel in the left putamen. Data acquired at 0.5 T with short echo time elliptical
excitation chemical shift imaging. Acquired as part of a repeatability study. Data shown in b was from the second session when the volunteer was inebriated.
Triplet resonance of ethanol marked ETOH.
experiments, the magnitude of the polarization is directly

proportional to magnetic field strength. The minuscule size
of the NMR signal causes the effect to be highly susceptible
to imperfections in the system. Thus, one of the factors
which has driven the evolution of NMR is the quest for
stronger magnets which increase polarization. The advent of
superconducting magnets has allowed the creation of fields
in excess of 23 T. These very large magnets have tiny bore
sizes that make them suitable only for conducting experiments on small samples. Designing high field strength magnets which can accommodate human subjects is constrained
by the tremendous mechanical forces created by large fields
and the critical current of the superconducting wire. Magnets
large enough for human use are now available with field
strengths up to 7 T, and one 9.4 T system has been built.3,4
Higher field strength magnets should allow one to detect
moieties at lower concentrations and to minimize resonance
FIG. 2. 1H spectrum from a leg showing water and lipid resonances in a.

This frequency difference manifests itself in routine imaging b as a spatial
shift in the readout gradient direction. Thick black arrow is muscle, predominantly water containing physically shifted from the surrounding adipose tissue.
overlap. In humans, all of the 1H resonances of interest lie

within an approximately 4 ppm range, resulting in the complete or partial obscuration of otherwise detectable resonances.
I.C. Technical considerations in MRS: Signal to noise
ratio and overlap
Improving MRS data is usually accomplished by improving the signal to noise ratio of the data and by increasing the
separation between resonances. Signal to noise ratio in MR
spectroscopy is a linear function of field strength and is
given by5
SNR =
B0KNNEXeTe/T21 eTr/T1
lwbw
where B0 is the strength of the main magnetic field, NEX is

the number of averages, T1 and T2 are longitudinal and
transverse relaxation times, lw is the linewidth of the resonance, N is the number of spins, bw is the receiver bandwidth, and K is a constant that accounts for other devicespecific factors. For all other equal factors, signal to noise
ratio SNR will increase linearly with B0. With increasing
B0, the dispersion difference in resonance frequency in
hertz between singlet resonances also increases linearly.
Therefore, closely spaced singlet resonances become more
distinct and quantifiable as field strength increases. This occurs despite the fact that the difference as measured in ppm
remains the same Fig. 3. One other complication arises in
considering these resonances. Many of these resonances are
not singlets a single resonance line. Nuclei with different
chemical shifts may exchange energy through the bonding
electron clouds in the molecule. Coupling can occur between
nuclei of the same element or a nucleus with a different
4532
4532
FIG. 3. Ethanol phantom 1H spectra demonstrating change of chemical shift with main magnetic field B0 strength. In a and c, spectra are displayed with
chemical shift in hertz with water resonance set to 0 Hz. In b and d, spectra are displayed in ppm showing that the centers of each multiplet are at the same
offset in ppm. Frequency splitting of the multiplets in d appears larger than in b because J coupling is invariant to change in field strength, whereas
chemical shift is linear with B0.
These effects are referred to as homo- and heteronuclear coupling, respectively. Also known as J coupling, it causes the
resonance to split into two or more resonances, depending on
the number of other nuclei involved in the coupling and the
magnitude of the coupling, as shown in Fig. 4. The total
amount of energy in the resonance is split between the individual resonances with an energy distribution which is dependent on the quantum states of the spins in the coupled
nuclei.6 Two coupled spins form a doublet with two resonances of equal amplitude 1:1 and three coupled spins form
a 1:2:1 pattern, while four create a 1:3:3:1 pattern. The magnitude of the coupling is independent of field strength, and is
measured in hertz, not ppm. The appearance of coupled resonances is a function of the ratio of / J, which is a function
of field strength. As B0 approaches zero, the coupled resonances degenerate into a singlet as the ratio / J approaches
zero.7 At very high values of B0, the individual resonances in
the multiplet may separate sufficiently to be individually deMedical Physics, Vol. 35, No. 10, October 2008
tected. Intermediate field strengths can cause problems where

the phase of the individual resonances in the multiplet destructively interfere. Figure 5 demonstrates this effect in the
methylene protons of glutamate. Even at very high field
strengths, resonances may still overlap.8 The resonance of
the two methylene groups in taurine is shown to collapse into
a singlet at 0.5 T. Establishing the identity of an unknown
coupled resonance may be accomplished by sampling in another dimension, allowing the J couplings to evolve in time.9
Spectra of J-coupled spins evolve with echo time as shown
in Fig. 6 for the doublet of lactate in a phantom. The J
coupling, under some circumstances, may be removed by
irradiating the spin to which the observed spin is coupled. In
doing so, the J coupling disappears and the resulting resonance area is the sum of all the areas in the multiplet, as
shown in Fig. 4b. In addition to the effects noted above, the
spins also experience the same types of relaxation parameters
well known in MR imaging. These include T1, T2, and T2*
4533
4533
FIG. 4. Homonuclear and heteronuclear spin J coupling in a and b, respectively, showing the paths of coupling. Note that the coupling is much smaller
for the homonuclear case. Removal of heteronuclear splittings by proton decoupling shown in top spectrum on the right in b. The area and amplitude of the
resulting resonance after decoupling is greater than the coupled case as energy split between resonances in the multiplet are summed when the J coupling is
removed.
relaxation. Each resonance has its own relaxation parameters

which are usually different than the water in the same voxel.
As a result, the choice of echo time TE and repetition time
TR, have an impact on the absolute and relative intensity of
each resonance. The effect of changing echo time in a brain
mimicking phantom can be appreciated in Fig. 6 where a 1H
spectrum was repeatedly acquired at different echo times.
The resonances of some moieties, such as glutamate and
myoinositol vanish at the longer echo times as the T2 of
these are short.
I.D. MRI and MRS: Past is prolog
New techniques in clinical MRI or MRS are first developed using phantom and animal models in preclinical, small
bore systems, followed by evolution of whole-body magnet
technology and proof of safety and efficacy through human
clinical trials. One of the first methods proposed for MR
imaging by Mansfield in 1977 was echo planar imaging
where the spin system is excited and the entire k-space plane
is acquired by a string of phase and frequency encoding graMedical Physics, Vol. 35, No. 10, October 2008
dient pulses.10 Until the late 1980s, gradient slew-rate and

magnetic field homogeneity were both insufficiently controlled in whole-body MRI systems to support echo planar
imaging. Additionally, gradient eddy currents swamped the
effects of the localizing gradients. Once the engineering
challenges had been met shielded gradient coils, stronger
gradient amplifiers, and rapid shimming methods, Mansfields original concept became an integral part of the routine
MRI exam. Another physical phenomenon, the diffusion of
water, first demonstrated in the laboratory in the 1960s, became practical in humans, but only after the same problems
preventing the application of echo planar imaging had been
solved.11 MR spectroscopy has developed along a similar
trajectory. Localized spectroscopy within organs, and lesions
within those organs, required that gradient eddy currents and
magnetic field homogeneity problems in whole-body MRI
systems be resolved. Localization is accomplished with
shaped radio frequency pulses e.g., slice selective sincshaped pulses combined with gradients to modify the magnetic field so as to allow only the spins close to the resonance
4534
4534
FIG. 5. Phantom study of the change of resonance appearance with field strength. The creatine resonances are singlets which do not change appearance with
field strength while coupled resonances of glutamate and taurine are strongly affected. Phantom composed of 20 mM each creatine, glutamate, and taurine in
water, titrated to physiologic pH.
frequency to be excited.12 These methods, now common,

were previously impractical due to the effect of gradient
eddy currents. Early MRS studies therefore accomplished localization by placing a surface radio frequency RF coil
over the area of interest.13 Many of these early studies were
of phosphorus and took place in skeletal muscle. To predict
the future of MR spectroscopy, one should examine those

applications presently in the analytical and small-bore MR
systems. While the concept of MRS was described in 1951,
four decades would elapse before this effect was reliably
reproduced in human subjects. In 1991, Frahm showed that
alterations in the area of resonances observed in the brain
FIG. 6. Data acquired at 1.5 T in a brain mimicking phantom at constant repetition time TR and multiple echo times TE. Note that the phase of the doublet
of lactate at 1.3 ppm changes phase with echo time. This forms the basis of J-coupled spectroscopy. Also it can be seen that each of the metabolites changes
amplitude with TE based on the T2 of that resonance. Therefore, observation of some metabolites such as glutamate requires short TE times.
4535
lesions in patients were due to histological changes associated with those lesions14 Since that time, 1H MRS has moved
into routine use in patients.
II. THE STATE OF THE ART
II.A. 1H MR Spectroscopy
Spectroscopy methods fall roughly into two categories.

The first are hydrogen based 1H and the second are all
other nuclei sometimes referred to as X nuclei. The preponderant clinical application is 1H MRS, which dominates
in part because additional additional hardware is not required. All of the same RF coils transmit and receive systems
used for MRI are applicable to 1H MRS. In fact, the resonances of all hydrogen-bearing moieties other than water and
the methyl/methelene groups are contained within the signals
received during a MRI image acquisition. However, these
resonances are not detectable for two reasons. The first is
that the concentrations of the molecules which give rise to
these resonances are far below that of water. For example,
water in brain parenchyma is approximately 72 Molar in
hydrogen, while the methyl resonance of the most abundant
of the metabolites of interest, n-acetyl aspartate NAA, is
30 mM.15 The concentration ratio, 1Hwater: 1HNAA is 2400:1.
In a typical MRI brain image with an SNR of 100, the voxel
volume is on the order of 5 mm3. The theoretical signal to
noise ratio for NAA in this same acquisition is 0.042. As a
result, MRS acquisitions utilize much larger voxels to provide a useful SNR. The volume of the MRS voxel must be
2400 times greater than that of MRI, for an equivalent signal
to noise ratio. This translates into a cube 13.4 times larger on
an edge than the MRI voxel. The smallest currently practical
MRS voxel in a conventional head RF coil is about 1 cm3.
Imaging gradients also broaden the resonance, lowering its
SNR and further rendering it undetectable. The second major
problem of 1H MRS is water suppression. The dispersion of
the 1H spectrum is small, with all the resonances of interest
in the human within 5 ppm of the water resonance
0 4.8 ppm. The line shape of the water resonance in vivo
yields a large base line signal in the 0 4.8 ppm range, overwhelming the resonances of interest. The method developed
to ameliorate this problem is water suppression, nominally
implemented by a series of RF pulses that selectively excite
just the water resonance with crusher gradients to eliminate
the residual signal. Large signals also arise from lipid in the
skull marrow and the scalp. These resonances, predominantly at 1.3 ppm can hide resonances arising in adjacent
parenchyma. Eliminating lipid signal is accomplished by
avoiding excitation of the scalp and skull marrow and/or by
placing spatial saturation pulses over the scalp and marrow
spaces.
II.B. Pulse sequences for acquisition
Two principle pulse sequences are used for 1H MRS. Both

use three orthogonal slice selective RF pulses. Of these, the
most commonly used sequence is point resolved spectroscopy PRESS, which consists of a 90 RF pulse followed by
4535
two 180 RF pulses to form a right rectangular prism of

excited tissue.16 The second most used sequence is stimulated echo acquisition mode STEAM, which consists of
three successive 90 pulses which create the same right rectangular prism of excited tissue.14 The virtue of STEAM is
that the magnetization is stored along Z between the second
and third RF pulses. This time does not count against the
echo time of the sequence. Shorter echo times can be
achieved by STEAM for the identical gradient and RF capabilities of the scanner, but at the cost of half of the SNR of
PRESS. The difference in signal is due to the fact that only
half of the magnetization in the transverse plane is rotated
down to Z by the second of the three RF pulses.17 All signals arising from spins in the voxel sum together to create
the received signal. Unless further steps are taken to produce
spatial encoding within the voxel, the data from the voxel
yield only one spectrum.
II.C. Multiple voxel acquisition methods: Chemical
shift imaging
Spatial encoding when applied across the voxel can create

chemical shift imaging CSI data where each voxel has associated with it a spectrum.18 The method of encoding is
usually performed using a phase-encoding gradient pulse
along each encoded dimension and a much larger excited
volume. The resultant data set from a two-dimensional CSI
acquisition is shown in Fig. 7b for the case of a patient with
a brain tumor. The duration of the acquisition is extended by
a factor which is a product of the number of phase encodings
in each direction. In the conventional CSI imaging method,
one or more dimensions must be orthogonally encoded with
its own individual value of phase encoding. Therefore, the
acquisition time for a CSI dataset with a resolution of 16 in
each of the possible three dimensions is at a minimum 4096
times as long as a single voxel acquisition. MR imaging
enjoys a greater efficiency because phase and frequency encoding can be simultaneously applied without loss of information. In MRS, applying a frequency encoding gradient
causes a phase twist along the chemical shift dimension
which cannot be removed except by incrementing the echo
time and acquiring data at multiple echo times. This technique phase encoded echo planar spectroscopic imaging
PEPSI has been applied using echo planar imaging.19
However, numerous artifacts are generated as a result of
eddy currents and resolution in the chemical shift dimension
is inversely dependent on the size of the echo time increment. In elliptical polarization illumination-based methods
demonstrated to date, the acquisition times are the same as
that used in conventional encoding for equivalent SNR.
Data acquired with CSI can be displayed as a grid of
spectra, or a pseudocolor image where the color is a function
of the fitted or integrated area of a particular resonance. The
slow nature of CSI limits the number of voxels which can be
sampled over the field of view. This in turn leads to undersampling errors and broadens the point spread function
which describes the shape of each voxel.20 The rectangular
nature of the sampling matrix means that the voxel will have
4536
4536
FIG. 7. Sum of spectra in a, from a single slice CSI acquisition in a patient with a newly appearing lesion shows a characteristic pattern for neoplasm
increased choline and decreased NAA. Plotting several spectra as shown in b demonstrates the necrotic core of the lesion with changing neoplastic features.
The use of this spatial data provides the actual tissue diagnosis, glioblastoma. Data acquired at 0.5 T using a short echo time CSI sequence.
4537
4537
FIG. 8. Coronal T2-weighted image of

a patient with a brainstem lesion. Acquired voxel is illustrated by the rectangle on the image. Accompanying 1H
spectrum demonstrates elevated choline, decreased to absent NAA and
strongly elevated methyl and methylene resonances. Lesion diagnosed as a
high-grade neoplasm.
a sinc sin x / x shape with positive and negative sidelobes.

These sidelobes, if they fall in the area of incompletely suppressed lipid, can generate a significant contribution to the
signal arising from the central lobe of the point spread function.
The SNR efficiency of conventionally encoded CSI
makes it the most used method.21 Figure 7 demonstrates the
utility of CSI. In Fig. 7a, a single large voxel has been
acquired in the brain of a patient with a lesion in the right
frontal lobe. The voxel covers not only the frontal lobe but
other areas as well. The spectrum does not appear normal
when compared to a spectrum from a healthy control subject.
However, no information is available about the difference in
tissue between what is very clearly, by MRS, abnormal, and
what is by MRI, apparently normal tissue. The same large
voxel, with phase encoding applied in dimensions transverse
to the slice Fig. 7b, yields spectra which demonstrate necrosis within the lesion elevated lipid and lactate resonances, while voxels adjacent become progressively less
neoplastic appearing with distance from the lesion, with the
choline eventually returning to normal levels in the voxels
furthest away from the lesion. The spatially encoded data
allow the lesion to be diagnosed as glioblastoma multiforme
for which this spectroscopic change with position is pathognomonic of this tumor. Judged only by the single spectrum as
shown in Fig. 7a, the lesion would have been diagnosed as
a glioma of lower grade.
II.D. Single voxel MRS
Despite the ability of CSI to acquire data from many voxels simultaneously, single voxel MRS methods retain significant diagnostic utility. Some smaller lesions in areas of the
brain which may be surrounded by air, bone, or adipose tissue are not amenable to exploration by CSI methods due to
contamination from adjacent voxels which may sample tissue outside of parenchyma. For example, the lesion shown in
Fig. 8 could only be sampled with a single voxel technique.
Another compelling reason to acquire data by the single
voxel method is that the entire acquisition may be completed

in less time than required by CSI for patients who are marginally cooperative.
II.E. X nucleus methods
In contrast to 1H MRS, relatively few clinical applications

exist for X nucleus MRS. While the X nuclei do not suffer
from the water and lipid suppression problems of 1H MRS,
there are two other significant issues that determine which
acquisition methods can be used. The first of these issues is
the lower sensitivity per nucleus. The smaller magnetic moment of any of the X nuclei produces less signal per nucleus
than in the case of hydrogen. The second issue is that the
required gradient amplitudes for slice selection, phase encoding, and readout all scale inversely with . The of phosphorus is 2.5 times smaller than that of hydrogen. Gradients
for localizing or encoding phosphorus signal must therefore
be 2.5 times greater. This represents a significant limitation
for most whole-body gradient systems. Single voxel excitation methods such as PRESS or STEAM are generally not
used for X nucleus work due both to gradient limitations and
the short T2 of many of the X nucleus moieties. Surface coil
localization combined with pulse and acquire methods or
slice selective methods with CSI phase encoding are among
the principal localization methods used.22 The smallest practical voxel volumes at 3 T are 8 cm3 or greater. An additional problem which further limits the utilization of X
nucleus methods is the need for additional MR system hardware. Present generation MRI systems have all been designed to operate only at the resonant frequency of hydrogen.
Resonant frequencies of the X nuclei are all far enough from
that of hydrogen so as to require additional hardware. The X
nuclei acquisitions require separate RF coils, preamplifiers,
power amplifiers, and transceiver hardware. Such hardware,
especially RF coils and preamplifiers are not typically supported by MRI manufacturers, requiring the user to acquire
custom built hardware from third party vendors. The cost of
the X-nucleus capable hardware is a significant fraction of
cost of the entire system.
4538
4538
II.F. Present usual and customary acquisition and

analysis techniques
Hydrogen MRS voxel sizes range from 0.5 to 8 cm2 or

more. The acquisition times are typically 5 30 min. The upper limits of acquisition time are set by patient tolerance.
Data are acquired using both PRESS and STEAM excitation
methods, although PRESS is predominant due to its factor of
2 advantage in SNR over STEAM. Single voxel methods are
the most used for human brain, while CSI methods, which
can range from one-dimensional to three-dimensional encoding, tend to be two-dimensional, single slice methods. This is
primarily due to time constraints. Acquisitions in prostate are
often 3D CSI to map the choline distribution. Single voxel
studies of isolated lesions are used in breast. Echo times in
brain range from 20 to 290 ms and in prostate or breast,
138 to 290 ms. These longer echo times are used to minimize the artifact from lipid moieties which are not nominally
a problem in brain parenchyma.
Spectra may be interpreted in several different ways.
Qualitative interpretation of resonance amplitudes is common in the clinical setting. Most of this interpretation is done
by radiologists and consists of comparing the relative amplitude, presence, or absence of resonances in the frequency
domain spectra. Qualitative assessment of spectra is prone to
two important errors. The first is that the entire spectrum is
reduced in amplitude due to partial volume averaging with
inert material. The second is that the changes in TR, TE, or
magnetic field strength changes relative amplitudes as a result of differential T1, T2 relaxation or the change in T1 and
T2 with field strength. Quantitative analysis of spectra, either
in the time domain or in the frequency domain will become
more broadly utilized in the future as the analyses become
more robust.23
The X nucleus localization methods can be more varied at
least in part due to the fact that no water suppression is
required. The localization method used is more a function of
the target organ or disease process studied than in the case of
1 24
H. Localization by RF coil sensitivity profile methods are
more common in X nucleus studies of skeletal muscle. Radio
frequency coils for X-nucleus spectroscopy will have some
manner of transmit capability and may be volume or surface
coils. Some of these coils will also have a second mode of
operation which allows operation at 1H frequencies. This
feature is important for localization imaging and shimming
B0 homogeneity adjustment, both of which require 1H
operation.25 The 1H channel of these dual coils can be used
for decoupling, but with the requirement of yet more hardware over and above that required for X nucleus acquisition.
III. CLINICAL APPLICATIONS
Clinical spectroscopy applications are almost all 1H acquisitions. Target organs include the brain, prostate, and
breast. In the brain, 1H MRS is used to differentiate neoplastic from non-neoplastic etiologies of lesions detected by
MRI.26 While MRI is exquisitely sensitive to the presence of
lesions, it is less specific as to etiology. Neoplastic processes
cause an up-regulation in the amount of free choline in the
FIG. 9. Images from two patients with similar lesions referred for 1H MRS
of newly appearing lesions. Images on the right have a false color overlay of
the ratio of choline to NAA. Lesion in the upper patient clearly demonstrates
neoplasm Cho/ NAA 1 while the image below shows that the lesion is
not neoplastic.
lesion.27 Choline is a precursor or breakdown product of the

phosphocholines which are a constituent of the cell membrane. Figure 9 demonstrates two patients with similar appearing lesions by MRI. The patient with the neoplastic lesion is easily identified by the elevated choline Cho
resonance in the 1H MRS. Similar results have been obtained
in the breast and prostate.28
Therapy planning has advanced furthest in prostate.29 Using an inserted endorectal RF coil, 3D CSI data sets are now
generated to delineate regions of neoplasm. As in the brain
and breast, choline is elevated. Healthy prostate tissue produces citrate and the diminution of this resonance is used in
conjunction with the elevation of choline to create maps of
neoplasm. Treatment planning using MRS in brain tumors is
beginning to see some utilization. As is shown in Fig. 10,
treatment plans which do not include neoplastic tissue which
remains occult by MRI, should fail sooner.30
Therapy monitoring by MRS is an increasingly important
aid in decision making in brain tumor patients.31 Radiation
induced necrosis and recurrent neoplasm are difficult to distinguish by MRI. With some caveats, 1H MRS detects recurrence by the elevation of choline. One such caveat is that
some temporal phases of radiation recovery in brain may
transiently show elevations in choline that are not due to
neoplastic activity.32
Studies of metabolic dysfunction, while less common than
tumor studies, have demonstrated clinical utility. One of the
very first clinical applications was the detection of elevated
NAA in Canavans disease.33 Canavans is an inborn defect
in the production of aspartoacylase, which catabolizes
NAA.34 Other diseases of deranged metabolism detectable
4539
FIG. 10. Chemical shift image of glutamate combined with the postcontrast
T1-weighted image of a patient with a glioblastoma. Also on this image is
the 98% dose contour line of the radiation therapy plan shown as the thin
red line. The site of eventual recurrence of this tumor was posterior to this
lesion and is predicted by the red elevated glutamate area Ref. 63.
by 1H MRS include the adrenoleukodystrophies, MELAS,

and phenylketonuria.35 Therapy monitoring by use of MRS
allows the evaluation of the efficacy of treatment strategies
in advance of clinical outcome. Other brain chemistry
changes which result from disease or trauma include hepatic
encephalopathy in which glutamine is found to be elevated
secondary to hepatic failure and the inability of the brain to
clear ammonia.36 In cases of near-death asphyxia, 1H MRS
has been shown to be prognosticative of recovery.37
The use of 1H MRS in drug studies has shown great
promise. One example of this is the reversal of NAA depletion as a marker of neuronal recovery in Alzheimers patients
receiving an acetocholinesterase inhibitor. These elevations
of NAA correlate well with recovery in minimental status
scores.38
Not all in vivo MRS studies are conducted using hydrogen. Other endogenous candidate nuclei are phosphorus
31P, and carbon 13C. These nuclei present several problems, among which are resonance frequency, abundance, detectability, chemical shift, and resonance overlap. Carbon
presents an additional complication. While a highly abundant
nucleus in living systems, 98.9% of the spins are of the isotope carbon-12, which has zero magnetic moment and is undetectable. Only 1.1% of carbon is carbon-13. Phosphorus
moieties in the body are spread over a chemical shift range
of approximately 25 ppm and carbon-13 over roughly
200 ppm. This range causes problems in localization sequences where the chemical shift of a moiety causes the
location of the excited tissue to be shifted relative to other
moieties. The resulting spectrum does not originate from a
single spatial location. This effect is similar to the spatial
displacement of Fig. 2b. An additional problem is resonance overlap, especially in phosphorus studies. These overlaps can be partially resolved while improving the SNR of
the spectrum by removing the J coupling between the observed nucleus and hydrogens to which these are bonded.
This method, well known to the analytical NMR community,
4539
is decoupling. It requires yet more hardware and is exquisitely sensitive to artifact. Proton decoupling of the 13C spectrum of ethanol is shown in Fig. 4b.
Phosphorus spectroscopy detects adenosine triphosphate
ATP, phosphocreatine PCr, inorganic phosphate Pi, the
phosphomono- PME, and phosphodiesters. Studies of brain
have shown alterations of phospholipid metabolism in psychiatric disorders.39 Cardiac function can also be studied by
31
P spectroscopy.40 Dynamic studies of muscle are presently
being used to assess muscle recovery during isometric or
dynamic exercise as shown in Fig. 11.
Glycogen in muscle has been studied in humans with the
use of the 13C.41 Glucose metabolism and its interaction with
neuronal function has been studied in brain using 13C labeled
glucose.42 Similar studies are giving insight into neuronastrocyte trafficking in glutamate and glutamine and the diseases associated with the dysregulation of this process.43
Other nuclei which are not otherwise present in appreciable abundance in the body have been the subject of MRS
studies. These include lithium in the brains of bipolar patients and fluorine in antipsychotic drugs in brain and chemotherapeutic drugs in liver.44
Given the additional hardware required and the additional
complications of acquisition, no X nucleus exams are in
widespread clinical use.
IV. THE IMPERILED FUTURE OF CLINICAL MRS

MR spectroscopy has been shown to have high sensitivity
and specificity in distinguishing neoplastic from nonneoplastic etiologies in brain lesions.45 This work has been
completed in a well controlled setting by MR physicists using methods which required manual adjustment for acquisition and postprocessing. Based on this result and others like
it, spectroscopy became a reimbursable procedure.46 Concurrent with the advent of reimbursement, manufacturers automated MRS acquisition and postprocessing, eventually offering packages which allowed the scanner operator to acquire
data independent of the site physicist. Usage and billing followed the ability to bill for the procedure. Unfortunately for
MRS, data quality and the specificity of the results have
suffered. Recently, Medicare has ruled that MRS is an experimental procedure and will no longer be billable within its
system.47 Many private insurers have quickly followed to
deny reimbursement. Yet, the technique is not worse than it
once was, instead, it has been used with less than sufficient
care for data quality. Thus, many radiology departments have
abandoned the clinical use of MRS. This state of affairs is
unfortunate because MRI system manufacturers lose interest
in supporting hardware for procedures which are not reimbursed. Historically, much of the scientific progress in MRI
and MRS depends on the diligence with which manufacturers attend to the technological as well as operational problems of the MR system.
4540
4540
FIG. 11. 31P MRS spectrum from the gastrocnemious muscle of a patient in an isometric contraction apparatus. 15 s contraction demonstrates the decrease of
PCr heavy arrow and the increase of Pi thin arrow.
V. FUTURE DEVELOPMENTS IN MR
SPECTROSCOPY
The brain will continue to be one of the most studied
organs by MRS. The ability to extract functional data noninvasively continues to be the driving force behind MRS. This
is likely to remain so in the future. Several benefits accrue to
1
H MRS from increasing field strength. These include the
ability to separate resonances that are too closely spaced at
lower field strengths, most notably 1.5 T. The increased signal to noise and chemical shift dispersion of increased field
strength also improves metabolite editing sequences. An example is the detection of the inhibitory neurotransmitter
GABA which nominally is obscured by the creatine
resonance.48 Increased SNR from operation at high field and
the use of phased array surface coils allow data acquisition
with either shorter total scan times or with higher spatial
resolution. Similar phase undersampling strategies used by
MRI for scan time reduction can also be applied to MRS.49
One further benefit of increased chemical shift dispersion is
the ease with which chemical shift saturation water saturaMedical Physics, Vol. 35, No. 10, October 2008
tion or spectral-spatial pulses may be applied.50 A greater

chemical shift between water and the aliphatic metabolites of
interest means that shorter duration saturation pulses can be
used. Therefore, more of the time in the pulse sequence can
be used to sample the signal. High speed echo-planar based
MRSI sequences, such as PEPSI, also benefit as fewer increments in echo time need to be collected to distinguish adjacent resonances. These future developments do not require
additional hardware, as the same hardware used for high field
MRI is also suitable for 1H MRS.
It should be noted that not all aspects of 1H MRS are
beneficiaries of the quest for higher field strength. There are
practical problems in clinical MRS studies which do not improve with field strength. The first is that human tissues are
intrinsically magnetically inhomogeneous. Numerous structures of interest in the brain, such as the basal ganglia, accumulate iron with age and disease.51 The deleterious effect of
the iron upon resonance linewidth increases with increasing
field strength. The second is the effect of other perturbations
of the magnetic field, such as the effect of the air/tissue in-
4541
4541
FIG. 12. Spectra from a 3.4 cm3 voxel acquired in the occipital lobe of a healthy volunteer acquired under identical conditions at three different magnetic field
strengths. Note the increased amplitude of the glutamate/glutamine resonance Glx at 0.5 T relative to 1.5 T. While the individual resonances in the multiplet
start to become visible at 3 T, they are still clearly overlapped. The prominence of the myoinositol resonance also decreases with increasing field strength.
terface above the sinuses and the imperfections of field homogeneity, both of which increase with field strength, cause
losses in the SNR. Third, many of the resonances which
overlap at lower field strengths continue to do so even at
field strengths above 7 T. As the MRS field matures, a more
balanced view may become predominant. Many disease
states will benefit from observation at lower field strengths.
Reduced linewidths and the collapse of multiplets into pseudosinglets favor operation at field strengths as low as
0.5 T.52 Figure 12 illustrates the change of spectral appearance with field strength in the same voxel of a volunteer
acquired at three different field strengths. Note that the methylene resonances of glutamate and glutamine at 2.3 ppm
increase in amplitude at 0.5 T. A similar effect is shown in
the myoinositol resonance at 3.56 ppm. Even taurine benefits, as was shown in Fig. 5. Magnet designs intended for
obese and/or claustrophobic patients with field strengths beMedical Physics, Vol. 35, No. 10, October 2008
low 1.5 T have also shown utility in performing clinical

MRS.53 It must be remembered that MRS is detecting surrogate markers of disease, not the disease itself. MRS does not
detect neoplasm, rather the elevated choline that is a marker
of the mitotic process. Knowing exactly which chemical
moiety is elevated in a disease becomes less important than
the observation that a surrogate marker is reliably altered in
that disease. Perhaps the most exciting future lies ahead for X
nucleus applications. Other nuclei which have a nonzero gyromagnetic ratio and have been used in human MR studies
are 13C, 31P, 19F, 23Na, 7Li, 129Xe, and 3He. Naturally abundant nuclei in the body are 13C, 31P, and 23Na. Exogenous
contrast agents include 13C, 129Xe, and 3He. Other nuclei that
can be detected by MRS in vivo include lithium and fluorine.
Lithium is used to treat bipolar disease and fluorine is
present in a number of psychoactive drugs and some chemotherapeutic agents. Sodium, like lithium, is only detectable in
4542
FIG. 13. Hyperpolarized gas images from the lungs of a healthy subject
comparing 3He to 129Xe. Images courtesy of Dr. John Mugler, University of
Virginia.
the body in ionic form. Having only one moiety, sodium can
be imaged using high speed MRI methods. Phosphorus-31
has 100% isotopic abundance unlike 13C which is only 1.1%
abundant. All of the X nuclei observed with natural polarization will benefit from higher field strength magnets. The increased field strength is needed as the intrinsic signals from a
normally polarized 13C or 31P nucleus are 0.016 and 0.066
that of a hydrogen nucleus. Additional transmit channels can
be used to eliminate heteronuclear J couplings of the X nuclei with 1H as shown in Fig. 4. The signal intensity of the X
nucleus resonance then increases as decoupling removes the
splittings. An additional effect to gain signal, utilizing the
same hardware, is nuclear Overhauser enhancement NOE
which can create an increase in the polarization. Decoupling
combined with NOE can provide two to fourfold signal
increases.54 Equipping a MRI system with both the additional transmit and receive channels, RF coils, and associated
hardware to achieve decoupling can cost a substantial fraction of the cost of the original MRI system. That is a large
investment for what is still a time consuming and difficult
study. There is now a method of increasing the polarization
of the nuclei of interest, called hyperpolarization.55 Helium-3
and xenon-129 can be polarized to 50% or more. This fact is
quite startling when one realizes that the polarization of 1H
in the body at 1.5 T is 1:100 000. The hyperpolarized gases
have T1 relaxation times ex vivo on the order of 10 min.4,56
4542
This extraordinarily long T1 means that the gas can be polarized, stored in a magnetic bottle, and carried to the scanner. Studies of lung function are now being carried out with
3
He and 129Xe as shown in Fig. 13. Xenon has been used for
cerebral perfusion studies in computed tomography where it
changes tissue x-ray attenuation. In MRI, xenon can be used
to study cerebral perfusion as it readily dissolves into blood
through alveolar exchange. The xenon resonance frequency
is sensitive to surrounding tissues. A chemical shift in excess
of 200 ppm is created by proximity to the alveolar surface
and thus can be used to study not just ventilation but lung
function as well.55 The prepolarization of the gas means that
a very high field magnet is not needed to yield signal.
Smaller, more open and patient-friendly magnets will be sufficient to create both imaging conditions and the 1H localization images.
The most exciting new development in spectroscopy is
the hyperpolarization of 13C. Achieved in a different manner
from hyperpolarization of the previously mentioned gases, a
large range of bioavailable chemicals can be created with the
polarized 13C in their structure. To date, two different methods for achieving high levels of polarization in carbon-13
bearing compounds have been proposed. The first is called
dynamic nuclear polarization.57 In this method, the 13C enriched compound in the solid state undergoes microwave irradiation in a 3.3 T magnet while the sample is held at 1 K.
Once polarized, the sample is rapidly dissolved in a suitable
solvent, purified, and injected into the subject who is lying in
the MRI magnet. The second method is polarization transfer
from parahydrogen PHIP in which the polarization from
the parahydrogen is transferred to a 13C containing molecule
through a rhodium catalyst.58 While naturally polarized
carbon-13 labeled glucose has been used to measure the rate
of the tricarboxilic acid cycle in brain,59 the technique requires long acquisition times and high magnetic field
strengths. Hyperpolarization carries the potential to create
metabolic tracers similar in function to those used in positron
emission tomography. Unlike fluorodeoxyglucose FDG, the
tracers created using hyperpolarized 13C are chemically identical to those in the body. In postitron emission tomography
FIG. 14. Location of the image slab in the rat and the corresponding transversal 1H-NMR image. The NMR signal distribution obtained simultaneously from
pyruvate, lactate, and alanine is calculated, and the color images representing the intensity of each metabolite are projected on the anatomical 1H image.
Alanine is most prominent in the skeletal muscle around the spinal cord, whereas the P22 tumor tissue is indicated by the highest signal for lactate. Note the
different scale. Images courtesy of Klaes Golman, Rene int Zandt, Mathilde Lerche, Rikard Pehrson, and Jan Henrik Ardenkjaer-Larsen, Amersham Health
R&D AB Part of GE Healthcare, Malm, Sweden.
4543
PET, FDG does not enter the glucose-hexokinase pathway,

and so the information available from PET using FDG is
limited to that of cellular uptake. In brain tumors this creates
some diagnostic confusion as the elevated metabolic rate of
low grade glial tumors is equivalent to that of gray matter.
Using hyperpolarized 13C labeled metabolic substrates will
allow direct imaging of neoplastic tissues.60 Also, the 13C
label changes chemical shifts as substrate is metabolized into
other moieties. A 13C image of a tumor in an animal model
which demonstrates the effect of the change in chemical shift
of the metabolized substrate is shown in Fig. 14. Two important caveats remain. The first is that the lifetime of the polarization state is limited which places constraints on the
duration of experiments seconds instead of minutes and
also constrains the location of the polarizing facility to be
near the magnet. Like all X-nucleus studies, considerable investment in hardware is required, even before the polarizing
equipment is acquired. However, the cost of the polarizing
equipment and associated lab may be far less than the cost of
having ones own cyclotron for PET isotope production. The
absence of a radiation dose also means that serial studies
using 13C MRS for screening, therapy monitoring, and disease progression can be done with minimal risk.
4543
a lesion and areas around it more practical. The study of

neurotransmitter-related and other diffuse processes will benefit from the increased chemical shift dispersion available at
higher field strength as well as the increased ease with which
spectral editing for otherwise hidden resonances such as
GABA may be accomplished.
Higher field strengths will be of greatest advantage to X
nucleus studies. The inherent low sensitivity and low abundance of the X nuclei make them perfect candidates for investigation at higher fields. The improved SNR that results
should make it practical to include X nuclei in clinical exams, where previously, acquisition times were far too long.
However, the biggest news in the X nucleus world is the
coming of hyperpolarized nuclei and the 45 orders of magnitude in additional signal they create.
Magnetic resonance spectroscopy is poised to make a
great leap forward. It is up to us as scientists and clinicians to
use it responsibly, while it is up to the manufacturers to
continue to support MR spectroscopy by making the MR
system better. To realize the benefit of MR spectroscopy for
all patients, we need to undertake larger, more organized
studies across platforms and across healthcare
organizations.62 The time to begin is now.
a
VI. CONCLUSIONS
In the past 15 years 1H MR spectroscopy has moved from
the laboratory into routine use for patient care. MR system
manufacturers continue to improve their systems to meet the
stringent demands which MRS places on machine performance. Higher field strength MR systems will improve many
facets of spectral data quality, but bring their own complications, such as a heightened sensitivity to brain iron and a
change in metabolite ratios. Most clinical spectra are still
qualitatively interpreted by the ratio of resonance amplitudes. Changing field strength changes what constitutes normal. These changes need to be measured across patient populations and used in patient data analysis. Improvements in
the analysis and interpretation of spectra, such as time domain, automated fitting methods in the frequency domain,
and feature-based algorithms to classify tissue types, will
address these field strength questions as well as improve the
utility and accessibility of 1H MRS to the clinician. Using 1H
MRS to not only distinguish tissue types but to predict recurrence or premorphologic tumor spread will secure a place
in for 1H MRS in the clinicians retinue of patient management tools.61
Future directions in 1H MRS include smaller voxel volumes which will improve the ability to distinguish healthy
from diseased tissue. Decreasing voxel volumes while maintaining clinically useful acquisition times will require higher
field strength magnets and improved RF coil designs, such as
phased array head coils. Increased use of multiple voxel acquisition methods such as CSI will make MRS less sensitive
to the skills of the operator and clinician in prescribing a
voxel location. A CSI study with small voxels, aided by increased SNR from the use of phased array coils and higher
field strength magnets will make the retrospective analysis of

[email protected]
F. Bloch, Phys. Rev. 70, 460 1946; N. Bloembergen, E. M. Purcell, and
R. V. Pound, ibid. 73, 679 1948.
2
J. T. Arnold, S. S. Dharmatti, and M. E. Packard, J. Chem. Phys. 19, 507
1951.
3
J. T. Vaughan et al., Magn. Reson. Med. 46, 24 2001.
4
I. C. Atkinson et al., J. Magn. Reson Imaging 26, 1222 2007.
5
C. J. Hardy et al., Magn. Reson. Med. 8, 104 1988.
6
A. Carrington and A. D. McLachlan, Introduction to Magnetic Resonance
with Applications to Chemistry and Chemical Physics Wiley, New York,
1979.
7
E. Becker, High Resolution NMR, Theory and Chemical Applications
Academic, Orlando, FL, 1980.
8
K. Behar and T. Ogino, Magn. Reson. Med. 17, 285 1991.
9
K. L. Behar et al., Magn. Reson. Med. 32, 294 1994.
10
P. Mansfield, J. Phys. C 10, 55 1977.
11
E. O. Stejskal and J. E. Tanner, J. Chem. Phys. 42, 288 1965; R. Turner
et al., Radiology 177, 407 1990.
12
T. H. Mareci and H. R. Brooker, J. Magn. Reson. 92, 229 1991.
13
J. J. Ackerman et al., Nature London 283, 167 1980.
14
J. Frahm et al., Magn. Reson. Med. 9, 79 1989.
15
R. Kreis, T. Ernst, and B. D. Ross, J. Magn. Reson., Ser. B 102, 9 1993.
16
P. Bottomley, U.S. Patent No. 4 480 228 1987.
17
J. Frahm et al., J. Magn. Reson. 64, 81 1985.
18
T. T. Brown, B. M. Kincaid, and K. Ugurbil, Proc. Natl. Acad. Sci. U.S.A.
79, 3523 1982.
19
S. Posse et al., Magn. Reson. Med. 33, 34 1995.
20
T. R. Brown, NMR Biomed. 5, 238 1992.
21
R. Pohmann, M. von Kienlin, and A. Haase, J. Magn. Reson. 129, 145
1997.
22
P. A. Bottomley et al., NMR Biomed. 2, 284 1989.
23
S. Mierisova and M. Ala-Korpela, NMR Biomed. 14, 247 2001; L.
Vanhamme et al., ibid. 14, 233 2001.
24
S. F. Keevil, Phys. Med. Biol. 51, 21 2006.
25
R. E. Lenkinski et al., Radiology 169, 201 1988.
26
L. Kwock et al., Laucet Oncol. 7, 859 2006.
27
K. K. Bhakoo et al., Cancer Res. 56, 4630 1996.
28
C. Mountford et al., J. Magn. Reson Imaging 24, 459 2006; P. J. Bolan
et al., Breast Cancer Res. Treat. 7, 149 2005.
29
A. Zapotoczna et al., Neoplasia 9, 455 2007.
30
R. Tarnawski et al., Int. J. Radiat. Oncol., Biol., Phys. 52, 1271 2002.
31
I. Park et al., Int. J. Radiat. Oncol., Biol., Phys. 69, 381 2007.
1
4544
L. L. Wald et al., J. Neurosurg. 87, 525 1997.

G. Bartalini et al., Childs Nerv. Syst. 8, 468 1992.
34
J. R. Moffett et al., Prog. Neurobiol. 81, 89 2007.
35
L. Chang et al., Neurology 48, 836 1997.
36
R. Kreis et al., Radiology 182, 19 1992.
37
R. Kreis et al., J. Clin. Invest. 97, 1142 1996.
38
K. R. Krishnan et al., Am. J. Psychiatry 160, 2003 2003.
39
B. K. Puri, Int. Rev. Psychiatry 18, 145 2006.
40
G. M. Pohost et al., Rays 26, 93 2001.
41
J. J. Prompers et al., NMR Biomed. 19, 927 2006.
42
F. Hyder et al., J. Cereb. Blood Flow Metab. 26, 865 2006.
43
C. Zwingmann and D. Leibfritz, NMR Biomed. 16, 370 2003.
44
R. A. Komoroski, NMR Biomed. 18, 67 2005; I. K. Lyoo and P. F.
Renshaw, Biol. Psychiatry 51, 195 2002; R. Martino, M. Malet-Martino,
and V. Gilard, Current Drug Metabolism 1, 271 2000.
45
S. D. Rand et al., AJNR Am. J. Neuroradiol. 18, 1695 1997.
46
X. Leclerc, T. A. Huisman, and A. G. Sorensen, Curr. Opin. Oncol. 14,
292 2002; J. Butzen et al., AJNR Am. J. Neuroradiol. 21, 1213 1213.
47
A. P. Lin, T. T. Tran, and B. D. Ross, NMR Biomed. 19, 476 2006; H.
32
33
4544
S. Jordan et al., AHQR 2007 2003.
G. Bielicki et al., NMR Biomed. 17, 60 2004.
49
X. Zhao et al., Magn. Reson. Med. 53, 30 2005.
50
C. H. Meyer et al., Magn. Reson. Med. 15, 287 1990.
51
J. Vymazal et al., J. Neurol. Sci. 134, 19 1995.
52
R. W. Prost et al., Magn. Reson. Med. 37, 615 1997.
53
T. G. Perkins et al., ISMRM, Berlin, Germany, 2007 unpublished.
54
D. W. Klomp et al., Magn. Reson. Med. 55, 271 2006.
55
M. S. Albert et al., Nature London 370, 199 1994.
56
K. Ruppert et al., Magn. Reson. Med. 44, 349 2000.
57
J. H. Ardenkjaer-Larsen et al., Proc. Natl. Acad. Sci. U.S.A. 100, 10158
2003.
58
M. Goldman et al., Magn. Reson. Imaging 23, 153 2005.
59
G. F. Mason et al., J. Neurochem. 100, 73 2007.
60
K. Golman et al., Cancer Res. 66, 10855 2006.
61
E. E. Graves et al., Neurosurgery 46, 319 2000.
62
W. Hollingworth et al., AJNR Am. J. Neuroradiol. 27, 1404 1404.
63
T. Takano et al., Nat. Med. 7, 1010 2001.
48
Breast cancer imaging: A perspective for the next decade

Andrew Karellasa and Srinivasan Vedantham
Department of Radiology, University of Massachusetts Medical School, Worcester, MA 01655
Received 9 February 2008; revised 24 June 2008; accepted for publication 26 August 2008;
Breast imaging is largely indicated for detection, diagnosis, and clinical management of breast
cancer and for evaluation of the integrity of breast implants. In this work, a prospective view of
techniques for breast cancer detection and diagnosis is provided based on an assessment of current
trends. The potential role of emerging techniques that are under various stages of research and
development is also addressed. It appears that the primary imaging tool for breast cancer screening
in the next decade will be high-resolution, high-contrast, anatomical x-ray imaging with or without
depth information. MRI and ultrasonography will have an increasingly important adjunctive role for
imaging high-risk patients and women with dense breasts. Pilot studies with dedicated breast CT
have demonstrated high-resolution three-dimensional imaging capabilities, but several technological barriers must be overcome before clinical adoption. Radionuclide based imaging techniques and
x-ray imaging with intravenously injected contrast offer substantial potential as a diagnostic tools
and for evaluation of suspicious lesions. Developing optical and electromagnetic imaging techniques hold significant potential for physiologic information and they are likely to be of most value
when integrated with or adjunctively used with techniques that provide anatomic information.
Experimental studies with breast specimens suggest that phase-sensitive x-ray imaging techniques
can provide edge enhancement and contrast improvement but more research is needed to evaluate
their potential role in clinical breast imaging. From the technological perspective, in addition to
improvements within each modality, there is likely to be a trend towards multi-modality systems
that combine anatomic with physiologic information. We are also likely to transition from a standardized screening, where all women undergo the same imaging exam mammography, to selection of a screening modality or modalities based an individual-risk or other classification. 2008
Key words: breast cancer, mammography, digital mammography, tomosynthesis, CT, ultrasound,
MRI, contrast agents, optical imaging
I. INTRODUCTION
II. MAMMOGRAPHY
Imaging of the breast is indicated almost exclusively for the

detection, diagnosis, and clinical management of cancer and
for the assessment of the integrity of breast implants. Commonly used imaging modalities include mammography, ultrasonography, magnetic resonance imaging MRI, scintimammography, single photon emission computed
tomography SPECT, and positron emission tomography
PET. The goal of this work is to provide a prospective view
of breast imaging, primarily for detection and diagnosis.
In addition to standard recommendations for imaging
based screening, self breast examination SBE has long
been recommended because it is not unusual for women to
feel anatomic changes in the breast that turn out to be malignant. Clinical breast examination CBE is commonly
practiced as a part of routine physical examination by health
providers or when prompted by a concern arising from SBE.
The scientific evidence is inconclusive on the effectiveness
of SBE and CBE in reducing breast cancer mortality. However, a recent study demonstrated a modest improvement in
sensitivity when combined with mammography.1
Mammography is essentially the only widely used imaging modality for breast cancer screening. Various forms of
radiographic imaging of the breast have been used for nearly
one hundred years but mammography with dedicated equipment and technique did not emerge as a screening tool until
the mid-1960s.2 The importance of physical compression of
the breast and the significant association between microcalcifications and breast carcinoma were recognized in the early
1950s by Leborgne.3 While Gershon-Cohen in the United
States and Gros in Europe were strong advocates for breast
cancer screening, the development of a low kVp, high mAs
technique that can be performed in a reproducible manner in
1960 by Egan enabled an organized breast cancer screening
program.4 In 1966, Compagnie Gnrale de Radiologie
CGR, France in collaboration with Gros developed the first
dedicated mammography unit. This unit featured a molybdenum anode with a nominal 0.7-mm focal spot. Subsequently,
several system manufacturers developed dedicated mammography units that facilitated its widespread availability.
Several large randomized clinical trials have shown that
mammography reduces mortality from breast cancer.510
4878
0094-2405/2008/3511/4878/20/$23.00
4878
4879
A. Karellas and S. Vedantham: Breast cancer imaging: A perspective
While clinical trials have been essential in establishing the

efficacy of mammography as a screening tool, the critical
role of physics in the evolution of mammography cannot be
overemphasized. Key developments in mammography include the invention of sensitive high-resolution imageintensifying screens, improved films, low x-ray absorption
cassettes, and dedicated mammography film processors that
enabled major dose reduction.11 The development of an x-ray
tube with molybdenum Mo target; molybdenum filter, and
beryllium Be window, and with a small focal spot12 that
was specifically designed for mammography represents another major technological breakthrough. The concept of the
projection geometry in which the central ray grazes the chest
wall and meets the image receptor close to the chest wall is
unique to mammography, and it represents a crucial design
element that enables maximum inclusion of breast tissue.
The development of low-ratio, high primary transmission
moving grids and subsequently cellular grids are also important contributors to the high image quality achieved by modern mammography systems.
Screen-film mammography has long been considered as a
gold standard for breast cancer screening. In addition to its
ability to provide adequate visualization of soft tissue abnormalities, its particular strength is the ability to depict subtle
calcifications. While screen-film mammography is a powerful tool for initial detection and subsequent follow-up of suspicious lesions, it has certain inherent limitations which are
difficult to overcome. The most important and widely acknowledged weaknesses of screen-film mammography are
associated with its limited dynamic range, contrast characteristics, susceptibility to suboptimal film processing conditions, and granularity. It also presents significant limitations
in detecting very subtle lesions, especially in the presence of
dense glandular tissue. These limitations were well elucidated during the preliminary development stages of digital
mammography.13 Additional investment to develop improved
screen-film technology is unlikely because of the emergence
of digital mammography that provides wide dynamic range
and offers the convenience of digital image manipulation,
communication, and archival.
The term digital mammography is used for any technology which employs a single or multiple detector assembly to
capture an electronic image of the x-rays transmitted through
the breast that can be displayed, stored, and communicated
electronically.14 Digital mammography is now a standard
tool for breast cancer imaging and is steadily replacing
screen-film mammography as the preferred tool for screening. As of this writing, approximately 36% of all mammography units in the United States are digital,15 and considering
the higher throughput of digital than screen-film, it is likely
that more than 36% of the examinations are performed with
digital mammography. The conversion to digital is likely to
accelerate and within the next three to five years digital
mammography will be the dominant modality. The zerospatial frequency detective quantum efficiency DQE0 of
digital mammography systems is in the range of 0.45 to 0.65,
which is substantially higher than 0.35 or lower for screenfilm systems.1618 Clinical trials comparing digital mammogMedical Physics, Vol. 35, No. 11, November 2008
4879
raphy with screen-film mammography in a screening population demonstrate equivalency for cancer detection,1921 and
digital mammography performed significantly better for preand perimenopausal women younger than 50 years with
dense breasts.22
Current technological approaches for digital mammography that are in clinical use can be broadly classified into
fixed-detector and flexible-detector installations. In a
fixed-detector installation, the dedicated detector is hardwired to the support column of the mammography system
and is not removable. Examples of such fixed-detector installations include indirect conversion hydrogenated amorphous silicon a-Si:H based detectors and direct conversion
amorphous selenium a-Se based detectors. In contrast, a
flexible-detector installation uses the x-ray cassette holder
of a mammography system to house the detector. An example of such a flexible-detector system is computed radiography CR technology. One advantage of the flexibledetector installation is the ease of conversion of an existing
screen-film mammography system to a digital system.
An indirect conversion system uses an intermediary stage,
typically a scintillator to convert the transmitted x-rays to
light photons, followed by detection of the converted light
photons using an optical sensor. The scintillator of choice for
indirect conversion systems is CsI:Tl, which, due to its columnar structure, suppresses lateral light diffusion and,
hence, results in better preservation of spatial resolution. Development of a-Si:H arrays was an important contributor for
adaptation to large field of view full-field imaging.23 Physical characterization indicate a substantial improvement in
DQE 0.55 at zero-spatial frequency characteristics compared to screen-film mammography.18,24
A direct conversion system does not use an intermediary
stage and conversion is made from x-rays to electrons after
interaction in a photoconductive layer, typically a-Se. While
a-Se was used more than four decades ago in xeroradiography and xeromammography, the development of advanced
charge readout methods enabled its adaptation to digital
mammography.25,26 The elimination of the intermediate scintillator layer with direct detection systems allows such systems to achieve high spatial resolution,27 and higher DQE
characteristics 0.54 to 0.64 at zero-spatial frequency compared to screen-film mammography.17 In addition, a slot-scan
photon-counting digital mammography system that uses silicon strip detectors is in clinical use in some countries. This
approach, in addition to providing efficient scatter rejection
due to the slot-scan geometry,28,29 substantially reduces electronic noise contribution to the acquired image resulting in
improved detective quantum efficiency.30
Computed radiography CR uses a photostimulable phosphor plate typically consisting of BaFBr, I : Eu2+ crystals on
a suitable substrate in the form of a portable cassette. When
exposed to x rays, f-centers are created in proportion to x-ray
exposure and remain stable for several hours. This latent
image is read out by stimulating the photostimulable phosphor plate with a laser beam, typically in a raster fashion,
resulting in the emission of light photons in the ultraviolet-
4880
blue region. The emitted light is detected by a photosensor,

typically a photomultiplier tube, to generate an electronic
image. While CR has been in clinical use for radiographic
applications for nearly three decades, the development of a
dual-side readout that uses a transparent phosphor substrate
and light collection from both sides of the photostimulable
phosphor plate31 was an important contributor for improved
DQE 0.54 at zero-spatial frequency of such systems.16,32
Outside the United States, CR based single-side read digital
mammography systems using BaFI-, BaSrFBrI-, and
BaFBr,I-photostimulable phosphor plates are also in clinical use.33 Recent reviews of technological approaches for
digital mammography have been published.14,34,35
While there is no consensus on spatial resolution requirements in digital mammography, current research suggests
that for indirect detection approaches, a low noise detector
with smaller pixel size than that used in current practice may
enhance visual perception of small objects such as
microcalcifications.36 Recently, it was observed that noise
had a more dominant effect than display resolution for the
tasks of detecting microcalcifications and discrimination of
masses.37
Research and development to improve digital mammography system performance characteristics and to develop
new detector technologies, are in progress. In indirect-type
a-Si:H based detectors, a recent important advancement is
the development of a continuous photodiode design rather
than the traditionally used discrete a-Si:H photodiode array.38
This would increase the fill factor, resulting in improved signal transfer characteristics and dose efficiency. Also, incorporation of compensation lines with a-Si:H based detectors
has been shown to reduce correlated electronic noise arising
from external and ambient sources.39 Methods to limit lateral
diffusion of light within the scintillator either by laser etching columnar CsI: Tl40 or by infusion of crystalline CsI:Tl
within a tungsten grid matrix41 hold promise for improved
spatial resolution, but have not yet been translated to practice.
Recent research with direct-conversion a-Se based detectors have been focused mainly on understanding the temporal imaging characteristics image lag and ghosting
properties.4244 Knowledge of these properties will be essential for emerging techniques of digital breast tomosynthesis
DBT and contrast enhanced digital mammography
CEDM that require fast image acquisition and readout. Ongoing investigations are aimed at improving the gain of a-Se
based detectors through avalanche multiplication process and
in understanding its implication on noise.45,46 The development of low-noise detectors with improved sensitivity would
also be beneficial to the promising technique of DBT that
requires low dose per projection view. Attempts to adapt materials such as gallium arsenide GaAs, cadmium zinc telluride CdZnTe, mercuric iodide HgI2 and lead iodide PbI2
that hold promise of improved quantum efficiency are yet to
bear fruition due to challenges such as charge-traps, higher
leakage current, and spatial nonuniformity.4750
4880
In CR technology, some of the important advances include the development of columnar stimulable phosphor
screens based on CsBr,51 development of linear readout line
scan technology for faster readout of the phosphor,52 and
integration of these approaches.53,54 However, we are not
aware of their transition to digital mammography as yet, but
that could likely occur in the near future.
Another area of substantial interest in detector technology
is the development of energy-resolving large-area photoncounting detectors, which has been challenging. While several approaches for photon-counting detectors have been proposed including scintillation detection using micro-channel
plates55 and gaseous detectors, it appears that semiconductorbased direct-conversion photon-counting systems may be
better suited for medical imaging.56 Some of the semiconductor materials that are considered promising for photoncounting detectors include Si, GaAs, CdTe, and CdZnTe,
typically bonded to a two-dimensional readout with application specific integrated circuits, and are commonly referred
to as hybrid detectors. Several collaborative research projects
to develop hybrid detectors exist,57,58 increasing the prospects for their availability in the next decade.
The advent of digital mammography has prompted reconsideration of several aspects of mammographic imaging that
translated from screen-film imaging. These include anode
material, x-ray spectra, technique factors, and radiation dose.
In addition to the common target-filter combinations of MoMo, Mo-Rh, and Rh-Rh, there is an increasing trend to use
other combinations such as WRh, WAg, and WAl with
digital mammography. These alternative W-target techniques
may allow for modest reduction in radiation dose with no
apparent reduction in image quality.5961 In addition, the exploration of techniques such as contrast-enhanced digital
mammography CEDM and DBT that may require x-ray
tubes with increased heat capacity and output also have an
influence on the choice of anode material. In addition, radiation dose reduction with photon counting detectors,62 a-Si:H
indirect conversion detectors,63,64 and CR65 have also been
reported. While phantom studies indicated the possibility of
50% dose reduction with digital mammography,66 studies
with clinical backgrounds indicate that such a drastic dose
reduction could adversely affect detection of microcalcifications and discrimination of masses substantially, and to a
lesser extent adversely affect detection of masses.67,68 Continuing technological improvements and technique refinements could lead to further dose reduction while maintaining
image quality.
Extensive investigations on radiation dose to the breast
and its dependence on breast composition, breast thickness,
and x-ray spectral characteristics have been documented.6972
While it was believed that radiation dose to other organs
including the uterus would be low during a mammography
exam, until recently quantitative estimates were
unavailable.73,74 A Monte Carlo computational study with an
anthropomorphic software phantom indicates that the dose to
the uterus during the first trimester, where a women may be
unaware of her pregnancy, is less than 0.03 Gy per imaged
breast during a bilateral two-view mammography exam.75
4881
The use of a lead-shield further reduces this dose by a factor

of 2 to 7, suggesting that if it is clinically necessary for a
woman in her first trimester of pregnancy to undergo mammography, then the radiation dose to the fetus would be
minimal.75
The advantages of vigorous physical compression of the
breast in terms of adequate tissue separation, x-ray scatter,
and radiation dose, for producing good mammographic image quality are well known. Unfortunately, to this date, the
scientific community has been unable to overcome the vigorous breast compression needs of mammography. In screenfilm mammography, it is well established that the use of
anti-scatter grids improves contrast for breasts of average
and above average compressed thickness. However, the usefulness of anti-scatter grid is not firmly established for thin
and mainly adipose breasts. Current practice is to perform
mammography with the anti-scatter grid in place for all imaging needs, except for magnification views. In spite of its
advantages, anti-scatter grids do suffer from deficiencies
such as incomplete suppression of scattered radiation, reduction in primary radiation reaching the detector, and
artifacts.29,76,77
With the advent of digital mammography that would allow for numerical scatter correction techniques, the possibility of imaging without the anti-scatter grid were studied and
showed promise of improved performance.7880 However,
another study indicated that removal of the anti-scatter grid
could provide a modest reduction of 8% in radiation dose
to the subject.81 Although scatter correction for images acquired without the anti-scatter grid using numerical techniques have been reported,8284 its impact on contrast, resolution, noise, and artifacts are yet to be studied. At present,
digital mammography, with the exception of slot-scanning
systems, uses anti-scatter grids for scatter reduction. The
emergence of DBT where the use of an anti-scatter grid is
impractical, could further advance the research on scatter
correction algorithms that may translate to digital mammography in the future.
We will continue to observe increased transition to digital
from screen-film mammography. Also, improvements in detector technology and transition to higher kVp techniques
will provide for substantial reduction in radiation dose. The
prospects of a large-area photon-counting detector with the
ability to provide energy resolution appears brighter.
III. ULTRASONOGRAPHY
The role of B-mode ultrasound in breast imaging has been
largely limited to applications such as distinguishing between cystic versus solid masses, evaluation of palpable
masses, and for needle core biopsy of masses. In recent
years, the number of indications has been greatly expanded
and breast ultrasonography is now an essential modality in
breast imaging. A center frequency above 10 MHz is now
recommended for adequate spatial resolution by the American College of Radiology ACR. B-mode imaging works
well for the established indications of breast ultrasonography,
4881
but its imaging ability is limited by the beam direction that is

fixed and perpendicular to the face of the transducer.
In compound ultrasonography, the beam is electronically
steered for improved sampling of the anatomy that may lie
parallel to the ultrasound beam, rendering a more complete
image. Harmonic imaging techniques take advantage of the
variation of the velocity of the ultrasonic wave through different tissues that produces harmonic frequencies. The processing of harmonic signals leads to better selection of signals by suppressing effects such as side-lobe artifacts, beam
defocusing, and reverberations, resulting in improved
images.8587 Color Doppler and more recently sensitive
power Doppler ultrasound has been used for further evaluation of the breast.
In 2002, Kolb et al. published a landmark article that
showed improved sensitivity 97% versus 74% when adjunctively used with mammography compared to physical
examination with mammography.88 They also observed an
improvement in sensitivity for dense breasts with ultrasound
compared to mammography. In addition, they noted that the
time taken to perform an ultrasound exam was comparable to
that of a clinical breast exam. However, he also observed an
increased in false positive rates with the use of ultrasound as
an adjunct to mammography. In recently published results of
an ACRIN trial that included 2637 women with heterogeneously dense breast in at least one quadrant observed that
the diagnostic accuracy improved from 0.78 to 0.91 when
ultrasound was adjunctively used with mammography.89
However, there was also a substantial decrease in positive
predictive value with mammography plus ultrasound 11.2%
compared to mammography 22.6% alone. Also, the median
time to perform a bilateral ultrasound exam was 19 min.
Hence, the improved diagnostic accuracy provided by adjunctive use of ultrasound to mammography has to be carefully weighed against the decrease in positive predictive
value and exam duration. From the technological perspective, it can be inferred that there may be a need for a combined dual-modality mammography-ultrasound system. Even
if ultrasound is not adapted for screening, such a system is
likely to be of benefit for diagnostic evaluation as it provides
for co-registered dual-modality images.
Development of a semi-automated scanning ultrasound
device integrated with a digital breast tomosynthesis DBT
system was recently reported.90 Figure 1 shows the combined DBT-US system. Performance characteristics of the
DBT component indicates minimal in-plane blur and better
management of artifacts with reconstruction using generalized filtered backprojection.91 Identification of compression
paddles suitable for both x-ray and ultrasound imaging,92 as
well as techniques to address coverage and motion issues
during ultrasound scan93 have been described. Recently, development of an ultrasound computed tomography system
that generates reflection, attenuation, and sound speed images has been reported.94 Spatial resolution measurements
indicate in-plane resolutions of 0.5 and 4 mm in reflection
and transmission modes, respectively. A pilot clinical study
with 50 subjects indicate the ability to routinely detect
masses greater than 15 mm in size.95
4882
FIG. 1. Integrated dual-modality digital breast tomosynthesis and scanning

ultrasound system. Provided by Paul L. Carson and Mitchell M. Goodsitt at
the University of Michigan and Kai Thomenius at GE Global Research,
Niskayuna, NY. This system was developed with support by NIH R01CA
91713 and Office of Naval Research grant MDA905-00-10041.
Technological improvements in transducer technology

and signal processing have contributed to the extensive
adoption of ultrasonography for diagnostic evaluation of the
breast and for biopsy guidance. There is a possibility that
ultrasound could have an adjunctive role to screening mammography for hetereogeneously dense breasts.
IV. MAGNETIC RESONANCE IMAGING
In vitro measurements of relaxation times of tumors and
normal tissue with nuclear magnetic resonance NMR was
reported by Damadian in 1971.96 Subsequent to the development and application of local gradient fields that made imaging NMR zeugmatography as proposed by Lauterbur97
feasible, Mansfield et al. reported on imaging with surgical
breast tissue samples.98 Ross et al. and El Yousef et al. reported on some of the earliest in vivo breast MRI studies in
the early 80s.99101 Development of a coil specifically designed for breast imaging was an important technological
advancement.102 The development of gadolinium diethylene
triamine pentaacetic acid Gd-DTPA contrast agent and its
application to human imaging was an important milestone in
the advancement of MRI.103108 Following the application of
Gd-DTPA for contrast-enhanced breast MRI,109 initial interpretation criteria were based on morphology and contrast enhancement that were obtained with long acquisition times.
However, this did not provide substantial contrast between
tumor and proliferative changes. Hence, further research was
focused on developing fast pulse sequences110113 that ushered the era of dynamic contrast enhancement studies. Another important advancement was the development of a bilateral breast coil that facilitated simultaneous imaging of
both breasts. This allowed comparison of enhancement patterns between the breasts at reduced cost and time.
Currently dynamic contrast enhanced breast MRI is clinically used to provide volumetric three-dimensional 3-D
anatomical information and physiologic information that are
indicative of increased vascular density and vascular permeability changes associated with angiogenesis.114 Clinical
breast MRI studies have demonstrated its ability to provide
accurate diagnosis, extent of disease and multi-centricity,115
4882
and the ability to detect mammographically occult cancers in

the contralateral breast.116119 Independent clinical trials for
women at high-risk of hereditary breast cancer indicate increased sensitivity with breast MRI than mammography but
with variable specificity.120131 These studies have prompted
the American Cancer Society to recommend the use of MRI
as a adjunct to screening mammography for women with a
lifetime risk of 20%25% or greater.132 A recent meta analysis that included 44 prior studies indicate that breast MRI has
a sensitivity of 0.9 and specificity of 0.72.133 Importantly,
this analysis indicated that variability in specificity was due
to cancer prevalence in the individual studies and due to
interpretative criteria used to differentiate benign and malignant lesions.133 With the development of lexicons for breast
MRI interpretation134 and standardization of procedures,
variation in specificity is likely to be reduced. A recent twopart review addresses the current status of breast MRI.135,136
Dynamic contrast enhanced MRI DCE-MRI is a very important tool for detection, diagnosis, and clinical management of breast cancer. However, it requires intravenous injection of Gd-contrast agent that entails some elevated
risk.137,138 Hence, by its nature contrast-enhanced MRI is
likely to be limited to a subgroup of patients.
Another promising technique that has garnered substantial
interest recently is proton 1H magnetic resonance spectroscopy MRS. This technique allows for quantitative characterization of total or composite choline concentration that has
been shown to be elevated in malignant tumors compared to
normal breast tissue.139 While initial works were focused on
31
P MRS,140143 recent research has been primarily on the
use of 1H MRS,144147 in part due to its higher sensitivity
compared to 31P MRS.148 MRS is a nonvasive technique that
does not require contrast injection; however, most of current
research has been on spectroscopic analysis of regions localized from prior DCE-MRI. Current studies have shown the
correlation between DCE-MRI and MRS,149,150 and demonstrated improved sensitivity and specificity when used as an
adjunct to breast MRI.151153 While there are interesting aspects of this technique, most of the studies were performed at
1.5 T with voxel volume of 1 cm3 indicating better suitability for large tumors. Since total choline concentration
may be indicative of cell replication,154 MRS is being actively investigated and show promise for early determination
of the effectiveness of neoadjuvant chemotherapy for locally
advanced breast cancer.155159 In addition, studies on MR
spectroscopic imaging MRSI primarily to improve specificity have been reported.160162 Importantly, one of the studies describe a method to improve signal-to-noise ratio and
reduce acquisition time with sub-centimeter voxel volume.162
If this technique can evolve to the point of replacing
contrast-enhanced MRI, then risks associated with contrast
media and injection can be eliminated. Although this technique is in its infancy, the concept of a screening MR exam
that provides high resolution morphological information with
MRSI to provide physiologic information without contrast
media is highly appealing. Recent reviews address the clini-
4883
cal prospects and challenges of MRS and allied

techniques.163166
MRI already has a very important role in breast imaging
as an adjunctive tool for screening and as a diagnostic tool
for imaging lesions that are indeterminate from other modalities. MRI is also widely used for assessing the structural
integrity of breast implants. Its role will continue to expand
with the development of novel image acquisition techniques
and technological improvements.
V. RADIONUCLIDE IMAGING
Radionuclide based imaging techniques such as scintimammography, single-photon emission computed tomography SPECT, positron emission mammography PEM, and
positron emission tomography PET are additional imaging
techniques that provide for physiologic information.
Scintimammography and SPECT typically use
99m
Tc-sestamibi or 99mTc-tetrofosmin for breast cancer imaging. Scintimammography is used for imaging mostly palpable lesions that were occult or indeterminate from other
imaging modalities. For palpable mass, a meta-analysis indicates that the sensitivity and specificity were 87.8% and
87.5%, which is high.167 However, for nonpalpable lesions
the sensitivity was reduced to 66.8%.167 In an another study,
the sensitivity reduced from 74.2% for larger lesions to
48.2% for lesions smaller than 1 cm.168 New technological
advances such as the development of dedicated cameras for
scintimammography with improved spatial resolution has renewed interest in this field.169 Characterization of a dedicated
scintimammography system was recently reported to achieve
detection of 7 mm spheres bearing 99mTc.170 A clinical
study to evaluate its potential with 100 subjects scheduled to
undergo biopsy of 2 cm or smaller masses indicated a substantial improvement in sensitivity for imaging lesions
smaller than 1 cm.171 In an another study in which 94 highrisk subjects with normal mammographic exam were evaluated with a dedicated scintimammography camera using
99m
Tc-sestamibi detected two malignancies, both of which
were smaller than 1 cm.172 However, false-positives due to
fibrocystic change, fibroadenoma, and fat necrosis were also
observed. A study comparing dedicated scintimammography
with conventional SPECT indicated no differences in sensitivity, but observed better performance for 1 cm or smaller
lesions with the dedicated scintimammography system.173
Further improvements, such as development of a dual-head
camera and with increased spatial resolution may improve
sensitivity for smaller lesions.
The primary advantage of SPECT is its ability to provide
3-D information. In SPECT imaging, the development a
dedicated system for breast imaging in a prone patient position with pendant breast geometry has been reported.174 This
system uses a scanning trajectory that closely follows the
breast contour. A more recent work described the development of a dedicated breast SPECT system using CdZnTe
detectors.175 Integration of the dedicated SPECT system with
a dedicated breast CT system to provide for co-registered
dual-modality imaging is also in progress.176
4883
For detection of primary tumors, 18F-FDG-PET have been

reported to have similar sensitivity as that of SPECT.177 A
meta-analysis of whole body 18F-FDG-PET that included 13
studies indicated an overall sensitivity of 89% and specificity
of 80%.178 However, the mean tumor size in those studies
ranged from 2 to 4 cm. Characterization of a dedicated
breast PET system using four detector heads comprising of
LYSO crystals coupled to position-sensitive photomultiplier
tubes PSPMTs with multi-angle tomographic image acquisition capabilities was reported to achieve 2 mm spatial
resolution.179 Development of PET detector modules with
lutetium oxyorthosilicate LSO crystals coupled to
PSPMTs,180 and more recently LSO coupled to position sensitive avalanche photodiodes with depth-of-interaction capabilities have been reported.181,182 These systems can play an
important role in monitoring effectiveness of therapeutic
regimens.
To overcome the limited sensitivity and spatial resolution
of whole-body PET systems, a PEM system was
developed183,184 and showed promising results in a pilot
clinical study.185 Development of PEM systems using gadolinium orthosilicate GSO and yttrium-aluminium perovskite YAP scintillators have been reported.186,187 A collaborative effort to develop a dedicated PEM system using
cerium- and yttrium-doped lutetium orthosilicate LYSO
scintillator elements coupled to avalanche photodiodes is
also underway.188 These systems offer the potential of improved resolution and sensitivity. A pilot clinical study evaluating a prototype system using lutetium gadolinium oxyorthosilicate LGSO scintillator elements showed the ability
to detect smaller lesions.189 All of the aforementioned dedicated PEM systems use stationary detector heads. A multiinstitutional clinical study of a commercially available PEM
system using translational detector heads that enrolled 94
subjects with known breast cancer or suspicious lesions
showed high sensitivity and specificity and the ability to detect small lesions.190
A recent review address the clinical aspects and limitations of various radionuclide-based techniques for breast
cancer imaging.191 With continued advancement of detector
technology, radionuclide-based imaging systems dedicated to
breast imaging are likely to play a significant role in therapeutic monitoring of known cancers and in diagnostic imaging of suspicious breast lesions.
VI. INVESTIGATIONAL MODALITIES
VI.A. Volumetric x-ray imaging techniques
While there have been considerable advances in mammography, there is one major inherent limitation in that the
mammographic image reduces the three-dimensional
anatomy of the breast into a two-dimensional image. This
resulting superposition of normal breast structures, which
causes a visually distracting mask, is often referred to as
anatomical noise and has been shown to impair lesion
detection.192194 In addition, tissue superposition can mimic
the presence of a lesion, resulting in increased recall rates
and in increased biopsy rates, that add to the inconvenience
4884
and anxiety of the recalled subject. Hence, there is a need to

develop techniques that provide depth information in breast
x-ray imaging. Currently, stereoscopic digital mammography
SDM, digital breast tomosynthesis DBT, and dedicated
breast computed tomography BCT are three modalities that
are being actively investigated.
VI.A.1. Stereoscopic digital mammography
In stereoscopic digital mammography, two projection images spaced a few degrees apart are acquired with a digital
mammography system. A dedicated stereoscopic workstation
displays orthogonally polarized images and are viewed by
the observer using passive cross-polarized glasses, so that
each eye visualizes one image to provide depth perception.
Studies evaluating the accuracy of depth perception indicate
that 0.2 mm depth discrimination can be achieved using projection pair images acquired at an angular separation of 6
and with geometric magnification.195,196 Observer studies
comparing standard monoscopic imaging with stereoscopic
imaging using breast biopsy specimens indicate an improvement in detection of masses and microcalcifications, and an
improved confidence for the estimation of likelihood of
malignancy.197 Interim results from an ongoing prospective
screening trial with high-risk women suggest that SDM reduces false-positive lesion detections by 39% and falsenegative lesion detections by 46% compared to digital mammography, providing for significant improvements in
sensitivity and specificity.198 However, the study used twice
the dose of mammography. While contrast-detail experiments with a phantom indicates that 10% more radiation
dose is needed for SDM compared to mammography,199 further research is needed to understand the dose requirements
with clinical backgrounds. If ongoing and future clinical trials demonstrate its efficacy, it has the potential for quick
adaptation due to the ready availability of digital mammography systems and dedicated stereoscopic displays.200
VI.A.2. Digital breast tomosynthesis
Digital breast tomosynthesis DBT, a technique that relies upon acquisition of multiple projection views over a limited angular range to reconstruct a volumetric image, has
gained substantial interest primarily due to the availability of
flat-panel detectors.91,201203 Subsequent to the landmark article in 1997 by Niklason et al. that demonstrated feasibility
by modification of a digital mammography system,204 a similar system was used to study contrast-detail characteristics of
several reconstruction algorithms.205,206
Although in principle, implementation of DBT with a
digital mammography system is straightforward, there are
some design aspects that have to be considered. In a welldesigned DBT system, the x-ray tube assembly has to avoid
interference with the patients head during rotational movement and the rotational mechanics must provide for minimal
vibration. Also, the x-ray beam collimation must adapt asymmetrically during the scan to ensure that the x-ray field does
not extend marginally beyond the lateral edges of the detector. The acquisition of multiple frames in 10 s to minimize
4884
the potential for patient motion requires a detector that has

rapid frame acquisition capability and with minimal image
lag.
Current implementations of clinical prototype DBT systems use indirect conversion a-Si:H detectors with a
100-m pixel pitch91 or direct conversion a-Se
detectors.202,207 Depending upon the manufacturer, a-Se
based DBT system use a detector with 70-m pixel pitch
operated in a 2 2 binned mode202 or a detector with
85-m pixel pitch operated in unbinned or 2 1 asymmetric
binned mode.207,208 Pixel binning approaches allow for faster
readout of the detector, albeit at reduced resolution. As with
digital mammography, the detector pixel pitch and the overall spatial resolution requirements for breast tomosynthesis
are not well defined. After reconstruction, there is some degradation of in-plane plane parallel to the detector spatial
resolution due to several factors including oblique x-ray
incidence,209,210 blur due to tube motion within the acquired
projection view,208 mechanical vibration of the gantry, and
reconstruction related blur such as that due to interpolation.
There is considerable loss of spatial resolution along the
depth-direction perpendicular to the detector plane due to
limited angular sampling. Hence, there is a some concern
about its ability to detect subtle structures such as amorphous
microcalcifications.
Until recently, estimates of glandular breast dose during
DBT were based on conversion factors for mammography. A
Monte Carlo study provided conversion factors for DBT that
showed variation in glandular dose with projection angle,211
facilitating the application towards advanced techniques such
as tube-current modulation with changing projection angle, if
needed. While the objective for a DBT acquisition is not to
exceed the dose of a two-view mammogram, in a 15 projection acquisition for example, the detector receives substantially less signal per projection compared to a mammographic view. To some extent, this can be compensated by
increasing the beam transmission though the breast with the
use of W anode and higher kVp than in conventional mammography, as suggested by modeling and phantom
studies,212214 but may change as more experience is gained
through clinical trials. Research in developing detectors with
performance characteristics suitable for DBT seems well justified, in particular if higher kVp is used in the future for
techniques such as contrast-enhanced DBT.
The adverse effects of scatter on projection imaging and
computed tomography are well known. However, its adverse
effects in DBT have not yet been quantified. Recently, comprehensive Monte Carlo based estimates of scatter and
hence, scatter-to-primary ratio SPR was reported.215 SPR
trends showed high dependence on compressed thickness
and weak dependence on x-ray spectrum and breast glandularity, similar to mammography. Importantly, substantial spatial variation of SPR at oblique projections was observed.
One interesting observation in that study215 was the increase
in SPR observed at breast periphery, which was apparently
due to the large component of x-ray scatter from the breast
support plate compared to the primary x-ray component under the breast tissue. Conventional anti-scatter grids are not
4885
4885
FIG. 2. A clinical case showing 1.6 cm low-grade invasive ductal carcinoma with minor ductal carcinoma in situ DCIS component. Better visualization of
the tumor is observed with DBT right, B compared to mammography left, A. Courtesy: Steven P. Poplack, Dartmouth Hitchcock Medical Center
used in DBT because of the required increase in dose bucky

factor and problems due to grid cut-off at oblique projection
angles. While it may be possible to develop anti-scatter grids
that are designed specifically for DBT, such designs have not
yet materialized. Hence, accurate characterization of x-ray
scatter during a DBT exam can be extremely useful in the
development of computational techniques for scatter correction.
DBT reconstruction algorithms either use variants of
back-projection technique or iterative techniques. Optimization of these reconstruction methods, quantification of their
effect on resolution, noise, and artifacts as well as comparative analysis of these methods have been the subject of ongoing research.216222 There are two important concerns with
DBT, viz., its ability to provide adequate visualization of
microcalcifications223 and out-of-plane artifacts that occur
due to limited angular sampling. With regards to microcalcification visibility, a technique that could provide for improved morphology using back-projection or its variants was
recently published.217 Also, improved image quality has been
reported with an iterative algebraic reconstruction
technique.218 Techniques specifically designed for suppressing artifacts are being actively investigated.219,224226
In spite of some challenges, DBT enables better separation of overlying tissues than mammography and it is likely
to play an important role in breast imaging. Breast tomosynthesis has the potential of detecting occult lesions, particularly soft tissue abnormalities that can be extremely difficult
to detect in the presence of dense glandular tissue. Currently,
several clinical studies with human subjects are being conducted to evaluate the clinical potential of DBT. Early indications from ongoing clinical studies suggest that DBT show
promise.223,227 In a recently published study on their initial

experience with DBT, most of the soft tissue abnormalities
were subjectively rated as either equivalent or superior to
mammography.223 Figure 2 shows a clinical case of a woman
with 1.6 cm low-grade invasive ductal carcinoma with ductal
carcinoma in situ DCIS, where the carcinoma is better visualized with DBT right than with mammography left.
Interim results after accrual of 1957 women from an ongoing
clinical trial indicate that DBT reduces recall rate to 4.4%
compared to 7.5% with digital mammography, and unlike
mammography does not appear to show dependence with
breast density.227 With further refinements, breast tomosynthesis is likely to become a clinical tool for screening, particularly for women with dense breasts.
VI.A.3. Dedicated breast computed tomography

Another technique that will overcome the tissue superposition problem is dedicated breast computed tomography
BCT, in which the subject is in a prone position and the
pendant breast is imaged. The concept of a dedicated BCT
system was envisioned in the 1970s228 and a system was
built. Clinical trials using the dedicated BCT system229 and
with a conventional CT scanner230 suggested that injected
contrast media enhanced CT could differentiate malignant
from benign lesions. However, it did not translate to clinical
practice due to the limitations with the technology available
at that time. Subsequently in the mid-1990s, images of surgical biopsy specimens acquired with a state-of-the-art scanner at that time were compared with specimen radiography
4886
FIG. 3. Clinical images of a soft tissue lesion with digital mammography

A: MLO view and multi-planar reconstructions B: Coronal, C: Axial, and
D: Sagittal views with dedicated BCT. Courtesy: Koning Corporation,
West Henrietta, NY
images, and indicated that CT imaging can significantly improve the confidence to detect mass but was inferior for
microcalcifications.231
In 2001, a landmark article by Boone et al. provided
quantitative estimates of radiation dose from a dedicated
BCT exam and image quality achieved with cadaveric
breast.232 The development of high-frame-rate flat-panel detectors accelerated further exploration of this technique.
Works addressing radiation dose to the imaged breast232234
and other organs,235 x-ray scatter,236,237 system design,238,239
imaging trajectory,240,241 image acquisition technique
factors,242244 resolution,245 noise,246 and artifacts 247 have
been published.
Independent studies with surgical mastectomy
specimens248 and with limited number of subjects249 suggest
that BCT provides for excellent anatomical detail and soft
tissue lesion visualization. Figure 3 shows multi-planar reconstructions acquired with a clinical prototype dedicated
BCT system and a MLO view projection image acquired
with a digital mammography system. Excellent 3-D visualization of the anatomy and soft tissue lesion is observed. A
recent study that compared visualization in BCT with lesions
detected in screen-film mammography in 65 subjects, suggest that BCT was superior for visualization of masses but
was inferior for visualization of microcalcifications.250
Some of the challenges with breast CT include simultaneous inclusion of medial and axillary aspects of the breast
during the scan and adequate visualization of
microcalcifications.250 Limitations in breast coverage with
dedicated breast CT has been inferential, and as yet there
have been no quantitative studies comparing tissue inclusion
with mammography. Improved breast coverage can be potentially achieved through system design including novel table
design251 and imaging trajectories,240,241 and with appropriate patient positioning. Current prototype systems used in
clinical studies utilize circular scan,250,252 and we are aware
of only one implementation using a non-circular trajectory
4886
for breast imaging with a laboratory prototype system.240

Further studies are needed to understand tissue coverage
with dedicated breast CT for the various source-detector trajectories.
Initial computer simulation study and independently a receiver operating characteristic study with computer simulated BCT images suggested the possibility of detecting microcalcifications in the range of 175 to 200 m.238,253
However, experimental studies with phantoms suggest that
visibility of microcalcifications may be limited to 300 m
with current technology.254,255 Considering that these studies
point to a noticeable improvement with increased dose or
reduced breast size,254,255 it is likely that image noise rather
than resolution may be a dominant factor for limited visibility of microcalcifications. Hence, techniques for noise
suppression246 may be beneficial. However, further research
is needed to study its effect on artifacts, resolution, and lesion visualization. Additional improvements in microcalcification visualization can be potentially achieved with the use
of high-resolution, low-noise detector technology such as
photon-counting detectors256 or electron-multiplying
CCDs,257 which are not yet available for large field of view
imaging. A recent review article addresses the potential advantages of BCT and the challenges that need to be overcome for routine clinical use.258
While the role of BCT and its efficacy need to be ascertained through clinical studies, BCT has one major advantage in that it does not require physical compression of the
breast, thus greatly alleviating patient discomfort. Ongoing
clinical studies have shown exquisite anatomical detail and
soft tissue lesion morphology. If visibility of microcalcifications is improved, breast CT allows for 3-D visualization of
its distribution that could potentially improve specificity.
While each of the aforementioned techniques that are targeted towards overcoming the tissue superposition problem
are in various stages of research, an observer study with
computer simulated images comparing digital mammography, DBT and BCT showed improved accuracy to detect
masses,259 highlighting the potential advantage of DBT and
BCT over digital mammography. Among the three techniques SDM, DBT, and BCT that are being investigated to
provide depth information, the body of knowledge is too
limited to obtain meaningful inferences about the superiority
of one technique versus the other.
VI.B. Contrast-enhanced x-ray imaging
While investigations into the applicability of digital subtraction angiography of the breast in the mid-1980s showed
the potential for differential diagnosis for malignant and benign tumors,260,261 the lack of appropriate imaging system
was a major limitation. The advent of digital mammography
has stimulated the exploration of quicker and cost-effective
alternative for breast MRI such as intravenously injected iodinated contrast media enhanced imaging of the breast for
angiogenesis imaging. Two approaches for contrastenhanced digital mammography are being explored. In dualenergy K-edge subtraction, two images at different energies
4887
that straddle the K-edge of the contrast media are acquired

after contrast administration and subtracted.262 In temporal
subtraction, an image acquired prior to contrast administration is subtracted from the images acquired after contrast
administration.263 The former method reduces the likelihood
of misregistration between the two images, while the latter is
better suited for dynamic contrast enhancement studies and
would require image processing for proper registration.
Multiple pilot clinical studies, each with a small number
of subjects who had suspicious findings and underwent
contrast-enhanced digital mammography, indicate enhancement of tumor and vasculature that are qualitatively similar
to breast MRI.262266 There have also been studies on optimization of the x-ray spectrum,267,268 on determination of the
minimum concentration of iodine required,267,269 and on investigation of elements other than iodine that may provide
for alternate contrast media.270272
A pilot clinical study evaluating the potential of contrast
enhanced DBT, showed characteristics that were concordant
with digital mammography and breast MRI.273 Contrastenhanced DBT is being actively investigated, particularly in
terms of imaging technique optimization and quantification
of iodine concentration.268,274,275 The role of contrastenhanced conventional CT in management of breast cancer
such as lymph node imaging, tumor staging, identifying tumor extent for breast conservative surgery, treatment planning for radiotherapy, and in monitoring of cancer therapy is
well established. A pilot study with contrast-enhanced dedicated BCT showed enhancement of malignant lesions similar
to MRI.250 One potential advantage of contrast-enhanced
x-ray imaging is that signal intensity changes are typically
linear with contrast agent uptake, unlike MRI.135 All of the
aforementioned x-ray-based contrast-enhanced imaging techniques are in various stages of research.
While there are promising aspects, all of these techniques
require intravenous injection, which entails a slightly elevated risk.137 While contrast media used with x-ray imaging
are considered relatively safe, risks including allergic reactions and contrast nephropathy, though uncommon, are of
concern. Hence, there is a need to develop contrast agents
with minimal or no adverse effects. Further improvements in
specificity are achievable with development of tumortargeted agents276,277 and with better understanding of the
criteria for determining malignancy such as morphology, tumor enhancement due to agent uptake, and agent uptake kinetics.
VI.C. Alternative modalities
X-ray phase sensitive imaging such as in-line phase contrast x-ray imaging278,279 and diffraction enhanced
imaging280283 are currently being investigated. The theoretical basis and design considerations for phase contrast
imaging284,285 and method to retrieve the phase map from
attenuation and phase contrast images286 have been well addressed. Development of a dual-detector system that uses
two CR cassettes for simultaneous acquisition of conventional attenuation and phase contrast images, from which
4887
the phase map is retrieved was recently published.287 Figure

4 shows conventional attenuation and phase contrast images of ACR-recommended accreditation phantom and of a
lumpectomy specimen acquired with a prototype system.
Substantial improvement in contrast and edge-enhancement
is observed. These techniques are at present limited to imaging specimens. Further research is needed to translate them
to clinical use.
Optical imaging of the breast such as diffuse optical tomography DOT, diffuse optical imaging, and diffuse optical spectroscopy is being investigated as an adjunct technique. This emerging technique uses light in the nearinfrared NIR regime to noninvasively image total
hemoglobin, oxygen saturation, water content, optical scattering, and lipid concentration.288290 While substantial variations in optical properties for normal tissues291 and tumors292
have been observed, a recent study indicates that normal and
tumor tissues can be differentiated based on total hemoglobin, hemoglobin oxygen saturation, and other parameters.293
Leaky, densely packed vasculature associated with tumor angiogenesis could contribute to increased total hemoglobin,
and increased metabolism in tumors could result in decreased oxygen saturation. NIRS has an inherent resolution
of 4 to 6 mm because of the highly scattering nature of
light. Ntziachristos et al.294 demonstrated that combining
DOT with MRI anatomical maps can improve spatial resolution. More accurate quantification of NIRS parameters
have been reported by incorporating structural priors from
MRI during NIRS reconstruction.295 In order to obtain coregistered dual-modality images, investigators at Dartmouth
have incorporated NIR spectroscopy instrumentation within
an open-frame MR breast coil Fig. 5. This combined approach has been used to demonstrate feasibility by imaging
more than 25 women with both normal296 and disease indications from mammography.297 Correlation between indocyanine green enhancement and Gd-enhanced MRI,298 and
its ability to differentiate cysts from solid tumors299 have
also been reported. Ongoing investigations indicate that
DOT in combination with MRI hold promise for early determination
of
the
effectiveness
of
neoadjuvant
chemotherapy.300302 In addition to NIRS-MRI systems, development of DOT imaging systems in combination with
DBT and ultrasound are also underway.303305 A recent article provides an assessment of its future role in breast
imaging.306
Electrical impedance spectroscopy EIS and microwave
imaging spectroscopy MIS are also being explored for potential use in breast cancer detection. A recent study suggests
that EIS could potentially be used as a risk assessment tool
for young women who are not routinely screened for breast
cancer.307 In vivo microwave imaging of women with normal
mammographic exam suggests that dielectric properties are
heterogeneous and correlate with radiographic density.308 A
recent two-part series on measurements of ultra-wideband
dielectric properties from normal and cancer tissue samples
indicate substantial variations in dielectric properties for normal tissues, and that the contrast in dielectric properties between malignant and adipose tissues could be as high as
4888
4888
FIG. 4. Conventional attenuation images A, C and phase contrast images B, D of accreditation phantom top row and lumpectomy specimen bottom
row. All images were acquired at 40 kVp with same entrance exposure for attenuation and phase contrast images. These images were acquired using
prototype systems developed at the University of Oklahoma Hong Liu, in collaboration with University of Alabama at Birmingham Xizeng Wu and
University of Iowa Laurie L. Fajardo. Courtesy: Hong Liu, University of Oklahoma
10:1, whereas that of malignant to fibroglandular tissue could

be no more than 10%.309,310 A pilot clinical study that evaluated EIS, MIS, and NIRS indicate substantial contrast difference between abnormal and normal breast tissue.311 A recent
article addresses the status of these imaging modalities.312
VII. CONCLUSION
Intense efforts are currently under way to improve the
technological aspects of the aforementioned modalities.
Those include advances in x-ray and gamma ray detector
technology, MRI coil design, MRI pulse sequences, tomographic image reconstruction, signal processing, computational speed, and mechanical design. Based on current understanding and results from ongoing research, it appears that
high-resolution, high-contrast, anatomical x-ray imaging either in 2-D mammography or with added depth information
will be the primary screening modality in the next decade.

However, MRI and ultrasound will have an increasingly important role for imaging high risk patients and for imaging
women with dense breasts. With increasing field strength and
improved homogeneity, MRS is likely to add value to dynamic contrast enhanced MRI studies. Contrast-enhanced
x-ray imaging, scintimammography with dedicated cameras,
dedicated SPECT, and dedicated PET offer great potential as
diagnostic evaluation tools and for clinical management.
While prototype BCT systems have been developed and pilot
clinical studies demonstrate the ability to provide highresolution 3-D anatomical information, there are some challenges that need to be addressed. NIR diffuse optics-based
imaging and electromagnetic imaging techniques such as EIS
and MIS could provide for physiologic information and is
likely to be of most value when integrated with or adjunc-
4889
4889
screening of high-risk subjects. This transformation to

individual-risk or other classification based screening is
likely to be of benefit to women. However this added layer,
which can be referred to as pre-screening, does place an onus
on the manner in which such classifications or risk assessments are performed and need to be carefully studied. In
summary, ongoing research and recent advances indicate that
the prospects of substantial improvements in early detection,
accurate diagnosis, and improved monitoring of therapeutic
response of breast cancer are highly promising.
ACKNOWLEDGMENTS
The authors would like to thank Dr. Paul L. Carson and
Dr. Mitchell M. Goodsitt at the University of Michigan, Dr.
Hong Liu from the University of Oklahoma, Dr. Brian W.
Pogue from Dartmouth College, Dr. Steven P. Poplack at
Dartmouth Hitchcock Medical Center, and Koning Corporation for providing images. This work was partially supported
by National Institutes of Health NIH Grant No. R01EB002123 and Grant No. R01-EB004015 from the National
Institute of Biomedical Imaging and Bioengineering
NIBIB. The contents are solely the responsibility of the
authors and do not necessarily represent the official views of
the NIH or the NIBIB. Mention of company names or products does not imply endorsement. The authors have collaborated in the past with GE Global Research Center.
a
FIG. 5. Dual-modality MRI-NIR system with insert showing the NIR

transmitter-receiver incorporated within the open breast coil. Courtesy:
Brian W. Pogue, Thayer School of Engineering, Dartmouth College
tively used with modalities that provide for anatomical structure such as x-ray, ultrasound, or MRI. At present, phasesensitive x-ray imaging that provides for improved contrast
and edge-enhancement has been demonstrated with specimens and would require substantial amount of translational
research to bring it to clinical use. From the technological
perspective, in addition to improvements with each modality,
we are likely to observe an increasing trend towards multimodality systems that combine the relative strengths of each
modality.
Importantly, we are likely to observe a paradigm shift in
the manner in which breast cancer screening will be performed in future. Breast cancer screening is currently performed in a standard manner for all women with mammography. However, this level of standardization is likely to be
replaced by a screening program where the selection of an
imaging modality or modalities would depend on an individuals risk and other classifications. Already elements of
this change can be observed with the recommendation by
American Cancer Society to adjunctively use breast MRI for

[email protected]. Telephone: 508 856-1238; Fax: 508
856-1860.
1
N. Oestreicher, C. D. Lehman, D. J. Seger, D. S. Buist, and E. White,
The incremental contribution of clinical breast examination to invasive
cancer detection in a mammography screening program, AJR, Am. J.
Roentgenol. 1842, 428432 2005.
2
L. W. Bassett and R. H. Gold, The evolution of mammography, AJR,
Am. J. Roentgenol. 1503, 493498 1988.
3
R. Leborgne, Diagnosis of tumors of the breast by simple roentgenography; calcifications in carcinomas, Am. J. Roentgenol., Radium Ther.
Nucl. Med. 651, 111 1951.
4
R. L. Egan, Experience with mammography in a tumor institution.
Evaluation of 1,000 studies, Radiology 75, 894900 1960.
5
S. Shapiro, P. Strax, and L. Venet, Periodic breast cancer screening in
reducing mortality from breast cancer, JAMA, J. Am. Med. Assoc.
21511, 17771785 1971.
6
E. L. Thurfjell and J. A. Lindgren, Breast cancer survival rates with
mammographic screening: similar favorable survival rates for women
younger and those older than 50 years, Radiology 2012, 421426
1996.
7
R. E. Hendrick, R. A. Smith, J. H. Rutledge III, and C. R. Smart, Benefit
of screening mammography in women aged 40-49: a new meta-analysis
of randomized controlled trials, J Natl Cancer Inst Monogr 22, 8792
1997.
8
L. Tabar, B. Vitak, H. H. Chen, M. F. Yen, S. W. Duffy, and R. A. Smith,
Beyond randomized controlled trials: organized mammographic screening substantially reduces breast carcinoma mortality, Cancer 919,
17241731 2001.
9
The Swedish Organized Service Screening Evaluation Group, Reduction
in breast cancer mortality from organized service screening with mammography: 1. Further confirmation with extended data, Cancer Epidemiol. Biomarkers Prev. 151, 4551 2006.
10
The Swedish Organized Service Screening Evaluation Group, Reduction
in breast cancer mortality from the organised service screening with
mammography: 2. Validation with alternative analytic methods, Cancer
Epidemiol. Biomarkers Prev. 151, 5256 2006.
11
J. L. Price and P. D. Bler, The reduction of radiation and exposure time
4890
in mammography, Br. J. Radiol. 43508, 251255 1970.

E. P. Muntz and W. W. Logan, Focal spot size and scatter suppression in
magnification mammography, AJR, Am. J. Roentgenol. 1333, 453459
1979.
13
R. M. Nishikawa, G. E. Mawdsley, A. Fenster, and M. J. Yaffe,
Scanned-projection digital mammography, Med. Phys. 145, 717727
1987.
14
A. Karellas, S. Vedantham, and S. Suryanarayanan, Digital Mammography Image Acquisition Technology, in RSNA Categorical Course in Diagnostic Radiology Physics: Advances in Breast ImagingPhysics, Technology and Clinical Applications, edited by A. Karellas and M. L. Giger
RSNA, Oak Brook, IL, 2004, p. 8799.
15
FDA US, MQSA Facility Score Card, January 2, 2008, Center for Devices
and
Radiological
Health,
http://www.fda.gov/CDRH/
MAMMOGRAPHY/scorecard-statistics.html Accessed: January 28,
2008.
16
K. A. Fetterly and B. A. Schueler, Performance evaluation of a dualside read dedicated mammography computed radiography system, Med.
Phys. 307, 18431854 2003.
17
R. S. Saunders, Jr., E. Samei, J. L. Jesneck, and J. Y. Lo, Physical
characterization of a prototype selenium-based full field digital mammography detector, Med. Phys. 322, 588599 2005.
18
S. Vedantham, A. Karellas, S. Suryanarayanan, D. Albagli, S. Han, E. J.
Tkaczyk, C. E. Landberg, B. Opsahl-Ong, P. R. Granfors, I. Levis, C. J.
DOrsi, and R. E. Hendrick, Full breast digital mammography with an
amorphous silicon-based flat panel detector: Physical characteristics of a
clinical prototype, Med. Phys. 273, 558567 2000.
19
J. M. Lewin, C. J. DOrsi, R. E. Hendrick, L. J. Moss, P. K. Isaacs, A.
Karellas, and G. R. Cutter, Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer,
AJR, Am. J. Roentgenol. 1793, 671677 2002.
20
P. Skaane, S. Hofvind, and A. Skjennald, Randomized trial of screenfilm versus full-field digital mammography with soft-copy reading in
population-based screening program: Follow-up and final results of Oslo,
II study, Radiology 2443, 708717 2007.
21
E. D. Pisano, C. Gatsonis, E. Hendrick, M. Yaffe, J. K. Baum, S.
Acharyya, E. F. Conant, L. L. Fajardo, L. Bassett, C. DOrsi, R. Jong, and
M. Rebner, Diagnostic performance of digital versus film mammography
for breast-cancer screening, N. Engl. J. Med. 35317, 17731783
2005.
22
E. D. Pisano, R. E. Hendrick, M. J. Yaffe, J. K. Baum, S. Acharyya, J. B.
Cormack, L. A. Hanna, E. F. Conant, L. L. Fajardo, L. W. Bassett, C. J.
DOrsi, R. A. Jong, M. Rebner, A. N. Tosteson, and C. A. Gatsonis,
Diagnostic accuracy of digital versus film mammography: Exploratory
analysis of selected population subgroups in DMIST, Radiology 2462,
376383 2008.
23
L. E. Antonuk, J. Boudry, W. Huang, D. L. McShan, E. J. Morton, J.
Yorkston, M. J. Longo, and R. A. Street, Demonstration of megavoltage
and diagnostic x-ray imaging with hydrogenated amorphous silicon arrays, Med. Phys. 196, 14551466 1992.
24
S. Suryanarayanan, A. Karellas, and S. Vedantham, Physical characteristics of a full-field digital mammography system, Nucl. Instrum. Methods Phys. Res. A 5333, 560570 2004.
25
J. A. Rowlands, D. M. Hunter, and N. Araj, X-ray imaging using amorphous selenium: a photoinduced discharge readout method for digital
mammography, Med. Phys. 183, 421431 1991.
26
W. Zhao, J. Law, D. Waechter, Z. Huang, and J. A. Rowlands, Digital
radiology using active matrix readout of amorphous selenium: detectors
with high voltage protection, Med. Phys. 254, 539549 1998.
27
W. Zhao, W. G. Ji, A. Debrie, and J. A. Rowlands, Imaging performance
of amorphous selenium based flat-panel detectors for digital mammography: Characterization of a small area prototype detector, Med. Phys.
302, 254263 2003.
28
G. T. Barnes and I. A. Brezovich, The design and performance of a
scanning multiple slit assembly, Med. Phys. 63, 197204 1979.
29
J. M. Boone, J. A. Seibert, C. M. Tang, and S. M. Lane, Grid and slot
scan scatter reduction in mammography: comparison by using Monte
Carlo techniques, Radiology 2222, 519527 2002.
30
M. Aslund, B. Cederstrom, M. Lundqvist, and M. Danielsson, Physical
characterization of a scanning photon counting digital mammography system based on Si-strip detectors, Med. Phys. 346, 19181925 2007.
31
S. Arakawa, W. Itoh, K. Kohda, and T. Suzuki, Improvement of image
quality in CR mammography by detection of emissions from dual sides of
12
4890
an imaging plate, Proc. SPIE 3977, 590600 2000.

J. A. Seibert, J. M. Boone, V. N. Cooper, and K. K. Lindfors, Cassettebased digital mammography, Technol. Cancer Res. Treat. 35, 413427
2004.
33
C. Lawinski, A. MacKenzie, H. Cole, P. Blake, I. Honey, and A. Pascoal,
Computed Radiography (CR) Systems for Mammography. A comparative
technical report. MHRA 04107 Medicines and Health Products Regulatory Agency MHRA, London, 2004.
34
E. D. Pisano and M. J. Yaffe, Digital mammography, Radiology 2342,
353362 2005.
35
M. J. Yaffe and J. G. Mainprize, Detectors for digital mammography,
Technol. Cancer Res. Treat. 34, 309324 2004.
36
S. Suryanarayanan, A. Karellas, S. Vedantham, I. Sechopoulos, and C. J.
DOrsi, Detection of simulated microcalcifications in a phantom with
digital mammography: Effect of pixel size, Radiology 2441, 130137
2007.
37
R. S. Saunders, Jr., J. A. Baker, D. M. Delong, J. P. Johnson, and E.
Samei, Does image quality matter Impact of resolution and noise on
mammographic task performance, Med. Phys. 3410, 39713981
2007.
38
Y. Ei-Mohri, L. E. Antonuk, Q. Zhao, Y. Wang, Y. Li, H. Du, and A.
Sawant, Performance of a high fill factor, indirect detection prototype
flat-panel imager for mammography, Med. Phys. 341, 315327 2007.
39
Y. El-Mohri, L. E. Antonuk, Q. Zhao, M. Maolinbay, X. Rong, K. W. Jee,
S. Nassif, and C. Cionca, A quantitative investigation of additive noise
reduction for active matrix flat-panel imagers using compensation lines,
Med. Phys. 278, 18551864 2000.
40
V. V. Nagarkar, S. V. Tipnis, V. B. Gaysinskiy, S. R. Miller, A. Karellas,
and S. Vedantham, New design of a structured CsITl screen for digital
mammography, Proc. SPIE 5030, 541546 2003.
41
A. Sawant, L. E. Antonuk, Y. El-Mohri, Q. Zhao, Y. Li, Z. Su, Y. Wang,
J. Yamamoto, H. Du, I. Cunningham, M. Klugerman, and K. Shah, Segmented crystalline scintillators: an initial investigation of high quantum
efficiency detectors for megavoltage x-ray imaging, Med. Phys. 3210,
30673083 2005.
42
A. W. Rau, L. Bakueva, and J. A. Rowlands, The x-ray time of flight
method for investigation of ghosting in amorphous selenium-based flat
panel medical x-ray imagers, Med. Phys. 3210, 31603177 2005.
43
W. Zhao, G. DeCrescenzo, S. O. Kasap, and J. A. Rowlands, Ghosting
caused by bulk charge trapping in direct conversion flat-panel detectors
using amorphous selenium, Med. Phys. 322, 488500 2005.
44
A. K. Bloomquist, M. J. Yaffe, G. E. Mawdsley, D. M. Hunter, and D. J.
Beideck, Lag and ghosting in a clinical flat-panel selenium digital mammography system, Med. Phys. 338, 29983005 2006.
45
D. C. Hunt, K. Tanioka, and J. A. Rowlands, X-ray imaging using avalanche multiplication in amorphous selenium: investigation of intrinsic
avalanche noise, Med. Phys. 3412, 46544663 2007.
46
D. L. Lee, Selenium detector with a grid for selenium charge gain,
Proc. SPIE 5745, 216222 2005.
47
J. G. Mainprize, N. L. Ford, S. Yin, E. E. Gordon, W. J. Hamilton, T. O.
Tumer, and M. J. Yaffe, A CdZnTe slot-scanned detector for digital
48
G. Zentai, L. D. Partain, R. Pavlyuchkova, C. H. Proano, M. M. Schieber,
and J. Thomas, Dark current, sensitivity, and image lag comparison of
mercuric iodide and lead iodide x-ray imagers, Proc. SPIE 5541, 171
178 2004.
49
Z. Su, L. E. Antonuk, Y. El-Mohri, L. Hu, H. Du, A. Sawant, Y. Li, Y.
Wang, J. Yamamoto, and Q. Zhao, Systematic investigation of the signal
properties of polycrystalline HgI2 detectors under mammographic, radiographic, fluoroscopic and radiotherapy irradiation conditions, Phys.
Med. Biol. 5012, 29072928 2005.
50
T. Sakellaris, G. Spyrou, G. Tzanakos, and G. Panayiotakis, Energy,
angular and spatial distributions of primary electrons inside photoconducting materials for digital mammography: Monte Carlo simulation
studies, Phys. Med. Biol. 5221, 64396460 2007.
51
P. J. R. Leblans, L. Struye, and P. Willems, New needle-crystalline CR
detector, Proc. SPIE 4320, 5967 2001.
52
R. Schaetzing, R. Fasbender, and P. Kersten, New high-speed scanning
technique for computed radiography, Proc. SPIE 4682, 511520 2002.
53
S. Rivetti, N. Lanconelli, M. Bertolini, G. Borasi, D. Acchiappati, and A.
Burani, Performance evaluation of a direct computed radiography system by means of physical characterization and contrast detail analysis,
Proc. SPIE 6510, 65104M 2007.
32
4891
54
C. Herrmann, J. Frankenberger, G. Reiser, and J. Lamotte, Optimization

of a CR system comprising line-scanning and needle image plate technology with respect to examinations of extremities, Proc. SPIE 6510,
65101B 2007.
55
P. M. Shikhaliev, T. Xu, H. Le, and S. Molloi, Scanning-slit photon
counting x-ray imaging system using a microchannel plate detector,
Med. Phys. 315, 10611071 2004.
56
J. Giersch, Medical quantum X-ray imaging with 2D detectors, Nucl.
Instrum. Methods Phys. Res. A 551, 125138 2005.
57
M. Bech, O. Bunk, C. David, P. Kraft, C. Bronnimann, E. F. Eikenberry,
and F. Pfeiffer, X-ray imaging with the PILATUS 100k detector, Appl.
Radiat. Isot. 664, 474478 2008.
58
J. Karg, D. Niederlohner, J. Giersch, and G. Anton, Using the Medipix2
detector for energy weighting, Nucl. Instrum. Methods Phys. Res. A 546,
306311 2005.
59
P. Bernhardt, T. Mertelmeier, and M. Hoheisel, X-ray spectrum optimization of full-field digital mammography: simulation and phantom study,
Med. Phys. 3311, 43374349 2006.
60
P. Toroi, F. Zanca, K. C. Young, C. van Ongeval, G. Marchal, and H.
Bosmans, Experimental investigation on the choice of the tungsten/
rhodium anode/filter combination for an amorphous selenium-based digital mammography system, Eur. Radiol. 179, 23682375 2007.
61
D. R. Dance, A. K. Thilander, M. Sandborg, C. L. Skinner, I. A. Castellano, and G. A. Carlsson, Influence of anode/filter material and tube
potential on contrast, signal-to-noise ratio and average absorbed dose in
mammography: A Monte Carlo study, Br. J. Radiol. 73874, 10561067
2000.
62
B. Heddson, K. Ronnow, M. Olsson, and D. Miller, Digital versus
screen-film mammography: A retrospective comparison in a populationbased screening program, Eur. J. Radiol. 643, 419425 2007.
63
G. Gennaro and C. di Maggio, Dose comparison between screen/film
and full-field digital mammography, Eur. Radiol. 1611, 25592566
2006.
64
K. P. Hermann, S. Obenauer, K. Marten, S. Kehbel, U. Fischer, and E.
Grabbe, Average glandular dose with amorphous silicon full-field digital
mammography-Clinical results, Rofo Fortschr Geb Rontgenstr Neuen
Bildgeb Verfahr 1746, 696699 2002.
65
S. Weigel, R. Girnus, J. Czwoydzinski, T. Decker, S. Spital, and W. Heindel, Digital mammography screening: average glandular dose and first
performance parameters, Rofo Fortschr Geb Rontgenstr Neuen Bildgeb
Verfahr 1799, 892895 2007.
66
G. Gennaro, L. Katz, H. Souchay, C. Alberelli, and C. di Maggio, Are
phantoms useful for predicting the potential of dose reduction in full-field
digital mammography?, Phys. Med. Biol. 508, 18511870 2005.
67
E. Samei, R. S. Saunders, Jr., J. A. Baker, and D. M. Delong, Digital
mammography;: effects of reduced radiation dose on diagnostic performance, Radiology 2432, 396404 2007.
68
M. Ruschin, P. Timberg, M. Bath, B. Hemdal, T. Svahn, R. S. Saunders,
E. Samei, I. Andersson, S. Mattsson, D. P. Chakrabort, and A. Tingber,
Dose dependence of mass and microcalcification detection in digital
mammography: Free response human observer studies, Med. Phys.
342, 400407 2007.
69
J. M. Boone, Normalized glandular dose DgN coefficients for arbitrary
X-ray spectra in mammography: computer-fit values of Monte Carlo derived data, Med. Phys. 295, 869875 2002.
70
D. R. Dance, Monte Carlo calculation of conversion factors for the estimation of mean glandular breast dose, Phys. Med. Biol. 359, 1211
1219 1990.
71
G. R. Hammerstein, D. W. Miller, D. R. White, M. E. Masterson, H. Q.
Woodard, and J. S. Laughlin, Absorbed radiation dose in mammography, Radiology 1302, 485491 1979.
72
X. Wu, E. L. Gingold, G. T. Barnes, and D. M. Tucker, Normalized
average glandular dose in molybdenum target-rhodium filter and rhodium
target-rhodium filter mammography, Radiology 1931, 8389 1994.
73
K. A. Hatziioannou, K. Psarrakos, E. Molyvda-Athanasopoulou, G. Kitis,
E. Papanastassiou, I. Sofroniadis, and O. Kimoundri, Dosimetric considerations in mammography, Eur. Radiol. 107, 11931196 2000.
74
R. H. Behrman, M. J. Homer, W. T. Yang, and G. J. Whitman, Mammography and fetal dose, Radiology 2432, 605 2007; author reply
2432, 605606 2007.
75
I. Sechopoulos, S. Suryanarayanan, S. Vedantham, C. J. DOrsi, and A.
Karellas, Radiation Dose to Organs and Tissues from Mammography:
Monte Carlo and Phantom Study, Radiology 2462, 434443 2008.
76
4891
L. W. Bassett, D. M. Farria, S. Bansal, M. A. Farquhar, P. A. Wilcox, and

S. A. Feig, Reasons for failure of a mammography unit at clinical image
review in the American College of Radiology Mammography Accreditation Program, Radiology 2153, 698702 2000.
77
P. S. Rezentes, A. de Almeida, and G. T. Barnes, Mammography grid
performance, Radiology 2101, 227232 1999.
78
D. P. Chakraborty, The effect of the antiscatter grid on full-field digital
mammography phantom images, J. Digit Imaging 121, 1222 1999.
79
D. P. Chakraborty, Effect of the antiscatter grid and target/filters in fullfield digital mammography, Proc. SPIE 3659, 878885 1999.
80
W. J. Veldkamp, M. A. Thijssen, and N. Karssemeijer, The value of
scatter removal by a grid in full field digital mammography, Med. Phys.
307, 17121718 2003.
81
G. Gennaro, L. Katz, H. Souchay, R. Klausz, C. Alberelli, and C. di
Maggio, Grid removal and impact on population dose in full-field digital
82
A. H. Baydush, and C. E. Floyd, Jr., Improved image quality in digital
mammography with image processing, Med. Phys. 277, 15031508
2000.
83
K. Nykanen and S. Siltanen, X-ray scattering in full-field digital mammography, Med. Phys. 307, 18641873 2003.
84
D. E. G. Trotter, J. E. Tkaczyk, J. Kaufhold, B. E. H. Claus, and J. W.
Eberhard, Thickness-dependent scatter correction algorithm for digital
mammography, Proc. SPIE 4682, 469478 2002.
85
C. M. Sehgal, S. P. Weinstein, P. H. Arger, and E. F. Conant, A review of
breast ultrasound, Journal of mammary gland biology and neoplasia
112, 113123 2006.
86
S. P. Weinstein, E. F. Conant, and C. Sehgal, Technical advances in
breast ultrasound imaging, Semin Ultrasound CT MR 274, 273283
2006.
87
F. Forsberg, C. W. Piccoli, D. A. Merton, J. J. Palazzo, and A. L. Hall,
Breast lesions: imaging with contrast-enhanced subharmonic USinitial
experience, Radiology 2443, 718726 2007.
88
T. M. Kolb, J. Lichy, and J. H. Newhouse, Comparison of the performance of screening mammography, physical examination, and breast US
and evaluation of factors that influence them: An analysis of 27,825 patient evaluations, Radiology 2251, 165175 2002.
89
W. A. Berg, J. D. Blume, J. B. Cormack, E. B. Mendelson, D. Lehrer, M.
Bohm-Velez, E. D. Pisano, R. A. Jong, W. P. Evans, M. J. Morton, M. C.
Mahoney, L. H. Larsen, R. G. Barr, D. M. Farria, H. S. Marques, and K.
Boparai, Combined screening with ultrasound and mammography vs
mammography alone in women at elevated risk of breast cancer, JAMA,
J. Am. Med. Assoc. 29918, 21512163 2008.
90
A. Kapur, P. L. Carson, J. Eberhard, M. M. Goodsitt, K. Thomenius, M.
Lokhandwalla, D. Buckley, M. A. Roubidoux, M. A. Helvie, R. C. Booi,
G. L. LeCarpentier, R. Q. Erkamp, H. P. Chan, J. B. Fowlkes, J. A.
Thomas, and C. E. Landberg, Combination of digital mammography
with semi-automated 3D breast ultrasound, Technol. Cancer Res. Treat.
34, 325334 2004.
91
J. W. Eberhard, P. Staudinger, J. Smolenski, J. Ding, A. Schmitz, J. McCoy, M. Rumsey, A. Al-Khalidy, W. Ross, C. E. Landberg, B. E. Claus,
P. Carson, M. M. Goodsitt, H. P. Chan, M. Roubidoux, J. A. Thomas, and
J. Osland, High-speed large angle mammography tomosynthesis system, Proc. SPIE 6142, 61420C 2006.
92
R. C. Booi, J. F. Krucker, M. M. Goodsitt, M. ODonnell, A. Kapur, G. L.
LeCarpentier, M. A. Roubidoux, J. B. Fowlkes, and P. L. Carson, Evaluating thin compression paddles for mammographically compatible ultrasound, Ultrasound Med. Biol. 333, 472482 2007.
93
S. P. Sinha, M. M. Goodsitt, M. A. Roubidoux, R. C. Booi, G. L. LeCarpentier, C. R. Thomenius, K. E. Thomenius, C. L. Chalek, and P. L.
Carson, Automated ultrasound scanning on a dual-modality breast imaging system: coverage and motion issues and solutions, J. Ultrasound
Med. 265, 645655 2007.
94
N. Duric, P. Littrup, A. Babkin, D. Chambers, S. Azevedo, R. Pevzner,
M. Tokarev, E. Holsapple, O. Rama, and R. Duncan, Development of
ultrasound tomography for breast imaging: technical assessment, Med.
Phys. 325, 13751386 2005.
95
N. Duric, P. Littrup, L. Poulo, A. Babkin, R. Pevzner, E. Holsapple, O.
Rama, and C. Glide, Detection of breast cancer with ultrasound tomography: First result with the Computed Ultrasound Risk Evaluation
CURE prototype, Med. Phys. 342, 773785 2007.
96
R. Damadian, Tumor detection by nuclear magnetic resonance, Science
171976, 11511153 1971.
4892
97
P. C. Lauterbur, Image formation by induced local interactions: Examples employing nuclear magnetic resonance, Nature London 242,
190191 1973.
98
P. Mansfield, P. G. Morris, R. Ordidge, R. E. Coupland, H. M. Bishop,
and R. W. Blamey, Carcinoma of the breast imaged by nuclear magnetic
resonance NMR, Br. J. Radiol. 52615, 242243 1979.
99
R. J. Ross, J. S. Thompson, K. Kim, and R. A. Bailey, Nuclear magnetic
resonance imaging and evaluation of human breast tissue: preliminary
clinical trials, Radiology 1431, 195205 1982.
100
S. J. El Yousef, R. H. Duchesneau, R. J. Alfidi, J. R. Haaga, P. J. Bryan,
and J. P. LiPuma, Magnetic resonance imaging of the breast. Work in
progress, Radiology 1503, 761766 1984.
101
S. J. El Yousef, R. J. Alfidi, R. H. Duchesneau, C. A. Hubay, J. R. Haaga,
P. J. Bryan, J. P. LiPuma, and A. E. Ament, Initial experience with
nuclear magnetic resonance NMR imaging of the human breast, J.
Comput. Assist. Tomogr. 72, 215218 1983.
102
C. B. Stelling, P. C. Wang, A. Lieber, S. S. Mattingly, W. O. Griffen, and
D. E. Powell, Prototype coil for magnetic resonance imaging of the
female breast. Work in progress, Radiology 1542, 457462 1985.
103
R. C. Brasch, H. J. Weinmann, and G. E. Wesbey, Contrast-enhanced
NMR imaging: animal studies using gadolinium-DTPA complex, AJR,
Am. J. Roentgenol. 1423, 625630 1984.
104
D. H. Carr, J. Brown, G. M. Bydder, R. E. Steiner, H. J. Weinmann, U.
Speck, A. S. Hall, and I. R. Young, Gadolinium-DTPA as a contrast
agent in MRI: Initial clinical experience in 20 patients, AJR, Am. J.
Roentgenol. 1432, 215224 1984.
105
D. H. Carr, J. Brown, G. M. Bydder, H. J. Weinmann, U. Speck, D. J.
Thomas, and I. R. Young, Intravenous chelated gadolinium as a contrast
agent in NMR imaging of cerebral tumours, Lancet 18375, 484486
1984.
106
M. Laniado, H. J. Weinmann, W. Schorner, R. Felix, and U. Speck, First
use of GdDTPA/dimeglumine in man, Physiol. Chem. Phys. Med. NMR
162, 157165 1984.
107
H. J. Weinmann, R. C. Brasch, W. R. Press, and G. E. Wesbey, Characteristics of gadolinium-DTPA complex: a potential NMR contrast agent,
108
H. J. Weinmann, M. Laniado, and W. Mutzel, Pharmacokinetics of
GdDTPA/dimeglumine after intravenous injection into healthy volunteers, Physiol. Chem. Phys. Med. NMR 162, 167172 1984.
109
S. H. Heywang, D. Hahn, H. Schmidt, I. Krischke, W. Eiermann, R.
Bassermann, and J. Lissner, MR imaging of the breast using gadoliniumDTPA, J. Comput. Assist. Tomogr. 102, 199204 1986.
110
W. A. Kaiser and E. Zeitler, MR imaging of the breast: fast imaging
sequences with and without Gd-DTPA. Preliminary observations, Radiology 1703 Pt 1, 681686 1989.
111
J. Frahm, A. Haase, and D. Matthaei, Rapid NMR imaging of dynamic
processes using the FLASH technique, Magn. Reson. Med. 32, 321
327 1986.
112
J. Frahm, A. Haase, and D. Matthaei, Rapid three-dimensional MR imaging using the FLASH technique, J. Comput. Assist. Tomogr. 102,
363368 1986.
113
A. Haase, D. Matthaei, W. Hanicke, and J. Frahm, Dynamic digital subtraction imaging using fast low-angle shot MR movie sequence, Radiology 1602, 537541 1986.
114
J. Folkman and M. Klagsbrun, Angiogenic factors, Science 2354787,
442447 1987.
115
F. L. Flanagan, J. G. Murray, P. Gilligan, J. P. Stack, and J. T. Ennis,
Digital subtraction in Gd-DTPA enhanced imaging of the breast, Clin.
Radiol. 5012, 848854 1995.
116
C. D. Lehman, C. Gatsonis, C. K. Kuhl, R. E. Hendrick, E. D. Pisano, L.
Hanna, S. Peacock, S. F. Smazal, D. D. Maki, T. B. Julian, E. R. DePeri,
D. A. Bluemke, and M. D. Schnall, MRI evaluation of the contralateral
breast in women with recently diagnosed breast cancer, N. Engl. J. Med.
35613, 12951303 2007.
117
L. Liberman, E. A. Morris, C. M. Kim, J. B. Kaplan, A. F. Abramson, J.
H. Menell, K. J. Van Zee, and D. D. Dershaw, MR imaging findings in
the contralateral breast of women with recently diagnosed breast cancer,
118
S. G. Lee, S. G. Orel, I. J. Woo, E. Cruz-Jove, M. E. Putt, L. J. Solin, B.
J. Czerniecki, and M. D. Schnall, MR imaging screening of the contralateral breast in patients with newly diagnosed breast cancer: Preliminary results, Radiology 2263, 773778 2003.
119
C. D. Lehman, J. D. Blume, D. Thickman, D. A. Bluemke, E. Pisano, C.
4892
Kuhl, T. B. Julian, N. Hylton, P. Weatherall, M. OLoughlin, S. J. Schnitt,

C. Gatsonics, and M. D. Schnall, Added cancer yield of MRI in screening the contralateral breast of women recently diagnosed with breast cancer: Results from the International Breast Magnetic Resonance Consortium IBMC trial, J. Surg. Oncol. 921, 915 2005; discussion 921,
1516 2005.
120
E. Warner, D. B. Plewes, R. S. Shumak, G. C. Catzavelos, L. S. Di
Prospero, M. J. Yaffe, V. Goel, E. Ramsay, P. L. Chart, D. E. Cole, G. A.
Taylor, M. Cutrara, T. H. Samuels, J. P. Murphy, J. M. Murphy, and S. A.
Narod, Comparison of breast magnetic resonance imaging, mammography, and ultrasound for surveillance of women at high risk for hereditary
breast cancer, J. Clin. Oncol. 1915, 35243531 2001.
121
M. J. Stoutjesdijk, C. Boetes, G. J. Jager, L. Beex, P. Bult, J. H. Hendriks,
R. J. Laheij, L. Massuger, L. E. van Die, T. Wobbes, and J. O. Barentsz,
Magnetic resonance imaging and mammography in women with a hereditary risk of breast cancer, J. Natl. Cancer Inst. 9314, 10951102
2001.
122
C. K. Kuhl, S. Schrading, C. C. Leutner, N. Morakkabati-Spitz, E. Wardelmann, R. Fimmers, W. Kuhn, and H. H. Schild, Mammography, breast
ultrasound, and magnetic resonance imaging for surveillance of women at
high familial risk for breast cancer, J. Clin. Oncol. 2333, 84698476
2005.
123
C. K. Kuhl, R. K. Schmutzler, C. C. Leutner, A. Kempe, E. Wardelmann,
A. Hocke, M. Maringa, U. Pfeifer, D. Krebs, and H. H. Schild, Breast
MR imaging screening in 192 women proved or suspected to be carriers
of a breast cancer susceptibility gene: Preliminary results, Radiology
2151, 267279 2000.
124
C. D. Lehman, J. D. Blume, P. Weatherall, D. Thickman, N. Hylton, E.
Warner, E. Pisano, S. J. Schnitt, C. Gatsonis, M. Schnall, G. A. DeAngelis, P. Stomper, E. L. Rosen, M. OLoughlin, S. Harms, and D. A.
Bluemke, Screening women at high risk for breast cancer with mammography and magnetic resonance imaging, Cancer 1039, 18981905
2005.
125
L. Liberman, E. A. Morris, C. L. Benton, A. F. Abramson, and D. D.
Dershaw, Probably benign lesions at breast magnetic resonance imaging:
Preliminary experience in high-risk women, Cancer 982, 377388
2003.
126
L. Liberman, E. A. Morris, D. D. Dershaw, A. F. Abramson, and L. K.
Tan, Ductal enhancement on MR imaging of the breast, AJR, Am. J.
Roentgenol. 1812, 519525 2003.
127
M. M. Tilanus-Linthorst, I. M. Obdeijn, K. C. Bartels, H. J. de Koning,
and M. Oudkerk, First experiences in screening women at high risk for
breast cancer with MR imaging, Breast Cancer Res. Treat. 631, 5360
2000.
128
E. A. Morris, L. Liberman, D. J. Ballon, M. Robson, A. F. Abramson, A.
Heerdt, and D. D. Dershaw, MRI of occult breast carcinoma in a highrisk population, AJR, Am. J. Roentgenol. 1813, 619626 2003.
129
M. Kriege, C. T. Brekelmans, C. Boetes, P. E. Besnard, H. M. Zonderland, I. M. Obdeijn, R. A. Manoliu, T. Kok, H. Peterse, M. M. TilanusLinthorst, S. H. Muller, S. Meijer, J. C. Oosterwijk, L. V. Beex, R. A.
Tollenaar, H. J. de Koning, E. J. Rutgers, and J. G. Klijn, Efficacy of
MRI and mammography for breast-cancer screening in women with a
familiar or genetic predisposition, N. Engl. J. Med. 3515, 427437
2004.
130
M. O. Leach, C. R. Boggis, A. K. Dixon, D. F. Easton, R. A. Eeles, D. G.
Evans, F. J. Gilbert, I. Griebsch, R. J. Hoff, P. Kessar, S. R. Lakhani, S.
M. Moss, A. Nerurkar, A. R. Padhani, L. J. Pointon, D. Thompson, and R.
M. Waren, Screening with magnetic resonance imaging and mammography of a UK population at high familial risk of breast cancer: A prospective multicentre cohort study MARIBS, Lancet 3659473, 1769
1778 2005.
131
F. Sardanelli, F. Podo, G. DAgnolo, A. Verdecchia, M. Santaquilani, R.
Musumeci, G. Trecate, S. Manoukian, S. Morassut, C. de Giacomi, M.
Federico, L. Cortesi, S. Corcione, S. Cirillo, V. Marra, A. Cilotti, C. Di
Maggio, A. Fausto, L. Preda, C. Zuiani, A. Contegiacomo, A. Orlacchio,
M. Calabrese, L. Bonomo, E. Di Cesare, M. Tonutti, P. Panizza, and A.
Del Maschio, Multicenter comparative multimodality surveillance of
women at genetic-familial high risk for breast cancer HIBCRIT study:
Interim results, Radiology 2423, 698715 2007.
132
D. Saslow, C. Boetes, W. Burke, S. Harms, M. O. Leach, C. D. Lehman,
E. Morris, E. Pisano, M. Schnall, S. Sener, R. A. Smith, E. Warner, M.
Yaffe, K. S. Andrews, and C. A. Russell, American Cancer Society
guidelines for breast screening with MRI as an adjunct to mammogra-
4893
phy, Ca-Cancer J. Clin. 572, 7589 2007.

N. H. Peters, I. H. Borel Rinkes, N. P. Zuithoff, W. P. Mali, K. G. Moons,
and P. H. Peeters, Meta-analysis of MR imaging in the diagnosis of
breast lesions, Radiology 2461, 116124 2008.
134
ACR, ACR Breast Imaging Reporting and Data Systems (BI-RADS):
Breast Imaging Atlas American College of Radiology, Reston, VA,
2003.
135
C. Kuhl, The current status of breast MR imaging. Part I. Choice of
technique, image interpretation, diagnostic accuracy, and transfer to clinical practice, Radiology 2442, 356378 2007.
136
C. K. Kuhl, Current status of breast MR imaging. Part 2. Clinical applications, Radiology 2443, 672691 2007.
137
M. F. Bellin, J. A. Jakobsen, I. Tomassin, H. S. Thomsen, S. K. Morcos,
H. S. Thomsen, S. K. Morcos, T. Almen, P. Aspelin, M. F. Bellin, W.
Clauss, H. Flaten, N. Grenier, J. M. Idee, J. A. Jakobsen, G. P. Krestin, F.
Stacul, and J. A. Webb, Contrast medium extravasation injury: Guidelines for prevention and management, Eur. Radiol. 1211, 28072812
2002.
138
FDA, US Public Health Advisory: Update on Magnetic Resonance Imaging MRI Contrast Agents Containing Gadolinium and Nephrogenic Fibrosing Dermopathy, Date created: December 22, 2006; updated May 23,
2007.
http://www.fda.gov/Cder/Drug/advisory/
gadolinium_agents_20061222.htm Accessed: April 29, 2008.
139
I. S. Gribbestad, B. Sitter, S. Lundgren, J. Krane, and D. Axelson, Metabolite composition in breast tumors examined by proton nuclear magnetic resonance spectroscopy, Anticancer Res. 193A, 17371746
1999.
140
A. Barzilai, A. Horowitz, A. Geier, and H. Degani, Phosphate metabolites and steroid hormone receptors of benign and malignant breast tumors. A nuclear magnetic resonance study, Cancer 6711, 29192925
1991.
141
J. S. Cohen, Phospholipid and energy metabolism of cancer cells monitored by 31P magnetic resonance spectroscopy: possible clinical significance, Mayo Clin. Proc. 6312, 11991207 1988.
142
R. Kalra, K. E. Wade, L. Hands, P. Styles, R. Camplejohn, M. Greenall,
G. E. Adams, A. L. Harris, and G. K. Radda, Phosphomonoester is
associated with proliferation in human breast cancer: a 31P MRS study,
Br. J. Cancer 675, 11451153 1993.
143
J. M. Park and J. H. Park, Human in-vivo 31P MR spectroscopy of
benign and malignant breast tumors, Korean J Radiol 22, 8086
2001.
144
K. A. Kvistad, I. J. Bakken, I. S. Gribbestad, B. Ehrnholm, S. Lundgren,
H. E. Fjosne, and O. Haraldseth, Characterization of neoplastic and normal human breast tissues with in vivo 1H MR spectroscopy, J. Magn.
Reson Imaging 102, 159164 1999.
145
J. R. Roebuck, K. M. Cecil, M. D. Schnall, and R. E. Lenkinski, Human
breast lesions: characterization with proton MR spectroscopy, Radiology
2091, 269275 1998.
146
I. S. Gribbestad, T. E. Singstad, G. Nilsen, H. E. Fjosne, T. Engan, O. A.
Haugen, and P. A. Rinck, In vivo 1H MRS of normal breast and breast
tumors using a dedicated double breast coil, J. Magn. Reson Imaging
86, 11911197 1998.
147
D. K. Yeung, H. S. Cheung, and G. M. Tse, Human breast lesions:
characterization with contrast-enhanced in vivo proton MR
spectroscopyinitial results, Radiology 2201, 4046 2001.
148
R. Katz-Brull, P. T. Lavin, and R. E. Lenkinski, Clinical utility of proton
magnetic resonance spectroscopy in characterizing breast lesions, J.
Natl. Cancer Inst. 9416, 11971203 2002.
149
H. M. Baek, H. J. Yu, J. H. Chen, O. Nalcioglu, and M. Y. Su, Quantitative correlation between 1H MRS and dynamic contrast-enhanced
MRI of human breast cancer, Magn. Reson. Imaging 264, 523531
2008.
150
M. Y. Su, H. M. Baik, H. J. Yu, J. H. Chen, R. S. Mehta, and O. Nalcioglu, Comparison of choline and pharmacokinetic parameters in breast
cancer measured by MR spectroscopic imaging and dynamic contrast
enhanced MRI, Technol. Cancer Res. Treat. 54, 401410 2006.
151
S. Meisamy, P. J. Bolan, E. H. Baker, M. G. Pollema, C. T. Le, F. Kelcz,
M. C. Lechner, B. A. Luikens, R. A. Carlson, K. R. Brandt, K. K. Amrami, M. T. Nelson, L. I. Everson, T. H. Emory, T. M. Tuttle, D. Yee, and
M. Garwood, Adding in vivo quantitative 1H MR spectroscopy to improve diagnostic accuracy of breast MR imaging: Preliminary results of
observer performance study at 4.0 T, Radiology 2362, 465475
2005.
133
152
4893
W. Huang, P. R. Fisher, K. Dulaimy, L. A. Tudorica, B. OHea, and T. M.

Button, Detection of breast malignancy: Diagnostic MR protocol for
improved specificity, Radiology 2322, 585591 2004.
153
L. Bartella, S. B. Thakur, E. A. Morris, D. D. Dershaw, W. Huang, E.
Chough, M. C. Cruz, and L. Liberman, Enhancing nonmass lesions in
the breast: Evaluation with proton 1H MR spectroscopy, Radiology
2451, 8087 2007.
154
E. O. Aboagye and Z. M. Bhujwalla, Malignant transformation alters
membrane choline phospholipid metabolism of human mammary epithelial cells, Cancer Res. 591, 8084 1999.
155
M. Kumar, N. R. Jagannathan, V. Seenu, S. N. Dwivedi, P. K. Julka, and
G. K. Rath, Monitoring the therapeutic response of locally advanced
breast cancer patients: Sequential in vivo proton MR spectroscopy study,
J. Magn. Reson Imaging 242, 325332 2006.
156
N. R. Jagannathan, M. Kumar, V. Seenu, O. Coshic, S. N. Dwivedi, P. K.
Julka, A. Srivastava, and G. K. Rath, Evaluation of total choline from
in-vivo volume localized proton MR spectroscopy and its response to
neoadjuvant chemotherapy in locally advanced breast cancer, Br. J. Cancer 848, 10161022 2001.
157
H. M. Baek, J. H. Chen, O. Nalcioglu, and M. Y. Su, Proton MR spectroscopy for monitoring early treatment response of breast cancer to neoadjuvant chemotherapy, Ann. Oncol. 195, 10221024 2008.
158
D. J. Manton, A. Chaturvedi, A. Hubbard, M. J. Lind, M. Lowry, A.
Maraveyas, M. D. Pickles, D. J. Tozer, and L. W. Turnbull, Neoadjuvant
chemotherapy in breast cancer: Early response prediction with quantitative MR imaging and spectroscopy, Br. J. Cancer 943, 427435
2006.
159
S. Meisamy, P. J. Bolan, E. H. Baker, R. L. Bliss, E. Gulbahce, L. I.
Everson, M. T. Nelson, T. H. Emory, T. M. Tuttle, D. Yee, and M. Garwood, Neoadjuvant chemotherapy of locally advanced breast cancer:
predicting response with in vivo 1H MR spectroscopy--a pilot study at 4
T, Radiology 2332, 424431 2004.
160
M. A. Jacobs, P. B. Barker, P. A. Bottomley, Z. Bhujwalla, and D. A.
Bluemke, Proton magnetic resonance spectroscopic imaging of human
breast cancer: a preliminary study, J. Magn. Reson Imaging 191,
6875 2004.
161
M. A. Jacobs, R. Ouwerkerk, A. C. Wolff, V. Stearns, P. A. Bottomley, P.
B. Barker, P. Argani, N. Khouri, N. E. Davidson, Z. M. Bhujwalla, and D.
A. Bluemke, Multiparametric and multinuclear magnetic resonance imaging of human breast cancer: current applications, Technol. Cancer Res.
Treat. 36, 543550 2004.
162
J. Hu, Y. Yu, Z. Kou, W. Huang, Q. Jiang, Y. Xuan, T. Li, V. Sehgal, C.
Blake, E. M. Haacke, and R. L. Soulen, A high spatial resolution 1H
magnetic resonance spectroscopic imaging technique for breast cancer
with a short echo time, Magn. Reson. Imaging 263, 360366 2008.
163
I. S. Haddadin, A. McIntosh, S. Meisamy, C. Corum, A. L. Snyder, N. J.
Powell, M. T. Nelson, D. Yee, M. Garwood, and P. J. Bolan, Metabolite
quantification and high-field MRS in breast cancer, NMR Biomed. to be
published.
164
G. M. Tse, D. K. Yeung, A. D. King, H. S. Cheung, and W. T. Yang, In
vivo proton magnetic resonance spectroscopy of breast lesions: An update, Breast Cancer Res. Treat. 1043, 249255 2007.
165
P. Stanwell and C. Mountford, In vivo proton MR spectroscopy of the
breast, Radiographics 27Suppl 1, S253S266 2007.
166
A. B. Wolbarst and W. R. Hendee, Evolving and experimental technologies in medical imaging, Radiology 2381, 1639 2006.
167
M. Liberman, F. Sampalis, D. S. Mulder, and J. S. Sampalis, Breast
cancer diagnosis by scintimammography: a meta-analysis and review of
the literature, Breast Cancer Res. Treat. 801, 115126 2003.
168
I. Khalkhali, J. Villanueva-Meyer, S. L. Edell, J. L. Connolly, S. J.
Schnitt, J. K. Baum, M. J. Houlihan, R. M. Jenkins, and S. B. Haber,
Diagnostic accuracy of 99mTc-sestamibi breast imaging: multicenter
trial results, J. Nucl. Med. 4112, 19731979 2000.
169
C. B. Hruska, M. K. OConnor, and D. A. Collins, Comparison of small
field of view gamma camera systems for scintimammography, Nucl.
Med. Commun. 265, 441445 2005.
170
B. Mueller, M. K. OConner, I. Blevis, D. J. Rhodes, R. Smith, D. A.
Collins, and S. W. Phillips, Evaluation of a small cadmium zinc telluride
detector for scintimammography, J. Nucl. Med. 444, 602609 2003.
171
M. K. OConnor, S. W. Phillips, C. B. Hruska, D. J. Rhodes, and D. A.
Collins, Molecular breast imaging: advantages and limitations of a scintimammographic technique in patients with small breast tumors, Breast
J. 131, 311 2007.
4894
172
R. F. Brem, J. A. Rapelyea, G. Zisman, K. Mohtashemi, J. Raub, C. B.

Teal, S. Majewski, and B. L. Welch, Occult breast cancer: scintimammography with high-resolution breast-specific gamma camera in women
at high risk for breast cancer, Radiology 2371, 274280 2005.
173
A. Spanu, P. Cottu, A. Manca, F. Chessa, D. Sanna, and G. Madeddu,
Scintimammography with dedicated breast camera in unifocal and
multifocal/multicentric primary breast cancer detection: a comparative
study with SPECT, Int. J. Oncol. 312, 369377 2007.
174
M. P. Tornai, J. E. Bowsher, C. N. Archer, J. Peter, R. J. Jaszczak, L. R.
MacDonald, B. E. Patt, and J. S. Iwanczyk, A 3D gantry single photon
emission tomograph with hemispherical coverage for dedicated breast
imaging, Nucl. Instrum. Methods Phys. Res. A 4971, 157167 2003.
175
C. N. Brzymialkiewicz, M. P. Tornai, R. L. McKinley, and J. E. Bowsher,
Evaluation of fully 3-D emission mammotomography with a compact
cadmium zinc telluride detector, IEEE Trans. Med. Imaging 247, 868
877 2005.
176
M. L. Bradshaw, R. L. McKinley, E. Samei, C. N. Archer, and M. P.
Tornai, Initial x-ray design considerations for application specific emission and transmission tomography ASETT of the breast, Proceedings
of the Annual Meeting of the Society of Nuclear Medicine, New Orleans,
Louisiana, J. Nucl. Med. 445, 287P 2003.
177
K. Yutani, E. Shiba, H. Kusuoka, M. Tatsumi, T. Uehara, T. Taguchi, S. I.
Takai, and T. Nishimura, Comparison of FDG-PET with MIBI-SPECT
in the detection of breast cancer and axillary lymph node metastasis, J.
Comput. Assist. Tomogr. 242, 274280 2000.
178
M. F. Smith, S. Majewski, A. G. Weisenberger, D. A. Kieper, R. R.
Raylman, and T. G. Turkington, Analysis of factors affecting positron
emission mammography PEM image formation., IEEE Trans. Nucl.
Sci. 50, 5359 2003.
179
R. R. Raylman, S. Majewski, M. F. Smith, J. Proffitt, W. Hammond, A.
Srinivasan, J. McKisson, V. Popov, A. Weisenberger, C. O. Judy, B.
Kross, S. Ramasubramanian, L. E. Banta, P. E. Kinahan, and K. Champley, The positron emission mammography/tomography breast imaging
and biopsy system PEM/PET: Design, construction and phantom-based
measurements, Phys. Med. Biol. 533, 637653 2008.
180
N. K. Doshi, Y. Shao, R. W. Silverman, and S. R. Cherry, Design and
evaluation of an LSO PET detector for breast cancer imaging, Med.
Phys. 277, 15351543 2000.
181
P. A. Dokhale, R. W. Silverman, K. S. Shah, R. Grazioso, R. Farrell, J.
Glodo, M. A. McClish, G. Entine, V. H. Tran, and S. R. Cherry, Performance measurements of a depth-encoding PET detector module based on
position-sensitive avalanche photodiode read-out, Phys. Med. Biol.
4918, 42934304 2004.
182
J. Zhang, P. D. Olcott, G. Chinn, A. M. Foudray, and C. S. Levine, Study
of the performance of a novel 1 mm resolution dual-panel PET camera
design dedicated to breast cancer imaging using Monte Carlo simulation,
Med. Phys. 342, 689702 2007.
183
C. J. Thompson, K. Murthy, I. N. Weinberg, and F. Mako, Feasibility
study for positron emission mammography, Med. Phys. 214, 529538
1994.
184
K. Murthy, M. Aznar, A. M. Bergman, C. J. Thompson, J. L. Robar, R.
Lisbona, A. Loutfi, and J. H. Gagnon, Positron emission mammographic
instrument: initial results, Radiology 2151, 280285 2000.
185
K. Murthy, M. Aznar, C. J. Thompson, A. Loutfi, R. Lisbona, and J. H.
Gagnon, Results of preliminary clinical trials of the positron emission
mammography system PEM-I: A dedicated breast imaging system producing glucose metabolic images using FDG, J. Nucl. Med. 4111,
18511858 2000.
186
R. R. Raylman, S. Majewski, R. Wojcik, A. G. Weisenberger, B. Kross, V.
Popov, and H. A. Bishop, The potential role of positron emission mammography for detection of breast cancer. A phantom study, Med. Phys.
278, 19431954 2000.
187
A. Motta, A. D. Guerra, N. Belcari, S. Moehrs, D. Panetta, S. Righi, and
D. Valentini, Fast 3D-EM reconstruction using Planograms for stationary
planar positron emission mammography camera, Comput. Med. Imaging
Graph. 298, 587596 2005.
188
P. Amaral, B. Carrio, M. Ferreira, R. Moura, C. Ortigo, P. Rodrigues, J.
C. Da Silva, A. Trindade, and J. Varela, Performance and quality control
of clear-PEM detector modules, Nucl. Instrum. Methods Phys. Res. A
5802, 11231126 2007.
189
E. L. Rosen, T. G. Turkington, M. S. Soo, J. A. Baker, and R. E. Coleman, Detection of primary breast carcinoma with a dedicated, largefield-of-view FDG PET mammography device: Initial experience, RadiMedical Physics, Vol. 35, No. 11, November 2008
4894
ology 2342, 527534 2005.

W. A. Berg, I. N. Weinberg, D. Narayanan, M. E. Lobrano, E. Ross, L.
Amodei, L. Tafra, L. P. Adler, J. Uddo, W. Stein, 3rd, and E. A. Levine,
High-resolution fluorodeoxyglucose positron emission tomography with
compression positron emission mammography is highly accurate in
depicting primary breast cancer, Breast J. 124, 309323 2006.
191
F. Benard and E. Turcotte, Imaging in breast cancer: Single-photon computed tomography and positron-emission tomography, Breast Cancer
Res 74, 153162 2005.
192
A. E. Burgess, F. L. Jacobson, and P. F. Judy, Human observer detection
experiments with mammograms and power-low noise, Med. Phys. 284,
419437 2001.
193
A. E. Burgess, X. Li, and C. K. Abbey, Visual signal detectability with
two noise components: Anomalous masking effects, J. Opt. Soc. Am. A
149, 24202442 1997.
194
D. P. Chakraborty and H. L. Kundel, Anomalous nodule visibility effects
in mammography images, Proc. SPIE 4324, 6876 2001.
195
H. P. Chan, M. M. Goodsitt, L. M. Hadjiiski, J. E. Bailey, K. Klein, K. L.
Darner, and B. Sahiner, Effects of magnification and zooming on depth
perception in digital stereomammography: An observer performance
study, Phys. Med. Biol. 4822, 37213734 2003.
196
M. M. Goodsitt, H. P. Chan, K. L. Darner, and L. M. Hadjiiski, The
effects of stereo shift angle, geometric magnification and display zoom on
depth measurements in digital stereomammography, Med. Phys. 2911,
27252734 2002.
197
H. P. Chan, M. M. Goodsitt, M. A. Helvie, L. M. Hadjiiski, J. T. Lydick,
M. A. Roubidoux, J. E. Bailey, A. Nees, C. E. Blane, and B. Sahiner,
ROC study of the effect of stereoscopic imaging on assessment of breast
lesions, Med. Phys. 324, 10011009 2005.
198
D. J. Getty et al., Improved accuracy of lesion detection in breast cancer
screening with stereoscopic digital mammography abstract, in 93rd
Scientific Assembly and Annual Meeting of the Radiological Society of
North America RSNA. RSNA Scientific Assembly and Annual Meeting
Program 2007, pp.381382.
199
A. D. Maidment, P. R. Bakic, and M. Albert, Effects of quantum noise
and binocular summation on dose requirements in stereoradiography,
Med. Phys. 3012, 30613071 2003.
200
D. J. Getty and P. J. Green, Clinical medical applications for stereoscopic 3D displays, J. Soc. Inf. Disp. 156, 377384 2007.
201
M. Bissonnette, M. Hansroul, E. Masson, S. Savard, S. Cadieux, P. Warmoes, D. Gravel, J. Agopyan, B. Polischuk, W. Haerer, T. Mertelmeier, J.
Y. Lo, Y. Chen, JTr. Dobbins, J. L. Jesneck, and S. Singh, Digital breast
tomosynthesis using an amorphous selenium flat panel detector, Proc.
SPIE 5745, 529540 2005.
202
B. Ren, C. Ruth, J. Stein, A. Smith, I. Shaw, and Z. Jing, Design and
performance of the prototype full field breast tomosynthesis system with
selenium based flat panel detector, Proc. SPIE 5745, 550561 2005.
203
A. D. Maidment, C. Ullberg, K. Lindman, L. Adelw, J. Egerstrm, M.
Eklund, T. Francke, U. Jordung, T. Kristoffersson, L. Lindqvist, D. Marchal, H. Olla, E. Penton, J. Rantanen, S. Solokov, N. Weber, and H.
Westerberg, Evaluation of a photon-counting breast tomosynthesis imaging system, Proc. SPIE 6142, 61420B 2006.
204
L. T. Niklason, B. T. Christian, L. E. Niklason, D. B. Kopans, D. E.
Castleberry, B. H. OpsahlOng, C. E. Landberg, P. J. Slanetz, A. A. Giardino, R. Moore, D. Albagli, M. C. DeJule, P. F. Fitzgerald, D. F. Fobare,
B. W. Giambattista, R. F. Kwasnick, J. Q. Liu, S. J. Lubowski, G. E.
Possin, J. F. Richotte, C. Y. Wei, and R. F. Wirth, Digital Tomosynthesis
in breast imaging, Radiology 2052, 399406 1997.
205
S. Suryanarayanan, A. Karellas, S. Vedantham, S. P. Baker, S. J. Glick, C.
J. DOrsi, and R. L. Webber, Evaluation of linear and nonlinear tomosynthetic reconstruction methods in digital mammography, Acad. Radiol. 83, 219224 2001.
206
S. Suryanarayanan, A. Karellas, S. Vedantham, S. J. Glick, C. J. DOrsi,
S. P. Baker, and R. L. Webber, Comparison of tomosynthesis methods
used with digital mammography, Acad. Radiol. 712, 10851097
2000.
207
W. Zhao, B. Zhao, P. R. Fisher, P. Warmoes, T. Mertelmeier, and J.
Orman, Optimization of detector operation and imaging geometry for
breast tomosynthesis, Proc. SPIE 6510, 65101M 2007.
208
B. Zhao and W. Zhao, Imaging performance of an amorphous selenium
digital mammography detector in a breast tomosynthesis system, Med.
Phys. 355, 19781987 2008.
209
A. Badano, I. S. Kyprianou, R. J. Jennings, and J. Sempau, Anisotropic
190
4895
imaging performance in breast tomosynthesis, Med. Phys. 3411, 4076

4091 2007.
210
J. G. Mainprize, A. K. Bloomquist, M. P. Kempston, and M. J. Yaffe,
Resolution at oblique incidence angles of a flat panel imager for breast
tomosynthesis, Med. Phys. 339, 31593164 2006.
211
I. Sechopoulos, S. Suryanarayanan, S. Vedantham, C. DOrsi, and A.
Karellas, Computation of the glandular radiation dose in digital tomosynthesis of the breast, Med. Phys. 341, 221232 2007.
212
J. Zhou, B. Zhao, and W. Zhao, A computer simulation platform for the
optimization of a breast tomosynthesis system, Med. Phys. 343, 1098
1109 2007.
213
S. J. Glick and X. Gong, Optimal spectra for indirect detector breast
tomosynthesis, Proc. SPIE 6142, 61421L 2006.
214
T. Wu, B. Liu, R. Moore, and D. B. Kopans, Optimal acquisition techniques for digital breast tomosynthesis screening, Proc. SPIE 6142,
61425E 2006.
215
I. Sechopoulos, S. Suryanarayanan, S. Vedantham, C. J. DOrsi, and A.
Karellas, Scatter radiation in digital tomosynthesis of the breast, Med.
Phys. 342, 564576 2007.
216
F. Diekmann, H. Meyer, S. Diekmann, S. Puong, S. Muller, U. Bick, and
P. Rogalla, Thick slices from tomosynthesis data sets: Phantom study for
the evaluation of different algorithms, J. Digit Imaging to be published.
217
Y. Chen, J. Y. Lo, and J. T. Dobbins, 3rd, Importance of point-by-point
back projection correction for isocentric motion in digital breast tomosynthesis: Relevance to morphology of structures such as microcalcifications, Med. Phys. 3410, 38853892 2007.
218
Y. Zhang, H. P. Chan, B. Sahiner, J. Wei, M. M. Goodsitt, L. M. Hadjiiski, J. Ge, and C. Zhou, A comparative study of limited-angle conebeam reconstruction methods for breast tomosynthesis, Med. Phys.
3310, 37813795 2006.
219
T. Wu, R. H. Moore, and D. B. Kopans, Voting strategy for artifact
reduction in digital breast tomosynthesis, Med. Phys. 337, 24612471
2006.
220
J. T. Rakowski and M. J. Dennis, A comparison of reconstruction algorithms for C-arm mammography tomosynthesis, Med. Phys. 338,
30183032 2006.
221
T. Mertelmeier, J. Orman, W. Haerer, and M. K. Dudam, Optimizing
filtered backprojection reconstruction for a breast tomosynthesis prototype device, Proc. SPIE 6142, 61420F 2006.
222
Y. Chen, J. Y. Lo, J. A. Baker, and J. Tr. Dobbins, Gaussian frequency
blending algorithm with matrix inversion tomosynthesis MITS and filtered back projection FBP for better digital breast tomosynthesis reconstruction, Proc. SPIE 6142, 61420E 2006.
223
S. P. Poplack, T. D. Tosteson, C. A. Kogel, and H. M. Nagy, Digital
breast tomosynthesis: initial experience in 98 women with abnormal digital screening mammography, AJR, Am. J. Roentgenol. 1893, 616623
2007.
224
C. Badea, Z. Kolitsi, and N. Pallikarakis, A wavelet-based method for
removal of out-of-plane structures in digital tomosynthesis, Comput.
Med. Imaging Graph. 224, 309315 1998.
225
Z. Kolitsi, G. Panayiotakis, and N. Pallikarakis, A method for selective
removal of out-of-plane structures in digital tomosynthesis, Med. Phys.
201, 4750 1993.
226
J. Ge, H. P. Chan, Y. Zhang, B. Sahiner, J. Wei, and L. M. Hadjiiski,
Digital tomosynthesis mammography: An interplane artifact reduction
method for microcalcifications on reconstructed slices based on 3D geometrical information, in 92nd Scientific Assembly and Annual Meeting
of the RSNA. RSNA Scientific Assembly and Annual Meeting Program
2006, p. 230.
227
R. H. Moore et al., Initial callback rates for conventional and digital
breast tomosynthesis mammography comparison in the screening setting, in 93rd Scientific Assembly and Annual Meeting of the Radiological Society of North America RSNA. Scientific Assembly and Annual
Meeting Program 2007, p. 381.
228
R. W. Redington and J. L. Henkes, Jr., Mammography, U.S. Patent, No.
3,973,126, General Electric Co., Shenectady, NY, August 3, 1976.
229
C. H. Chang, J. L. Sibala, S. L. Fritz, S. J. Dwyer, 3rd, A. W. Templeton,
F. Lin, and W. R. Jewell, Computed tomography in detection and diagnosis of breast cancer, Cancer 464 Suppl, 939946 1980.
230
C. H. Chang, D. E. Nesbit, D. R. Fisher, S. L. Fritz, S. J. Dwyer, 3rd, A.
W. Templeton, F. Lin, and W. R. Jewell, Computed tomographic mammography using a conventional body scanner, AJR, Am. J. Roentgenol.
1383, 553558 1982.
231
4895
V. Raptopoulos, J. K. Baum, M. Hochman, A. Karellas, M. J. Houlihan,

and C. J. DOrsi, High resolution CT mammography of surgical biopsy
specimens, J. Comput. Assist. Tomogr. 202, 179184 1996.
232
J. M. Boone, T. R. Nelson, K. K. Lindfors, and J. A. Seibert, Dedicated
breast CT: radiation dose and image quality evaluation, Radiology
2213, 657667 2001.
233
J. M. Boone, N. Shah, and T. R. Nelson, A comprehensive analysis of
DgNCT coefficients for pendant-geometry cone-beam breast computed
234
S. C. Thacker and S. J. Glick, Normalized glandular dose DgN coefficients for flat-panel CT breast imaging, Phys. Med. Biol. 4924, 5433
5444 2004.
235
I. Sechopoulos, S. Vedantham, S. Suryanarayanan, C. J. DOrsi, and A.
Karellas, Monte Carlo and phantom study of the radiation dose to the
body from dedicated CT of the breast, Radiology 2471, 98105 2008.
236
A. L. Kwan, J. M. Boone, and N. Shah, Evaluation of x-ray scatter
properties in a dedicated cone-beam breast CT scanner, Med. Phys.
329, 29672975 2005.
237
B. Liu, S. J. Glick, and C. Groiselle, Characterization of scatter radiation
in cone beam CT mammography, Proc. SPIE 5745, 818827 2005.
238
B. Chen and R. Ning, Cone-beam volume CT breast imaging: feasibility
study, Med. Phys. 295, 755770 2002.
239
J. M. Boone, Breast CT: Its prospect for breast cancer screening and
diagnosis, in Advances in breast imaging: Physics, Technology and Clinical Applications, Categorical Course in Diagnostic Radiology Physics,
edited by A. Karellas and M. L. Giger Radiological Society of North
America RSNA, Oak Brook, IL, 2004.
240
R. L. McKinley, C. N. Bryzmialkiewicz, P. Madhav, and M. P. Tornai,
Investigation of cone-beam acquisitions implemented using a novel
dedicated mammotomography system with unique arbitrary orbit capability, Proc. SPIE 5745, 609617 2005.
241
K. Zeng, H. Yu, L. L. Fajardo, and G. Wang, Cone-beam mammocomputed tomography from data along two tilting arcs, Med. Phys.
3310, 36213633 2006.
242
J. M. Boone, A. L. Kwan, J. A. Seibert, N. Shah, K. K. Lindfors, and T.
R. Nelson, Technique factors and their relationship to radiation dose in
pendant geometry breast CT, Med. Phys. 3212, 37673776 2005.
243
S. J. Glick, S. Thacker, X. Gong, and B. Liu, Evaluating the impact of
X-ray spectral shape on image quality in flat-panel CT breast imaging,
Med. Phys. 341, 524 2007.
244
R. L. McKinley, M. P. Tornai, E. Samei, and M. L. Bradshaw, Simulation study of a quasi-monochromatic beam for x-ray computed mammotomography, Med. Phys. 314, 800813 2004.
245
A. L. Kwan, J. M. Boone, K. Yang, and S. Y. Huang, Evaluation of the
spatial resolution characteristics of a cone-beam breast CT scanner, Med.
Phys. 341, 275281 2007.
246
J. Zhong, R. Ning, and D. Conover, Image denoising based on multiscale singularity detection for cone beam CT breast imaging, IEEE
247
M. C. Altunbas, C. C. Shaw, L. Chen, C. Lai, X. Liu, T. Han, and T.
Wang, A post-reconstruction method to correct cupping artifacts in cone
beam breast computed tomography, Med. Phys. 347, 31093118
2007.
248
W. T. Yang, S. Carkaci, L. Chen, C. J. Lai, A. Sahin, G. J. Whitman, and
C. C. Shaw, Dedicated cone-beam breast CT: Feasibility study with
surgical mastectomy specimens, AJR, Am. J. Roentgenol. 1896, 1312
1315 2007.
249
J. M. Boone, A. L. Kwan, K. Yang, G. W. Burkett, K. K. Lindfors, and T.
R. Nelson, Computed tomography for imaging the breast, Journal of
mammary gland biology and neoplasia 112, 103111 2006.
250
K. K. Lindfors, J. M. Boone, T. R. Nelson, K. Yang, A. L. Kwan, and D.
F. Miller, Dedicated breast CT: Initial clinical experience, Radiology
2463, 725733 2008.
251
D. J. Crotty, P. Madhav, R. L. McKinley, and M. P. Tornai, Investigating
novel patient bed designs for use in a hybrid dual modality dedicated 3D
breast imaging system, Proc. SPIE 6510, 65101H 2007.
252
R. Ning, D. Conover, Y. Yu, Y. Zhang, W. Cai, R. Betancourt-Benitez,
and X. Lu, A novel cone beam breast CT scanner: system evaluation,
Proc. SPIE 6510, 651030 2007.
253
X. Gong, A. A. Vedula, and S. J. Glick, Microcalcification detection
using cone-beam CT mammography with a flat-panel imager, Phys.
Med. Biol. 4911, 21832195 2004.
254
C. J. Lai, C. C. Shaw, L. Chen, M. C. Altunbas, X. Liu, T. Han, T. Wang,
4896
W. T. Yang, G. J. Whitman, and S. J. Tu, Visibility of microcalcification

in cone beam breast CT: Effects of X-ray tube voltage and radiation
dose, Med. Phys. 347, 29953004 2007.
255
A. L. Kwan, K. Yang, K. K. Lindfors, T. R. Nelson, and J. M. Boone,
Visualization of micro-calcifications in a prototype breast CT abstract,
Med. Phys. 33, 1990 2006.
256
P. M. Shikhaliev, Tilted angle CZT detector for photon counting/energy
weighting x-ray and CT imaging, Phys. Med. Biol. 5117, 42674287
2006.
257
S. Rudin, A. T. Kuhls, G. K. Yadava, G. C. Josan, Y. Wu, R. N. Chityala,
H. S. Rangwala, N. C. Ionita, K. R. Hoffmann, and D. R. Bednarek, New
light-amplifier-based detector designs for high spatial resolution and high
sensitivity CBCT mammography and fluoroscopy, Proc. SPIE 6142,
61421R 2006.
258
S. J. Glick, Breast CT, Annu. Rev. Biomed. Eng. 9, 501526 2007.
259
X. Gong, S. J. Glick, B. Liu, A. A. Vedula, and S. Thacker, A computer
simulation study comparing lesion detection accuracy with digital mammography, breast tomosynthesis, and cone-beam CT breast imaging,
Med. Phys. 334, 10411052 2006.
260
A. C. Watt, L. V. Ackerman, P. C. Shetty, M. Burke, M. Flynn, C.
Grodsinsky, G. Fine, and S. Wilderman, Differentiation between benign
and malignant disease of the breast using digital subtraction angiography
of the breast, Cancer 566, 12871292 1985.
261
A. C. Watt, L. V. Ackerman, J. P. Windham, P. C. Shetty, M. W. Burke,
M. J. Flynn, C. Grodinsky, G. Fine, and S. J. Wilderman, Breast lesions:
Differential diagnosis using digital subtraction angiography, Radiology
1591, 3942 1986.
262
J. M. Lewin, P. K. Isaacs, V. Vance, and F. J. Larke, Dual-energy
contrast-enhanced digital subtraction mammography: feasibility, Radiology 2291, 261268 2003.
263
R. A. Jong, M. J. Yaffe, M. Skarpathiotakis, R. S. Shumak, N. M. Danjoux, A. Gunesekara, and D. B. Plewes, Contrast-enhanced digital mammography: Initial clinical experience, Radiology 2283, 842850
2003.
264
F. Diekmann, S. Diekmann, F. Jeunehomme, S. Muller, B. Hamm, and U.
Bick, Digital mammography using iodine-based contrast media: Initial
clinical experience with dynamic contrast medium enhancement, Invest.
Radiol. 407, 397404 2005.
265
F. Diekmann, S. Diekmann, M. Taupitz, U. Bick, K. J. Winzer, C. Huttner, S. Muller, F. Jeunehomme, and B. Hamm, Use of iodine-based
contrast media in digital full-field mammographyinitial experience,
Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1753, 342345
2003.
266
C. Dromain, C. Balleyguier, S. Muller, M. C. Mathieu, F. Rochard, P.
Opolon, and R. Sigal, Evaluation of tumor angiogenesis of breast carcinoma using contrast-enhanced digital mammography, AJR, Am. J.
Roentgenol. 1875, W528537 2006.
267
M. Skarpathiotakis, M. J. Yaffe, A. K. Bloomquist, D. Rico, S. Muller, A.
Rick, and F. Jeunehomme, Development of contrast digital mammography, Med. Phys. 2910, 24192426 2002.
268
M. Saito, Dual-energy approach to contrast-enhanced mammography using the balanced filter method: spectral optimization and preliminary
phantom measurement, Med. Phys. 3411, 42364246 2007.
269
P. Baldelli, A. Bravin, C. Di Maggio, G. Gennaro, A. Sarnelli, A. Taibi,
and M. Gambaccini, Evaluation of the minimum iodine concentration
for contrast-enhanced subtraction mammography, Phys. Med. Biol.
5117, 42334251 2006.
270
F. Diekmann, A. Sommer, R. Lawaczeck, S. Diekmann, H. Pietsch, U.
Speck, B. Hamm, and U. Bick, Contrast-to-noise ratios of different elements in digital mammography: evaluation of their potential as new contrast agents, Invest. Radiol. 425, 319325 2007.
271
R. Lawaczeck, F. Diekmann, S. Diekmann, B. Hamm, U. Bick, W. R.
Press, H. Schirmer, K. Schon, and H. J. Weinmann, New contrast media
designed for x-ray energy subtraction imaging in digital mammography,
Invest. Radiol. 389, 602608 2003.
272
S. Suryanarayanan, I. Sechopoulos, S. Vedantham, and A. Karellas, Feasibility of low dose x-ray contrast enhanced digital mammography with
gold nanoparticles, Med. Phys. 346, 2360 2007.
273
S. C. Chen, A. K. Carton, M. Albert, E. F. Conant, M. D. Schnall, and A.
D. Maidment, Initial clinical experience with contrast-enhanced digital
breast tomosynthesis, Acad. Radiol. 142, 229238 2007.
274
S. Puong, X. Bouchevreau, N. Duchateau, R. Iordache, and S. Muller,
Optimization of beam parameters and iodine quantification in dualMedical Physics, Vol. 35, No. 11, November 2008
4896
energy contrast enhanced digital breast tomosynthesis, Proc. SPIE 6913,

69130Z 2008.
275
R. Saunders, E. Samei, C. Badea, H. Yuan, K. Ghaghada, Y. Qi, L. W.
Hedlund, and S. Mukundan, Optimization of dual energy contrast enhanced breast tomosynthesis for improved mammographic lesion detection and diagnosis, Proc. SPIE 6913, 69130Y 2008.
276
C. T. Badea, E. Samei, K. Ghaghada, R. Saunders, H. Yuan, Y. Qi, L. W.
Hedlund, and S. Mukundan, Utility of a prototype liposomal contrast
agent for x-ray imaging of breast cancer: A proof of concept using
micro-CT in small animals, Proc. SPIE 6913, 691303 2008.
277
G. H. Simon, Y. Fu, K. Berejnoi, L. S. Fournier, V. Lucidi, B. Yeh, D. M.
Shames, and R. C. Brasch, Initial computed tomography imaging experience using a new macromolecular iodinated contrast medium in experimental breast cancer, Invest. Radiol. 409, 614620 2005.
278
S. W. Wilkins, T. E. Gureyev, D. Gao, A. Pogany, and A. W. Stevenson,
Phase-contrast imaging using polychromatic hard X-rays, Nature London 384, 335338 1996.
279
Y. Suzuki, N. Yagi, and K. Uesugi, X-ray refraction-enhanced imaging
and a method for phase retrieval for a simple object, J. Synchrotron
Radiat. 9Pt 3, 160165 2002.
280
D. Chapman, W. Thomlinson, R. E. Johnston, D. Washburn, E. Pisano, N.
Gmur, Z. Zhong, R. Menk, F. Arfelli, and D. Sayers, Diffraction enhanced x-ray imaging, Phys. Med. Biol. 4211, 20152025 1997.
281
M. O. Hasnah, Z. Zhong, O. Oltulu, E. Pisano, R. E. Johnston, D. Sayers,
W. Thomlinson, and D. Chapman, Diffraction enhanced imaging contrast mechanisms in breast cancer specimens, Med. Phys. 2910, 2216
2221 2002.
282
E. D. Pisano, R. E. Johnston, D. Chapman, J. Geradts, M. V. Iacocca, C.
A. Livasy, D. B. Washburn, D. E. Sayers, Z. Zhong, M. Z. Kiss, and W.
C. Thomlinson, Human breast cancer specimens: diffraction-enhanced
imaging with histologic correlationimproved conspicuity of lesion detail compared with digital radiography, Radiology 2143, 895901
2000.
283
S. Fiedler, A. Bravin, J. Keyrilainen, M. Fernandez, P. Suortti, W. Thomlinson, M. Tenhunen, P. Virkkunen, and M. Karjalainen-Lindsberg, Imaging lobular breast carcinoma: comparison of synchrotron radiation
DEI-CT technique with clinical CT, mammography and histology, Phys.
Med. Biol. 492, 175188 2004.
284
X. Wu and H. Liu, A new theory of phase-contrast x-ray imaging based
on Wigner distributions, Med. Phys. 319, 23782384 2004.
285
X. Wu and H. Liu, Clinical implementation of x-ray phase-contrast imaging: theoretical foundations and design considerations, Med. Phys.
308, 21692179 2003.
286
X. Wu and H. Liu, Clarification of aspects in in-line phase-sensitive
x-ray imaging, Med. Phys. 342, 737743 2007.
287
F. Meng, A. Yan, G. Zhou, X. Wu, and H. Liu, Development of a dualdetector X-ray imaging system for phase retrieval study, Nucl. Instrum.
Methods Phys. Res. B 2542, 300306 2007.
288
A. Corlu, R. Choe, T. Durduran, K. Lee, M. Schweiger, S. R. Arridge, E.
M. C. Hillman, and A. G. Yodh, Diffuse optical tomography with spectral constraints and wavelength optimization, Appl. Opt. 4411, 2082
2093 2005.
289
S. Srinivasan, B. W. Pogue, B. Brooksby, S. Jiang, H. Dehghani, C.
Kogel, W. A. Wells, S. Poplack, and K. D. Paulsen, Near-infrared characterization of breast tumors in-vivo using spectrally-constrained reconstruction, Technol. Cancer Res. Treat. 45, 513526 2005.
290
S. Srinivasan, B. W. Pogue, S. Jiang, H. Dehghani, and K. D. Paulsen,
Spectrally constrained chromophore and scattering near-infrared tomography provides quantitative and robust reconstruction, Appl. Opt. 4410,
18581869 2005.
291
N. Shah, A. E. Cerussi, D. Jakubowski, D. Hsiang, J. Butler, and B. J.
Tromberg, Spatial variations in optical and physiological properties of
healthy breast tissue, J. Biomed. Opt. 93, 534540 2004.
292
R. L. van Veen, H. J. Sterenborg, A. W. Marinelli, and M. MenkePluymers, Intraoperatively assessed optical properties of malignant and
healthy breast tissue used to determine the optimum wavelength of contrast for optical mammography, J. Biomed. Opt. 96, 11291136 2004.
293
C. Zhou, R. Choe, N. Shah, T. Durduran, G. Yu, A. Durkin, D. Hsiang, R.
Mehta, J. Butler, A. Cerussi, B. J. Tromberg, and A. G. Yodh, Diffuse
optical monitoring of blood flow and oxygenation in human breast cancer
during early stages of neoadjuvant chemotherapy, J. Biomed. Opt. 125,
051903 2007.
294
V. Ntziachristos, A. G. Yodh, M. D. Schnall, and B. Chance, MRI-
4897
guided diffuse optical spectroscopy of malignant and benign breast lesions, Neoplasia 44, 347354 2002.
295
B. Brooksby, S. Jiang, H. Dehghani, B. W. Pogue, K. D. Paulsen, J. B.
Weaver, C. Kogel, and S. P. Poplack, Combining near-infrared tomography and magnetic resonance imaging to study in vivo breast tissue:
Implementation of a Laplacian-type regularization to incorporate magnetic resonance structure, J. Biomed. Opt. 105, 051504 2005.
296
B. Brooksby, B. W. Pogue, S. Jiang, H. Dehghani, S. Srinivasan, C.
Kogel, T. D. Tosteson, J. Weaver, S. P. Poplack, and K. D. Paulsen,
Imaging breast adipose and fibroglandular tissue molecular signatures by
using hybrid MRI-guided near-infrared spectral tomography, Proc. Natl.
Acad. Sci. U.S.A. 10323, 88288833 2006.
297
C. M. Carpenter, B. W. Pogue, S. Jiang, H. Dehghani, X. Wang, K. D.
Paulsen, W. A. Wells, J. Forero, C. Kogel, J. B. Weaver, S. P. Poplack,
and P. A. Kaufman, Image-guided optical spectroscopy provides
molecular-specific information in vivo: MRI-guided spectroscopy of
breast cancer hemoglobin, water, and scatterer size, Opt. Lett. 328,
933935 2007.
298
V. Ntziachristos, A. G. Yodh, M. Schnall, and B. Chance, Concurrent
MRI and diffuse optical tomography of breast after indocyanine green
enhancement, Proc. Natl. Acad. Sci. U.S.A. 976, 27672772 2000.
299
X. Gu, Q. Zhang, M. Bartlett, L. Schutz, L. L. Fajardo, and H. Jiang,
Differentiation of cysts from solid tumors in the breast with diffuse
optical tomography, Acad. Radiol. 111, 5360 2004.
300
B. J. Tromberg, A. Cerussi, N. Shah, M. Compton, A. Durkin, D. Hsiang,
J. Butler, and R. Mehta, Imaging in breast cancer: diffuse optics in breast
cancer: detecting tumors in pre-menopausal women and monitoring neoadjuvant chemotherapy, Breast Cancer Research 76, 279285 2005.
301
N. Shah, J. Gibbs, D. Wolverton, A. Cerussi, N. Hylton, and B. J. Tromberg, Combined diffuse optical spectroscopy and contrast-enhanced
magnetic resonance imaging for monitoring breast cancer neoadjuvant
chemotherapy: a case study, J. Biomed. Opt. 105, 051503 2005.
302
D. B. Jakubowski, A. E. Cerussi, F. Bevilacqua, N. Shah, D. Hsiang, J.
Butler, and B. J. Tromberg, Monitoring neoadjuvant chemotherapy in
breast cancer using quantitative diffuse optical spectroscopy: A case
study, J. Biomed. Opt. 91, 230238 2004.
303
G. Boverman, Q. Fang, S. A. Carp, E. L. Miller, D. H. Brooks, J. Selb, R.
H. Moore, D. B. Kopans, and D. A. Boas, Spatio-temporal imaging of
the hemoglobin in the compressed breast with diffuse optical tomogra-
4897
phy, Phys. Med. Biol. 5212, 36193641 2007.

C. Li, H. Zhao, B. Anderson, and H. Jiang, Multispectral breast imaging
using a ten-wavelength, 64 64 source/detector channels silicon
photodiode-based diffuse optical tomography system, Med. Phys. 333,
627636 2006.
305
Q. Zhu, N. Chen, and S. H. Kurtzman, Imaging tumor angiogenesis by
use of combined near-infrared diffusive light and ultrasound, Opt. Lett.
28, 337339 2003.
306
B. J. Tromberg, B. W. Pogue, K. D. Paulsen, A. G. Yodh, D. A. Boas, and
A. E. Cerussi, Assessing the future of diffuse optical imaging technologies for breast cancer management, Med. Phys. 356, 24432451
2008.
307
A. Stojadinovic, A. Nissan, C. D. Shriver, E. A. Mittendorf, M. D. Akin,
V. Dickerson, S. Lenington, L. D. Platt, T. Stavros, S. R. Goldstein, O.
Moskovitz, Z. Gallimidi, S. I. Fields, A. Yeshaya, T. M. Allweis, R.
Manassa, T. Pappo, R. X. Ginor, R. B. DAgostino, and D. Gur, Electrical impedance scanning as a new breast cancer risk stratification tool
for young women, J. Surg. Oncol. 972, 112120 2008.
308
P. M. Meaney, M. W. Fanning, T. Raynolds, C. J. Fox, Q. Fang, C. A.
Kogel, S. P. Poplack, and K. D. Paulsen, Initial clinical experience with
microwave breast imaging in women with normal mammography, Acad.
Radiol. 142, 207218 2007.
309
M. Lazebnik, L. McCartney, D. Popovic, C. B. Watkins, M. J. Lindstrom,
J. Harter, S. Sewall, A. Magliocco, J. H. Booske, M. Okoniewski, and S.
C. Hagness, A large-scale study of the ultrawideband microwave dielectric properties of normal breast tissue obtained from reduction surgeries,
Phys. Med. Biol. 5210, 26372656 2007.
310
M. Lazebnik, D. Popovic, L. McCartney, C. B. Watkins, M. J. Lindstrom,
J. Harter, S. Sewall, T. Ogilvie, A. Magliocco, T. M. Breslin, W. Temple,
D. Mew, J. H. Booske, M. Okoniewski, and S. C. Hagness, A large-scale
study of the ultrawideband microwave dielectric properties of normal,
benign and malignant breast tissues obtained from cancer surgeries,
Phys. Med. Biol. 5220, 60936115 2007.
311
S. P. Poplack, T. D. Tosteson, W. A. Wells, B. W. Pogue, P. M. Meaney,
A. Hartov, C. A. Kogel, S. K. Soho, J. J. Gibson, and K. D. Paulsen,
Electromagnetic breast imaging: results of a pilot study in women with
abnormal mammograms, Radiology 2432, 350359 2007.
312
R. J. Brenner and Y. Parisky, Alternative breast-imaging approaches,
Radiol. Clin. North Am. 455, 907923 2007.
304
PET/CT in radiation oncology

Tinsu Pana and Osama Mawlawi
Department of Imaging Physics, M. D. Anderson Cancer Center, The University of Texas, Houston, Texas
77030
Received 27 January 2008; revised 3 June 2008; accepted for publication 26 August 2008;
PET/CT is an effective tool for the diagnosis, staging and restaging of cancer patients. It combines
the complementary information of functional PET images and anatomical CT images in one imaging session. Conventional stand-alone PET has been replaced by PET/CT for improved patient
comfort, patient throughput, and most importantly the proven clinical outcome of PET/CT over that
of PET and that of separate PET and CT. There are over two thousand PET/CT scanners installed
worldwide since 2001. Oncology is the main application for PET/CT. Fluorine-18 deoxyglucose is
the choice of radiopharmaceutical in PET for imaging the glucose uptake in tissues, correlated with
an increased rate of glycolysis in many tumor cells. New molecular targeted agents are being
developed to improve the accuracy of targeting different disease states and assessing therapeutic
response. Over 50% of cancer patients receive radiation therapy RT in the course of their disease
treatment. Clinical data have demonstrated that the information provided by PET/CT often changes
patient management of the patient and/or modifies the RT plan from conventional CT simulation.
The application of PET/CT in RT is growing and will become increasingly important. Continuing
improvement of PET/CT instrumentation will also make it easier for radiation oncologists to integrate PET/CT in RT. The purpose of this article is to provide a review of the current PET/CT
technology, to project the future development of PET and CT for PET/CT, and to discuss some
issues in adopting PET/CT in RT and potential improvements in PET/CT simulation of the thorax
in radiation therapy. 2008 American Association of Physicists in Medicine.
DOI: 10.1118/1.2986145
Key words: PET/CT, 4D CT, cine CT, MIP CT, average CT
I. INTRODUCTION
PET/CT was developed in 1998.1 It was not until 2001 when
the first commercial PET/CT scanner became available.2
Prior to the technology of PET/CT, the CT and the PET data
were acquired in two different scanners. Fusion of the PET
and CT data was performed with software techniques.3 Registration of the PET and CT data was more successful for the
brain studies with rigid transformation4 but less accurate for
the other regions of the body, in particular in the thorax and
the abdomen, due to the difficulty in repositioning the patient
in two separate sessions and the nonrigid nature of the
organs.1,5 Average error in the fusion of separate PET and CT
brain images was in the order of 2 3 mm.3 This error increased to 5 11 mm when fusing separate PET and CT body
images of the thorax.6 The advent of PET/CT scanners has
facilitated the hardware fusion of PET and CT data sets by
transporting the patient between the PET and CT components of the scanner without repeating patient setup. It has
been shown that hardware fusion is more accurate than software fusion in diagnosis, staging and restaging of many cancer types, in particular in the area of tumor infiltration of
adjacent structures that could not be conclusively assessed
using the separate CT and PET data.7
The fast scan speed of CT has shortened a normal imaging session of about 1 h with a stand-alone PET to less than
30 min with a PET/CT. This time savings has a major impact
on patient comfort in particular for patients who may need to
4955
raise their arms over the head during the PET/CT scan. The
CT images are adapted for attenuation correction and tumor
localization of the PET data.1 There are over 2000 PET/CT
scanners installed worldwide, and stand-alone PET scanners
have not been in production since 2006.5 Fluorine-18 deoxyglucose 18F-FDG is the major radiopharmaceutical in
PET/CT and has been approved for diagnosis and staging of
nonsmall-cell lung cancer NSCLC, colorectal cancer,
esophageal cancer, head and neck cancer, lymphoma, and
melanoma since 1998.8 In addition, PET/CT has been approved for staging and restaging and for therapeutic monitoring of breast cancer.
The average 18F-FDG PET sensitivity and specificity
across all oncology application are estimated at 84% based
on 18 402 patient studies and 88% based on 14 264 patient
studies, respectively, according to Gambhir et al.9 from a
collection of 419 articles from 1993 to 2000. Specifically, the
sensitivity of PET ranged from 84% to 86%, the specificity
ranged from 88% to 93%, and the accuracy ranged from 87%
to 90%.9 The addition of PET significantly improves the diagnosis of lung cancer with the sensitivity of 96% and specificity of 73% as compared to sensitivity of 67% for CT
alone, and the staging with the sensitivity of 83% and specificity of 91%, respectively as compared to sensitivity of
64% and specificity of 74% for CT alone.
The additional clinical values of PET/CT to PET alone or
separate PET and CT have also been documented. The first
0094-2405/2008/3511/4955/12/$23.00
4955
4956
T. Pan and O. Mawlawi: PET/CT in radiation oncology
4956
study was published by Kluetz et al.10 using the first prototype PET/CT scanner. It was found that hardware registration
enables physicians to more precisely discriminate between
physiologic and malignant 18F-FDG uptake and more accurately localize lesions. Hany et al.11 compared the data of
PET/CT and PET alone in 53 patients for the diagnosis or
suspicion of malignancy. Of 287 lesions, the sensitivity,
specificity, and accuracy increased from 90%, 93%, and 91%
for PET alone, to 98%, 99%, and 98% for PET/CT. BarShalom et al.12 evaluated 204 patients with 586 suspicious
lesions by interpreting PET and CT data together or PET and
CT data separately. They found that interpretation of PET
and CT together provided additional information over the
separate interpretation of PET and CT in 99 patients 49%
with 178 lesions 30%. PET/CT improved characterization
of equivocal lesions as definitely benign in 10% of sites and
as definitely malignant in 5% of sites. The results of PET/CT
had an impact on the management of 28 patients 14%.
Lardinois et al.13 compared the diagnostic accuracy of
PET/CT with that of CT alone, that of PET alone, and that of
conventional visual correlation of PET and CT in determining the stage of disease in 50 patients with proven or suspected NSCLC. PET/CT provided additional information in
20 of 49 patients. PET/CT had better diagnostic accuracy and
more accurate nodal staging than CT alone, PET alone, or
visual correlation of PET and CT.
Over 50% of all patients with cancer receive radiation
therapy RT during the course of their disease
management.14 Applications of PET to RT have been reported in head and neck,1526 lung,19,2740 breast,41
esophageal,42,43
non-Hodgkins
lymphoma,30,44
19,45
46,47
19
gynecologic,
rectal,
anal,
and prostate with
18
F-choline.48 Studies have also found that PET has advantages over CT in the standardization of volume
delineation,29,49,50 in the reduction of the risk for geometrical
misses,36 in the reduction of mediastinal nodal failure rate,51
in the minimization of radiation dose to the nontarget
organs,19,26,52 and in the assessment of tumor burden, blood
flow, tissue inflammation, and hypoxia.5358 Integration of
functional PET data with anatomical CT data should be a
standard in RT.59
There are some challenges in PET/CT imaging for RT.
The most challenging one is with reimbursement. Many patients are referred to a PET scan for diagnosis before receiving RT. The time between diagnosis and RT is too short for a
second PET to be reimbursable. In this situation, the diagnostic images of PET or PET/CT are fused using software
with the simulation CT images for RT. As of today, there is
no reimbursement for a simulation PET/CT. Unlike a whole
body PET/CT for diagnosis, a simulation PET/CT only needs
to be performed with a limited coverage much like in a simulation CT. A simulation PET/CT may be performed or embedded in a whole body PET/CT scanner for staging, restaging, or assessment of the treatment response in the course of
treatment. Integration of PET/CT in RT also involves at least
the following two disciplines of people working together:
Nuclear medicine and radiation oncology. Patient throughput
is an important concern in the practice of nuclear medicine,
as opposed to the accurate setup of patients in the treatment

position with external laser lights and a flat table support
needed for radiation oncology. Collaboration between the
two disciplines is critical to the success of PET/CT for RT.
Application of PET/CT in nuclear medicine is a task of
detection based on the standardized uptake value SUV.60,61
The physicians can sometimes read through the artifacts such
as the ones caused by respiratory motion,6265 implanted
metal such as dental filling,6668 pace makers,69,70
prosthetics,71 and surgical clips, and intravenous72 or oral
contrast media.73 Radiation oncologists may not be familiar
with the images in PET/CT imaging and their associated artifacts. They generally depend on the diagnosis and staging
of the nuclear medicine physicians for RT planning.7476
Both the location and extent of tumor are critical in RT and
may not be fully captured in the nuclear medicine report. In
addition, there is still no standard for delineating the gross
tumor volume GTV in PET images. The workstations used
for evaluating PET/CT images in nuclear medicine are usually not equipped with planning software utilities required
for RT. Reproduction of SUV may be difficult once the
PET/CT images are transferred to planning workstations in
RT.77,78 It is critical to ensure the PET/CT scan data are
acquired, processed and transferred correctly for RT.
Migration from stand-alone PET to PET/CT has improved
registration between the CT and PET data from hardware
fusion, in particular in the area outside the thorax and the
abdomen such us the head and neck area. However, PET/CT
has also introduced a new problem, namely the misalignment
between the PET and CT data in the thorax and the abdomen
due to patient respiration and the different scan speeds of CT
and PET.6265,7989 The first clinical investigation addressing
the impact of misalignment was conducted by Osman et al.89
They observed that curvelinear cold artifacts paralleling the
dome of the diaphragm at the lung bases were frequently
noted in PET/CT images. However, clinically they reported
that this artifact resulted in significant inaccurate localization
of lesions for only 2% in a study of 300 patients. Gould et
al.86 reported in a cohort of 259 patients for cardiac PET/CT
imaging, 40% of the patient data exhibited false positive defects due to misalignment. It is evident from these two studies that the severity of misalignment is dependent on the area
of interest. Misalignment between PET and CT may not be a
significant issue in a whole body PET/CT scan when the
lesions may not be in the lungs, liver, and esophageal and not
close to the diaphragm. However, it becomes a significant
issue in cardiac PET/CT imaging when the heart is right
above the diaphragm. Coaching patients to breath-hold in an
effort to mitigate the artifact was not very successful. In a
study of 100 patients undergoing PET/CT with breath-hold at
midexpiration CT by Pan et al.,80 50% of the patients exhibited the curvelinear cold artifacts in the PET images caused
by respiratory motion. In a recent study of 216 consecutive
patients undergoing free breathing CT in PET/CT by Chi et
al.,90 68% of the patient data exhibited either the curvelinear
cold artifacts or tumor misalignment. Misalignment due to
breathing has been shown to cause variation in SUV,80 critical to delineation of gross tumor volume in the PET data. We
4957
4957
TABLE I. Properties of major PET detectors.
Scintillator
Relative light output

NaITl = 100
Decay time ns
Thickness for 90%

efficiency at 511 keV cm
BGO
GSO
LSO, LYSO
15
25
80
300
60
40
2.4
3.3
2.7
will highlight and project the technology advancement of

PET/CT, propose imaging techniques that will help improve
registration of the PET and CT data in the thorax, and emphasize the particular requirements of PET/CT and for radiation oncology.
II. PET DETECTOR
The PET of a PET/CT scanner today is distinguished by
the scintillator materials used for the detection of 511 keV
annihilation photons. The major materials are bismuth germinate oxide BGO, gadolinium oxyorthosilicate GSO, lutetium oxyorthosilicate LSO, and LYSO LSO doped with
a small amount of yttrium.9193,9597 BGO has a slow decay
time and a low light output leading to relatively poor timing
and energy resolution. It has a high stopping power to capture most of the photons reaching the detector. The new materials of LSO and LYSO have a fast decay time and a high
light output leading to improved timing and energy resolution. Their stopping power is slightly inferior to that of
BGO. Intermediate to both is GSO, which has a higher light
output and a faster decay time than BGO and yet is not as
efficient as BGO or LSO/LYSO in stopping the 511 keV
photons. Table I lists the attributes of different detector material used in commercially available PET/CT scanners. It is
clear that LSO/LYSO has and will become the detector of
choice in the future PET/CT as manufacturers have chosen
either LSO or LYSO as their new PET scintillation material.
III. SPATIAL RESOLUTION OF PET
The limit spatial resolution of clinical PET scanners with
a ring diameter of 80 cm is about 2 mm.96 There are basically five factors that determine the spatial resolution: 1
positron range: the distance between the point of emission of
the positron to the point of annihilation of the positron with
the electron. The higher the kinetic energy of the positron,
the longer is the distance to the site of annihilation. This
distance is depend on the isotope; 2 noncolinearity: deviation from the assumed emissions of two 511 keV photons at
180 deg from the point of annihilation. The larger the ring
diameter is; the more non-colinearity becomes; 3 detector
geometry: smaller detector element of size d leads to better
spatial resolution of size d / 2; 4 photon interaction in the
detector: Compton scattering is the dominate interaction between the 511 keV photon and the detector. Multiple interactions are required to stop the photon in the detector. Some
interactions may even spread over to adjacent detector elements. Typical detector elements are in the size of 4 6 mm
square and 2 3 cm thick.94 The result is some uncertainty
for identifying the first interaction of the photon and the

detector; and 5 reconstruction algorithm that models the
degradation from emission of the positron to detection of the
photons and that is capable of deconvolving the degradation
to improve the resolution. Noncolinearity and detector size
are the predominant factors contributing to the degradation
of spatial resolution for the clinical PET scanner.5 Some preclinical PET scanners have been shown to achieve spatial
resolution of 0.4 mm with a ring diameter of 2 cm and a
detector size of 0.25 mm.98 The spatial resolution of the
clinical PET scanner is about 5 8 mm,99 and can still be
improved in the future. The impact of improving spatial resolution may be more significant in diagnosis than in RT. This
is mainly due to the clinical target volume and planning target volume margins that are added to the GTV thereby reducing the advantage of PET imaging resolution.77
IV. TIME-OF-FLIGHT PET
Conventional reconstruction of PET images assumes that
the annihilation event of a positron and an electron detected
by a pair of detector elements 180-deg apart, is equally likely
to occur at any point along the line defined by this pair of
detector elements. This assumption has served well for many
years when detectors with short decay time, high light output, and fast electronics were not available. The short decay
time of new detector material such as LSO and LYSO combined with fast electronics has enabled time-of-flight TOF
PET. The timing resolution to locate an annihilation event to
within 5 mm of PET resolution is approximately 30 ps.96
The current commercial TOF PET has the timing resolution
of about 600 ps Ref. 100 and can locate an annihilation
event to within 18 cm which in turn improves contrast. The
gain in signal to noise ratio in PET images with TOF is
proportional to the object size and inversely proportional to
the timing resolution.100 Current technology of 600 ps timing
resolution shows improvement of signal to noise ratio for
large patients. The future improvement of timing resolution
will further reduce noise and improve the signal to noise
ratio for all patients. This benefit can also be translated into
shorter scan time or reduced injected radiopharmaceutical
dose to the patient and reduced radiation exposure to the
technologist.
V. TWO-DIMENSIONAL AND THREE-DIMENSIONAL
PET IMAGING
PET data can be acquired in two-dimensional 2D or
three-dimensional 3D mode. In 2D mode, a collimation
septa is positioned between detector rings in the axial di-
4958
4958
rection. The collimation minimizes the scatter and provides a

uniform sensitivity profile along the cranial-caudal direction.
An overlap of only 1 cm is needed to compensate for the
lower sensitivity of the edge slices between two bed positions in order to provide a uniform sensitivity along the axial
direction in a whole body PET/CT scan. On the other hand,
3D PET imaging does not utilize septa during data acquisition. It has higher sensitivity but more randoms and scatters
than 2D PET. Due to its nonuniform sensitivity profile high
at the center of the detector with a linear drop to the edge of
the detector, it normally requires up to 50% detector overlap
between two bed positions. With the improvement of 3D
reconstruction techniques101 and better modeling of the randoms and scatters,102 it has been shown that 3D PET results
in better image quality than 2D PET with less injection of
18
F-FDG.103 This has resulted in lower radiation exposure to
the patients, nursing staff and technologists. Therefore, 3D
PET imaging is bound to become the standard mode of imaging in the future. Some manufacturers of PET/CT scanners
provide 3D only capabilities on their systems.
motion. 4D CT may be needed for accurate quantification of

4D PET data as each phase of PET data may need its own
CT data for attenuation correction.108 Although, it is still
cumbersome to acquire 4D PET data due to its prolonged
acquisition time, it is expected that with the new LSO/LYSO
detector, large detector coverage for higher sensitivity, and
the advent of 3D image reconstruction to better cope with the
increased noise from randoms and scatters in 3D data acquisition, the limitation in acquisition time will be minimized
and 4D PET will become a clinically feasible solution to
improving the quantification accuracy of tumor in motion.
VI. FOUR-DIMENSIONAL PET IMAGING

Four-dimensional 4D PET was first developed for cardiac imaging to assess myocardial motion and to obtain ejection fraction.104 It has been adopted for tumor imaging of the
thorax in the last several years.105107 When data are acquired
in 4D or gated mode, the data are split into several exclusive
bins. For example, there can be 8 bins of 500 ms for an
average respiratory cycle of 4 s. Because of insufficient statistics of photons obtained in PET imaging of 3 6 min for
the 8 bins in each bed location, the duration of a 4D PET
scan is normally prolonged to over 10 min to compensate for
the fewer photons recorded in each bin. Image reconstruction
is performed on the data of each bin and the result is a set of
3D PET images over a respiratory cycle for the assessment
of tumor motion and quantification. Even though, the number of photons in each bin is small, resulting in higher noise
in the 4D images; the 4D images can potentially be used for
accurate assessment of FDG uptake.108 4D PET has not taken
off due to its complexity and limited workflow in which 4D
PET had to be performed in a separate session after a routine
PET/CT scan on most of the PET/CT scanners, prolonging
the acquisition time and impacting patient comfort. In our
opinion, an ideal PET/CT imaging session should take less
than 20 min. Any duration beyond that will increase the potential for patient motion.
PET/CT scanners have recently been equipped with listmode data acquisition whereby events from each coincidence
pair of 511 keV photons are stored in a list stream for later
reconstruction. It has been demonstrated that the list-mode
data acquisition can be performed with either cardiac or respiratory triggering during a normal static image
acquisition.109 This functionality offers the capability of retrospectively sorting the coincidence events into multiple
phases/bins for the reconstruction of 4D PET images. Current PET scanners can be configured to acquire data which
can produce both static and 4D PET data to freeze tumor
VII. LARGE BORE PET/CT

Most current commercially available PET/CT scanners
have a 70 cm bore size for both PET and CT scanners. Early
models had a smaller bore size in PET than in CT.110 However, it is clear that the bore size will have to increase to
accommodate PET/CT simulation of the breast patients and
some large patients. The CT scanner of 85 cm bore size was
introduced in 2000 to address this issue.111 Today most of the
PET/CT scanners have 50 cm standard field of view in CT
image reconstruction with up to 60 65 cm extended field of
view reconstruction.110 This was accomplished by extrapolating the truncated projection data in CT normally in the
anterior-posterior views to match the total attenuation measured in the none-truncated projection data in the lateral
views.112114 This approach may not provide accurate CT
numbers for RT. A correct solution to the truncation problem
would be to increase the detector fan angle to avoid truncation by increasing the number of detector elements. One
manufacturer has increased the detector size to better control
the truncation problem up to 70 cm, and one manufacturer
has offered a PET/CT with 85 cm bore size for both PET and
CT. It is expected that bore size of 85 cm should become a
standard and the CT detector size should be increased to
accommodate up to 70 cm CT image reconstruction without
truncation for RT.
VIII. MULTISLICE CT
PET needs CT data for tumor localization, attenuation
correction, and quantification. PET acquires data over
3 6 min for a 15-cm superior-inferior coverage.115 With a
clinical whole-body scan from base of the skull to midthigh
of about 100 cm, the acquisition time for PET data is about
18 21 min for 3 min/bed in 2D data acquisition. To improve
the CT image quality of the thorax and the abdomen, the
patient is normally asked to raise his/her arms over the head
during data acquisition of the CT and the PET data. In this
position, an average person may not be able to remain stationary for over 20 min. Figure 1 demonstrates an example
of patient motion in the middle of a scan. If patient motion is
not accounted for, the PET images may not be reliable for
diagnosis. Incorporation of CT has greatly improved the
throughput of PET/CT scan and patient comfort when a CT
scan of 100 cm can be performed in less than 20 s on a
modern 16-slice PET/CT with good image quality.
4959
4959
FIG. 1. Effect of a patient motion at the 9th min during a single bed acquisition of 15 min. The CT before PET scan and PET images of the first 9 min were
in a and b, respectively. The CT after PET scan and PET images of the last 6 min were in c and d, respectively. The mean SUVs of the region of
interest pointed by the arrow were 2.7 in b, 1.2 in d and 2.1 in e of total 15 min acquisition.
The first prototype PET/CT was with a single-slice CT

SSCT,1 which was the major CT scanner in diagnostic radiology at the time. A CT scan would take 5 10 min with
this prototype PET/CT.10 Today almost all the PET/CT scanners are with multislice CT MSCT, and the number of detector channels or slices has increased from 2 to 64 in the last
several years. In general, 16-slice CT is sufficient for tumor
imaging and 64-slice CT is necessary for coronary artery CT
imaging of the heart. A 64-slice CT can provide high spatial
resolution of 0.5 mm in 3D, fast temporal resolution of
100 200 ms at a gantry rotation cycle time of less than
0.35 s, and short breath-hold time of less than 10 s, critical
to coronary artery imaging. One vendor has made available
320-slice CT of 16 cm coverage to image the heart in a
single gantry rotation. Perfusion CT, which requires a cine
scan of over 30 50 s at the same location for assessing the
function of tumor by analyzing the dynamics of injected iodine contrast in and out of the tumor, may require more than
64 slices for its expanded coverage of more than 4 cm. A
16-slice CT covers only 2 2.4 cm per gantry rotation. Incorporation of MSCT in PET/CT has and will have a profound
impact on improving the registration between the CT and the
PET data, and the assessment of tumor motion in 4D CT
imaging. Both applications are critical to RT.

IX. AVERAGE CT
The issue of potential misalignment between the CT and
the PET data in the thorax and the abdomen was discovered
soon after PET/CT was introduced, and has become the most
widely addressed issue in PET/CT.5,65,79,80,85,86,89,116119
MSCT normally acquires data at a very high temporal resolution of less than 1 s, while PET acquires data in several
min.65,85 Mismatch in temporal resolution leads to a potential
mismatch of the tumor positions between the CT and the
PET data, and compromises the quantification of the PET
data and the determination of PET GTV. Current design of
PET/CT only matched the spatial resolutions of CT and PET
by blurring the CT images to match the PET images in spatial resolution.94 No attempt has been made to match the
temporal resolutions of CT and PET.
Since the PET data is time averaged, it was recognized
that PET data can be a more accurate representation of the
3D volume encompassing the tumor motion.120 However,
quantification of the PET data relies on the CT data. If the
CT data are not matched with the PET data in position, quan-
4960
4960
(A)
(A)
(B)
(B)
FIG. 2. Effect of misalignment to the assessment of lymph node 18F-FDG

uptake. The CT, PET corrected with the CT, and fused PET-CT images with
misalignment were in a from left to right. The ACT, PET corrected with
the ACT, and fused PET-CT images in b. The maximum SUV of the lymph
node identified by a cross was 4.6 in a and 7.5 in b. The misalignment
caused the change of maximum SUV by 63%.
tification of the PET data will be compromised. The impact

becomes severe when the tumor is near the diaphragm such
as some lung, liver, and esophageal tumors.
Mismatch between the CT and PET data can be identified
by a curvelinear cold artifact paralleling the dome of the
diaphragm at the lung bases or photopenic region in the PET
images. We normally spend more time in exhale than in inhale. The PET data averaged over several minutes is closer to
the end-expiration than end-inspiration phase. If the CT data
are acquired in or near the end-inspiration phase, the inflated
lung due to inspiration will be bigger than the deflated lung.
The larger area of the inflated lungs in CT renders less attenuation correction in the reconstruction of the PET data
resulting in a photopenic region identified as the misaligned
region. Several measures have been proposed to mitigate this
problem. Coaching patient to hold-breath at midexpiration
during CT acquisition was suggested,62 and the outcomes
were mixed because coaching patient to hold breath at certain state may not be reliable both from the patient and the
technologist operating the PET/CT scanner perspectives. Another approach to improve registration between the CT and
the PET data is to bring the temporal resolution of the CT
images to that of the PET data.80 Recognizing the fact that
PET is averaged over many breath cycles, an average CT
ACT image of about one breath cycle has been shown to
improve registration between the CT and the PET data. Figures 2 and 3 show two examples of misalignemnt between
the CT and the PET data corrected by ACT to improve the
assessment of lymph node 18F-FDG uptake and the accuracy
of tumor targeting for RT. This concept has also been shown
effective in improving registration between the CT and the
PET data in cardiac PET/CT imaging.79
Acquiring ACT can be accomplished by scanning at the
same slice location over a breath cycle at a very high speed
gantry rotation for a subsecond temporal resolution. The CT
images, almost free of motion artifacts, in a respiratory cycle
are averaged for ACT. The conventional approach with slowscan CT for ACT is ineffective and should be
(C)
FIG. 3. The PET/CT images of a 69 year-old female patient with an esophageal tumor after an induction chemotherapy. a shows an axial slice of the
fused clinical CT and PET image at the level of the esophageal tumor left
and the PET image in coronal view right. The radiology report indicated
the patient had a positive response to the chemotherapy. After removal of
misalignment by the AVG CT, the tumor reappeared in the same PET data
set in b. The arrows point to the tumor location. The gross target volumes
drawn in the images of a and b are shown in blue and in green, respectively, in c. The patient was treated with the tumor volume in green, and
the radiology report was corrected by the AVG CT.
discouraged.79,121 Figure 4 shows a clinic example of the

same patient scanned with slow-scan CT of 4 s and ACT of
fast gantry rotation of 0.5 s for 4 s. Cine-CT and low-pitch
helical CT pitch 0.1 scans can be adopted to obtain ACT
and both have been utilized in 4D CT imaging.122124 However, it is not clear whether the setup of 4D CT imaging for
the assessment of tumor motion is ideal for obtaining ACT
when the majority of PET/CT scanners are in nuclear medicine and are without a respiratory monitoring device for 4D
CT. A practical approach is to acquire ACT without a respiratory monitoring device to improve quantification of the
PET data.79,125 The additional radiation dose for ACT was
estimated to be from 5 to 50 mGy and the additional scan
time is less than a couple of minutes and will not significantly impact the overall scan time of a PET/CT procedure.
ACT can also be derived from 4D CT, incorporation of ACT
for attenuation correction of the PET data should be made
available in the future. ACT can also be used for dose calculation due to its time-averaging of CT numbers.126 Selection
of optimum parameters for acquiring ACT with cine-CT or
low-pitch helical CT has also been investigated.88
MSCT can acquire the ACT data over the chest and the
abdomen in less than a couple of minutes due to its large
detector coverage over SSCT. It can also provide a sequence
of almost motion-free cine CT images at the same slice location due to its subsecond gantry rotation. The marriage of
MSCT and PET in PET/CT thus has provided a platform for
4961
4961
FIG. 4. The RACT and the SSCT images of a patient with an average breath cycle of 4 s. The SSCT images in a were taken with one single CT gantry
rotation of 4 s, and the two images were 2.5 mm apart and 2.5 mm thick. The corresponding RACT images in b, obtained by averaging the cine CT images,
were averaged from 4 s of data collection over eight gantry rotations. The ACT images were almost free of reconstruction artifacts, which were observed in
the SSCT images reproduced from Fig. 11 of Ref. 79.
registration of the CT and the PET data, directly impacting

RT.
X. MAXIMUM INTENSITY PROJECTION CT

Maximum intensity projection MIP CT images can be
derived by finding the maximum pixel value at each pixel
location from all the phases of 4D CT Ref. 123 or cine CT
images.125 It has been shown that MIP CT images are effective in depicting the extent of tumor motion.127,128 For peripheral lung tumors surrounded by the lower density air in
the lungs, MIP CT images can be used to assist the determination of the tumor target volume and to avoid ambiguity
in using a threshold of SUV to determine the target volume
in PET. Any 18F-FDG uptake in the lungs should be supported by tissues with or without motion. Any PET GTV
determined with a threshold should not exceed the boundary
in the MIP images. Both MIP and ACT images can be derived without gating.79,125 Their application for treatment
planning for stereotactic RT has been demonstrated.129 Their
applications for PET/CT in RT are expected to become important as they can be applicable to most of the PET/CT
scanners without the hardware setup for 4D CT. MIP CT can
help determine the tumor volume in the thorax. Figure 5
shows an example of determining the extent of PET GTV
with MIP CT. A similar concept has been attempted with
regular helical CT data for peripheral NSCLC by Biehl et
al.,130 when the boarder of the tumor can be identified by the
helical CT data.
XI. 4D CT IMAGING
4D CT Refs. 122124 has found its acceptance in RT for
providing the gated CT images of multiple phases over a
respiratory cycle to assist contouring the extent of tumor
motion. 4D CT takes less time in acquisition than 4D PET
does. It normally takes less than 2 min like ACT to cover
40 cm in the superior-inferior direction on an 8- or 16-slice
CT. Two data acquisition modes have been used in 4D CT.
One is cine CT and the other one is low-pitch helical CT.
Both acquisitions scan any point in space for over one breath
cycle plus one or 2 / 3 gantry rotation cycle. Cine CT uses
less radiation and generates thinner slices than low-pitch helical CT. On the other hand, low-pitch helical CT has an
advantage over cine CT in acquisition time.131 The cine CT
based 4D CT is available on 4, 8, 16, and 64-slice CT scanners, and the low pitch helical CT based 4D CT is only
available on newer 16- and 64-slice CT scanners. Multislice
CT has significantly improved the speed of 4D CT data acquisition and made 4D CT clinically acceptable. It is expected that 4D CT will become an integral part of PET/CT
imaging of the thorax and 4D CT will be incorporated in the
quantification of 4D PET as each phase of gated PET image
needs its own gated CT image for attenuation
correction.108,132134 The ACT and MIP images can be derived from 4D CT or directly from cine CT.125
XII. DETERMINATION OF TUMOR CONTOUR IN
PET
Determination of tumor volume in PET image is normally
performed with SUV thresholds. This is an important yet
4962
4962
FIG. 5. The MIP CT and PET images of a patient. The MIP CT in a is displayed with window, level = 400, 40 and PET in b with the highest
intensity= 40% of the maximum SUV in the tumor. The images of a and b are displayed again with 1000, 700 in c and the highest density= 20% in
d, respectively. The tumor contour in c is superimposed in a, b, and d to demonstrate the effect on tumor size by display parameters.
controversial issue, in particular for the lung tumors impacted by the respiratory motion. To determine the gross tumor volume in PET, it is generally based on the visual interpretation by the experienced nuclear medicine physician.35
Some suggested an absolute threshold of 2.5,37,135 or a
threshold between 15% and 50% of the maximum SUV
mSUV.27,35,37,136138 Recent studies by Biehl et al.130 and
Nestle et al.78 suggested that no single SUV can reliably be
used for segmentation of the PET tumor volume, and that an
individualized threshold should be derived in consideration
of the size, location, nonuniform distribution of 18F-FDG
activity of the tumor. Most of the discussions have been with
the data of stand-alone PET. The issue may be further complicated in PET/CT. Any misalignment can further compromise the quantification of the PET data. Erdi et al.132 reported that different phases of 4D CT image can cause up to
24% variation of mSUV in the PET data. Pan et al.80 observed a variation of more than 50% mSUV between the
PET data with ACT and the same PET data with helical CT.
Chi et al.90 investigated the effects of misalignment to the
determination of PET GTV and found that GTV centroid
location could shift by 2.4 mm and GTV volume change
could be as high as 154% for the tumors of less than 50 cm3.
Partial volume also impacts the determination of SUV.
Partial volume is caused by the limited spatial resolution of
PET scanners 5 8 mm, and renders a blurry look of the
PET image. Partial volume mostly affects smaller tumor
sizes of less than 1 1.6 cm or 2 times the PET resolution. In
addition, studies have shown that SUV is a function of uptake time between injection of 18F-FDG and data
acquisition.61,139,140 Care should be taken to ensure the same
uptake time for the same patient imaged in a multiple of
PET/CT sessions during the course of diagnosis, treatment,
and staging. Tumor volume determined by PET may be more
indicative of the GTV plus motion since PET data are acquired over several min. PET GTV delineation has been and
will be a very active area of research. Input from MIP and

4D CT images will be very important and complementary to
this endeavor. Contouring a tumor in PET/CT images should
be an integral process with complementary inputs of both
PET and CT images.
XIII. WORKFLOW
PET/CT simulation of the chest and abdomen poses a
challenge for registration of the CT and the PET data. We
can acquire the list mode PET data with the physiologic triggers of respiratory or cardiac nature. Two sets of PET data
can be reconstructed: one to generate the static PET data for
an initial whole-body diagnosis; and the other to generate a
gated or 4D PET data over a certain anatomical region using
the list mode data and physiologic triggers. Static PET data
should be attenuation corrected with ACT. 4D PET data
should be corrected with 4D CT data. The images for radiation therapy should be free of misalignment and truncation
artifacts which can be accomplished using a large bore
PET/CT and a large detector CT, respectively. All the data
acquisition should be performed in 20 min to improve patient comfort and avoid potential patient motion.
XIV. SUMMARY
There has been a tremendous growth in the use of
PET/CT scanners over the past several years. The combination of faster detector of LSO/LYSO and faster electronics
has made the TOF PET/CT feasible in improving the image
quality of PET. The improvement in 3D reconstruction with
accurate modeling of randoms and scatters has also made 3D
PET more attractive than 2D PET. The benefits of 3D include
the reduction of both the injected radiopharmaceutical dose
to the patient and the radiation exposure to the nursing staff
and the technologists. The large bore PET/CT and the large
detector CT will make the positioning of a large patient in
4963
4963
the scanner much easier and minimize the truncation of CT

images, respectively. Application of PET/CT for radiation
therapy is expected to grow albeit there are some challenges.
Reimbursement would be feasible when more clinical data to
support PET/CT simulation become available. A similar
challenge was experienced during the adaptation of CT simulation in the 1990s when the radiation oncologists were not
familiar with GTV delineation on the CT images. With the
development of 4D CT, the difficulties in the simulation of
the thorax and abdomen have largely been minimized. Further improvement in registration of the PET and CT data by
ACT, MIP, 4D CT, and 4D PET will improve the registration
in the thorax and the abdomen. Furthermore, data acquisition
should be less than 20 min to simulate the patient condition
in a treatment session.
Davis, U. M. Lutolf, T. F. Hany, and I. F. Ciernik, PET/CT staging

followed by intensity-modulated radiotherapy IMRT improves treatment
outcome of locally advanced pharyngeal carcinoma: A matched-pair comparison, Radiat. Oncol. Invest. June 222 2007.
17
D. Wang et al., Initial experience of FDG-PET/CT guided IMRT of
head-and-neck carcinoma, Int. J. Radiat. Oncol., Biol., Phys. 65, 143
151 2006.
18
A. C. Riegel, A. M. Berson, S. Destian, T. Ng, L. B. Tena, R. J. Mitnick,
and P. S. Wong, Variability of gross tumor volume delineation in headand-neck cancer using CT and PET/CT fusion, Int. J. Radiat. Oncol.,
Biol., Phys. 65, 726732 2006.
19
I. F. Ciernik, E. Dizendorf, B. G. Baumert, B. Reiner, C. Burger, J. B.
Davis, U. M. Lutolf, H. C. Steinert, and G. K. Von Schulthess, Radiation
treatment planning with an integrated positron emission and computer
tomography PET/CT: A feasibility study, Int. J. Radiat. Oncol., Biol.,
Phys. 57, 853863 2003.
20
M. R. Vernon, M. Maheshwari, C. J. Schultz, M. A. Michel, S. J. Wong,
B. H. Campbell, B. L. Massey, J. F. Wilson, and D. Wang, Clinical
outcomes of patients receiving integrated PET/CT-guided radiotherapy
for head and neck carcinoma, Int. J. Radiat. Oncol., Biol., Phys. 70,
678684 2008.
21
H. Ashamalla, A. Guirgius, E. Bieniek, S. Rafla, A. Evola, G. Goswami,
R. Oldroyd, B. Mokhtar, and K. Parikh, The impact of positron emission
tomography/computed tomography in edge delineation of gross tumor
volume for head and neck cancers, Int. J. Radiat. Oncol., Biol., Phys. 68,
388395 2007.
22
P. H. Ahn and M. K. Garg, Positron emission tomography/computed
tomography for target delineation in head and neck cancers, Semin Nucl.
Med. 38, 141148 2008.
23
B. Dietl and J. Marienhagen, The therapeutic impact of 18F-FDG
whole body PET. A radiooncologists view, Nuklearmedizin 44, 814
2005.
24
X. K. Zheng, L. H. Chen, Q. S. Wang, H. B. Wu, H. M. Wang, Y. Q.
Chen, W. P. Yan, Q. S. Li, and Y. K. Xu, Influence of FDG-PET on
computed tomography-based radiotherapy planning for locally recurrent
nasopharyngeal carcinoma, Int. J. Radiat. Oncol., Biol., Phys. 69, 1381
1388 2007.
25
R. H. Ireland, K. E. Dyker, D. C. Barber, S. M. Wood, M. B. Hanney, W.
B. Tindale, N. Woodhouse, N. Hoggard, J. Conway, and M. H. Robinson,
Nonrigid image registration for head and neck cancer radiotherapy treatment planning with PET/CT, Int. J. Radiat. Oncol., Biol., Phys. 68,
952957 2007.
26
D. L. Schwartz et al., FDG-PET/CT-guided intensity modulated head
and neck radiotherapy: A pilot investigation, Head Neck 27, 478487
2005.
27
L. J. Vanuytsel, J. F. Vansteenkiste, S. G. Stroobants, P. R. De Leyn, W.
De Wever, E. K. Verbeken, G. G. Gatti, D. P. Huyskens, and G. J.
Kutcher, The impact of 18F-fluoro-2-deoxy-D-glucose positron emission tomography FDG-PET lymph node staging on the radiation treatment volumes in patients with non-small cell lung cancer, Radiother.
Oncol. 55, 317324 2000.
28
J. Y. Chang, L. Dong, H. Liu, G. Starkschall, P. Balter, R. Mohan, Z.
Liao, J. D. Cox, and R. Komaki, Image-guided radiation therapy for
non-small cell lung cancer, J. Thorac. Oncol. 3, 177186 2008.
29
H. Ashamalla, S. Rafla, K. Parikh, B. Mokhtar, G. Goswami, S. Kambam,
H. Abdel-Dayem, A. Guirguis, P. Ross, and A. Evola, The contribution
of integrated PET/CT to the evolving definition of treatment volumes in
radiation treatment planning in lung cancer, Int. J. Radiat. Oncol., Biol.,
Phys. 63, 10161023 2005.
30
E. Brianzoni et al., Radiotherapy planning: PET/CT scanner performances in the definition of gross tumour volume and clinical target volume, Eur. J. Nucl. Med. Mol. Imaging 32, 13921399 2005.
31
D. De Ruysscher et al., Effects of radiotherapy planning with a dedicated combined PET-CT-simulator of patients with non-small cell lung
cancer on dose limiting normal tissues and radiation dose-escalation: A
planning study, Radiother. Oncol. 77, 510 2005.
32
A. H. Klopp, J. Y. Chang, S. L. Tucker, E. P. Sulman, P. A. Balter, H. H.
Liu, M. K. Bucci, H. A. Macapinlac, R. Komaki, and J. D. Cox, Intrathoracic patterns of failure for non-small-cell lung cancer with
positron-emission tomography/computed tomography-defined target delineation, Int. J. Radiat. Oncol., Biol., Phys. 69, 14091416 2007.
33
C. Greco, K. Rosenzweig, G. L. Cascini, and O. Tamburrini, Current
status of PET/CT for tumour volume definition in radiotherapy treatment
Author to whom correspondence should be addressed. Present address:

The University of Texas, M.D. Anderson Cancer Center, Department of
Imaging PhysicsUnit 56, 1515 Holcome Blvd., Houston, TX 77030.
[email protected]
1
T. Beyer, D. W. Townsend, T. Brun, P. E. Kinahan, M. Charron, R.
Roddy, J. Jerin, J. Young, L. Byars, and R. Nutt, A combined PET/CT
scanner for clinical oncology, J. Nucl. Med. 41, 13691379 2000.
2
D. W. Townsend and T. Beyer, A combined PET/CT scanner: The path
to true image fusion, Br. J. Radiol. 75, S24S30 2002.
3
B. F. Hutton and M. Braun, Software for image registration: Algorithms,
accuracy, efficacy, Semin Nucl. Med. 33, 180192 2003.
4
J. West et al., Comparison and evaluation of retrospective intermodality
brain image registration techniques, J. Comput. Assist. Tomogr. 21, 554
566 1997.
5
D. W. Townsend, Positron emission tomography/computed tomography, Semin Nucl. Med. 38, 152166 2008.
6
R. L. Wahl, L. E. Quint, R. D. Cieslak, A. M. Aisen, R. A. Koeppe, and
C. R. Meyer, Anatometabolic tumor imaging: Fusion of FDG PET with
CT or MRI to localize foci of increased activity, J. Nucl. Med. 34,
11901197 1993.
7
P. Reinartz, F. J. Wieres, W. Schneider, A. Schur, and U. Buell, Side-byside reading of PET and CT scans in oncology: Which patients might
profit from integrated PET/CT?, Eur. J. Nucl. Med. Mol. Imaging 31,
14561461 2004.
8
E. M. Rohren, T. G. Turkington, and R. E. Coleman, Clinical applications of PET in oncology, Radiology 231, 305332 2004.
9
S. S. Gambhir, J. Czernin, J. Schwimmer, D. H. Silverman, R. E. Coleman, and M. E. Phelps, A tabulated summary of the FDG PET literature, J. Nucl. Med. 42, 1S93S 2001.
10
P. G. Kluetz, C. C. Meltzer, V. L. Villemagne, P. E. Kinahan, S. Chander,
M. A. Martinelli, and D. W. Townsend, Combined PET/CT imaging in
oncology. Impact on patient management, Clin Positron Imaging 3, 223
230 2000.
11
T. F. Hany, H. C. Steinert, G. W. Goerres, A. Buck, and G. K. von
Schulthess, PET diagnostic accuracy: Improvement with in-line PET-CT
system: Initial results, Radiology 225, 575581 2002.
12
R. Bar-Shalom, N. Yefremov, L. Guralnik, D. Gaitini, A. Frenkel, A.
Kuten, H. Altman, Z. Keidar, and O. Israel, Clinical performance of
PET/CT in evaluation of cancer: Additional value for diagnostic imaging
and patient management, J. Nucl. Med. 44, 12001209 2003.
13
D. Lardinois, W. Weder, T. F. Hany, E. M. Kamel, S. Korom, B. Seifert,
G. K. von Schulthess, and H. C. Steinert, Staging of non-small-cell lung
cancer with integrated positron-emission tomography and computed tomography, N. Engl. J. Med. 348, 25002507 2003.
14
V. Gregoire, K. Haustermans, X. Geets, S. Roels, and M. Lonneux, PETbased treatment planning in radiotherapy: A new standard?, J. Nucl.
Med., Suppl. 48, 68S77S 2007.
15
F. Paulsen, J. Scheiderbauer, S. M. Eschmann, K. Brechtel, M. Klein, C.
Pfannenberg, A. Mondry, T. Hehr, C. Belka, and M. Bamberg, First
experiences of radiation treatment planning with PET/CT, Strahlenther.
Onkol. 182, 369375 2006.
16
S. Rothschild, G. Studer, B. Seifert, P. Huguenin, C. Glanzmann, J. B.
4964
4964
planning for non-small cell lung cancer NSCLC, Lung Cancer 57,
125134 2007.
34
G. L. Ceresoli, G. M. Cattaneo, P. Castellone, G. Rizzos, C. Landoni, V.
Gregorc, R. Calandrino, E. Villa, C. Messa, A. Santoro, and F. Fazio,
Role of computed tomography and 18F fluorodeoxyglucose positron
emission tomography image fusion in conformal radiotherapy of nonsmall cell lung cancer: A comparison with standard techniques with and
without elective nodal irradiation, Tubercle 93, 8896 2007.
35
U. Nestle et al., 18F-deoxyglucose positron emission tomography FDGPET for the planning of radiotherapy in lung cancer: High impact in
patients with atelectasis, Int. J. Radiat. Oncol., Biol., Phys. 44, 593597
1999.
36
Y. E. Erdi, K. Rosenzweig, A. K. Erdi, H. A. Macapinlac, Y. C. Hu, L. E.
Braban, J. L. Humm, O. D. Squire, C. S. Chui, S. M. Larson, and E. D.
Yorke, Radiotherapy treatment planning for patients with non-small cell
lung cancer using positron emission tomography PET, Radiother. Oncol. 62, 5160 2002.
37
J. Bradley, W. L. Thorstad, S. Mutic, T. R. Miller, F. Dehdashti, B. A.
Siegel, W. Bosch, and R. J. Bertrand, Impact of FDG-PET on radiation
therapy volume delineation in non-small-cell lung cancer, Int. J. Radiat.
Oncol., Biol., Phys. 59, 7886 2004.
38
F. M. Kong, K. A. Frey, L. E. Quint, R. K. Ten Haken, J. A. Hayman, M.
Kessler, I. J. Chetty, D. Normolle, A. Eisbruch, and T. S. Lawrence, A
pilot study of 18Ffluorodeoxyglucose positron emission tomography
scans during and after radiation-based therapy in patients with non smallcell lung cancer, J. Clin. Oncol. 25, 31163123 2007.
39
M. P. Mac Manus, R. J. Hicks, D. L. Ball, V. Kalff, J. P. Matthews, E.
Salminen, P. Khaw, A. Wirth, D. Rischin, and A. McKenzie, F-18 fluorodeoxyglucose positron emission tomography staging in radical radiotherapy candidates with nonsmall cell lung carcinoma: Powerful correlation with survival and high impact on treatment, Cancer 92, 886895
2001.
40
M. T. Munley, L. B. Marks, C. Scarfone, G. S. Sibley, E. F. Patz, Jr., T. G.
Turkington, R. J. Jaszczak, D. R. Gilland, M. S. Anscher, and R. E.
Coleman, Multimodality nuclear medicine imaging in three-dimensional
radiation treatment planning for lung cancer: Challenges and prospects,
Lung Cancer 23, 105114 1999.
41
E. C. Ford, W. C. Lavely, D. A. Frassica, L. T. Myers, F. Asrari, R. L.
Wahl, and R. C. Zellars, Comparison of FDG-PET/CT and CT for delineation of lumpectomy cavity for partial breast irradiation, Int. J. Radiat. Oncol., Biol., Phys. 71, 595602 2008.
42
C. Everitt and T. Leong, Influence of F-fluorodeoxyglucose-positron
emission tomography on computed tomography-based radiation treatment
planning for oesophageal cancer, Australas Radiol. 50, 271274 2006.
43
T. Leong et al., A prospective study to evaluate the impact of FDG-PET
on CT-based radiotherapy treatment planning for oesophageal cancer,
44
M. Hutchings, A. Loft, M. Hansen, A. K. Berthelsen, and L. Specht,
Clinical impact of FDG-PET/CT in the planning of radiotherapy for
early-stage Hodgkin lymphoma, Eur. J. Haematol. 78, 206212 2007.
45
J. Esthappan, S. Chaudhari, L. Santanam, S. Mutic, J. Olsen, D. M. Macdonald, D. A. Low, A. K. Singh, and P. W. Grigsby, Prospective clinical
trial of positron emission tomography/computed tomography imageguided intensity-modulated radiation therapy for cervical carcinoma with
positive para-aortic lymph nodes, Int. J. Radiat. Oncol., Biol., Phys. in
press.
46
I. F. Ciernik, M. Huser, C. Burger, J. B. Davis, and G. Szekely, Automated functional image-guided radiation treatment planning for rectal
cancer, Int. J. Radiat. Oncol., Biol., Phys. 62, 893900 2005.
47
D. A. Patel, S. T. Chang, K. A. Goodman, A. Quon, B. Thorndyke, S. S.
Gambhir, A. McMillan, B. W. Loo, Jr., and A. C. Koong, Impact of
integrated PET/CT on variability of target volume delineation in rectal
cancer, Technol. Cancer Res. Treat. 6, 3136 2007.
48
I. F. Ciernik, D. W. Brown, D. Schmid, T. Hany, P. Egli, and J. B. Davis,
3D-segmentation of the 18F-choline PET signal for target volume definition in radiation therapy of the prostate, Technol. Cancer Res. Treat. 6,
2330 2007.
49
R. J. Steenbakkers et al., Reduction of observer variation using matched
CT-PET for lung cancer delineation: A three-dimensional analysis, Int. J.
50
J. L. Fox, R. Rengan, W. OMeara, E. Yorke, Y. Erdi, S. Nehmeh, S. A.
Leibel, and K. E. Rosenzweig, Does registration of PET and planning
CT images decrease interobserver and intraobserver variation in delineat-
ing tumor volumes for non-small-cell lung cancer?, Int. J. Radiat. Oncol., Biol., Phys. 62, 7075 2005.
51
D. De Ruysscher et al., Selective mediastinal node irradiation based on
FDG-PET scan data in patients with non-small-cell lung cancer: A prospective clinical study, Int. J. Radiat. Oncol., Biol., Phys. 62, 988994
2005.
52
A. van Baardwijk, B. G. Baumert, G. Bosmans, M. van Kroonenburgh, S.
Stroobants, V. Gregoire, P. Lambin, and D. De Ruysscher, The current
status of FDG-PET in tumour volume definition in radiotherapy treatment
planning, Cancer Treat Res. 32, 245260 2006.
53
R. D. Stewart and X. A. Li, BGRT: Biologically guided radiation
therapyThe future is fast approaching, Med. Phys. 34, 37393751
2007.
54
C. C. Ling, J. Humm, S. Larson, H. Amols, Z. Fuks, S. Leibel, and J. A.
Koutcher, Towards multidimensional radiotherapy MD-CRT: Biological imaging and biological conformality, Int. J. Radiat. Oncol., Biol.,
Phys. 47, 551560 2000.
55
C. C. Ling and X. A. Li, Over the next decade the success of radiation
treatment planning will be judged by the immediate biological response
of tumor cells rather than by surrogate measures such as dose maximization and uniformity, Med. Phys. 32, 21892192 2005.
56
K. S. Chao, W. R. Bosch, S. Mutic, J. S. Lewis, F. Dehdashti, M. A.
Mintun, J. F. Dempsey, C. A. Perez, J. A. Purdy, and M. J. Welch, A
novel approach to overcome hypoxic tumor resistance: Cu-ATSM-guided
intensity-modulated radiation therapy, Int. J. Radiat. Oncol., Biol., Phys.
49, 11711182 2001.
57
Z. Lin, J. Mechalakos, S. Nehmeh, H. Schoder, N. Lee, J. Humm, and C.
C. Ling, The influence of changes in tumor hypoxia on dose-painting
treatment plans based on 18F-FMISO positron emission tomography,
58
I. Serganova, J. Humm, C. Ling, and R. Blasberg, Tumor hypoxia imaging, Clin. Cancer Res. 12, 52605264 2006.
59
K. Mah, C. B. Caldwell, Y. C. Ung, C. E. Danjoux, J. M. Balogh, S. N.
Ganguli, L. E. Ehrlich, and R. Tirona, The impact of 18FDG-PET on
target and critical organs in CT-based treatment planning of patients with
poorly defined non-small-cell lung carcinoma: A prospective study, Int.
J. Radiat. Oncol., Biol., Phys. 52, 339350 2002.
60
K. R. Zasadny and R. L. Wahl, Standardized uptake values of normal
tissues at PET with 2-fluorine-18-fluoro-2-deoxy-D-glucose: Variations
with body weight and a method for correction, Radiology 189, 847850
1993.
61
J. W. Keyes, Jr., SUV: Standard uptake or silly useless value?, J. Nucl.
Med. 36, 18361839 1995.
62
G. W. Goerres, C. Burger, M. R. Schwitter, T. N. Heidelberg, B. Seifert,
and G. K. von Schulthess, PET/CT of the abdomen: Optimizing the
patient breathing pattern, Eur. Radiol. 13, 734739 2003.
63
T. Beyer, G. Antoch, T. Blodgett, L. F. Freudenberg, T. Akhurst, and S.
Mueller, Dual-modality PET/CT imaging: The effect of respiratory motion on combined image quality in clinical oncology, Eur. J. Nucl. Med.
Mol. Imaging 30, 588596 2003.
64
G. W. Goerres, C. Burger, E. Kamel, B. Seifert, A. H. Kaim, A. Buck, T.
C. Buehler, and G. K. Von Schulthess, Respiration-induced attenuation
artifact at PET/CT: Technical considerations, Radiology 226, 906910
2003.
65
T. Beyer, S. Rosenbaum, P. Veit, J. Stattaus, S. P. Muller, F. P. Difilippo,
H. Schoder, O. Mawlawi, F. Roberts, A. Bockisch, and H. Kuhl, Respiration artifacts in whole-body 18F-FDG PET/CT studies with combined
PET/CT tomographs employing spiral CT technology with 1 to 16 detector rows, Eur. J. Nucl. Med. Mol. Imaging 32, 14291439 2005.
66
G. W. Goerres, T. F. Hany, E. Kamel, G. K. von Schulthess, and A. Buck,
Head and neck imaging with PET and PET/CT: Artefacts from dental
metallic implants, Eur. J. Nucl. Med. Mol. Imaging 29, 367370 2002.
67
G. W. Goerres, D. T. Schmid, and G. K. Eyrich, Do hardware artefacts
influence the performance of head and neck PET scans in patients with
oral cavity squamous cell cancer?, Dentomaxillofac Radiol. 32, 365371
2003.
68
E. M. Kamel, C. Burger, A. Buck, G. K. von Schulthess, and G. W.
Goerres, Impact of metallic dental implants on CT-based attenuation
correction in a combined PET/CT scanner, Eur. Radiol. 13, 724728
2003.
69
S. Bujenovic, F. Mannting, R. Chakrabarti, and D. Ladnier, Artifactual
2-deoxy-2-18Ffluoro-D-glucose localization surrounding metallic objects in a PET/CT scanner using CT-based attenuation correction, Mol.
4965
4965
Imaging Biol. 5, 2022 2003.

F. P. DiFilippo and R. C. Brunken, Do implanted pacemaker leads and
ICD leads cause metal-related artifact in cardiac PET/CT?, J. Nucl. Med.
46, 436443 2005.
71
G. W. Goerres, S. I. Ziegler, C. Burger, T. Berthold, G. K. Von Schulthess, and A. Buck, Artifacts at PET and PET/CT caused by metallic hip
prosthetic material, Radiology 226, 577584 2003.
72
A. K. Berthelsen, S. Holm, A. Loft, T. L. Klausen, F. Andersen, and L.
Hojgaard, PET/CT with intravenous contrast can be used for PET attenuation correction in cancer patients, Eur. J. Nucl. Med. Mol. Imaging 32,
11671175 2005.
73
C. Cohade, M. Osman, Y. Nakamoto, L. T. Marshall, J. M. Links, E. K.
Fishman, and R. L. Wahl, Initial experience with oral contrast in PET/
CT: Phantom and clinical studies, J. Nucl. Med. 44, 412416 2003.
74
M. P. Mac Manus, R. J. Hicks, J. P. Matthews, A. McKenzie, D. Rischin,
E. K. Salminen, and D. L. Ball, Positron emission tomography is superior to computed tomography scanning for response-assessment after radical radiotherapy or chemoradiotherapy in patients with non-small-cell
lung cancer, J. Clin. Oncol. 21, 12851292 2003.
75
R. J. Hicks, M. P. Mac Manus, J. P. Matthews, A. Hogg, D. Binns, D.
Rischin, D. L. Ball, and L. J. Peters, Early FDG-PET imaging after
radical radiotherapy for non-small-cell lung cancer: Inflammatory
changes in normal tissues correlate with tumor response and do not confound therapeutic response evaluation, Int. J. Radiat. Oncol., Biol., Phys.
60, 412418 2004.
76
M. Macmanus, I. DCosta, S. Everitt, J. Andrews, T. Ackerly, D. Binns,
E. Lau, D. Ball, L. Weih, and R. J. Hicks, Comparison of CT and positron emission tomography/CT coregistered images in planning radical
radiotherapy in patients with non-small-cell lung cancer, Australas. Radiol. 51, 386393 2007.
77
J. Brunetti, A. Caggiano, B. Rosenbluth, and C. Vialotti, Technical aspects of positron emission tomography/computed tomography fusion
planning, Semin Nucl. Med. 38, 129136 2008.
78
U. Nestle, S. Kremp, A. Schaefer-Schuler, C. Sebastian-Welsch, D. Hellwig, C. Rube, and C. M. Kirsch, Comparison of different methods for
delineation of 18F-FDG PET-positive tissue for target volume definition
in radiotherapy of patients with non-Small cell lung cancer, J. Nucl.
Med. 46, 13421348 2005.
79
T. Pan, O. Mawlawi, D. Luo, H. H. Liu, P. C. Chi, M. V. Mar, G. Gladish,
M. Truong, J. Erasmus, Jr., Z. Liao, and H. A. Macapinlac, Attenuation
correction of PET cardiac data with low-dose average CT in PET/CT,
Med. Phys. 33, 39313938 2006.
80
T. Pan, O. Mawlawi, S. A. Nehmeh, Y. E. Erdi, D. Luo, H. H. Liu, R.
Castillo, R. Mohan, Z. Liao, and H. A. Macapinlac, Attenuation correction of PET images with respiration-averaged CT images in PET/CT, J.
Nucl. Med. 46, 14811487 2005.
81
J. A. Nye, F. Esteves, and J. R. Votaw, Minimizing artifacts resulting
from respiratory and cardiac motion by optimization of the transmission
scan in cardiac PET/CT, Med. Phys. 34, 19011906 2007.
82
T. Ishimori, T. Saga, Y. Nagata, Y. Nakamoto, T. Higashi, M. Mamede, T.
Mukai, Y. Negoro, T. Aoki, M. Hiraoka, and J. Konishi, 18F-FDG and
11C-methionine PET for evaluation of treatment response of lung cancer
after stereotactic radiotherapy, Ann. Nucl. Med. 18, 669674 2004.
83
G. J. Cook, E. A. Wegner, and I. Fogelman, Pitfalls and artifacts in
18FDG PET and PET/CT oncologic imaging, Semin Nucl. Med. 34,
122133 2004.
84
A. Martinez-Moller, M. Souvatzoglou, N. Navab, M. Schwaiger, and S.
G. Nekolla, Artifacts from misaligned CT in cardiac perfusion PET/CT
studies: Frequency, effects, and potential solutions, J. Nucl. Med. 48,
188193 2007.
85
T. Beyer, G. Antoch, S. Muller, T. Egelhof, L. S. Freudenberg, J. Debatin,
and A. Bockisch, Acquisition protocol considerations for combined
PET/CT imaging, J. Nucl. Med., Suppl. 45, 25S35S 2004.
86
K. L. Gould, T. Pan, C. Loghin, N. P. Johnson, A. Guha, and S. Sdringola,
Frequent diagnostic errors in cardiac PET/CT due to misregistration of
CT attenuation and emission PET images: A definitive analysis of causes,
consequences, and corrections, J. Nucl. Med. 48, 11121121 2007.
87
K. L. Gould, T. Pan, C. Loghin, N. P. Johnson, and S. Sdringola, Reducing radiation dose in rest-stress cardiac PET/CT by single poststress cine
CT for attenuation correction: Quantitative validation, J. Nucl. Med. in
press.
88
P. C. Chi, O. Mawlawi, S. A. Nehmeh, Y. E. Erdi, P. A. Balter, D. Luo, R.
Mohan, and T. Pan, Design of respiration averaged CT for attenuation
correction of the PET data from PET/CT, Med. Phys. 34, 20392047
2007.
89
M. M. Osman, C. Cohade, Y. Nakamoto, L. T. Marshall, J. P. Leal, and R.
L. Wahl, Clinically significant inaccurate localization of lesions with
PET/CT: Frequency in 300 patients, J. Nucl. Med. 44, 240243 2003.
90
P. M. Chi, O. Mawlawi, D. Luo, Z. Liao, H. A. Macapinlac, and T. Pan,
Effects of respiration-averaged CT on PET/CT quantification and its
potential impact on GTV delineation, Int. J. Radiat. Oncol., Biol., Phys.
accepted.
91
D. W. Townsend, A combined PET/CT scanner: The choices, J. Nucl.
Med. 42, 533534 2001.
92
D. W. Townsend, Physical principles and technology of clinical PET
imaging, Ann. Acad. Med. Singapore 33, 133145 2004.
93
D. W. Townsend, T. Beyer, and T. M. Blodgett, PET/CT scanners: A
hardware approach to image fusion, Semin Nucl. Med. 33, 193204
2003.
94
D. W. Townsend, J. P. Carney, J. T. Yap, and N. C. Hall, PET/CT today
and tomorrow, J. Nucl. Med., Suppl. 45, 4S14S 2004.
95
D. W. Townsend and S. R. Cherry, Combining anatomy and function:
The path to true image fusion, Eur. Radiol. 11, 19681974 2001.
96
S. R. Cherry, The 2006 Henry N. Wagner lecture: Of mice and men and
positronsAdvances in PET imaging technology, J. Nucl. Med. 47,
17351745 2006.
97
T. M. Blodgett, C. C. Meltzer, and D. W. Townsend, PET/CT: Form and
function, Radiology 242, 360385 2007.
98
J. R. Stickel and S. R. Cherry, High-resolution PET detector design:
modelling components of intrinsic spatial resolution, Phys. Med. Biol.
50, 179195 2005.
99
O. Mawlawi, D. A. Podoloff, S. Kohlmyer, J. J. Williams, C. W. Stearns,
R. F. Culp, and H. Macapinlac, Performance characteristics of a newly
developed PET/CT scanner using NEMA standards in 2D and 3D
modes, J. Nucl. Med. 45, 17341742 2004.
100
S. Surti, A. Kuhn, M. E. Werner, A. E. Perkins, J. Kolthammer, and J. S.
Karp, Performance of Philips Gemini TF PET/CT scanner with special
consideration for its time-of-flight imaging capabilities, J. Nucl. Med.
48, 471480 2007.
101
P. E. Kinahan and J. S. Karp, Figures of merit for comparing reconstruction algorithms with a volume-imaging PET scanner, Phys. Med. Biol.
39, 631642 1994.
102
R. Leahy and C. Byrne, Recent developments in iterative image reconstruction for PET and SPECT, IEEE Trans. Med. Imaging 19, 257260
2000.
103
V. Bettinardi, P. Mancosu, M. Danna, G. Giovacchini, C. Landoni, M.
Picchio, M. C. Gilardi, A. Savi, I. Castiglioni, M. Lecchi, and F. Fazio,
Two-dimensional vs three-dimensional imaging in whole body oncologic PET/CT: A discovery-STE phantom and patient study, Q. J. Nucl.
Med. Mol. Imaging 51, 214223 2007.
104
E. J. Hoffman, M. E. Phelps, G. Wisenberg, H. R. Schelbert, and D. E.
Kuhl, Electrocardiographic gating in positron emission computed tomography, J. Comput. Assist. Tomogr. 3, 733739 1979.
105
S. A. Nehmeh, Y. E. Erdi, C. C. Ling, K. E. Rosenzweig, H. Schoder, S.
M. Larson, H. A. Macapinlac, O. D. Squire, and J. L. Humm, Effect of
respiratory gating on quantifying PET images of lung cancer, J. Nucl.
Med. 43, 876881 2002.
106
S. A. Nehmeh, Y. E. Erdi, C. C. Ling, K. E. Rosenzweig, O. D. Squire, L.
E. Braban, E. Ford, K. Sidhu, G. S. Mageras, S. M. Larson, and J. L.
Humm, Effect of respiratory gating on reducing lung motion artifacts in
PET imaging of lung cancer, Med. Phys. 29, 366371 2002.
107
S. A. Nehmeh, Y. E. Erdi, K. E. Rosenzweig, H. Schoder, S. M. Larson,
O. D. Squire, and J. L. Humm, Reduction of respiratory motion artifacts
in PET imaging of lung cancer by respiratory correlated dynamic PET:
Methodology and comparison with respiratory gated PET, J. Nucl. Med.
44, 16441648 2003.
108
S. A. Nehmeh et al., Four-dimensional 4D PET/CT imaging of the
thorax, Med. Phys. 31, 31793186 2004.
109
P. E. Kinahan, H. Vesselle, L. MacDonald, A. M. Alessio, S. Kohlmyer,
and T. Lewellen, Whole-body respiratory gated PET/CT, J. Nucl. Med.
Meeting Abstracts, May 2006; 47, 187P.
110
O. Mawlawi, T. Pan, and H. A. Macapinlac, PET/CT imaging techniques, considerations, and artifacts, J. Thorac. Imaging 21, 99110
2006.
111
J. L. Garcia-Ramirez, S. Mutic, J. F. Dempsey, D. A. Low, and J. A.
Purdy, Performance evaluation of an 85-cm-bore x-ray computed tomog-
70
4966
raphy scanner designed for radiation oncology and comparison with current diagnostic CT scanners, Int. J. Radiat. Oncol., Biol., Phys. 52, 1123
1131 2002.
112
T. Beyer, A. Bockisch, H. Kuhl, and M. J. Martinez, Whole-body 18FFDG PET/CT in the presence of truncation artifacts, J. Nucl. Med. 47,
9199 2006.
113
J. S. Maltz, S. Bose, H. P. Shukla, and A. R. Bani-Hashemi, CT truncation artifact removal using water-equivalent thicknesses derived from
truncated projection data, Conf. Proc. IEEE Eng. Med. Biol. Soc. 2007,
29072911 2007.
114
O. Mawlawi, J. J. Erasmus, T. Pan, D. D. Cody, R. Campbell, A. H. Lonn,
S. Kohlmyer, H. A. Macapinlac, and D. A. Podoloff, Truncation artifact
on PET/CT: Impact on measurements of activity concentration and assessment of a correction algorithm, AJR, Am. J. Roentgenol. 186, 1458
1467 2006.
115
K. R. Britton, G. Starkschall, S. L. Tucker, T. Pan, C. Nelson, J. Y. Chang,
J. D. Cox, R. Mohan, and R. Komaki, Assessment of gross tumor volume regression and motion changes during radiotherapy for non-smallcell lung cancer as measured by four-dimensional computed tomography, Int. J. Radiat. Oncol., Biol., Phys. in press.
116
W. V. Vogel, J. A. van Dalen, B. Wiering, H. Huisman, F. H. Corstens, T.
J. Ruers, and W. J. Oyen, Evaluation of image registration in PET/CT of
the liver and recommendations for optimized imaging, J. Nucl. Med. 48,
910919 2007.
117
M. M. Osman, C. Cohade, Y. Nakamoto, and R. L. Wahl, Respiratory
motion artifacts on PET emission images obtained using CT attenuation
correction on PET-CT, Eur. J. Nucl. Med. Mol. Imaging 30, 603606
2003.
118
G. W. Goerres, E. Kamel, B. Seifert, C. Burger, A. Buck, T. F. Hany, and
G. K. Von Schulthess, Accuracy of image coregistration of pulmonary
lesions in patients with non-small cell lung cancer using an integrated
PET/CT system, J. Nucl. Med. 43, 14691475 2002.
119
G. W. Goerres, E. Kamel, T. N. Heidelberg, M. R. Schwitter, C. Burger,
and G. K. von Schulthess, PET-CT image co-registration in the thorax:
influence of respiration, Eur. J. Nucl. Med. Mol. Imaging 29, 351360
2002.
120
C. B. Caldwell, K. Mah, M. Skinner, and C. E. Danjoux, Can PET
provide the 3D extent of tumor motion for individualized internal target
volumes? A phantom study of the limitations of CT and the promise of
PET, Int. J. Radiat. Oncol., Biol., Phys. 55, 13811393 2003.
121
P. J. Keall et al., The management of respiratory motion in radiation
oncology report of AAPM Task Group 76, Med. Phys. 33, 38743900
2006.
122
D. A. Low et al., A method for the reconstruction of four-dimensional
synchronized CT scans acquired during free breathing, Med. Phys. 30,
12541263 2003.
123
T. Pan, T. Y. Lee, E. Rietzel, and G. T. Chen, 4D-CT imaging of a
volume influenced by respiratory motion on multi-slice CT, Med. Phys.
31, 333340 2004.
124
P. J. Keall, G. Starkschall, H. Shukla, K. M. Forster, V. Ortiz, C. W.
Stevens, S. S. Vedam, R. George, T. Guerrero, and R. Mohan, Acquiring
4D thoracic CT scans using a multislice helical method, Phys. Med.
Biol. 49, 20532067 2004.
125
T. Pan, X. Sun, and D. Luo, Improvement of the cine-CT based 4D-CT
imaging, Med. Phys. 34, 44994503 2007.
4966
126
M. A. Admiraal, D. Schuring, and C. W. Hurkmans, Dose calculations

accounting for breathing motion in stereotactic lung radiotherapy based
on 4D-CT and the internal target volume, Radiother. Oncol. 86, 5560
2008.
127
R. W. Underberg, F. J. Lagerwaard, B. J. Slotman, J. P. Cuijpers, and S.
Senan, Use of maximum intensity projections MIP for target volume
generation in 4DCT scans for lung cancer, Int. J. Radiat. Oncol., Biol.,
Phys. 63, 253260 2005.
128
J. D. Bradley, A. N. Nofal, I. M. El Naqa, W. Lu, J. Liu, J. Hubenschmidt,
D. A. Low, R. E. Drzymala, and D. Khullar, Comparison of helical,
maximum intensity projection MIP, and averaged intensity AI 4D CT
imaging for stereotactic body radiation therapy SBRT planning in lung
cancer, Radiother. Oncol. 81, 264268 2006.
129
A. C. Riegel, J. Y. Chang, S. S. Vedam, V. Johnson, P. M. Chi, and T. Pan,
Cine CT without a respiratory surrogate in planning of stereotactic radiotherapy for non-small cell lung cancer, Int. J. Radiat. Oncol., Biol.,
Phys. in press.
130
K. J. Biehl, F. M. Kong, F. Dehdashti, J. Y. Jin, S. Mutic, I. El Naqa, B.
A. Siegel, and J. D. Bradley, 18F-FDG PET definition of gross tumor
volume for radiotherapy of non-small cell lung cancer: Is a single standardized uptake value threshold approach appropriate?, J. Nucl. Med.
47, 18081812 2006.
131
T. Pan, Comparison of helical and cine acquisitions for 4D-CT imaging
with multislice CT, Med. Phys. 32, 627634 2005.
132
Y. E. Erdi et al., The CT motion quantitation of lung lesions and its
impact on PET-measured SUVs, J. Nucl. Med. 45, 12871292 2004.
133
S. A. Nehmeh et al., Quantitation of respiratory motion during 4DPET/CT acquisition, Med. Phys. 31, 13331338 2004.
134
C. C. Nagel, G. Bosmans, A. L. Dekker, M. C. Ollers, D. K. De Ruysscher, P. Lambin, A. W. Minken, N. Lang, and K. P. Schafers, Phased
attenuation correction in respiration correlated computed tomography/
positron emitted tomography, Med. Phys. 33, 18401847 2006.
135
A. C. Paulino and P. A. Johnstone, FDG-PET in radiotherapy treatment
planning: Pandoras box?, Int. J. Radiat. Oncol., Biol., Phys. 59, 45
2004.
136
Y. E. Erdi, O. Mawlawi, S. M. Larson, M. Imbriaco, H. Yeung, R. Finn,
and J. L. Humm, Segmentation of lung lesion volume by adaptive positron emission tomography image thresholding, Cancer 80, 25052509
1997.
137
T. R. Miller and P. W. Grigsby, Measurement of tumor volume by PET
to evaluate prognosis in patients with advanced cervical cancer treated by
radiation therapy, Int. J. Radiat. Oncol., Biol., Phys. 53, 353359 2002.
138
J. D. Kiffer, S. U. Berlangieri, A. M. Scott, G. Quong, M. Feigen, W.
Schumer, C. P. Clarke, S. R. Knight, and F. J. Daniel, The contribution
of 18F-fluoro-2-deoxy-glucose positron emission tomographic imaging to
radiotherapy planning in lung cancer, Lung Cancer 19, 167177 1998.
139
S. Beaulieu, P. Kinahan, J. Tseng, L. K. Dunnwald, E. K. Schubert, P.
Pham, B. Lewellen, and D. A. Mankoff, SUV varies with time after
injection in 18F-FDG PET of breast cancer: Characterization and
method to adjust for time differences, J. Nucl. Med. 44, 10441050
2003.
140
L. M. Hamberg, G. J. Hunter, N. M. Alpert, N. C. Choi, J. W. Babich, and
A. J. Fischman, The dose uptake ratio as an index of glucose metabolism: Useful parameter or oversimplification?, J. Nucl. Med. 35, 1308
1312 1994.
Planning and delivery of intensity-modulated radiation therapy

Cedric X. Yua
Department of Radiation Oncology, University of Maryland School of Medicine, 22 South Greene Street,
Baltimore, Maryland 21201
Christopher J. Amies
Siemens Medical Solutions USA, Inc., Oncology Care Systems Group, 4040 Nelson Avenue, Concord,
California 94520
Michelle Svatos
Translational Research, Varian Medical Systems, 3100 Hansen Way M/S E263 Palo Alto,
California 94304-1038
Received 9 May 2008; revised 11 August 2008; accepted for publication 18 September 2008;
Intensity modulated radiation therapy IMRT is an advanced form of external beam radiation
therapy. IMRT offers an additional dimension of freedom as compared with field shaping in threedimensional conformal radiation therapy because the radiation intensities within a radiation field
can be varied according to the preferences of locations within a given beam direction from which
the radiation is directed to the tumor. This added freedom allows the treatment planning system to
better shape the radiation doses to conform to the target volume while sparing surrounding normal
structures. The resulting dosimetric advantage has shown to translate into clinical advantages of
improving local and regional tumor control. It also offers a valuable mechanism for dose escalation
to tumors while simultaneously reducing radiation toxicities to the surrounding normal tissue and
sensitive structures. In less than a decade, IMRT has become common practice in radiation oncology. Looking forward, the authors wonder if IMRT has matured to such a point that the room for
further improvement has diminished and so it is pertinent to ask what the future will hold for IMRT.
This article attempts to look from the perspective of the current state of the technology to predict
the immediate trends and the future directions. This article will 1 review the clinical experience of
IMRT; 2 review what we learned in IMRT planning; 3 review different treatment delivery
techniques; and finally, 4 predict the areas of advancements in the years to come. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.3002305
Key words: radiation therapy, intensity modulation, treatment planning, radiotherapy delivery,
IMRT
I. OVERVIEW
Intensity modulated radiation therapy IMRT has been
widely adopted as a new tool in radiation therapy to deliver
high doses of radiation to the tumor while providing maximal sparing of surrounding critical structures. What facilitated the quick dissemination of IMRT technology was not
hard clinical evidence but rather the individually calculated
and visually illustrated dosimetric advantages. Because such
dosimetric advantage was obvious, wide clinical adoption of
IMRT preceded any randomized clinical trials. In the U.S.,
the increased reimbursement for the technology also fueled
the speed of clinical dissemination. Early studies indicated
that with IMRT, radiation doses to sensitive structures could
be reduced significantly while maintaining sufficient dose
coverage to the targeted tumorous tissues.19 Both rotational
and gantry-fixed IMRT techniques have been implemented
clinically
using
dynamic
multileaf
collimation
DMLC.1,36,1014 In gantry-fixed IMRT, multiple coplanar
and noncoplanar beams at different orientations, each with
spatially modulated beam intensities, are used.1,3,12 Rotational IMRT, as it is practiced today, mainly employs temporally modulated fan beams,5,6 commonly known as tomo5233
therapy. The use of overlapping cone-beam arcs13,14 to

deliver modulated beam intensities around the patient, referred to as intensity modulated arc therapy IMAT, was
also proposed but has not been widely adopted for clinical
use yet.
With nearly 10 years of clinical application, both shortterm and long-term clinical results of IMRT treatments are
emerging. Its ability to shape the high dose volume to conform to the shape of the target tissues allows the decrease of
irradiation related sequelae by limiting the dose delivered to
the surrounding normal tissues. The same dose shaping capability also allows the physicians to escalate doses to certain tumors to enhance local control. Both dosimetric and
clinical advantages have been demonstrated for almost all
common anatomical sites,15 though the largest number of
applications has been for prostate cancer and cancers of the
head and neck region.15 IMRT has allowed the physicians to
escalate the dose to the prostate while reducing the toxicities
to the rectum and bladder, resulting in improved local control
and reduced complications as compared with conventional
three-dimensional 3D conformal therapy.1619 IMRT has
also shown greater capability in sparing salivary functions in
0094-2405/2008/3512/5233/9/$23.00
5233
5234
Yu, Amies, and Svatos: IMRT planning and delivery
patients receiving radiation therapy for head and neck

cancers.2023 Rather than delivering the prescribed dose to
the entire pelvis, IMRT was able to spare the small bowel,
the bladder and the rectum, resulting in significantly lower
GI toxicities.2428 The use of IMRT instead of wedge pairs
for tangential whole breast irradiation has resulted in improved dose uniformity,2932 which in turn resulted in significantly reduced acute and chronic toxicities.33 IMRT has also
been combined with stereotactic localization for delivering
radiosurgery treatments to intracranial and extracranial sites
using linear accelerators.9,34,35 Other sites of application include lung,3638 gastrointestinal tumors,3841 pediatric
tumors,4244 and soft tissue sarcoma.45
These data indicate that for certain clinical cases, there is
a direct translation between dosimetric advantage and improved clinical outcome. With more emerging clinical data
pointing to the same conclusion, IMRT as a technology will
continue to be an important technique in external beam radiation therapy.
The way IMRT spares critical structures is by redistributing the normal tissue dose to less critical regions and to
reducing the high dose volume to just cover the target. For a
given integral dose to the target, the integral dose to the
surrounding structures is roughly constant as dictated by the
physics of dose deposition.46 In many cases, the use of IMRT
results in a greater volume of surrounding normal tissues
receiving a lower dose as compared with traditional threedimensional conformal therapy. This phenomenon raises significant concerns for pediatric applications of IMRT.47,48
While the long term clinical results of using IMRT for inoperable nonsmall lung cancer are promising,38 its use for mesothelioma following extrapleural pneumonectomy showed a
high rate of fatal pulmonary toxicity.49,50 A high mean lung
dose and a higher percentage of lung volume receiving
20 Gy were clear predictors of fatal pulmonary toxicity.
Another issue of using IMRT for treating tumors in the
thorax and abdominal regions is the breathing induced target
motion. As illustrated by Yu et al.,51 the interplay between
the target motion and the movement of the beam to achieve
intensity modulation could cause unintended hot and cold
spots in excess of 100% in a given fraction of treatment.
Different motion management strategies have been proposed
or applied clinically, including gating the radiation beam52
and dynamically tracking the tumor with DMLC53 or with
couch counter motion.54
Clinical use of IMRT has also brought many changes in
clinical practice. Owing to its ability to reduce dose to surrounding structures, IMRT has allowed physicians to escalate
dose to the prostate gland to improve local control.5557 The
reduced toxicity also encouraged the use of hypofractionation schemes.5861 The ability of IMRT to paint more complex dose distributions also allowed the increased use of concomitant boost and the treatment of target within targets.60,61
The purpose of this article is not to provide an extensive
review of IMRT technology. Such reviews can be found in
recent books62,63 and review articles.6467 We attempt to predict from what we learned in the last 10 years of development and clinical practice of IMRT the potential direction of
5234
further advancement in the near future. IMRT technology

consists of planning methods and delivery technologies. If
IMRT continues to advance, improvements must be made in
these two areas. It is therefore crucial to summarize what we
have learned in these two key aspects of the technology.
II. IMRT TREATMENT PLANNING

The key technology for IMRT planning is computer optimization. Although relatively new in radiation oncology,
computer optimization is not new and has been used for
operations research and other fields of theoretical and applied
research. Almost all the algorithms that have been used in
other fields, including simulated annealing,7,46 gradient
search,68,69 genetic algorithms,70,71 and linear and nonlinear
programming,7274 have been applied to plan IMRT treatments.
There have also been significant efforts on how the optimization problem is formulated. Universal to all the optimization methods, the objectives of each treatment plan are
reduced to a single value by the objective function or cost
function. The objectives of the treatment can be expressed
according to the desired dose distributions7,46,75 or in terms
of biological objectives.68,76 Essentially, when the objective
function is formulated to closely depict the desirable features
of the radiation dose distribution, the quality of the candidate
plans is judged more accurately during the optimization process. Therefore, as long as the objectives are defined to appropriately reflect what is really clinically optimal, the results are generally clinically meaningful and acceptable.
Different commercial treatment planning systems use different optimization algorithms, and slightly different dose or
biological dose, in the form of equivalent uniform dose, objectives. In general and except for certain limitations in the
implementation and planner experience, there is no clinically
meaningful difference in plan quality when using different
approaches.
Complex treatment planning problems often involve multiple targets with many surrounding normal structures. The
objectives specified for these structures are often in conflict.
In such complex situations, it is hard if not impossible to
optimally describe what is truly desired before a plan is optimized. In the situation where one cannot achieve optimality
for all the objectives simultaneously, the optimal solution
becomes the optimal trade-off among these objectives. There
have been several approaches proposed by different researchers for dealing with such conflicting objectives. One of
the approaches is to specify the treatment objective as a
probability density function rather than as a rigid dose
objective.77 If the most desired dose is not attainable by the
optimizer, other doses are acceptable but less preferred. Consequently, during the optimization process, the prescription
dose is allowed to deviate, with a certain preference level,
from the most desired dose. Another approach is to build a
database of plans by varying the emphasis of different objectives. Using an interactive plan navigation tool, the trade-offs
between different objectives can be made explicitly based on
5235
the users clinical goals.78 The resultant plans from this approach are also called Pareto-optimal plans.
One of the tasks in treatment planning for external beam
radiation therapy is to determine the number of beams to use
and their orientations. For conventional treatments, beam
angles used for different treatment sites are well established.
With IMRT, the task of beam angle selection is more complicated and less intuitive because of more complex interactions and mutual compensations among the different beam
angles as the result of computer optimized beam intensities.
As the result, beam angle optimization has been one of the
areas of intensive investigation. Haas et al.79 employed genetic algorithms to search for the best beam orientations.
Rowbottom et al.80 and Stein et al.81 used simulated annealing algorithms to perform beam angle optimization by comparing thousands of sets of fixed number of beams sampled
from a constrained or unconstrained space of beam orientations. Pugachev et al.82 also reported different schemes of
beam angle optimization with the simulated annealing algorithm. Although all these studies demonstrated some effects
of beam angle selection on plan quality, the degree of improvements has been small. It is thus reasonable to conclude
that the mutual interactions and compensations among the
beams under intensity optimization also made beam angle
selection less influential on plan quality as compared with
three-dimensional conformal therapy.
In spite of years of refinements in IMRT planning, arguably, the quality of treatment plans has not improved much
from that of the early days. The explanation may be rooted in
the basic principles of inverse planning. For each patient, the
anatomy dictates a unique set of preferred beam orientations
and, in each beam orientation and radiation field, the preferred locations through which radiation is directed to the
tumor. If more freedom is given to make use of these intrinsic preferences in treatment planning, either performed by a
human planner or by a computer optimization system, better
plans can be generated. IMRT provides an additional freedom as compared with 3D conformal radiation therapy because the radiation intensities within a radiation field can be
varied according to the location preferences. However, there
is a fundamental limit to which the quality of treatment plans
can be improved. This limit is defined by the physics of the
radiation dose deposition and by the degrees of freedom that
are given to the computer optimization routine. There are
indications that the required degree of intensity variation for
achieving the optimal plan quality for a large majority of
cases is relatively small. For example, Ludlum et al.83 optimized IMRT plans for different challenging cancer sites using different number of intensity levels. They found that reducing the number of intensity levels from 10 to 3 only cause
minor degradations in dose distribution. Comparisons among
different treatment planning and delivery approaches8487
have not yield any approach with clinically meaningful superiority. These studies and our own experience led us to
believe that with the current state of the radiation delivery
systems and without new advances to provide additional degrees of freedom, IMRT has approached its limits in the
quality of treatment plans that can be physically achieved. As
5235
a result, the residual room for improvements in plan quality

under existing delivery systems has diminished over the
years of refinements in treatment planning algorithms and
methods.
III. IMRT TREATMENT DELIVERY

Most IMRT treatment deliveries require the use of the
multileaf collimator MLC, which was originally developed
for shaping radiation fields. The ability of the MLC to easily
change and dynamically vary the field shapes was quickly
explored for IMRT delivery. Over the years, linear accelerator vendors have not only improved the reliability of MLCs
but also made the leaf width smaller. Smaller leaf widths
allow the treatment planning system to use finer beamlet size
during optimization for achieving better plan quality.88
IMRT treatments are delivered using either fixed beam
angles or rotational beams. Both forms of delivery were proposed and developed at the start of the IMRT technology.89
Although the technology is termed intensity modulation,
the intensity variation is actually achieved by temporal
modulation, allowing different beamlets to be irradiated for
different durations.
Most of the IMRT treatments delivered today use a few
fixed beam directions. The treatment planning system has
converted the optimal intensity distributions into deliverable
overlapping field segments. Because the conversion is not
within the optimization process and MLC motion constraints
have to be considered during such conversion, it may introduce differences between the treatment plans using the optimal intensity distributions and that achievable with overlapping field segments.11 These field segments, also called
subfields, can be delivered dynamically, i.e., MLC transitioning through the shapes of the subfields dynamically when the
beam is on. They can also be delivered one segment at a time
such that the MLC is not moving during irradiation.
Up to now, rotational IMRT treatments have primarily
been delivered using tomotherapy,5,6 an approach where a
binary collimator open/close is used to control the amount
of exposure time of a small width of the fan beam, or a
beamlet. The fraction of time a leaf is in the open position
at a given beam angle determines the relative intensity of
radiation at that angle. Because the fan beam can only irradiate a slice of the patient at a given time, larger tumors are
treated by either stepping the table one slice at a time5 or
continuously.6 The latter is called helical tomotherapy because the trajectory of the radiation source relative to the
patient is a helix. Tomotherapy Madision, WI offers a
turn-key approach to IMRT implementation with the planning system specifically designed for the delivery unit. Tomotherapy utilizes all coplanar beam angles and the intensity
variations of the beamlets are not constrained by the mechanical limits of the binary MLC. Therefore, theoretically
speaking, tomotherapy provides a planning system with the
freedom to use highly modulated beams to create more conformal treatment plans then MLC delivery with fixed cone
beams.
5236
Another rotational IMRT approach, called IMAT, was

suggested by Yu.13 Instead of rotating fan beams around the
patient, IMAT uses cone beams shaped with conventional
MLCs. IMAT delivers optimized intensity distributions for
large number of beams spaced every 5 10 deg around the
patient. Optimized intensity distributions are translated into a
stack of superimposed irregular fields of uniform beam intensities and delivered by overlapping arcs with synchronized gantry rotation and field shape variations. As the gantry is rotating around the patient and the radiation beam is
on, it is important that the subfields of adjacent beam angles
do not require the MLC leaves to travel very long distances.
Ensuring such connectedness of adjacent subfields for
smooth leaf motion is of great concern in the leaf sequencing
algorithm for IMAT.13,14 Effective planning tools for IMAT
have only been developed recently.84,90
Comparisons between IMRT plans for different delivery
methods have been conducted. In a comparison of tomotherapy and MLC delivery, Mavroidis et al. found that linear
accelerator delivery with a MLC has slight advantage over
tomotherapy for most sites other than head and neck.85 Similar results have been found by Muzik et al.86 Cao et al.87
compared the treatment plan quality of IMAT plans and tomotherapy plans for ten cases including head and neck, lung,
brain, and prostate. It was found that these two kinds of
rotational delivery methods are also equivalent for most
cases. For cases where noncoplanar beams are desirable,
such as for intracranial tumors and some head and neck
cases, the use of partial noncoplanar arcs in IMAT was found
to be more advantageous.87 Shepard et al. compared IMAT
plans with IMRT 84 and found that the employment of rotational IMRT is advantageous for most of the cases.
It is important to note that there are many other issues
besides plan quality that are associated with different delivery techniques. These include the efficiency of planning, delivery, and quality assurance, the complexity and reliability
of delivery, and the total monitor units required to deliver the
prescribed dose.
IV. FUTURE DIRECTIONS OF ADVANCEMENTS
OF IMRT
Revolutionary advances in biology and genomics have
brought a wave of new molecular therapeutic agents for
fighting cancer and the continued acceleration in the direction will mark a new era of cancer therapies. However, molecular agents have not been shown to eliminate local therapies. Therefore, we do not see that the role of radiation
therapy will diminish in this new era. The dose shaping capabilities of IMRT make it an essential tool for better integration of radiation with chemotherapy and molecular therapeutic agents.
The areas in which IMRT technology can be improved
include 1 the quality of treatment plans, 2 the accuracy of
treatment delivery, and 3 the efficiency of both planning
and delivery of IMRT treatments. We predict that the advances in these areas will be driven by factors including
on-line and real-time image guidance, individualized apMedical Physics, Vol. 35, No. 12, December 2008
5236
proaches to metabolic imaging, and the emergence of a new

generation of radiation treatment delivery systems.
IV.A. Quality improvements
Based on 10 years of experience with IMRT, we have

learned that the opportunities in improving plan quality are
limited within the constraint of present linac/MLC delivery.
To improve the quality of IMRT treatment plans, we must
inject new degrees of freedom. This may require an overhaul
of existing technologies. A new generation of treatment delivery systems is emerging. One of such machines is proposed by Kamino et al.91 Besides the IMRT capability and
improved image guidance features, a notable feature is that
the linear accelerator head can be pivoted. By offering noncoplanar beams easily without couch rotations, such machine
provides new degrees of freedom into the solutions, and,
therefore, has the potential of improving the qualities of
treatment plans.
Several new machines currently under development incorporate image guidance from Magnetic resonance imaging
MRI. The design of the Renaissance ViewRay, Inc.,
Cleveland, OH incorporates real-time MR imaging and multihead Co-60 rotational delivery. The use of MRI for image
guidance may provide improved soft tissue contrast for improved targeting accuracy.
These new technologies are aimed for delivering radiation
treatments to all sites. As the room for quality improvements
is further squeezed, technologies specific for particular treatment sites may also emerge to take advantage of the anatomy
and biology of such specific sites.
As machines increase in complexity, it becomes harder for
a generic treatment planning system to fully utilize the machines capabilities. Diversity in delivery systems forces the
planning system to be more specifically tailored to the treatment machine. Treatment planning systems designed specifically for a delivery device as a turn-key solution will become
the most common configuration.
IMRT is not limited to the use of photons. It has been
demonstrated that the IMRT principle can also be applied to
electrons92 and protons.93,94 The fast introduction of proton
beams in radiation therapy would allow us to push the limit
of achievable plan quality higher than using only photons.
IV.B. Efficiency improvements
The fact that treatment plans of different complexity designed for different delivery methods can achieve similar
quality of treatments suggest that there is substantial room
for improving the workflow and efficiency. Past attempts in
improving treatment efficiency include the use of direct aperture optimization95 and the development of single arc
IMAT.96100
Instead of optimizing intensity maps and then translating
the intensity maps into deliverable segments, Shepard et al.95
proposed and developed a method called direct aperture optimization DAO that directly optimizes the shapes and
weights of MLC segments simultaneously in one step. The
planner determines the number of beams to use and the num-
5237
ber of apertures desired at each of the beams. The planning

system optimizes the aperture shape and aperture weights
simultaneously. Physical constraints of the MLC are considered in the optimization process. With a small number of
apertures, the resulting plan quality generally rivals that
based on intensity optimization which requires large number
of apertures. Earl et al.101 and Kim et al.102 also demonstrated that even with overlapping rectangular fields formed
by independent jaws, IMRT type of dose distribution can be
achieved for prostate, breast, and even head and neck cases,
if the aperture shapes and weights are optimized with DAO.
Earl et al. illustrated that DAO can also be applied for the
optimization of IMAT plans.103 The possibility of varying
dose rates during gantry rotation and irradiation also opens
the opportunity for delivering IMRT-like plans with a single
arc rotation. In proposing the IMAT idea, Yu13 has predicted
that, with reasonable number of beam angles, the quality of
an IMRT or IMAT plan depends on the number of total segments, or aperture variations. Given enough aperture variations, a single arc should be able to achieve IMRT-like plan
quality. Beam intensity modulation is not a fundamental requirement for achieving optimal treatment plans. The ability
to take advantage of the preferred angles and locations,
through which the radiation is directed to the tumor, is the
key to achieve optimal dose distributions. For example, the
CyberKnife system104 Accuray, Sunnyvale, CA, which
uses small circular x-ray beams generated by an x-band linear accelerator mounted on a robot to deliver the radiation to
the target, does not explicitly modulate the intensity of a
beam, but is able to deliver highly conformal treatments.
The feasibility of delivering IMRT with a single arc was
first demonstrated by Cameron et al.98 Using a DAO scheme
as Earl et al. but employing a more efficient way for selecting the initial aperture shapes, Ulrich et al. also showed that
IMRT-like dose distributions can be achieved with a single
arc.97 When optimizing a large number of beam apertures
from large number of beam angles, the scheme by Ulrich and
Earl can take a long time for the optimization to converge.
Otto devised a coarse-to-fine optimization scheme that starts
with a small number of beams with large angular spacing and
gradually inserts new beam angles to be optimized.96 The
computer optimizes the aperture shapes and weights simultaneously as in DAO. Tang et al.99 illustrated that a single arc
IMAT plan can be derived from a multiarc IMAT plan by
spacing the stacked apertures to the neighboring angles. Using graph algorithms, Wang et al.100 have shown that IMRT
plans optimized with 36 fields can be sequenced into and
delivered with a single arc. The pressure for more efficient
IMRT delivery will encourage linear accelerator vendors to
offer different single arc IMAT solutions.
In addition to rotational delivery, the use of higher dose
rates, either by the improvements in acceleration technology
or by eliminating the flattening filter, could also be explored
for efficiency gains.105 As IMRT becomes easier to plan and
to deliver, the need for traditional beam modifiers such as the
use of physical wedges will diminish.
5237
IV.C. Improving the accuracy of IMRT delivery with

image guidance
Besides improving efficiency, great opportunities also exist for improving the accuracy of treatment delivered by using on-board image guidance. The advent of on-board cone
beam computed tomography CBCT and in-room CT with
high soft tissue contrast opens new opportunities for highprecision radiation therapy.106 The success of the image
guided radiation therapy IGRT lies not only in the ability to
acquire the images but largely in how we use the images to
achieve our goal of controlling more cancers while decreasing the normal tissue toxicities. In many cases, the tumor not
only changes its location but also its shape and the relative
geometrical relationship with its surrounding normal structures. Consequently, simply shifting the patient does not
guarantee optimal treatments. Therefore, reoptimizing the
IMRT plan or adapting an existing plan to the anatomy of
the day will increasingly become an active area of research.
Recognizing the impracticality of a full-fledged daily replanning with current technology, several researchers have
recently proposed methods for incorporating the deformable
component into an online correction strategy. Wu et al.107
have shown that for tomotherapy delivery, intensities can be
modified online with given transformation parameters or
with a quick reoptimization. Mohan et al.108 proposed to deform the intensity patterns as an online correction strategy.
Feng et al.109 have proposed a scheme that deforms the aperture shapes of the IMRT segments by using the deformation matrix derived from the planning image set and the images of the day. Ahunbay et al.110 developed a similar
scheme but added aperture weight optimization after aperture
deformation. These schemes have shown to be able to approach the plan quality resulting from reoptimization using
the images of the day.
As computer technology continue to advance, such short
cuts as described above may no longer be needed. Replanning using the images of the day will become a reality. The
advance of IGRT will make imaging, planning and treatment
delivery all performed in a single session. Under the new
image-plan-treat process, initial imaging and planning will
become merely a first look into what degree of dose conformity is achievable. With experience and confidence, the initial planning may even be eliminated. Such a procedure,
which has been the norm for radiosurgery, will allow further
shrinkage of PTV margins and encourage radiation oncologists to revise the current dose fractionation schemes. There
are three technologies that will enable for this image-plantreat scheme.
1 Imaging systems integrated with delivery machines will
be faster and provide greater image quality before and
during treatment. Currently, the gantry speed is setting
the limit on the speed of image acquisition for imaging
systems mounted on the treatment gantry. The image
quality achievable with CBCT also does not match that
of fan beam CT units because of the limitations of the
reconstruction algorithms for CBCT and the added radiation scatter contributions. These limitations will be
5238
corrected in the long-term by the emergence of new designs of on-line imaging systems, new imaging reconstruction and scatter subtraction algorithms, and/or new
on-line imaging strategies.
2 Dose calculation and optimization can be performed
within a couple of minutes. Computational speeds have
been increasing rapidly according to Moores law. Other
innovations, such as the use of special hardware for dose
computation,111 further push the envelope of planning
speed.
3 Fast image registration will allow the structures delineated by the physicians on a previously acquired image
set to be transferred to the images of the day in very
short time. Although three-dimensional deformable image registration has been theoretically solved for decades, practical applications of these algorithms in radiation therapy are new and relatively primitive. The room
of improvements in both the speed and the accuracy of
registration is significant. As the need for speed and accuracy continue to rise, innovative solutions will continue to emerge. It will be possible within the next
5 years for a deformable image registration to be performed in less than a minute. With fast deformable image registration, the accumulation of doses to different
parts of the target and to critical structures will be performed more accurately.
IV.D. Dose painting with molecular and functional

imaging guidance
The integration of molecular and functional imaging with

radiation therapy will shift the current image guidance from
merely geometrical to biological. The scenario of dose
painting112 will become reality. Presently, our goals are generally to deliver a high uniform dose to the target while sparing surrounding critical structures. With the knowledge of
molecular signatures of the tumor and its microenvironment
revealed by the new imaging modalities, it will be more desirable to paint a nonuniform dose distribution to best attack
the tumor while protecting its surrounding structures. Being
able to use computer optimization to take advantage of the
angular and positional preferences, IMRT will be well positioned in molecular and functional image guided dose painting.
V. CONCLUSION
We have reviewed and described the current state of
IMRT and identified that IMRT is a well established technology offering dosimetric advantages for target coverage and
normal tissue avoidance. We believe that the treatment plan
quality has reached a limit imposed by the physics of photon
dose deposition and the room for further improvement in
plan quality is small. We also noticed that this intrinsic limit
is also not difficult to approach with different IMRT
schemes. This fact has allowed efficiency improvements
5238
such as DAO and single arc IMAT. The rapid adaptation of

IGRT will encourage the integration of IMRT planning and
delivery with on-line imaging. Strategies of on-line plan adaptation and on-line replanning will be the emphasis in the
near future. The adoption of biology and molecular imaging
guidance will push IMRT into dose painting paradigm in the
long term.
a

[email protected]
1
A. Brahme, Optimization of stationary and moving beam radiation
therapy techniques, Radiother. Oncol. 12, 129140 1988.
2
T. Bortfeld, D. Kahler, T. Waldron, and A. Boyer, X-ray field compensation with multileaf collimators, Int. J. Radiat. Oncol., Biol., Phys.
283, 723730 1994.
3
A. Boyer, and C. Yu, Intensity-modulated radiation therapy with dynamic multileaf collimators, Semin. Radiat. Oncol., 91, 4859 1999.
4
C. Burman, C. Chui, G. Kutcher, S. Leibel, M. Zelefsky, T. LoSasso, S.
Spirou, Q. Wu, J. Yang, J. Stein, R. Mohan, Z. Fuks, and C. Ling Planning, delivery, and quality assurance of intensity-modulated radiotherapy
using dynamic multileaf collimator: A strategy for large-scale implementation for the treatment of carcinoma of the prostate, Int. J. Radiat. Oncol., Biol., Phys., 394, 863873 1997.
5
M. Carol, Integrated 3D conformal planning/multivane intensity modulating delivery system for radiotherapy, in 3D Radiation Treatment Planning and Conformal Therapy, edited by J. A. Purdy and B. Emami Medical Physics, Madison, WI, 1995, pp. 435445.
6
T. Mackie, T. Holmes, and S. Swerdloff Tomotherapy: A new concept
for the delivery of dynamic conformal radiotherapy, Med. Phys. 206,
17091719 1993.
7
S. Webb Optimizing the planning of intensity-modulated radiotherapy,
Phys. Med. Biol. 3912, 22292246 1994.
8
Q. Wu, M. Manning, R. Schmidt-Ullrich, and R. Mohan The potential
for sparing of parotids and escalation of biologically effective dose with
intensity-modulated radiation treatments of head and neck cancers: A
treatment design study, Int. J. Radiat. Oncol., Biol., Phys. 461, 195
205 2000.
9
S. Y. Woo et al., A comparison of intensity modulated conformal therapy
with a conventional external beam stereotactic radiosurgery system for
the treatment of single and multiple intracranial lesions, Int. J. Radiat.
Oncol., Biol., Phys. 35, 593597 1996.
10
D. J. Convery and M. E. Rosenbloom The generation of intensitymodulated fields for conformal radiotherapy by dynamic collimation,
Phys. Med. Biol. 376, 13591374 1992.
11
P. Xia and L. J. Verhey Mutileaf collimation leaf sequencing algorithm
for intensity modulated beams with multiple static segments, Med. Phys.
258, 14241434 1998.
12
C. X. Yu et al., A method for implementing dynamic photon beam intensity modulation using independent jaws and multileaf collimator,
Phys. Med. Biol. 40, 769787 1995.
13
C. X. Yu Intensity-modulated arc therapy with dynamic multileaf collimation: An alternative to tomotherapy, Phys. Med. Biol., 40, 14351449
1995.
14
C. X. Yu, X. A. Li, L. Ma, D. J. Chen, S. Naqvi, D. Shepard, M. Sarfaraz,
T. W. Holmes, M. Suntharalingam, and C. M. Mansfield Clinical Implementation of intensity-modulated arc therapy, Int. J. Radiat. Oncol.,
Biol., Phys. 532, 453463 2002.
15
L. K. Mell, A. K. Mehrotra, and A. J. Mundt Intensity-modulated radiation therapy use in the U. S., 2004, Cancer N.Y., 1046, 12961303
2005.
16
A. Pollack et al., Prostate cancer radiation dose response: Results of the
MD Anderson phase III randomized trial, Int. J. Radiat. Oncol., Biol.,
Phys. 535, 10971105 2002.
17
M. J. Zelefsky et al., High-dose intensity modulated radiation therapy
for prostate cancer: early toxicity and biochemical outcome in 772 patients, Int. J. Radiat. Oncol., Biol., Phys. 535, 11111116 2002.
18
C. M. Nutting et al., Reduction of small and large bowel irradiation
using an optimized intensity-modulated pelvic radiotherapy technique in
patients with prostate cancer, Int. J. Radiat. Oncol., Biol., Phys. 48,
649656 2000.
5239
19
A. A. Martinez et al., Improvement in dose escalation using the process

of adaptive radiotherapy combined with three-dimensional conformal or
intensity-modulated beams for prostate cancer, Int. J. Radiat. Oncol.,
Biol., Phys. 50, 12261234 2001.
20
K. S. C. Chao et al., A prospective study of salivary function sparing in
patients with head-and-neck cancers receiving intensity-modulated or
three-dimensional radiation therapy: initial results, Int. J. Radiat. Oncol.,
Biol., Phys. 494, 907916 2001.
21
A. Eisbruch et al., Dose, volume, and function relationships in parotid
salivary glands following conformal and intensity modulated irradiation
of head and neck cancer, Int. J. Radiat. Oncol., Biol., Phys. 45, 577587
1999.
22
C. M. Nutting et al., Improvements in target coverage and reduced spinal
cord irradiation using intensity-modulated radiotherapy IMRT in patients with carcinoma of the thyroid gland, Radiother. Oncol. 60, 163
172 2001.
23
K. Sultanem, H. K. Shu, P. Xia, C. Akazawa, J. M. Quivey, L. J. Verhey,
and K. K. Fu Three-dimensional intensity-modulated radiotherapy in the
treatment of nasopharyngeal carcinoma: The University of California-San
Francisco experience, Int. J. Radiat. Oncol., Biol., Phys. 483, 711722
2000.
24
J. C. Roeske, A. Lujan, J. Rotmensch, S. E. Waggoner, D. Yamada, and
A. J. Mundt Intensity-modulated whole pelvic radiation therapy in patients with gynecologic malignancies, Int. J. Radiat. Oncol., Biol., Phys.
485, 16131621 2000.
25
A. J. Mundt et al., Intensity-modulated whole pelvic radiotherapy in
women with gynecologic malignancies, Int. J. Radiat. Oncol., Biol.,
Phys. 52, 13301337 2002.
26
L. Portelance et al., Intensity-modulated radiation therapy IMRT reduces small bowel, rectum, and bladder doses in patients with cervical
cancer receiving pelvic and para-aortic irradiation, Int. J. Radiat. Oncol.,
Biol., Phys. 51, 261266 2001.
27
B. D. Kavanagh et al., Clinical application of intensity-modulated radiotherapy for locally advanced cervical cancer, Semin. Radiat. Oncol. 12,
260271 2002.
28
C. Brixey et al., Impact of intensity modulated whole pelvic radiation
therapy on acute hematologic toxicity in patients with gynecologic malignancies, Int. J. Radiat. Oncol., Biol., Phys. 54, 13881396 2002.
29
L. Hong et al., Intensity-modulated tangential beam irradiation of the
intact breast, Int. J. Radiat. Oncol., Biol., Phys. 44, 11551164 1999.
30
Y. C. Lo et al., Intensity modulation for breast treatment using static
multileaf collimators, Int. J. Radiat. Oncol., Biol., Phys. 46, 187194
2000.
31
B. van Asselen et al., An improved breast irradiation technique using
three-dimensional geometrical information and intensity modulation,
32
L. L. Kestin et al., Intensity modulation to improve dose uniformity with
tangential breast radiotherapy: Initial clinical experience, Int. J. Radiat.
Oncol., Biol., Phys. 48, 15591568 2000.
33
A. Harsolia, L. Kestin, I. Grills, M. Wallace, S. Jolly, C. Jones, M. Lala,
A. Martinez, S. Schell, and F. A. Vicini Intensity-modulated radiotherapy results in significant decrease in clinical toxicities compared with
conventional wedge-based breast radiotherapy, Int. J. Radiat. Oncol.,
Biol., Phys. 685, 13751380 2007.
34
R. M. Cardinale et al., A comparison of three stereotactic radiotherapy
techniques: ARCS vs. noncoplanar fixed fields vs. intensity modulation,
35
D. D. Leavitt et al., Intensity-modulated radiosurgery/radiotherapy using
a micromultileaf collimator, Med. Dosim. 26, 143150 2001.
36
S. Derycke et al.,Conformal radiotherapy of stage III non-small cell lung
cancer: a class solution involving non-coplanar intensity-modulated
beams, Int. J. Radiat. Oncol., Biol., Phys. 41, 771777 1998.
37
J. van Sornsen de Koste et al., An evaluation of two techniques for beam
intensity modulation in patients irradiated for stage III non-small cell lung
cancer, Lung Cancer 32, 145153 2001.
38
S. Sura, V. Gupta, E. Yorke, A. Jackson, H. Amols, and K. E. Rosenzweig
Intensity-modulated radiation therapyIMRT for inoperable non-small
cell lung cancer: the Memorial Sloan-Kettering Cancer Center MSKCC
experience, Radiother. Oncol. 871, 1723 2008.
39
M. T. Milano et al., Intensity-modulated radiotherapy in treatment of
pancreatic and bile duct malignancies: Toxicity and clinical outcome,
5239
40
J. C. Landry et al., Treatment of pancreatic cancer tumors with intensitymodulated radiation therapy IMRT using the volume at risk approach
VARA: Employing dose-volume histogram DVH and normal tissue
complication probability NTCP to evaluate small bowel toxicity, Med.
Dosim. 27, 121129 2002.
41
C. M. Nutting et al., A comparison of conformal and intensitymodulated techniques for esophageal radiotherapy, Radiother. Oncol. 61,
157163 2001.
42
R. B. Kuppersmith et al., The use of intensity modulated radiotherapy
for the treatment of extensive and recurrent juvenile angiofibroma, Int. J.
Pediatr. Otorhinolaryngol. 52, 261268 2000.
43
A. C. Paulino and M. Skwarchuk Intensity-modulated radiation therapy
in the treatment of children, Med. Dosim. 27, 115120 2002.
44
E. Huang et al., Intensity-modulated radiation therapy for pediatric
medulloblastoma: Early report on the reduction of ototoxicity, Int. J.
45
L. Hong, K. M. Alektiar, M. Hunt, E. Venkatraman, and S. A. Leibel,
Intensity-modulated radiotherapy for soft tissue sarcoma of the thigh,
Int. J. Radiat. Oncol., Biol., Phys. 593,752975 2004.
46
D. M. Shepard, G. Olivera, L. Angelos, O. Sauer, P. Reckwerdt, and T. R.
Mackie, A simple model for examining issues in radiotherapy optimization, Med. Phys. 267, 12121221 1999.
47
E. J. Hall and C. Wuu, Radiation-induced second cancers: The impact of
3D-CRT and IMRT, Int. J. Radiat. Oncol., Biol., Phys. 56, 8388 2003.
48
D. B. Mansur, E. E. Klein, and B. P. Maserang, Measured peripheral
dose in pediatric radiation therapy: A comparison of intensity-modulated
and conformal techniques, Radiother. Oncol. 822, 179184 2007.
49
A. M. Allen et al., Fatal pneumonitis associated with intensitymodulated radiation therapy for mesothelioma, Int. J. Radiat. Oncol.,
Biol., Phys. 65, 640645 2006.
50
D. C. Rice, W. R. Smythe, Z. Liao, T. Guerrero, J. Y. Chang, M. F.
McAleer, M. D. Jeter, A. Correa, A. A. Vaporciyan, H. H. Liu, R. Komaki, K. M. Forster, and C. W. Stevens, Dose-dependent pulmonary
toxicity after postoperative intensity-modulated radiotherapy for malignant pleural mesothelioma, Int. J. Radiat. Oncol., Biol., Phys. 692,
350357 2007.
51
C. X. Yu, D. A. Jaffray, and J. W. Wong, Effects of intra-fraction organ
motion on the delivery of dynamic intensity modulation, Phys. Med.
Biol. 43, 91104 1998.
52
H. D. Kubo and B. C. Hill, Respiration-gated radiotherapy treatment: A
technical study, Phys. Med. Biol. 41, 8391 1996.
53
P. J. Keall, V. R. Kini, S. S. Vedam, and R. Mohan, Motion adaptive
x-ray therapy: A feasibility study, Phys. Med. Biol. 46, 110 2001.
54
W. D. DSouza, S. A. Naqvi, and C. X. Yu, Real-time intra-fraction
motion tracking using the treatment couch: A feasibility study, Phys.
Med. Biol. 50, 40214033 2005.
55
D. A. Kuban, S. L. Tucker, L. Dong, G. Starkschall, E. H. Huang, M. R.
Cheung, A. K. Lee, and A. Pollack, Long-term results of the M. D.
Anderson randomized dose-escalation trial for prostate cancer, Int. J.
56
A. Pollack, A. Hanlon, E. M. Horwitz, S. Feigenberg, R. G. Uzzo, and R.
A. Price, Radiation therapy dose escalation for prostate cancer: A rationale for IMRT, World J. Urol. 4, 200208 2003.
57
M. J. Zelefsky, Z. Fuks, and S. A. Leibel, Intensity-modulated radiation
therapy for prostate cancer, Semin. Radiat. Oncol. 3, 229237 2002.
58
S. Junius, K. Haustermans, B. Bussels, R. Oyen, B. Vanstraelen, T.
Depuydt, J. Verstraete, S. Joniau, and H. Van Poppel, Hypofractionated
intensity modulated irradiation for localized prostate cancer, results from
a phase I/II feasibility study, Radiother. Oncol. 2, 29 2007.
59
K. K. Fu et al., A Radiation Therapy Oncology Group RTOG phase III
randomized study to compare hyperfractionation and two variants of accelerated fractionation to standard fractionation radiotherapy for head and
neck squamous cell carcinomas: First report of RTOG 9003, Int. J. Radiat. Oncol., Biol., Phys. 48, 716 2000.
60
G. M. Freedman, N. J. Meropol, E. R. Sigurdson, J. Hoffman, E. Callahan, R. Price, J. Cheng, S. Cohen, N. Lewis, D. Watkins-Bruner, A.
Rogatko, and A. Konski, Phase I trial of preoperative hypofractionated
intensity-modulated radiotherapy with incorporated boost and oral
capecitabine in locally advanced rectal cancer, Int. J. Radiat. Oncol.,
Biol., Phys. 675, 13891393 2007.
61
T. Iuchi, K. Hatano, Y. Narita, T. Kodama, T. Yamaki, and K. Osato,
Hypofractionated high-dose irradiation for the treatment of malignant
astrocytomas using simultaneous integrated boost technique by IMRT,
5240

A. J. Mundt and J. C. Roeske, Intensity Modulated Radiation Therapy-A
Clinical Perspective Decker, Toronto, 2005.
63
K. S. C. Chao, S. Apisarnthanarax, Go. Ozyigit, Practical Essentials of
Intensity Modulated Radiation Therapy Lippincott Williams & Wilkins,
New York, 2004.
64
L. Veldeman, I. Madani, F. Hulstaert, G. De Meerleer, M. Mareel, and W.
Neve De, Evidence behind use of intensity-modulated radiotherapy: A
systematic review of comparative clinical studies, Lancet Oncol. 94,
367375 2008.
65
K. Hatano, H. Araki, M. Sakai, T. Kodama, N. Tohyama, T. Kawachi, M.
Imazeki, T. Shimizu, T. Iwase, M. Shinozuka, and H. Ishigaki, Current
status of intensity-modulated radiation therapy IMRT, J. Clin. Oncol.
126, 408415 2007.
66
C. Yu, D. Shepard, M. Earl, D. Cao, S. Luan, C. Wang, and D. Z. Chen,
New developments in intensity modulated radiation therapy, Technol.
Cancer Res. Treat. 55, 451464 2006.
67
T. Bortfeld, IMRT: A review and preview, Phys. Med. Biol. 5113,
R363R379 2006.
68
A. Olafsson, R. Jeraj, and S. J. Wright, Optimization of intensitymodulated radiation therapy with biological objectives, Phys. Med. Biol.
5022, 53575379 2005.
69
X. Zhang, H. Liu, X. Wang, L. Dong, Q. Wu, and R. Mohan, Speed and
convergence properties of gradient algorithms for optimization of IMRT,
Med. Phys. 315, 11411152 2004.
70
C. Cotrutz and L. Xing, Segment-based dose optimization using a genetic algorithm, Phys. Med. Biol. 4818, 29872998 2003.
71
Y. Li, J. Yao, and D. Yao, Automatic beam angle selection in IMRT
planning using genetic algorithm, Phys. Med. Biol. 4910, 19151932
2004.
72
S. P. Goldman, J. Z. Chen, and J. J. Battista, Feasibility of a fast inverse
dose optimization algorithm for IMRT via matrix inversion without negative beamlet intensities, Med. Phys. 329, 30073016 2005.
73
E. K. Lee, T. Fox, and I. Crocker, Simultaneous beam geometry and
intensity map optimization in intensity-modulated radiation therapy, Int.
J. Radiat. Oncol., Biol., Phys. 641,301320 2006.
74
H. E. Romeijn, R. K. Ahuja, J. F. Dempsey, A. Kumar and J. G, Li, A
novel linear programming approach to fluence map optimization for intensity modulated radiation therapy treatment planning, Phys. Med. Biol.
4821, 35213542 2003.
75
T. Bortfeld, Optimized planning using physical objectives and constraints, Semin. Radiat. Oncol. 91, 2034 1999.
76
Q. Wu, D. Djajaputra, Y. Wu, J. Zhou, H. H. Liu, and R. Mohan,
Intensity-modulated radiotherapy optimization with gEUD-guided dosevolume objectives, Phys. Med. Biol. 483, 279291 2003.
77
J. Lian, C. Cotrutz, and L. Xing, Therapeutic treatment plan optimization
with probability density-based dose prescription, Med. Phys. 304, 655
666 2003.
78
D. Craft, T. Halabi, H. A. Shih, and T. Bortfeld, An approach for practical multiobjective IMRT treatment planning, Int. J. Radiat. Oncol.,
Biol., Phys. 695, 16001607 2007.
79
L. C. O. Haas, K. J. Burnham, and J. A. Mills, Optimization of beam
orientation in radiotherapy using planar geometry, Phys. Med. Biol. 43,
21792193 1998.
80
C. G. Rowbottom, S. Webb, and M. Oldham, Beam orientation customization using an artificial neural network, Phys. Med. Biol., 44, 2251
2262 1999.
81
J. Stein, R. Mohan, X. Wang, T. Bortfeld, Q. Wu, K. Preiser, C. Ling, and
W. Schlegel, Number and orientations of beams in intensity-modulated
radiation treatments, Med. Phys. 242, 149160 1997.
82
A. Pugachev, A. Boyer, and L. Xing, Beam orientation optimization in
intensity-modulated radiation treatment planning, Med. Phys. 276,
12381245 2000.
83
E. Ludlum and P. Xia, Comparison of IMRT planning with two-step and
one-step optimization: A way to simplify IMRT, Phys. Med. Biol. 53,
807821 2008.
84
D. M. Shepard, D. Cao, M. K. Afghan, and M. A. Earl, An arcsequencing algorithm for intensity modulated arc therapy, Med. Phys.
34, 464470 2007.
85
P. Mavroidis, B. C. Ferreira, C. Shi, B. K. Lind, and N. Papanikolaou,
Treatment plan comparison between helical tomotherapy and MLCbased IMRT using radiobiological measures, Phys. Med. Biol. 5213,
38173836 2007.
62
5240
86
J. Muzik, M. Soukup, and M. Alber Comparison of fixed beam IMRT,

helical tomotherapy, and IMPT for selected cases, Med. Phys. 354
15801592 2008.
87
D. Cao, T. W. Holmes, M. Afghan, and D. M. Shepard Comparison of
plan quality provide by intensity modulated arc therapy IMAT and helical tomotherapy, Int. J. Radiat. Oncol., Biol., Phys., 691, 240250
2007.
88
G. Zhang, Z. Jiang, D. Shepard, M. Earl, and C. Yu, Effect of beamlet
step-size on IMRT plan quality, Med. Phys. 3211, 34483454 2005.
89
J. M. Galvin, G. Ezzell, A. Eisbrauch, C. Yu, B. Butler, Y. Xiao, I. Rosen,
J. Rosenman, M. Sharpe, L. Xing, P. Xia, T. Lomax, D. A. Low, and J.
Palta, Implementing IMRT in clinical practice: A joint document of the
American Society for Therapeutic Radiology and Oncology and the
American Association of Physicists in Medicine, Int. J. Radiat. Oncol.,
Biol., Phys. 585, 16161634 2004.
90
S. Luan, C. Wang, D. Cao, D. Z. Chen, D. M. Shepard, and C. X. Yu,
Leaf-sequencing for intensity-modulated arc therapy using graph algorithms, Med. Phys. 351, 6169 2008.
91
Y. Kamino, K. Takayama, M. Kokubo, Y. Narita, E. Hira, N. Kawawda, T.
Mizowaki, Y. Nagata, T. Nishidai, and M. Hiraoka, Development of a
four-dimensional image-guided radiotherapy system with a gimble x-ray
head, Int. J. Radiat. Oncol., Biol., Phys. 661, 271278 2006.
92
C. M. Ma, T. Pawlicki, M. C. Lee, S. B. Jiang, J. S. Li, J. Deng, B. Yi, E.
Mok, and A. L. Boyer, Energy- and intensity-modulated electron beams
for radiotherapy, Phys. Med. Biol. 458, 22932311 2000.
93
J. J. Wilkens and U. Oelfke, Optimization of radiobiological effects in
intensity modulated proton therapy, Med. Phys. 322, 455465 2005.
94
A. J. Lomax, E. Pedroni, H. Rutz, and G. Goitein, The clinical potential
of intensity modulated proton therapy, Z. Med. Phys. 143, 147152
2004.
95
D. M. Shepard, M. A. Earl, X. A. Li, S. Naqvi, and C. X. Yu, Direct
aperture optimization: A turnkey solution for step-and-shoot IMRT,
Med. Phys. 296, 10071018 2002.
96
K. Otto, Volumetric modulated arc therapy: IMRT in a single gantry
arc, Med. Phys. 351, 310317 2008.
97
S. Ulrich, S. Nill, and U. Oelfke, Development of an optimization concept for arc-modulated cone beam therapy, Phys. Med. Biol. 5214,
40994119 2007.
98
C. Cameron, Sweeping-window arc therapy: an implementation of rotational IMRT with automatic beam-weight calculation, Phys. Med. Biol.
50, 43174336 2005.
99
G. Tang, M. Earl, S. Luan, S. Naqvi, and C. X. Yu, Converting multiplearc Intensity-modulated arc therapy into a single arc for efficient delivery, Int. J. Radiat. Oncol., Biol., Phys. 693, S673 2007.
100
C. Wang, S. Luan, G. Tang, D. Z. Chen, M. Earl, and C. Yu, Arcmodulated radiation therapy AMRT: A single-arc form of intensitymodulated arc therapy, Phys. Med. Biol. 5322, 62916303 2008.
101
M. A. Earl, M. K. N. Afghan, C. X. Yu, Z. Jiang, and D. M. Shepard,
Jaws-only IMRT using direct aperture optimization, Med. Phys. 341,
307314 2007.
102
Y. Kim, L. J. Verhey, and P. Xia, A feasibility study of using conventional jaws to deliver IMRT plans in the treatment of prostate cancer,
Phys. Med. Biol. 528, 21472156 2007.
103
M. A. Earl et al., Inverse planning for intensity-modulated arc therapy
using direct aperture optimization, Phys. Med. Biol. 48, 10751089
2003.
104
W. Hara, S. G. Soltys, and I. C. Gibbs, CyberKnife robotic radiosurgery
system for tumor treatment, Expert Review Anticancer Therapy 711,
15071515 2007.
105
U. Titt, O. N. Vassiliev, F. Ponisch, L. Dong, H. Liu, and R. Mohan, A
flattening filter free photon treatment concept evaluation with Monte
Carlo, Med. Phys. 336, 15951602 2006.
106
D. A. Jaffray, Emergent technologies for 3-dimensional image-guided
radiation delivery, Semin. Radiat. Oncol. 153, 208216 2005.
107
R. Jeraj, C. Wu, G. H. Olivera, and T. R. Mackie, Re-optimization in
adaptive radiotherapy Phys. Med. Biol. 47, 31813195 2002.
108
R. Mohan, X. Zhang, H. Wang, Y. Kang, X. Wang, H. Liu, K. K. Ang, D.
Kuban, and L. Dong, Use of deformed intensity distributions for on-line
modification of image-guided IMRT to account for interfractional anatomic changes, Int. J. Radiat. Oncol., Biol., Phys. 614, 12581266
2005.
109
Y. Feng, C. Castro-Pareja, R. Shekhar, and C. Yu, Direct aperture
5241
deformationAn inter-fraction image guidance strategy, Med. Phys.

3312, 44904498 2006.
110
E. Ahunbay, G. Chen, C. Peng, C. Yu, S. Narayanan, X. A. Li, and C.
Lawton, An on-line adaptive planning strategy for inter-fraction image
guidance Med. Phys. 358, 36073615 2008.
111
K. Whitton, X. Hu, C. Yu, and D. Z. Chan, An FPGA solution for
5241
radiation dose calculation, IEEE Symposium on Field-Programmable
custom Computing Machines FCCM April 2006.
112
C. C. Ling, J. Humm, S. Larson, H. Amols, Z. Fuks, S. Leibel, and J. A.
Koutcher, Towards multidimensional radiotherapy MD-CRT: Biological imaging and biological conformality, Int. J. Radiat. Oncol., Biol.,
Phys. 473, 551560 2000.
The future of heavy ion radiotherapy

Oliver Jkela
Department of Medical Physics in Radiation Oncology, German Cancer Research Center,
Im Neuenheimer Feld 280, 69120 Heidelberg, Germany and Heidelberg Ion Beam Therapy Center (HIT),
Heidelberg, Germany
Christian P. Karger
Department of Medical Physics in Radiation Oncology, German Cancer Research Center,
Im Neuenheimer Feld 280, 69120 Heidelberg, Germany
Jrgen Debus
Department of Radiation Oncology and Radiotherapy, University Clinic Heidelberg, Heidelberg, Germany
and Heidelberg Ion Beam Therapy Center (HIT), Heidelberg, Germany
Received 5 February 2008; revised 19 September 2008; accepted for publication 22 September 2008;
Currently, there is an increasing interest in heavy ion radiotherapy RT and a number of new
facilities are being installed in Europe and Japan. This development is accompanied by intensive
technical, physical, and clinical research. The authors identify six research fields where progress is
likely and propose a thesis on the expected achievements for each of the fields: 1 Synchrotrons
with active energy variation and three-dimensional beam scanning will be the standard in ion beam
RT. 2 Common standards for precise measurement, prescription, and reporting of dose will be
available. 3 Intensity-modulated particle therapy will be state-of-the-art. 4 Time-adaptive treatments of moving targets will be feasible. 5 Therapeutic effectiveness of heavy ions will be known
for the most important indications while cost effectiveness will remain to be shown. 6 The
potential of high-linear energy transfer radiation will be known. The rationale for each of these
theses is described. 2008 American Association of Physicists in Medicine.
DOI: 10.1118/1.3002307
Key words: radiotherapy, heavy ions, heavy charged particles, relative biological effectiveness
RBE, high-LET
I. BACKGROUND
The interest in heavy ion radiotherapy RT is currently increasing worldwide.14 Similar to protons, heavy ions exhibit
an inverted depth dose profile, the so-called Bragg curve,
which concentrates the dose at the end of the particle range.
By superimposing several monoenergetic Bragg curves, the
Bragg peak can be spreadout to the extension of the tumor in
depth, while still preserving an advantageous dose ratio between tumor and normal tissue. The degree of conformality
of charged particles was further enhanced by introducing active scanning techniques, which allow dose conformation not
only at the distal but also at the proximal edge of the tumor
for each field.5
As the linear energy transfer LET of heavy ions increases significantly towards the end of their range, heavy
ions are biologically more effective in the peak- than in the
plateau region, if the same dose is applied to the same tissue.
This behavior is considered by introducing the relative biological effectiveness RBE, which is defined as the ratio of a
photon dose to the corresponding iso-effective ion dose.5
Therefore, dose has to be prescribed in terms of biologically
equivalent dose =absorbed dose RBE rather than absorbed dose. It is the difference of the increased RBE between the peak and the plateau regions which potentially
increases the therapeutic effectiveness of heavy ion therapy
relative to proton therapy.
5653
To arrive at a biologically homogeneous dose distribution,

a RBE model has to be included into the treatment planning
system. This model has to consider the rather complex dependence of the RBE on LET, dose per fraction, the amount
of projectile fragmentation, the cell or tissue type irradiated,
as well as on the selected biological endpoint. The dependence of the RBE on biological properties means that the
therapeutic benefit may strongly depend on the combination
of the normal- and tumor-tissue type and on the regarded
biological end point e.g., early versus late effects. Experience gained for one tumor entity may not be transferred to
another and the biological properties of heavy ions may even
be disadvantageous for some tumors.
Pioneering work in heavy ion RT has been performed at
the Lawrence Berkley Laboratory USA, where 2054 patients were treated with helium ions between 1957 and
1992.6,7 Between 1975 and 1992, 433 patients have been
treated with heavier ions Ne, N, O, C, Si, Ar.7,8 Currently,
heavy ion RT is performed at three centers in Japan HIMAC, Chiba9,10 and HIBMC, Hyogo11 and Germany GSI,
Darmstadt. These centers all use carbon ions and have
treated approximately 3500 patients 3000 patients at HIMAC until 12/2006, 430 at GSI until 8/2008, and 93 at
HIBMC until 02/2007. The next heavy ion facilities, which
will be brought into clinical operation, are the Heidelberg Ion
Therapy facility HIT, Germany Fig. 1,12,13 the CNAO fa-
0094-2405/2008/3512/5653/11/$23.00
5653
5654
Jkel, Karger, and Debus: The future of heavy ion radiotherapy
5654
FIG. 1. Three-dimensional view of the new HIT facility at the University

Hospital Heidelberg. The path of the ions is outlined by the red line from the
two ion sources on the left for protons and carbon ions, through the linac
injector into the synchrotron upper left part, following the high energy
beam-lines towards the three treatment rooms equipped with two fixed horizontal beam lines and an isocentric gantry. Courtesy by Stern, Gruner
+ Jahr AG & CO KG, Germany.
cility in Pavia Italy,14,15 and the Gunma University Heavy

Ion Medical Center GHMC, Japan. An experimental treatment facility at a research laboratory has also been set up in
Lanzhou China and in November 2006 the treatment of
patients with superficially-seated tumors has started a maximum energy of 100 MeV/ u has been used16. Currently,
heavy ion RT is a matter of intensive technical, physical, and
clinical research and our knowledge about feasibility, safety,
and effectiveness will certainly increase within the next decade.
FIG. 2. For passive beam delivery a the dose distribution is shaped by four
components: the range shifter, the modulator wheel or alternatively a ridge
filter, the collimator, and the compensator. The modulation depth is constant over the tumor cross section. For active beam delivery b, a monoenergetic pencil beam is scanned over the tumor cross section. After one slice
is irradiated the energy of the beam is switched actively or passively, in
case of a range shifter to the next energy. Reprinted from Ref. 4 with
permission of Springer.
II. CURRENT STATUS

Currently, the only available sources for high energy ion
beams in radiotherapy are synchrotrons, which due to their
size and complexity are the main reason for the large capital
investment necessary to build a clinical ion beam facility.
Another aspect of the currently used beam delivery systems
is that they all rely on fixed beam lines: either horizontal
HIMAC and GSI, vertical HIMAC and HIBMC, or inclined by 45 HIBMC, rather than gantry systems.
Both Japanese heavy ion beam facilities rely on a socalled passive beam delivery system similar to the one developed in Berkeley Fig. 2a. This system uses a beam
which is extracted from a synchrotron at a fixed energy and
is then spread laterally as well as in depth, using passive
beam shaping techniques: a wobbler magnet in combination
with a scattering system is used to produce a broad beam,
which is then tailored to the target volume by a collimation
system; the depth dose modulation called the spread-out
Bragg peak is achieved by a ridge filter, variable range
shifter plates, and a patient specific compensator. In contrast,
the newly designed European ion facilities all rely on the
method of active energy variation and pencil beam scanning,
which was developed in the 1990s at the GSI facility Fig.
2b.
It should be noted that a more flexible beam delivery system based on the elements of passive systems has been developed and is explored at HIMAC. This so-called layerstacking method10 uses a sequence of different modulators
each with only a small modulation depth, which are then
stacked in depth to achieve the desired depth dose distribution. A multileaf collimator is used to adapt the field size to
the target in each layer. This method leads to a much better
conformation of the dose to the target volume, similar to a
scanning system.
To efficiently measure dose distributions generated with
the dynamic beam scanning system, multichannel detector
systems are used rather than single ion chamber measurements, which are a standard for passive techniques. Although
the dosimetry of heavy ion beams was already harmonized
by the introduction of the Code of Practice published by the
International Atomic Energy Agency TRS-39817, the overall uncertainty of dose determination in ion beams is still
considerably larger than in conventional radiotherapy. Moreover, there is no common standard for the reporting of biological effective dose and RBE in ion beam therapy.
For a passively modulated beam, the variation of the RBE
with depth has to be taken into account in the design of the
5655
ridge filters or modulators i.e., the hardware. With the selection of a certain modulator, the depth modulation, and
hence, the RBE profile is thus fixed. As the RBE profile
depends not only on the treatment schedule e.g., dose per
fraction, number of daily applied fields but also the type of
tumor and normal tissue or the optimization technique multiple fields versus single field optimization, the flexibility of
passive systems is limited. The realization of an intensity
modulated particle beam therapy IMPT was also never realized with a passive beam application system although in
principle the above mentioned layer-stacking system offers
this possibility.
While it is the current standard in ion beam therapy to
detect interfractional movements of bony structures or metal
markers by orthogonal x rays and to correct for thereafter,
the management of intrafractional movement is more difficult. Due to the quasistatic treatment fields obtained with
passive beam shaping techniques, the treatment of moving
organs is not much different from that in conventional radiotherapy with photons. It has been shown at HIMAC that
respiratory gating of the beam, e.g., by optical motion detection, can be used for ion beam therapy.18 The only additional
problem is the increased treatment time due to the pulsed
beam. For the current beam scanning systems, however, the
treatment of moving organs is a much greater challenge due
to interplay effects between organ motion and the scanned
beam for a detailed discussion see Ref. 19.
The clinical experience gained with ion beams so far is
very limited. The effectiveness of ion beams was demonstrated only for a limited number of tumors and with limited
patient numbers an overview of clinical data is found in
Refs. 9 and 2022. Moreover, the potential advantages of
high LET radiation versus low LET could only be demonstrated by comparison with historical clinical results and not
by directly comparing proton and ion beams in the same
setting. The available clinical data are mainly obtained using
carbon ions and are very scarce for other high LET ions.8
III. WHERE WE WILL BE IN 10 YEARS

Although it is difficult to make predictions about the future, current research activities point to six fields, in which
progress is likely on the one hand, but also mandatory on the
other hand to fully exploit the potential of heavy ion RT. In
the following, we start with a thesis related to each of the six
fields accompanied by a brief description of what we think
the major achievements will be.
1 Beam production and delivery: Synchrotrons with active
energy variation and three-dimensional 3D beam scanning will be the standard in ion beam RT.
Although different compact designs of ion beam accelerators are being investigated today, none of them
will be available for clinical use within the next decade.
The design of compact and maybe superconducting synchrotrons will be optimized in terms of performance,
stability, and costs. The techniques used to precisely
control the position and intensity of an ion beam will
advance strongly and will make 3D pencil beam scanMedical Physics, Vol. 35, No. 12, December 2008
5655
ning for ion beams the standard technique for all new
clinical ion beam facilities. Furthermore, the development of gantry systems for ions will continue and at
least some new facilities will install gantries.
Dosimetry: Common standards for precise measurement, prescription, and reporting of dose will be
available.
The overall uncertainty in determining the absorbed
dose to water in an ion beam will be at the same level as
for megavoltage x rays today. The main improvements
will come from a better knowledge of the underlying
physical parameters. Water calorimetry will serve as reference dosimetry and multichannel dosimetry systems
will facilitate the measurement of two-dimensional 2D
and 3D dose distribution. There will be standardized
methods to report the biological effective dose and the
RBE.
Treatment planning: Intensity-modulated particle
therapy will be state-of-the-art.
The advancements of IMRT in conventional and proton radiotherapy will immediately be introduced in ion
beam therapy. The principles of IMPT based on the optimization of biological effective dose have been developed and, due to the rapid increase of computing power,
will be incorporated in any treatment planning system
for ion beams. IMPT can and will further improve the
conformality of the applied dose distributions. Since
beam delivery by active beam scanning is intrinsically
controlling and modulating the intensity, IMPT will be
much more straightforward to be implemented as it was
for conventional RT with open fields and therefore requires little to no additional effort for quality assurance.
Moving organs: Time-adaptive treatments of moving targets will be feasible.
Although the principal problem is the same for photons and ions, it has to be considered that organ movements do not only change the lateral position of the target, but also its radiological depth, and hence, the
required range of the ions. For heavy ions, changes in
the depth modulation may also require a repeated optimization of the biologically effective dose. In the future,
interfractional movements will be detected by on-board
imaging and will be compensated by setup corrections
and if necessary also by adaptation of the treatment plan.
Intrafractional organ movements will be monitored in
real time and will be compensated to some extent by
using gating or tracking techniques. Improved imaging
technology like on-board cone beam computed tomography CT will also allow direct visualization of the
tumor and soft tissue structures, rather than just bony
structures or markers.
Clinical application: The therapeutic effectiveness of
heavy ions will be known for the most important indications. The cost effectiveness of heavy ion therapy,
however, will still have to be shown.
With the upcoming of new heavy ion centers, the
most important question is related to the safety and effectiveness of this new treatment modality. Both aspects
5656
may depend on the RBE model used for optimization of

the biologically effective dose as well as on the treated
indications. Due to the uncertainties involved in the
RBE models this especially includes the empirical determination of the optimal biologically effective dose. Considerable clinical experience in using heavy ion beams
has been gained in Berkeley and Chiba using passive
beam delivery techniques,6,8,9,11 as well as for active systems at GSI.2023 The number of patients treated with
heavy ions will increase significantly within the next
decade and the indications for which heavy ion therapy
is most promising will be identified.
The cost for a proton or an ion beam facility is considerably higher than for a conventional megavoltage
x-ray accelerator. This is due to the machine costs itself,
but also due to the larger buildings, shielding requirements as well as technical equipment, and additional
manpower needed to operate such a facility. This raises
the question if proton or ion beam therapy can ever be
cost effective. This question is highly complex as its
answer requires reasonable input data on both, therapeutic effectiveness as well as socioeconomic costs.
6 Radiobiology: The potential of high-LET radiation will
be known.
The simpler question of the effectiveness of heavy
ion therapy has to be separated from the more difficult
question, if there is an increased effectiveness due to the
high-LET component of heavy ions. Although the physical dose distributions are quite similar for heavy ions
and protons, heavy ions are expected to be more efficient in the Bragg peak than in the plateau region. Although there are radiobiological as well as some clinical
data supporting this assumption,2,5,8,9 this has to be validated in clinical trials and further experimental studies.
These investigations will lead to an improved understanding of the high-LET effect and the increased biological effectiveness of heavy ions will be provenor
disprovenat least for some indications. Most likely,
these activities will be focused on the comparison of
carbon ions and protons. Other ions such as helium or
oxygen will primarily be investigated within feasibility
studies.
IV. THE DEVELOPMENTS IN THE NEXT DECADE

IV.A. Beam production and delivery
Two new designs of accelerators are currently being discussed, which would be more compact, and hence, cheaper
than current solutions.
The first one uses laser induced acceleration of ions. Here,
a very high-intensity laser incidents on a thin metal target
and causes ionization. An electromagnetic shockwave causes
electrons to be accelerated in the target, which then ionize
the molecules on the backsurface of the target. The cloud of
electrons is then emitted from the backsurface, causing a
very strong electrostatic field 1013 V m1 that accelerates
protons and ions from the backlayer of the target thickness
5656
10 nm. These beams are very intense 1012 protons per

pulse and well collimated. In several research institutions,
protons have been accelerated up to 58 MeV in one institution, when lasers with intensities of about 1020 W cm2
were applied.24 Probably the intensity has to be further increased. The ion pulses are very short in the picosecond
range and intense several picocoulomb charge. The energy
of the heavy charged particles, however, is distributed over a
broad thermal spectrum with only a small intensity at high
energies.
Another concept that has recently been developed is based
on a new class of insulators called high gradient insulators
HGI, which feature significantly improved voltage holding
ability over conventional insulators. Using these HGI, a new
type of induction linac was designed: the so-called dielectric
wall accelerator DWA. A conventional induction linac has
an accelerating field only in the gap of the accelerating cells,
which represent only a small fraction of the length. By replacing the conducting beam pipe by an insulating wall, accelerating fields can be applied uniformly over the entire
length of the accelerator yielding a much higher gradient
around 100 MeV/ m allowing the design of much more
compact linear accelerators. The pulsed acceleration field is
developed by a series of so-called asymmetric Blumlein
structures incorporated into the insulator.25,26 At the
Lawrence Livermore National Lab, a prototype of a protonDWA is currently being built and design studies for a
250 MeV proton accelerator with an overall length of less
than 3 m have been presented.27 This DWA will produce a
100 mA beam current in a short pulse of 1 ns with variable
energy and focus. Such a compact accelerator mounted directly on a gantry would allow the design of much more
compact treatment facilities.
Both systems lead to a much more compact design of the
treatment machines which would drastically reduce the overall costs of the facilities. Consequently, there is a strong financial interest, which might push the development of these
machines, so that these systems may become available for
proton RT at the end of the next decade, for ion beams,
however, they will not be available within this time frame.
The first reason is that due to the higher energy needed, e.g.,
for a 12C beam, the DWA structure would be increased by at
least a factor of 2 in length. Using laser induced accelerators,
even with much increased laser power, the intensities of the
ion beam at these high energies will still be too small. Another unsolved problem is if and how beam scanning can be
realized with the very short and high intensity pulses from a
DWA or laser induction.
Although a single DWA would probably be cheaper than a
synchrotron, it can be used to operate only a single treatment
room, while a synchrotron can easily be operated with 35
treatment rooms. So unless the price of a DWA is significantly lower than for a synchrotron around 20 M, e.g., for
the synchrotron at the HIT facility, this will not be a cost
effective solution. With this respect, laser induction may be
more efficient, as a single laser could in principle be operated
in combination with several treatment rooms.
5657
The research needed to solve all the technical problems

encountered in a compact ion accelerator based on a DWA or
laser induction are still enormous and it is very unlikely that
these will be solved within the next 1 or 2 decades. Rather,
the existing accelerator designs will be optimized using, e.g.,
superconducting magnets to build more compact synchrotrons. For gantries also the use of superconducting fixed field
alternating gradient magnets is discussed to dramatically reduce the overall weight of a carbon gantry from around
600 tons to only 1 2 tons referring to the mass of the
beam-line only, the mass of the mechanical structure is possible in the same range.28
Compact superconducting cyclotrons with diameters of
less than 6 m have also been proposed for ion beam
therapy.29 Since cyclotrons do not allow for an active energy
variation, it has yet to be shown, if these machines will be a
serious alternative to synchrotrons. A more interesting development is the combination of a commercial cyclotron used,
e.g., for isotope production as an injector for a compact
linear accelerator of about 20 m length, acting as a booster.
The new system is called cyclinac15 and has the advantage,
that no extraction or injection is needed. Thus, a continuously extracted beam could be produced with a 1 s pulse
per millisecond and fast energy variation also within a millisecond. Especially this latter feature would be useful for the
treatment of moving organs.
IV.B. Dosimetry
The Code of Practice for dosimetry TRS-39817 is used

in both, the Japanese and German ion beam facilities and can
be considered a standard today. Accordingly, the dose in
heavy ion fields can already be measured with an uncertainty
of 3% and further improvements will be rather small. Nevertheless, the basic quantities, like stopping powers for ions,
fragmentation spectra, and w values, but also the chamber
specific correction factors have to be determined more precisely to arrive at the same level of accuracy as for photon
therapy. Moreover, the beam application using dynamic raster scanning asks for multichannel detector systems that still
have to be developed.
Determination of absorbed dose to water currently relies
of ionization chambers calibrated in a 60Co beam. In the long
term, water calorimeters will allow a more direct measurement of absorbed dose to water.30 The only correction appearing in calorimetry is the so-called heat deficiency, which
quantifies the amount of radiation energy converted into
chemical binding energies rather than into heat. This correction typically is less than 1% andaccording to current
knowledgedepends only little on the beam quality.31 If
these preliminary findings will be confirmed, it is very likely
that calorimetry can be used as primary standard for ion
beams in the same way as already done for megavoltage
x-rays today in some countries.
Another important gain of knowledge of physical parameters will come from Monte Carlo simulations which are just
being established as a new tool for dosimetry of ion beams.
This will affect especially the knowledge of secondary parMedical Physics, Vol. 35, No. 12, December 2008
5657
ticle spectra and the chamber specific correction factors.

There are two major directions that will be followed here:
one is the improvement of nuclear physics models to calculate fragmentation cross sections and secondary particle
spectra and the other one is the inclusion of secondary electrons produced by this spectrum of charged particles and
which is important to understand the chamber specific corrections.
In combination with the broader introduction of dynamic
beam scanning systems, the development of efficient multichannel dosimetry systems will continue. There are already a
number of detector systems developed for particle physics
applications that are currently investigated in view of their
use in proton and ion beam dosimetry. This includes socalled gas electron multiplier GEM chambers, silicon pixel
detectors, or polycrystalline planar diamond detectors.
GEM chambers32 can be used as large area tracking detectors with high temporal and spatial resolution 1 ms and
40 m, respectively. In combination with a scintillator for
residual energy measurements they were used already for
time resolved radiography or as beam profile monitors.33
Silicon pixel detectors with pixel sizes around 50 m allow measuring single particle events with little to no
background.34,35 These so-called active nuclear emulsions
have been used for x ray and neutron tomography to resolve
tiny anatomical structures, e.g., in insects.36 The detector signal is proportional to the energy loss in the detector and may
be used to discriminate particles and thus allow for a direct
dose determination.
Finally, extremely thin less than 50 m polycrystalline
diamond detectors with millimeter resolution can yield a
100% efficiency for particle counting even at high count
rates, which makes them interesting for monitoring of the
beam position, intensity, and profile during application while
minimizing the interaction with the beam.37
Finally, the prescription, recording, and reporting of biological effective doses and RBE will be standardized when
more clinical facilities become operational. This especially
includes specification of the applied RBE model, its version,
the physical and biological parameters used for the RBE calculation in tumor and normal tissues as well as the biological
end point to which the RBE-calculation refers to. This problem is currently being addressed by a new committee of the
International Commission on Radiation Units.38
IV.C. Treatment planning
The prototype system for radiotherapy planning for a

scanned carbon ion beam developed for the GSI facility already incorporates all features needed for intensity modulated ion beam therapy. To achieve a homogenous biological
effect even for a single field, it is necessary to optimize and
control the number of ions at each scan spot. This intrinsically implies a modulation of the intensity or rather fluence
of the ion beam at each beam spot. Since the depth dose
modulation in that case is varying throughout the radiation
field, a detailed biological modeling of the RBE had to be
included in the treatment planning system.39
5658
5658
FIG. 3. Distribution of the biological effective dose for the irradiation of a skull base tumour, using two nearly opposing lateral fields of carbon ion ions. The
treatment plans on the left and right side were generated with a separate and a simultaneous optimization of the dose distributions in both fields, respectively.
The target volume thick line is shown together with the isodose lines at 90%, 80%, 70%, 50%, 30%, and 10% of the prescribed dose, respectively.
Simultaneous optimization of multiple fields is an important feature of IMPT and its technical realization is in principle just a numerical problem, since the particle numbers of
all beam spots of all fields have to be optimized at the same
time, rather than separately. Due to the large number of beam
spots per field typically 50 000 for a skull base tumor, the
required computer memory and computing time is currently
considerably typically 1.2 GB and 4000 s on a 1.2 GHz
Power4-CPU, respectively. With the expected increase in
computing power, however, this will not be a serious restriction in the near future. IMPT has already been clinically
applied at GSI with the existing scanning technology and
without additional efforts for the quality assurance.40,41 Starting from this basis, clinical application of IMPT will continuously be developed. Figure 3 shows an example of an
IMPT plan including optimization of biological effective
doses.
Due to the sharp dose gradients especially at the distal end
of the Bragg peak, setup errors, organ movements, and range
uncertainties may have significant impact on the quality of
the dose distributions. Resulting over- and undershooting has
to be considered by adequate safety margins, which may be
determined, e.g., by simulated target point displacements or
modified CT-number/range-calibration curves. By introducing IMPT, it has been shown that setup errors do not lead to
worse treatment plans than without IMPT and therefore, robustness against setup-related range uncertainties will be included as an additional constraint in the dose optimization
process.
Since the number of fields applied in ion beam therapy is
generally smaller than in photon IMRT, the selection of
proper beam angles becomes more important. As the range
uncertainty may vary significantly for different beam angles,

automatic optimization algorithms for the beam angles
should be designed to reduce these uncertainties to a minimum.
IV.D. Moving organs
IV.D.1. Interfractional motion

In upcoming heavy ion facilities,15 on-board x-ray imaging devices will be state-of-the-art.12,13 These devices may be
used to detect translational or rotational setup errors between
fractions by planar or cone-beam imaging. If no significant
deformations are present, translational setup corrections can
be automatically applied to the treatment couch. In case of a
robotic couch, rotational corrections around three axes will
also be feasible.
If deformations are significant, however, an adaptive approach would need a quasi-instantaneous modification of
contours and reoptimization of the treatment plan. For heavy
ion RT, however, reoptimization must also include the biologically effective dose and for the near future, it seems not
very likely that this can be realized in an acceptable time. As
an alternative, different treatment plans for different representative geometries may be prepared. In the actual treatment
situation, the most suitable plan can then be selected.
IV.D.2. Intrafractional motion
The most significant source of intrafractional movements
is respiratory motion,42,43 although motion of heart44 or prostate induced, e.g., by peristaltic motion and variation of organ filling45,46 may also be important. To compensate for
respiratory motion, several concepts have been developed.
5659
Increasing margins as recommended for photon RT Ref.

47 may assure target coverage. In contrast to photon RT,
however, motion-related changes of the required particle
range may also lead to large margins in beam direction,48 and
hence, to additional high dose contributions in normal tissue.
If dose is delivered actively, target coverage may be disturbed by interplay effects between beam scanning and target
movements, which may lead to hot or cold spots,50 and
hence, cannot be considered by increased margins. To compensate for such effects, a rescanning technique was
suggested,50 which averages the effect of irradiations at different breathing phases.
More advanced techniques attempt to compensate respiratory motion by gating on a stable breathing phase10,18 or by
tracking of the tumor motion.19 For passive beam delivery
systems, gating is already clinically applied.10,18 For active
systems, however, irradiation time may increase significantly
due to the interplay between respiratory motion and duty
cycle of the synchrotron. This delivery time may be reduced
to a certain extent by gating also the extraction of the beam
from the synchrotron. Remaining interplay effects due to residual motion of the target may be compensated by
rescanning.50
Successful tracking of tumor motion would lead to optimal dose coverage of the target while minimizing the dose to
normal tissue. This approach is only feasible with a scanning
beam system and some developments have already been
made.49,51,52 Here, the change of the lateral tumor position is
compensated by the scanner magnets in real time. Changes
of radiological depth are additionally corrected by a
computer-controlled range shifter consisting of opposing
PMMA polymethyl methacrylate wedges.52 Depending on
the actual breathing phase, the beam delivery system
switches to one out of several treatment plans, each of them
optimized on a different phase of a four-dimensional
4D-CT.51
The necessity to calculate the range on a 4D-CT may
involve additional uncertainties as 4D-CT protocols differ
considerably from conventional CT protocols, and hence, the
relation between Hounsfield numbers and range may also be
different. This aspect has to be investigated thoroughly prior
to clinical application 4D-CTs for ion beam therapy.
Although technical aspects of tracking have already been
developed and the potential benefit for the patient has been
demonstrated,51,52 the most severe obstacle is the real-time
determination of the actual tumor position. Using surrogate
signals from external markers or breathing belts during treatment planning CT and the actual treatment may not be sufficient as the correlation of these signals with the tumor position is not always reproducible.5355 Therefore, it seems
likely that additional fluoroscopic images are required to assure that the tumor position is as expected. This, however,
requires real-time image analysis and, in contrast to photon
therapy, implantation of metallic fiducials as landmarks may
have a larger impact on the range of the particles and thus the
dose distribution depending on the size and location of the
marker.
In view of these problems, it is likely that only gating will
5659
be clinically implemented in the near future as this technique

solely requires reproducibility of a single breathing phase
rather than that of the complete breathing cycle.
IV.E. Clinical application
IV.E.1. Therapeutic effectiveness

A relatively large number of indications has already been
investigated in numerous clinical trials with carbon ion RT
using the passive beam delivery and a lot of experience on
safety and effectiveness of this method has been gained.9
Also for active scanning systems a number of clinical trials
have been performed but the experience is limited to very
few indications, where organ motion is not important.20,21,23
As the depth modulation for active techniques varies over the
tumor cross section, the RBE has to be calculated locally at
each point in the treatment field rather than by using fixed
RBE profiles. As compared to passive techniques, this may
introduce an additional source of uncertainty.
With upcoming of the new hospital-based facilities with
active beam delivery systems additional indications will be
investigated. These will include other malignant brain tumors, prostate carcinomas, pancreatic tumors and, if motion
compensation techniques for scanned beams are better understood, also lung tumors. For some indications, the use of
ion beams only may be problematic due to the involved
range uncertainties. For these indications, a combined treatment with photon IMRT and an ion beam boost may be of
significant impact. To find the optimal dose, the experience
obtained with passive beam delivery systems can be valuable.
It is likely that initial studies will be performed with carbon ions using relatively conservative fractionation regimes
as they have been applied, e.g., with the scanned beam at
GSI for most of the patients with doses of 3 3.5 GyE per
fraction. For the new indications, dose escalation studies
have to be performed to maximize local control at acceptable
normal tissue toxicity rates. Hypofractionated treatment
schedules may also be of interest. These, however, have to be
invented very carefully as the fractionation effect of heavy
ions in the plateau region is known to be similar as for photons, and hence, hypofractionation may lead to high biologically effective doses in the normal tissue. Some experience
may be gained from the clinical trials at HIMAC, where
several hypofractionated schedules were investigated.9
When scanned beams are applied to moving organs, the
potential interplay between the scanned beam and organ motion requires a new indication-related assessment of safety
and efficiency. To achieve this, additional clinical phase I/II
trials have to be performed. Similarly, if new or modified
RBE models are to be introduced in clinical routine, they
may lead to changes in the delivered biologically effective
dose, which have to be evaluated in clinical studies.
Of special interest for ion beam therapy is the treatment of
radioresistant tumors. Therefore, a central issue is the investigation of the effectiveness for hypoxic tumors, which are
known to be highly radioresistant against low-LET radiation.
The increase of the required dose for hypoxic tumors is de-
5660
5660
pediatric patients. For scanning techniques, the neutron dose

is estimated to be 4.3 mSv/ GyE,57 but further measurements
as well as clinical long term follow-up of patients is needed
to quantify the risk of the neutron component. Generally, a
larger number of neutrons is produced in a carbon as compared to a proton beam. Even more important may be the
fact, that the amount of neutrons produced in a passive beam
delivery system is significantly larger than for active beam
delivery.57 Especially for pediatric patients, active techniques
should be preferred.
IV.E.2. Cost effectiveness
FIG. 4. Distribution of OER at 10% survival level for V79 and HSG cells
exposed to helium He: , carbon 12C: , or neon 20Ne: ions as a
function of the dose-averaged LET. Reprinted from Ref. 63 with permissions of the author and Radiation Research.
scribed by the so-called oxygen enhancement ratio OER,

which has a value of 1 for well oxygenated tumors and values of up to 3 for hypoxic tumors. There is some experimental evidence that the OER decreases, if LET is increased
Fig. 4. This would indicate that the increased radioresistance of hypoxic tumors is significantly smaller for heavy ion
than for photon RT. This potential advantage, however, has
to be shown clinically and up to now, only very few data on
the comparison of the effectiveness of heavy ions for hypoxic and normoxic tumors are available.56 As the treatment
of hypoxic tumors is still a major challenge in radiotherapy,
future studies with heavy ions have to focus on these tumor
types.
To address the biological question, which ion type is clinically most effective, safety and efficiency has also to be
proven also for other ion types e.g., oxygen. Moreover, as
local control is only one aspect of oncologic diseases, interactions of heavy ion RT with additional chemotherapy, hormone, or targeted therapies have to be investigated. This is
especially important in cases, where a significant improvement in local control can be achieved with ion beams, but
where overall survival is still limited because of distant metastases like it was shown, e.g., for adenoidcystic carcinoma, see Ref. 23. Also the possibility of synergistic effects
between high LET radiation and other therapies has to be
considered.
Apart from therapeutic effectiveness, the possibility of
secondary malignancy due to neutrons produced by fragmentations is an important issue, especially for the treatment of
Reliable data on the costs as well as on the increased

therapeutic effectiveness relative to conventional RT is already very limited for proton and even more for carbon ion
RT. It has been shown in 2003 by Goitein and Jermann that
the relative cost for proton therapy versus conventional
therapy currently is around 2.5 times higher,58 if a recovery
of the capital investments is required. It was estimated that
until 2013 this ratio may decrease to 2.11.7. Without a recovery of investment costs, the ratio could be lowered to a
value between 1.6 and 1.3. Concerning the costs of carbon
ion beam therapy, several European studies concluded that
the costs for a single fraction of carbon beam therapy currently is around 1000 based on a 20 fraction treatment, see
Ref. 59, and references therein. Interestingly, this is nearly
identical to the costs per fraction given in Ref. 58 for proton
therapy.
A cost-effectiveness analysis for proton radiotherapy has
been performed for various tumors breast, prostate, head &
neck, and childhood medulloblastoma in Sweden,60 using a
Markov simulation and calculating the gain in quality adjusted life years. The results indicated that proton therapy for
these tumors was cost effective, if appropriate risk groups
were chosen. However, also the general lack of data and
large uncertainties in the analysis were mentioned. The study
can also be questioned, since many assumptions on the outcome of proton therapy had to be made without being based
on solid clinical data see Ref. 61 for a detailed discussion.
Also recently a literature review on the cost effectiveness has
been published,61 but could not draw any firm conclusions on
the efficacy of proton or carbon beam therapy.
The only published cost-effectiveness analysis for carbon
ion beam therapy has been performed by our group59 for
chordoma patients. For a limited collective of ten patients,
the overall treatment costs have been calculated and were
compared to the costs using conventional therapy. It has been
concluded, that carbon beam therapy is cost effective, if local
tumor control used as indicator for therapy outcome of
more than 70% can be achieved. The main cost driver in the
treatment of skull base tumors that was identified here was
the neurosurgical resection of recurrent tumors. The results
can therefore not be generalized to other tumors and further
investigations are needed.
Given the poor data base for costs and efficacy even for
proton radiotherapy, it cannot be expected that sufficient data
5661
on the cost effectiveness of ion beam therapy become available before solid clinical data on the clinical effectiveness
are available.
IV.F. Radiobiology
Although the clinical prove of safety and effectiveness of

heavy ion RT is highly important, this question has to be
clearly separated from the question, whether the high-LET
component of heavy ion RT is of additional advantage for the
patient. To answer this question, thoroughly designed clinical
trials supplemented by systematic experimental studies are
required.
IV.F.1. Clinical trials
Preliminary results of carbon ion RT are impressive,9,2023
however, they do not allow definitive conclusion on superiority of heavy ions over protons. For skull-base chordomas,
e.g., local control rates have been found to be comparable to
those of protons or even better, if dose was escalated.21 This
comparison, however, may be biased since heavy ion treatments were performed with scanning techniques, while most
of the proton studies used passive beam delivery techniques.
Moreover, the base line characteristics of patients and tumors
may not be comparable between these studies. A quasirandomized phase I/II trial comparing carbon ions versus photon IMRT showed significantly improved local control rates
for adenoidcystic carcinomas treated with a carbon ion
boost.23 In this study, randomization was implicitly attained
by the temporal availability of this irradiation modality during the year. For definitive conclusions, however, prospective
randomized-controlled phase III trials have to be performed,
which compare local control rates at the same toxicity level
using identical beam delivery techniques for protons and
heavy ions, respectively. For combined treatments of photon
IMRT and ion beam boosts, such comparisons must also include comparisons against a pure photon treatment. In both
cases, the already performed phase I/II trials will serve as a
basis for such investigations and adenoidcystic carcinoma,
chordoma, and chondrosarcoma, as well as advanced prostate carcinomas are certainly good candidates for the investigation of the potential benefit of heavy ions for patients.
While the direct comparison of proton and ion beams became possible for the first time already at the HIBMC facility in Hyogo Japan,11 such trials will also be in the focus of
the HIT facility,12,13 where a gantry will be available for the
first time, followed by other upcoming heavy ion centers.
IV.F.2. Experimental studies
Clinical trials are strongly limited with respect to selection of dose and toxicity level. They must therefore be
supplemented by systematic studies in animals to investigate
the predictive power of the underlying RBE model. These
studies have to include measurements of RBE and / values for normal tissue as well as tumor systems.62
While the investigation of the tumor response must
clearly be focused on the spreadout Bragg-peak region, norMedical Physics, Vol. 35, No. 12, December 2008
5661
mal tissue studies have to investigate both, the Bragg peak as

well as the plateau region. Especially, measurements in the
region proximal to the Bragg peak are highly important as
combinations of high absorbed dose and increased RBE values may be found here. Since side effects in normal tissues
typically occur near to the target volume, the key question is
whether the related biologically effective dose is lower or
higher than that obtained with proton irradiations.
Other relevant issues refer to the impact of hypofractionated treatments on normal tissue tolerance and the influence
of the oxygen status on tumor response. Although the tolerance data obtained from animal studies cannot directly be
transferred to patients, these studies offer the chance to principally prove a biological advantage of heavy ions. Apart
from dose response studies, additional investigations on a
molecular level are necessary to improve the understanding
of the underlying mechanisms of high-LET irradiations.
V. CONCLUSIONS
In the past decade, the fundamentals of heavy ion radiotherapy have been developed and clinical application has
shown to be feasible and effective. Synchrotrons in combination with 3D-scanning systems are and will continue to be
state-of-the-art. In the next decade the accuracy of absorbed
dose determination will be increased to the level of photon
therapy and intensity modulated ion beam therapy will advance to a routinely applied treatment technique for selected
indications. The most challenging aspects of heavy ion
therapy in the future will be to identify these indications and
to prove its safety and efficiency as well as of its additional
benefit with respect to proton therapy. Cost effectiveness of
ion beam therapy has to be investigated further. It is, however, not likely that substantial data will be available before
its therapeutic benefit has been quantified. For some indications, effectiveness strongly relies on, successful compensation of inter- and intrafractional motion. Form a strictly scientific view, the future role of heavy ion therapy in the
management of oncologic diseases is still unclear, from our
view today, however, it appears to be very promising.
a
Author to whom correspondence should be addressed. Present address:

Dept. of Medical Physics in Radiation Oncology E040, German Cancer
Research Center, Im Neuenheimer Feld 280, 69120 Heidelberg, Germany. Telephone: 49-6221-42-2596; 49-6221-56-37668; Fax: 496221-42-2572. Electronic mail: [email protected]
1
O. Jkel, D. Schulz-Ertner, C. P. Karger, A. Nikoghosyan, and J. Debus,
Heavy ion therapy: Status and perspectives, Technol. Cancer Res.
Treat. 2, 377387 2003.
2
D. Schulz-Ertner, O. Jkel, and W. Schlegel, Radiation therapy with
charged particles, Semin. Radiat. Oncol. 16, 249259 2006.
3
O. Jkel, M. Krmer, D. Schulz-Ertner, P. Heeg, C. P. Karger, B.
Didinger, A. Nikoghosyan, and J. Debus, Treatment planning for carbon
ion radiotherapy in Germany: Review of clinical trials and treatment planning studies, Radiother. Oncol. 73, S86S91 2004.
4
C. P. Karger and O. Jkel, Current status and new developments in ion
therapy, Strahlenther. Onkol. 183, 295300 2007.
5
G. Kraft, Tumor therapy with heavy charged particles, Prog. Part. Nucl.
Phys. 45, S473S544 2000.
6
J. R. Castro, D. E. Linstadt, J. P. Bahary, P. L. Petti, I. Daftari, J. M.
Collier, P. H. Gutin, G. Gauger, and T. L. Phillips, Experience in charged
particle irradiation of tumors of the skull base: 19771992, Int. J. Radiat.
Oncol., Biol., Phys. 29, 647655 1994.
5662
Particle Newsletter 36, edited by J. Sisterson Harvard Cyclotron Laboratory, Harvard University, Boston, 2005, www.ptcog.com.
8
J. R. Castro, Results of heavy ion radiotherapy, Radiat. Environ. Biophys. 34, 4548 1995.
9
H. Tsujii et al., Clinical results of carbon ion radiotherapy at NIRS, J.
Radiat. Res. Tokyo 48 Suppl A, A1A13 2007.
10
M. Torikoshi, S. Minohara, N. Kanematsu, M. Komori, M. Kanazawa, K.
Noda, N. Miyahara, H. Itoh, M. Endo, and T. Kanai, Irradiation System
for HIMAC, J. Radiat. Res. Tokyo 48 Suppl A, A15A25 2007.
11
Y. Hishikawa, Y. Oda, H. Mayahara, A. Kawaguchi, K. Kagawa, M.
Murakami, and M. Abe, Status of the clinical work at Hyogo, Radiother. Oncol. 73 Suppl 2, S38S40 2004.
12
P. Heeg, H. Eickhoff, and T. Haberer, Conception of heavy ion beam
therapy at Heidelberg University HICAT, Z. Med. Phys. 14, 1724
2004.
13
T. Haberer, J. Debus, H. Eickhoff, O. Jkel, D. Schulz-Ertner, and U.
Weber, The Heidelberg Ion Therapy Center, Radiother. Oncol. 73
Suppl 2, S186S190 2004.
14
U. Amaldi, CNAOThe Italian Centre for Light-Ion Therapy, Radiother. Oncol. 73 Suppl 2, S191S201 2004.
15
U. Amaldi and G. Kraft, European developments in radiotherapy with
beams of large radiobiological effectiveness, J. Radiat. Res. Tokyo 48
Suppl A, A27A41 2007.
16
Q. Li, Z. Dai, Z. Yan, X. Jin, X. Liu, and G. Xiao, Heavy-ion conformal
irradiation in the shallow-seated tumor therapy terminal at HIRFL, Med.
Biol. Eng. Comput. 4511, 10371043 2007.
17
P. Andreo, D. T. Burns, K. Hohlfeld, M. S. Huq, T. Kanai, F. Laitano, V.
Smyth, and S. Vynckier, Absorbed dose determination in external beam
radiotherapy, Technical Report Series No. 398 2000.
18
S. Minohara, T. Kanai, M. Endo, K. Noda, and M. Kanazawa, Respiratory gated irradiation system for heavy-ion radiotherapy, Int. J. Radiat.
Oncol., Biol., Phys. 47, 10971103 2000.
19
C. Bert, N. Saito, A. Schmidt, N. Chaudhri, D. Schardt, and E. Rietzel,
Target motion tracking with a scanned particle beam, Med. Phys. 34,
47684771 2007.
20
D. Schulz-Ertner, A. Nikoghosyan, H. Hof, B. Didinger, S. E. Combs, O.
Jkel, C. P. Karger, L. Edler, and J. Debus, Carbon ion radiotherapy of
skull base chondrosarcomas, Int. J. Radiat. Oncol., Biol., Phys. 67, 171
177 2007.
21
D. Schulz-Ertner, C. P. Karger, A. Feuerhake, A. Nikoghosyan, S. E.
Combs, O. Jkel, L. Edler, M. Scholz, and J. Debus, Effectiveness of
carbon ion radiotherapy in the treatment of skull-base chordomas, Int. J.
22
D. Schulz-Ertner and H. Tsujii, Particle radiation therapy using proton
and heavier ion beams, J. Clin. Oncol. 25, 953964 2007.
23
D. Schulz-Ertner, A. Nikoghosyan, B. Didinger, M. Munter, O. Jkel, C.
P. Karger, and J. Debus, Therapy strategies for locally advanced adenoid
cystic carcinomas using modern radiation therapy techniques, Cancer
104, 338344 2005.
24
K. W. Ledingham, W. Galster, and R. Sauerbrey, Laser-driven proton
oncologyA unique new cancer therapy?, Br. J. Radiol. 80, 855858
2007.
25
G. J. Caporaso, 1994 US-CERN-Japan Joint Accelerator School: Topical
Course: Frontiers Of Accelerator Technology, 39 Nov 1994, Maui, Hawaii, edited by M. M. S. T. S. I. Kurokawa World Scientific, New Jersey,
1996, pp. 594615.
26
S. E. Sampayan, G. J. Caporaso, D. M. Sanders, R. D. Stoddard, D. O.
Trimble, J. Elizondo, M. L. Krogh, and T. F. Wieskamp, Particle Accelerator Conference, 1997. Proceedings of the 1997 IEEE, Lawrence Livermore Nat. Lab., 1997, pp. 13081310.
27
Y.-J. Chen and A. C. Paul, Proceedings of PAC07, Albuquerque, New
Mexico, USA, edited by S. Hardage and C. Petit-Jean-Genaz IEEE, Piscataway, NJ, 2007, pp. 17871789.
28
D. Trbojevic, D. Parker, E. Keil, and A. M. Sessler, Carbon/proton
therapy: A novel gantry design, Phys. Rev. ST Accel. Beams 2007,
053503-1053503-6 2007.
29
Y. Jongen et al., Design Studies of the Compact Superconducting Cyclotron for Hadron Therapy, Abstract TUPLS078, Proceedings of EPAC,
Edinburgh, Scotland, 16781680 2006.
30
H. Palmans, J. Seuntjens, F. Verhaegen, J. M. Denis, S. Vynckier, and H.
Thierens, Water calorimetry and ionization chamber dosimetry in an
85-MeV clinical proton beam, Med. Phys. 23, 643650 1996.
31
H. J. Brede, K. D. Greif, O. Hecker, P. Heeg, J. Heese, D. T. Jones, H.
7
5662
Kluge, and D. Schardt, Absorbed dose to water determination with ionization chamber dosimetry and calorimetry in restricted neutron, photon,
proton and heavy-ion radiation fields, Phys. Med. Biol. 51, 36673682
2006.
32
F. Sauli, GEM: A new concept for electron amplification in gas detectors, Nucl. Instrum. Methods Phys. Res. A 386, 531534 1997.
33
G. T. Y. Chen, S. Mori, A. Chu, and M. Folkert, Uncertainties in image
guided radiotherapy of lung tumors, Proceedings of NIRS-CNAO Joint
Symposium on Carbon Ion Radiotherapy, Milan, Italy, 2006, pp. 192193
2006.
34
C. Zhang, P. Lechner, G. Lutz, M. Porroe, R. Richter, J. Treisa, L.
Strder, and S. N. Hang, Development of DEPFET Macropixel detectors, Nucl. Instrum. Methods Phys. Res. A 568, 207216 2008.
35
J. Jakubek, T. Holy, E. Lehmann, S. Pospisil, J. Uher, D. Vacik, and D.
Vavrik, Spatial resolution of Medipix-2 device as neutron pixel detector, Nucl. Instrum. Methods Phys. Res. A 546, 164169 2005.
36
J. Jakubek et al., Neutron imaging and tomography with Medipix2 and
dental micro-roentgenography, Nucl. Instrum. Methods Phys. Res. A
569, 205209 2006.
37
M. Rebisz, B. Voss, A. Heinz, E. Usenko, and M. Pomorski, CVD diamond dosimeters for heavy ion beams, Diamond Relat. Mater. 16, 1070
1073 2007.
38
A. Wambersie, J. H. Hendry, P. Andreo, P. M. DeLuca, R. Gahbauer, H.
Menzel, and G. Whitmore, The RBE issues in ion-beam therapy: Conclusions of a joint IAEA/ICRU working group regarding quantities and
units, Radiat. Prot. Dosim. 122, 463470 2006.
39
M. Krmer and M. Scholz, Rapid calculation of biological effects in ion
radiotherapy, Phys. Med. Biol. 51, 19591970 2006.
40
M. Ellerbrock, O. Jkel, M. Krmer, A. Nikoghosyan, and D. SchulzErtner, Clinical implementation of intensity modulated heavy ion
therapy, Radiother. Oncol. 84, 129 2007.
41
M. Ellerbrock, O. Jkel, M. Krmer, A. Nikoghosyan, and D. SchulzErtner, Multiple field optimization in heavy ion treatment planning
clinical aspects, in Abstracts der Frhjahrstagung in Heidelberg Deutsche Physikalische Gesellschaft DPG ST 6.4, Heidelberg, 2006.
42
M. Engelsman, E. Rietzel, and H. M. Kooy, Four-dimensional proton
treatment planning for lung tumors, Int. J. Radiat. Oncol., Biol., Phys.
64, 15891595 2006.
43
E. Rietzel, S. J. Rosenthal, D. P. Gierga, C. G. Willet, and G. T. Chen,
Moving targets: Detection and tracking of internal organ motion for
treatment planning and patient set-up, Radiother. Oncol. 73 Suppl 2,
S68S72 2004.
44
S. Mori, G. T. Chen, and M. Endo, Effects of intrafractional motion on
water equivalent pathlength in respiratory-gated heavy charged particle
beam radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 69, 308317
2007.
45
P. Kupelian, T. Willoughby, A. Mahadevan, T. Djemil, G. Weinstein, S.
Jani, C. Enke, T. Solberg, N. Flores, D. Liu, D. Beyer, and L. Levine,
Multi-institutional clinical experience with the Calypso System in localization and continuous, real-time monitoring of the prostate gland during
external radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 67, 10881098
2007.
46
D. W. Litzenberg, J. M. Balter, S. W. Hadley, H. M. Sandler, T. R.
Willoughby, P. A. Kupelian, and L. Levine, Influence of intrafraction
motion on margins for prostate radiotherapy, Int. J. Radiat. Oncol., Biol.,
Phys. 65, 548553 2006.
47
ICRU Report 62, Prescribing, Recording and Reporting Photon Beam
Therapy supplement to ICRU report 50 International Commission on
Radiation Units and Measurements, Washington, DC, 1999.
48
S. Minohara, M. Endo, T. Kanai, H. Kato, and H. Tsujii, Estimating
uncertainties of the geometrical range of particle radiotherapy during respiration, Int. J. Radiat. Oncol., Biol., Phys. 56, 121125 2003.
49
C. Bert, S. O. Grozinger, and E. Rietzel, Quantification of interplay
effects of scanned particle beams and moving targets, Phys. Med. Biol.
53, 22532265 2008.
50
T. Furukawa, T. Inaniwa, S. Sato, T. Tomitani, S. Minohara, K. Noda, and
T. Kanai, Design study of a raster scanning system for moving target
irradiation in heavy-ion radiotherapy, Med. Phys. 34, 10851097 2007.
51
C. Bert and E. Rietzel, 4D treatment planning for scanned ion beams,
52
S. O. Grozinger, E. Rietzel, Q. Li, C. Bert, T. Haberer, and G. Kraft,
Simulations to design an online motion compensation system for
scanned particle beams, Phys. Med. Biol. 51, 35173531 2006.
5663
53
C. Bert, K. G. Metheany, K. P. Doppke, A. G. Taghian, S. N. Powell, and

G. T. Chen, Clinical experience with a 3D surface patient setup system
for alignment of partial-breast irradiation patients, Int. J. Radiat. Oncol.,
Biol., Phys. 64, 12651274 2006.
54
J. D. Hoisak, K. E. Sixel, R. Tirona, P. C. Cheung, and J. P. Pignol,
Prediction of lung tumour position based on spirometry and on abdominal displacement: accuracy and reproducibility, Radiother. Oncol. 78,
339346 2006.
55
J. D. Hoisak, K. E. Sixel, R. Tirona, P. C. Cheung, and J. P. Pignol,
Correlation of lung tumor motion with external surrogate indicators of
respiration, Int. J. Radiat. Oncol., Biol., Phys. 60, 12981306 2004.
56
T. Nakano, Y. Suzuki, T. Ohno, S. Kato, M. Suzuki, S. Morita, S. Sato, K.
Oka, and H. Tsujii, Carbon beam therapy overcomes the radiation resistance of uterine cervical cancer originating from hypoxia, Clin. Cancer
Res. 12, 21852190 2006.
57
K. Gunzert-Marx, D. Schardt, and R. S. Simon, The fast neutron component in treatment irradiations with 12C beam, Radiother. Oncol. 73
Suppl 2, S92S95 2004.
58
M. Goitein and M. Jermann, The relative costs of proton and X-ray
5663
radiation therapy, Clin. Oncol. R Coll. Radiol 15, S37S50 2003.

O. Jkel, B. Land, S. E. Combs, D. Schulz-Ertner, and J. Debus, On the
cost-effectiveness of carbon ion radiation therapy for skull base chordoma, Radiother. Oncol. 83, 133138 2007.
60
J. Lundkvist, M. Ekman, S. R. Ericsson, B. Jonsson, and B. Glimelius,
Proton therapy of cancer: Potential clinical advantages and costeffectiveness, Acta Oncol. 44, 850861 2005.
61
M. Lodge, M. Pijls-Johannesma, L. Stirk, A. J. Munro, R. D. De, and T.
Jefferson, A systematic literature review of the clinical and costeffectiveness of hadron therapy in cancer, Radiother. Oncol. 83, 110122
2007.
62
C. P. Karger, P. Peschke, R. Sanchez-Brandelik, M. Scholz, and J. Debus,
Radiation tolerance of the rat spinal cord after 6 and 18 fractions of
photons and carbon ions: Experimental results and clinical implications,
63
Y. Furusawa, K. Fukutsu, M. Aoki, H. Itsukaichi, K. Eguchi-Kasai, H.
Ohara, F. Yatagai, T. Kanai, and K. Ando, Inactivation of aerobic and
hypoxic cells from three different cell lines by accelerated 3He-, 12C-,
and 20Ne-ion beams, Radiat. Res. 154, 485496 2000.
59
Locating and targeting moving tumors with radiation beams

Sonja Dieterich
Radiation Oncology, Stanford University, Stanford, California 94305 and Radiation Oncology,
Georgetown University Hospital, Washington, DC 20007
Kevin Cleary
Imaging Science and Information Systems, Georgetown University Hospital, Washington, DC 20007
Warren DSouza
Radiation Oncology, University of Maryland, College Park, Maryland 20742
Martin Murphy
Radiation Oncology, Virginia Commonwealth University, Richmond, Virginia 23284
Kenneth H. Wong
Imaging Science and Information Systems, Georgetown University Hospital, Washington, DC 20007
Paul Kealla
Radiation Oncology, Stanford University, Stanford, California 94305
Received 18 March 2008; revised 17 October 2008; accepted for publication 17 October 2008;
The current climate of rapid technological evolution is reflected in newer and better methods to
modulate and direct radiation beams for cancer therapy. This Vision 20/ 20 paper focuses on part of
this evolution, locating and targeting moving tumors. The two processes are somewhat independent
and in principle different implementations of the locating and targeting processes can be interchanged. Advanced localization and targeting methods have an impact on treatment planning and
also present new challenges for quality assurance QA, that of verifying real-time delivery. Some
methods to locate and target moving tumors with radiation beams are currently FDA approved for
clinical useand this availability and implementation will increase with time. Extensions of current
capabilities will be the integration of higher order dimensionality, such as rotation and deformation
in addition to translation, into the estimate of the patient pose and real-time reoptimization and
adaption of delivery to the dynamically changing anatomy of cancer patients. 2008 American
Key words: respiratory motion, adaptive radiotherapy, tracking
I. INTRODUCTION
This Vision 20/ 20 paper focuses on the evolution of locating
and targeting moving tumors. The rationale for using this
technology in the clinic is to significantly improve tumor
control while reducing critical organ damage for tumors
that move with respiration, resulting in a better quality of
life, or beneficence for cancer patients. Beneficence is
one of the three basic principles of the Belmont
report, http://www.hhs.gov/ohrp/humansubjects/guidance/
belmont.htm#xbenefit one of the most important documents
in the field of medical ethics. This report states The term
beneficence is often understood to cover acts of kindness
or charity that go beyond strict obligation. In this document,
beneficence is understood in a stronger sense, as an obligation. Two general rules have been formulated as complementary expressions of beneficent actions in this sense: (1) Do
not harm and (2) maximize possible benefits and minimize
possible harms. Beneficence, as it applies to image-guided
therapy, can be achieved by reducing the difference between
the planned and delivered dose, reducing treatment uncertainties and therefore margins, and improving dose conformation to the tumor. Our patients and the treatment team
should be demanding this beneficence. As with many
5684
physics-based advances in radiotherapy, the magnitude of

benefit may be hard to quantify in a series of site-specific
randomized trials.1 However, the fundamental basis of radiotherapy is that increased treatment accuracy improves tumor
control, and reduced normal tissue dose reduces treatmentrelated complications. Any improvements in this area will
therefore improve the beneficence of radiation treatment.
Beneficence applies to locating and targeting moving tumors with radiation beams. The concept of do no harm
means that this technology should only be applied when motion trackinga change from traditional radiotherapy in several aspectsis demonstrably safe and appropriate quality
assurance is performed. The concept of maximizing possible
benefits and minimizing possible harm can be achieved by
having the radiation beam aligned as closely as possible with
the moving tumor at all times during radiation delivery, appropriately avoiding critical structures. Therefore the burden
of responsibility for systems that locate and target moving
tumors is not necessarily the demonstration of a perfectly
accurate systemit is sufficient to demonstrate safety and
that the use of target tracking is an improvement over methods without target tracking.
0094-2405/2008/3512/5684/11/$23.00
5684
5685
Dieterich et al.: Locating and targeting moving tumors with radiation beams
Real time motion

correctioncomponents
5685
Treatment
unit
Compute and actuate

beam/ target alignment
Correct for
reaction time
Treatment
beam
Detect
motion
Moving
Target (tumor)
FIG. 2. A generalized flowchart showing the main components of a real-time

motion correction system.
II. DESCRIPTION OF CURRENT STATUS
FIG. 1. Observed variation in lung tumor motion with time in three dimensions from Cyberknife Synchrony data. Ref. 113 Notable is that there is
significant motion in three dimensions, there are variations in the baseline
position of the tumor in all three dimensions and variations in the cycle-tocycle period, respiratory cycle shape, and range of motion.
The multitude of new technologies in radiation oncology

has created a large amount of potentially confusing or contradictory terminology. Before we start discussing the stateof-the art, development, and future outlook on locating and
targeting moving tumors with radiation beams, we need to
clarify the general concepts and define the scope of the technology.
II.A. Methods for real-time tumor localization
Effectively treating a tumor that moves during respiration

has become the paradigm for motion-adaptive radiotherapy.
This Vision 20/ 20 article will address issues of respiratory
motion correction during treatment with emphasis on the two
key componentslocating the tumor and targeting the tumor. The complexity of the problem of targeting moving
tumors is shown in Fig. 1, where there is tumor motion in all
three dimensions, varying baseline position and also varying
peak-to-trough motion, period, and trajectory shape within
each motion cycle. A generalized schematic of real-time motion correction is shown in Fig. 2.
Respiratory gating is often seen as an alternative to motion tracking. However, gating controls whether the beam is
on or off. Tracking controls the alignment of the tumor and
the beam. Therefore, these two modalities are independent
and can coexist. The combination of both gating and tracking
will be at least as good as either modality alone. There are
also many other methods of respiratory motion management,
such as motion inclusive methods, breath hold, and abdominal compression. For a broader discussion on respiratory motion management, readers are directed to the AAPM Task
Group 76 report and references therein,2 as well as the Seminars in Radiation Oncology special issue on intrafraction
organ motion and its management edited by Bortfeld and
Chen 141, 2004.
The first task in a motion-adaptive radiotherapy system is

spatial localization of the tumor. If we were to conceptualize
the ideal localization system, we can readily imagine certain
capabilities. First, it would have real-time performance sufficient to match any patient changes such as respiration. Second, it would be capable of imaging both structure and function e.g., cancer vs. inflammation. Third, it would have a
large field of view, enabling localization of both the planned
treatment volume and surrounding body structures. Finally, it
would be non-invasive and present only minimal additional
risk to the patient, either during the course of radiation
therapy or afterwards. Although this ideal localization system has not yet been created, there are several localization
methods in current use or active development. Table I compares some methods for real-time tumor localization. Each of
these methods emphasizes a particular property, such as temporal resolution or field-of-view, which in turn creates advantages and limitations. It is helpful to group these approaches into two broad classes for the purpose of these
definitions, we will use target as a generic term for any
object of interest in the body.
Direct methods detect the target itself. Within this class
are modalities such as x-ray imaging,39 MRI,1013
ultrasound,1417 and electromagnetic tracking.1820 Direct
methods can identify either artificially implanted points fiducials or anatomical landmarks. As a note of caution:
Some direct methods detect target surrogates, which although placed within or near the target, may or may not
5686
5686
TABLE I. Some methods to locate moving targets.

Localization method
Data dimensionality
Invasive?
Availability
Imaging frequency
Electromagnetic Refs. 1820

Fluoroscopic Refs. 37
Fluoroscopic/optical Refs. 2629
Monoscopic/optical Refs. 8,9
MRI Ref. 10
Ultrasound Ref. 14
Points
Points
Points
Points
Volume
Volumes
Yes
Yes
Noa and Yes
Yes
No
No
Commercially available for prostate

Translated to patient treatments
Commercially available
Under development
Under development
Under development
10 Hz
30 Hz
Before every beam/30 Hz
30 Hz theoretical
N/A
25 Hz
Lung tumor projection method Xsight Lung available from Accuray.
move or deform in a direct linear relationship with the target.

Indirect methods observe a surrogate quantity and infer
localization information from the surrogate. The most common example of this method is optical tracking to monitor
skin motion, although respiratory belts21,22 and airflow
monitors23,24 also fall into this category. Indirect methods in
the absence of internal imaging provide limited utility.25
Combining direct localization methods with indirect methods, either before or during the treatment, in order to model
the relationship between the surrogate and the target leads to
hybrid tracking methods.8,9,26,27
Clearly, if we are pursuing the ideal localization system, a
direct method is best. Using surrogates always introduces an
extra source of error that can change quickly and is very
difficult to characterize in advance, even though in some
patients this error may be small see Table III in Ref. 2.
However, indirect methods do not require fiducial implantation, are nontoxic, and have excellent temporal resolution.
The challenge then is to develop direct methods which incorporate many of the benefits of indirect methods. Fortunately,
efforts toward this goal are well under way, e.g., in the ongoing development of electromagnetic fiducial tracking
methods and MRI techniques.
It is important to note that all real-world systems involve
design compromises. Thus, localization systems need to be
carefully matched to the capabilities of the other components
of the radiation therapy system. For example, one might be
able to continuously acquire four-dimensional 4D CT data
from the patient, but if the data cannot be analyzed fast
enough to direct changes in the therapy delivery, then we
have a surfeit of information that produces little impact.
Similarly, if the only response of the system is to gate the
beam on or off, that determination could probably be made
with much less information than a full 4D CT. Finally, although on-line target localization is appealing from the technical point of view and may produce clinical benefits in
some cases, the ultimate test of its value will be its impact on
patient care.
II.B. Methods for target tracking
The heart of a real-time tumor tracking system is the control loop that connects the tumor detection system with the
targeting alignment system. The control loop must receive
target coordinate measurements, test that the target identification is authentic so as not to go off chasing false motion,
filter the target data to avoid disruptive transients, and then
send the delivery system a prediction of where the tumor will
be when the targeting adjustment has been completed. Tumor
tracking requires motors that can guide the beam or couch
faster than most respiratory motion and a fast secondary
feedback mechanism for each motor used to control the
beam or patient placement. Table II compares several motion tracking methods.
The various motion adaptation systems in use and under
consideration involve two fundamentally different types of
control loop. The CyberKnife robotic treatment system,2629
the multi-leaf collimator,3042 gimbaled linac,43 and proton
and heavy ion treatment systems44,45 all use an open-loop
control design. In an open loop system, the corrective response has no effect on the target behavior that signaled the
response. In other words, there is no feedback between the
corrective signal and the adaptive response. In contrast, a
couch compensation system4649 has a closed-loop control
system. When the tumor moves within the patient, the detection system triggers a compensating movement of the patient
TABLE II. Delivery methods designed to hit moving targets.
Tracking method
Block Ref. 112
Couch Refs. 4648
DMLC Refs. 3040
Gimbaled linac Ref. 43
Magnetic/mechanical particle beam Refs. 44 and 45
Robotic linac Refs. 2629
Motion capable of compensation

in principal
Approximate system
delay ms
Translation
N/A
Translation, Rotation
Translation, Rotation, Deformation
80
160
N/A
80
110
Availability
Translated to patient treatments
at one center
Under development
Under development
Under development
Under development
Commercially available
5687
and thus the tumor via the treatment couch. Thus, there is
feedback through which the adaptive response may influence
the target movement.
Real-time systems must react to change at least as fast as
it occurs. A systems reaction time depends on a number of
factors. Time is required to detect a change in the target
position, process and filter the position data, communicate
with the beam alignment system, and complete the repositioning of either the beam or patient. If a single motion detection method fluoroscopic imaging, for example is used
in a first order open loop system, then all of these time delays
combine linearly in series to give the total reaction time. On
the other hand, if two different motion detection systems are
used in parallel to provide partially redundant target position
data, then only the faster systems delay adds into the reaction time of an open loop system. For example, the CyberKnife uses an optical tracking system to provide fast, continuous estimates of a tumors respiratory induced motion,
with the estimate updated by periodic x-ray imaging.27 This
is an example of a hybrid detection method. The total system
reaction time includes the optical processing time while the
time spent processing the x-ray images can be done outside
the duty cycle of the control loop. This allows the targeting
system to invest more time in sophisticated image processing
without a corresponding degradation of response time.
Closed loop systems such as a movable couch can have
higher order dynamics that involve multiple system response
time constants in addition to the dead time before the couch
begins its corrective response.46 In an open loop system the
total reaction time must be compensated by predicting the
future position of the target so that the beam reaches the
future target position at the same time as the tumor. A closed
control loop needs to predict the dead time before the system
begins to respond to a change in the tumor position.46
The future displacement of a breathing motion can be
predicted in several ways: 1 With a mathematical model of
the breathing signal,50 2 with a bio-mechanical model of
the breathing process,51 or 3 with heuristic learning algorithms that mimic the breathing process as it is observed.52,53
The last category includes neural network prediction algorithms, which have been studied by several groups.5255 Neural networks can adapt to changes in the breathing pattern as
they occur, making them an attractive way to deal with the
nonstationary character of real breathing.
II.C. Treatment planning
Conventional treatment planning has traditionally used a

planning or safety margin around the clinical target volume CTV that implicitly accommodates organ motion as
well as setup uncertainties. However, the International Commission on Radiation Units and Measurement ICRU Report
62 explicitly defines a margin to account for physiological
changes i.e., internal margin IM. For tumors affected by
normal respiration, an internal target volume ITV is defined
such that it includes the addition of the IM to the CTV. The
advent of 4D CT has enabled the delineation of the tumor
volume as a function of the respiration cycle leading to adMedical Physics, Vol. 35, No. 12, December 2008
5687
vances in treatment planning and 4D dose computation. The

IM is typically defined by the excursion of the tumor on a 4D
CT. While this approach involves a patient-specific motion
margin, it assumes that the tumor occupies the entire spatial
domain within the ITV. However, it does not consider the
variability in the respiration patterns spanning a single treatment fraction duration or interfraction respiration variability.
This assumption results in a larger than necessary volume of
tissue being irradiated. As motion management assumes
greater significance for highly conformal radiation delivery
and stereotactic radiosurgery, and with advances in treatment
delivery technologies for real-time tumor tracking, more sophisticated planning approaches are warranted.
Broadly, three advanced approaches have been proposed
for 4D treatment planning: 1 Statistically based inverse
planning approaches motion kernel approach,5663 2 4D
CT and dose registration based on deformable image
registration,6468 and 3 inverse planning that explicitly considers organ-at-risk constraints during real-time tumor
tracking.56,69 These approaches represent a significant increase in sophistication over conventional planning approaches. 4D planning is computationally expensive and requires additional resources such as 4D optimization as well
as image registration software. 4D planning also involves
additional sources of error, such as data sorting and image
registration, and is only one short timespan of a patients
anatomy that can change during the course of treatment. Additionally, in a 50-patient study artifacts of at least 4 mm
were observed in 90% of 50 4D CT scans.70 Therefore the
4D data should be used with caution.
In the motion kernel approach, the idealized pencil beam
dose distribution in the presence of a moving target is calculated either by time-averaging the dose matrices calculated
for different instances of the anatomy corresponding to different parts of the respiration cycle, or by convolving the
pencil beam kernel with the probability density function
PDF that describes the 3D target trajectory.56 This approach
results in a smeared out pencil beam dose distribution prior
to dose optimization in comparison with the conventional
pencil beam optimization approach in the absence of organ
motion. Treatment plans generated with the motion kernel
approach are less sensitive to the tumor motion compared
with optimization approaches that do not account for motion.
Current real-time motion correction strategies involve the
tracking of the target volume without consideration of the
organs-at-risk. As a result, while tracking considers adequate
coverage of the tumor, it may result in unintended irradiation
of organs-at-risk. This unintended irradiation occurs due to
differences in the trajectories of motion between the tumor
and the organs-at-risk. In an approach applied to robotic
radiosurgery,69 the motion path is discretized into steps and
the time-fraction of each step is determined. First, a set of
beam orientations is chosen. For these beam orientations, the
number of monitor units associated with this beam at a step
along the motion path is determined by considering its dose
contribution to tumor and organ-at-risk voxels. The total
dose is summed over all the discretized steps along the mo-
5688
5688
tion path of the tumor. The problem is formulated as a standard linear program and solved by minimizing the total
monitor units.
The most widely reported 4D planning method involves
the coregistration of dose distributions calculated on individual 3D CT data sets corresponding to various phases of
the respiration cycle with a single reference 3D CT data set
typically end-inhale or end-exhale.6468,7177 This approach
relies on the image registration algorithm to provide displacement vectors that show the trajectory of each voxel
from one image set to another. The dose distribution calculated on each CT data set is then weighted by the timeweight of the CT phase in the respiration cycle and accumulated in the reference image. In this way one can see
differences between the 4D computed dose and the dose calculated on individual 3D CT data. This method has been
applied for 4D proton treatment planning as well.78,79
The Radiological Physics Center RPC http://

rpc.mdanderson.org/rpc/ has responded to these study requirements by developing phantoms for IMRT and SBRT
treatments. The lung and thorax phantoms include motion
platforms preprogrammed with sinusoidal and patient specific motion patterns. The phantoms have been designed to
be used on a variety of treatment delivery platforms such as
external beam machines and the Cyberknife.
II.D. Quality assurance
III.A. Methods for real-time tumor localization
Quality assurance is a methodical system to assess the

quality of the product or service. In radiation therapy, the
quality assurance responsibilities of the medical physicists
are the technical functionality and safety of the equipment
under all clinical conditions, and the safety of information
workflow pertaining to treatment planning and accurate treatment delivery. State regulations and AAPM task group reports form the framework and basis of the quality assurance
program, which then has to be customized to the specific
work environment and technology.
The traditional QA approach has been to anticipate for
equipment or procedure failure and establish safety procedures to prevent technical errors. Relatively simple tools and
phantoms were used for functionality testing. This approach
has drawbacks in the age of rapidly developing, complex
technology such as real-time motion adaptive radiation
therapy. Recently, several nonstationary phantoms have been
published8083 as well as developed commercially, with more
becoming available, to measure the precision and accuracy
of respiratory motion compensation. Matching the complexity of human respiration, driven by multiple muscle groups
and involving significant deformation, remains a challenge
for phantom development.
National Cancer Institute NCI sponsored protocols are
required to address respiratory motion when intensity modulated radiation therapy IMRT is used for intra-thoracic tumors and other locations. Institutions participating in Radiation Therapy Oncology Group RTOG studies are required
to demonstrate the capability to fulfill study requirements
and must submit the data to the physics and image guidance
committees. RTOG-0618, a stereotactic lung radiotherapy
trial, states Special considerations must be made to account
for the effect of internal organ motion e.g., breathing on
target positioning and reproducibility. Acceptable maneuvers
include reliable abdominal compression, accelerator beam
gating with the respiratory cycle, tumor tracking and active
breath-holding techniques.
Real-time tumor tracking will become more widespread

in the next 10 years. This growth will stimulate two further
areas of research and development. The first problem is how
to account for deformation. There is both tumor deformation
and motion with respect to the normal structures in a manner,
which can change from day to day and cycle to cycle for
thoracic and abdominal tumors. Multiple targets can move
with different trajectories, such as in a lung treatment where
the primary gross tumor volume GTV and mediastinal
nodes can move independently from the spinal cord, esophagus, and heart. The second and somewhat related problem is
the integration of target tracking radiotherapy with online
and even real-time plan adaptation. Majorbut not
insurmountablebarriers to implementation of volumetric
real-time adaptation include 1 methods to measure or estimate the position of the evolving volumetric anatomy with
time, 2 real-time deformable image registration, and 3
fast 4D optimization algorithms. Some solutions under development to measure the evolving volumetric anatomy with
time are integrated MR-linacs or real-time ultrasound guidance discussed under the Direct methods section. Combined measurement/estimates of the evolving anatomy could
involve generalizations of the internal/external correlation
models to estimate the entire anatomy as opposed to a single
point. Graphical processing units GPUs can offer near realtime speed for deformable algorithms.8488 Fast 4D optimization algorithms will also directly benefit from ongoing advancements in processor technology.
III. PROJECTION AHEAD

Todays current technological status of locating and targeting moving tumors clearly leads the clinical status, and
therefore we can expect the controlled introduction of several
promising technologies into the clinic. The commercial
translation of technical developments into clinical tools is
not guaranteed for any invention and involves a complex mix
of variables including development costs, risk analysis, competitive advantage, priority, timing, and market demand.
III.B. Methods for target tracking
Future developments in real-time control will aim to reduce system reaction time Table II while improving temporal prediction to compensate for delays. Reaction time can be
reduced by increasing the speed of each stage of target acquisition, signal processing, communication, and adaptive
alignment response. It can also be reduced by developing
5689
parallel control architectures in which those elements of the

target localization process that are time-intensive are moved
partly or completely out of the serial part of the control loop
so that they do not add to the reaction time. Ideally the improved predictors will not add computational cost that increases system response time. There will, however, always
be some lag time left over. Future developments will correlate the characteristics of each patients breathing with the
performance of the prediction and tracking algorithms so that
the algorithms can be customized to each patient for the best
results. These developments will also provide resources to
estimate a patients trackability before treatment, so that
appropriate margins for the residual alignment uncertainty
can be tailored for the individual treatment plan.
A challenge for target tracking systems that move the
beam in response to respiratory motion the open loop systems described above is that the beam modifying devices
need to be able to adjust the beam with sufficient acceleration and velocity to follow the motion for a reasonable fraction of the treatment to maintain high delivery efficiency. If
mechanically the motion cannot keep up, a beam hold should
be asserted. This is easiest for magnetic steering of particle
beams, but more difficult where mechanical motion is
needed. A particular challenge is using the DMLC to shape
fields in real time, particularly in the presence of large motion perpendicular to the leaf motion direction. Delivery efficiencies of lower than 25% for SMLC and DMLC IMRT
fields in the presence of high frequency motion perpendicular to the leaf motion direction have been measured.41 For
such situations, several mean position estimation algorithms
have been proposed to maintain high delivery efficiency, in
which the average position of the tumor motion is followed,
and the respiratory variations about this average are
ignored.8991 These methods all but eliminate systematic error in tracking, with the remaining motion included as random error. With little systematic error a moderate random
error has little impact on margins,92 though there are of
course other errors such as tumor delineation uncertainty,
tumor rotation, and tumor deformation that may have a significant impact on margins.
Extensions of current delivery capabilities will be the integration of higher order dimensionality, such as rotational
and deformational information in addition to translation, into
the estimate of the patient pose and real-time reoptimization
and adaption of delivery to the dynamically changing
anatomy of cancer patients.
III.C. Treatment planning
Future 4D treatment planning approaches will have to

consider intra- and interfraction variations in tumor and
organs-at-risk motion patterns. While 4D planning is currently not routine from a clinical perspective, its adoption on
a routine basis will become essential to complement technological advances in 4D delivery techniques. Eventually, realtime volumetric imaging may also be incorporated in adaptive planning strategies such that the optimal plan is
delivered to the temporally evolving anatomy of the patient.
5689
III.D. Quality assurance
The complexity of these real-time motion adaptive technologies has made it very difficult for a single person to
understand in detail where potential failures might occur.
Current pathways to disseminate QA information are slow
compared to the speed at which new products are implemented in clinical practice. However, procedures and equipment need to be evaluated thoroughly in a clinical setting to
ensure that a comprehensive quality control QC and QA
program can be developed. The knowledge of experts in the
field of safety such as industrial process engineers has
scarcely been tapped in the health care environment.
The traditional QA approach had drawbacks in the age of
rapidly developing, complex technology such as real-time
motion adaptive radiation therapy. As a first step, the use of
nonstationary phantoms which simulate skin and tumor motion separately is becoming a standard tool in the QA process
to measure the precision and accuracy of respiratory motion
compensation. The major limitation of these phantoms is the
dependency of the measurement result on the simulated motion pattern used. Because we still do not have ground
truth data of real-time tumor motion and surrogate marker
for the time intervals required for treatment maximum 150 s
of data vs. 15 60 min treatment length, the results will not
truly reflect patient treatment results. In addition, simulating
rotation and deformation is still challenging. A minor, but
significant, issue may be the effects on the dose distribution
due to the change of the dose calculation parameters
throughout the respiratory cycle e.g., SSD, surface correction, density changes of lung. This potential change of dose
distribution can only be simulated and measured in anthropomorphic phantoms or very sophisticated software which
takes into account the body deformation in correlation with
the timing of the respective beam delivery with the corresponding respiratory phase.
Accuracy measurements with this new generation of
phantoms will give the medical physicist the answer to the
question of technical accuracy of the real-time motion adaptive treatment delivery system.93 Indirect tracking methods
such as hybrid skin/tumor tracking models and/or 4DCT
based simulations of the correlation between internal motion
and a surrogate respiratory marker do not provide real-time
position information. This in turn introduces an additional
error, namely the uncertainly from the correlation model between external surrogate and tumor position.29 Phantoms are
not sufficiently anthropomorphic, and real-time in vivo dosimetry can so far only provide a few checkpoints of the
delivery accuracy of a 3D treatment plan.
As treatments move towards real-time feedback and realtime adaptation, QA processes will need to account for the
increased delivery complexity and shorter timescales of
beam adjustment. This will involve the addition of real-time
measurement as appropriate on the radiation delivery devices
themselves and synchronization of measured motion and
dose delivery data streams to facilitate dose reconstruction to
estimate the dose delivered to the patient. Also, phantoms
5690
FIG. 3. The number of Medline-indexed articles published in the past 10

completed years with the search terms tumor tracking radiotherapy.
mimicking the complexity of multidimensional motion and

daily anatomic changes observed during human treatments
will form an integral part of system tests.
IV. EVIDENCE AND RATIONALE FOR FUTURE
PROJECTION
The future projections for locating and targeting moving
tumors can be broadly summarized as follows.
1 Technology and methods will continue to be developed
and improve.
2 Clinical availability and implementation will increase.
3 Efficiency of treatment delivery will increase.
4 Clinical outcomes will improve.
Evidence for these projections are given below.
Technology and method will continue to be developed and
improve/Clinical availability and implementation will increase: Stating that radiotherapy technology will improve is
obvious, given funding from governments, philanthropic organizations, individual departments, and intramural and extramural vendor research and development support. This projection is also based on the large increase in scientific interest
in explicit motion management. The number of papers with
keywords tumor tracking radiotherapy has grown from a
few in the mid-to-late 1990s to over 50 in 2007. This growth
is shown in Fig. 3.
The number of available systems for locating and targeting tumors has increased with time. When clinically available, tumor tracking is used. Cyberknife system users have
embraced the tumor tracking approach. Experimentally, the
use of the MLC for tracking has been demonstrated on three
different vendor platforms, Varian,38,41,42 Siemens personal
communication, Martin Tacke, DKFZ, June 2008, and Tomotherapy Personal communication, Gustavo Olivera, Tomotherapy. July 2008, though none are clinically available
yet.
Efficiency of treatment delivery will increase: Integrating
tumor localization and targeting can not only improve accuracy and precision, but also efficiency. The development of
techniques to rapidly account for variations in the external
pose of the patient and internal target and normal tissue moMedical Physics, Vol. 35, No. 12, December 2008
5690
tion could allow for significantly reduced patient-treatment

time, in addition to high treatment accuracy. Avoiding the
need to reposition the patient prior to or during treatment and
reducing the overall treatment time are hypothesized to reduce the post-couch-movement secondary skeletal motion
induced by couch motion or fatigue.
Clinical outcomes will improve: Anticipating improved
clinical outcomes is always controversial in the absence and
unlikely chance of a randomized trial comparing radiotherapy where in one arm real-time target information is not
acquired or ignored, and a second arm where real-time locating and tumor targeting occurs, however it is the driving
factor in the development of improved technology in radiotherapy. The entire field of radiotherapy is based on the tenets that increased tumor dose will increase local control and
possibly survival, and that lowering normal tissue dose decreases toxicity. For lung cancer there is ample evidence of
the benefits of dose escalation94100 and toxicity
reduction,101108 which can be simultaneously achieved by
more accurate and precise radiotherapy.
Intrafraction motion is not just a concern in the thorax and
abdomen. For the prostate an on-line target motion monitoring and correction strategy is necessary to implement hypofractionated SBRT with intensity modulated beams for
prostate cancer treatments concludes a recent Medical Physics paper109 investigating the dosimetric concerns of intrafraction motion for the prostatewhere motion is present
but significantly smaller than for thoracic and abdominal
sites. Perhaps the most compelling, albeit indirect, clinical
evidence for image guided radiotherapy is in a retrospective
review: Rectal distension on the treatment-planning CT
scan decreased the probability of biochemical control, local
control, and rectal toxicity in patients who were treated without daily image-guided prostate localization, presumably because of geographic misses.41 The rectal distension was
found to be a greater prognostic indicator than high risk disease.
V. STRATEGY FOR MOVING FROM THE PRESENT
TO THE FUTURE
Changing the practice of radiotherapy from predominantly static to real-time dynamic is a paradigm shift for the
treatment team. As in many other areas of medical physics
and particularly radiation oncologymany of these exciting
areas are highlighted in the Vision 20/ 20 seriesmoving
from the present to the future will involve continued innovation and research and development in academia and small
and large companies. Medical physicists will play key roles
in the clinical practice and guidelines. Locating and targeting
moving tumors with radiation beams is currently available
from one vendor, Accuray, via the Cyberknife platform. Thus
there is a wealth of clinical and medical physics expertise
and experience in this area, and this should be leveraged as
other real-time locating and targeting solutions are brought
on line. It may well be that different vendors are used for the
locating and targeting tasks, with real-time software integration. An example is electromagnetic localization Calypso
5691
integrated with an MLC tracking system Varian that has

been experimentally investigated in a phantom study.110,111
VI. CONCLUSIONS
In conclusion, we are seeing rapid progress in the development and clinical implementation of active target tracking.
These developments combine imaging, planning, and treatment delivery with motion tracking. The authors anticipate in
the future there will be a larger choice of real-time target
localization systems, all of the major linear accelerator
manufacturers will offer some form of motion tracking, and
the fraction of radiotherapy treatments using motion tracking
will steadily increase. Given the dose-response relationship
for tumor and normal tissues that the field of radiation oncology is founded upon, more accurate radiation therapy will
hopefully improve tumor control and reduce treatmentrelated toxicity.
ACKNOWLEDGMENTS
The authors acknowledge the organizers and sponsors of
the Motion Adaptive Radiotherapy Workshop held at Georgetown University in 2007 where many of the concepts described above were discussed. The major sponsors were Siemens Medical Solutions, Accuray Inc., Calypso Medical,
and Elekta Oncology Systems.
a

[email protected]
1
S. M. Bentzen, Randomized controlled trials in health technology assessment: Overkill or overdue?, Radiother. Oncol. 86, 142147 2008.
2
P. Keall, G. Mageras, J. Balter, R. Emery, K. Forster, S. Jiang, J. Kapatoes, D. Low, M. Murphy, C. Ramsey, M. Van Herk, S. Vedam, J. Wong,
and E. Yorke, The management of respiratory motion in radiation oncology report of AAPM Task Group 76, Med. Phys. 33, 38743900 2006.
3
S. Shimizu, H. Shirato, K. Kitamura, S. Ogura, H. Akita-Dosaka, U.
Tateishi, Y. Watanabe, K. Fujita, T. Shimizu, and K. Miyasaka, Fluoroscopic real-time tumor-tracking radiation treatment RTRT can reduce
internal margin IM and set-up margin SM of planning target volume
PTV for lung tumors, Int. J. Radiat. Oncol., Biol., Phys. 48, 166167
2000.
4
S. Shimizu, H. Shirato, K. Kitamura, N. Shinohara, T. Harabayashi, T.
Tsukamoto, T. Koyanagi, and K. Miyasaka, Use of an implanted marker
and real-time tracking of the marker for the positioning of prostate and
bladder cancers, Int. J. Radiat. Oncol., Biol., Phys. 48, 15911597
2000.
5
S. Shimizu, H. Shirato, S. Ogura, H. Akita-Dosaka, K. Kitamura, T. Nishioka, K. Kagei, M. Nishimura, and K. Miyasaka, Detection of lung tumor movement in real-time tumor-tracking radiotherapy, Int. J. Radiat.
Oncol., Biol., Phys. 51, 304310 2001.
6
H. Shirato, S. Shimizu, T. Kunieda, K. Kitamura, M. van Herk, K. Kagei,
T. Nishioka, S. Hashimoto, K. Fujita, H. Aoyama, K. Tsuchiya, K. Kudo,
and K. Miyasaka, Physical aspects of a real-time tumor-tracking system
for gated radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 48, 11871195
2000.
7
H. Shirato, S. Shimizu, T. Shimizu, T. Nishioka, and K. Miyasaka, Realtime tumour-tracking radiotherapy, Lancet 353, 13311332 1999.
8
R. I. Berbeco, S. B. Jiang, G. C. Sharp, G. T. Chen, H. Mostafavi, and H.
Shirato, Integrated radiotherapy imaging system IRIS: Design considerations of tumour tracking with linac gantry-mounted diagnostic x-ray
systems with flat-panel detectors, Phys. Med. Biol. 49, 243255 2004.
9
B. Cho, Y. Suh, S. Dieterich, and P. J. Keall, A monoscopic method for
real-time tumour tracking using combined occasional x-ray imaging and
continuous respiratory monitoring, Phys. Med. Biol. 53, 28372855
2008.
10
B. W. Raaymakers, A. J. Raaijmakers, A. N. Kotte, D. Jette, and J. J.
5691
Lagendijk, Integrating a MRI scanner with a 6 MV radiotherapy accelerator: Dose deposition in a transverse magnetic field, Phys. Med. Biol.
49, 41094118 2004.
11
J. J. Lagendijk, B. W. Raaymakers, A. J. Raaijmakers, J. Overweg, K. J.
Brown, E. M. Kerkhof, R. W. van der Put, B. Hardemark, M. van Vulpen,
and U. A. van der Heide, MRI/linac integration, Radiother. Oncol. 86,
2529 2008.
12
A. J. Raaijmakers, B. W. Raaymakers, and J. J. Lagendijk, Experimental
verification of magnetic field dose effects for the MRI-accelerator, Phys.
Med. Biol. 52, 42834291 2007.
13
C. Kirkby, T. Stanescu, S. Rathee, M. Carlone, B. Murray, and B. G.
Fallone, Patient dosimetry for hybrid MRI-radiotherapy systems, Med.
Phys. 35, 10191027 2008.
14
A. Hsu, N. R. Miller, P. M. Evans, J. C. Bamber, and S. Webb, Feasibility of using ultrasound for real-time tracking during radiotherapy,
Med. Phys. 32, 15001512 2005.
15
J. Wu, O. Dandekar, D. Nazareth, P. Lei, W. DSouza, and R. Shekhar,
Effect of ultrasound probe on dose delivery during real-time ultrasoundguided tumor tracking, Conf. Proc. IEEE Eng. Med. Biol. Soc. 1, 3799
3802 2006.
16
A. Sawada, K. Yoda, M. Kokubo, T. Kunieda, Y. Nagata, and M. Hiraoka,
A technique for noninvasive respiratory gated radiation treatment system
based on a real time 3D ultrasound image correlation: A phantom study,
Med. Phys. 31, 245250 2004.
17
M. Fuss, J. Boda-Heggemann, N. Papanikolau, and B. J. Salter, Imageguidance for stereotactic body radiation therapy, Med. Dosim. 32, 102
110 2007.
18
T. R. Willoughby, P. A. Kupelian, J. Pouliot, K. Shinohara, M. Aubin, M.
Roach, 3rd, L. L. Skrumeda, J. M. Balter, D. W. Litzenberg, S. W. Hadley, J. T. Wei, and H. M. Sandler, Target localization and real-time
tracking using the Calypso 4D localization system in patients with localized prostate cancer, Int. J. Radiat. Oncol., Biol., Phys. 65, 528534
2006.
19
J. M. Balter, J. N. Wright, L. J. Newell, B. Friemel, S. Dimmer, Y. Cheng,
J. Wong, E. Vertatschitsch, and T. P. Mate, Accuracy of a wireless localization system for radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 61,
933937 2005.
20
P. Kupelian, T. Willoughby, A. Mahadevan, T. Djemil, G. Weinstein, S.
Jani, C. Enke, T. Solberg, N. Flores, D. Liu, D. Beyer, and L. Levine,
Multi-institutional clinical experience with the Calypso System in localization and continuous, real-time monitoring of the prostate gland during
external radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 67, 10881098
2007.
21
L. Dietrich, S. Jetter, T. Tucking, S. Nill, and U. Oelfke, Linacintegrated 4D cone beam CT: First experimental results, Phys. Med.
Biol. 51, 29392952 2006.
22
N. Wink, C. Panknin, and T. D. Solberg, Phase versus amplitude sorting
of 4D-CT data, Med. Phys. 7, 7785 2006.
23
J. W. Wong, M. B. Sharpe, D. A. Jaffray, V. R. Kini, J. M. Robertson, J.
S. Stromberg, and A. A. Martinez, The use of active breathing control
ABC to reduce margin for breathing motion, Int. J. Radiat. Oncol.,
Biol., Phys. 44, 911919 1999.
24
J. Hanley, M. M. Debois, D. Mah, G. S. Mageras, A. Raben, K. Rosenzweig, B. Mychalczak, L. H. Schwartz, P. J. Gloeggler, W. Lutz, C. C.
Ling, S. A. Leibel, Z. Fuks, and G. J. Kutcher, Deep inspiration breathhold technique for lung tumors: The potential value of target immobilization and reduced lung density in dose escalation, Int. J. Radiat. Oncol.,
Biol., Phys. 45, 603611 1999.
25
S. S. Korreman, T. Juhler-Nottrup, and A. L. Boyer, Respiratory gated
beam delivery cannot facilitate margin reduction, unless combined with
respiratory correlated image guidance, Radiother. Oncol. 86, 6168
2008.
26
M. J. Murphy, Tracking moving organs in real time, Semin. Radiat.
Oncol. 14, 91100 2004.
27
A. Schweikard, H. Shiomi, and J. Adler, Respiration tracking in radiosurgery, Med. Phys. 31, 27382741 2004.
28
C. Ozhasoglu, M. J. Murphy, G. Glosser, M. Bodduluri, A. Schweikard,
K. Forster, D. P. Martin, and J. R. Adler, Real-time tracking of the tumor
volume in precision radiotherapy and body radiosurgeryA novel approach to compensate for respiratory motion, presented at the Computer
Assisted Radiology and Surgery, San Francisco, 2000.
5692
29
Y. Seppenwoolde, R. I. Berbeco, S. Nishioka, H. Shirato, and B. Heijmen,

Accuracy of tumor motion compensation algorithm from a robotic respiratory tracking system: A simulation study, Med. Phys. 34, 27742784
2007.
30
J. R. McClelland, S. Webb, D. McQuaid, D. M. Binnie, and D. J. Hawkes,
Tracking differential organ motion with a breathing multileaf collimator: Magnitude of problem assessed using 4D CT data and a motioncompensation strategy, Phys. Med. Biol. 52, 48054826 2007.
31
D. McQuaid and S. Webb, IMRT delivery to a moving target by dynamic
MLC tracking: Delivery for targets moving in two dimensions in the
beams eye view, Phys. Med. Biol. 51, 48194839 2006.
32
S. Webb and D. M. Binnie, A strategy to minimize errors from differential intrafraction organ motion using a single configuration for a breathing multileaf collimator, Phys. Med. Biol. 51, 45174531 2006.
33
S. Webb, Quantification of the fluence error in the motion-compensated
dynamic MLC DMLC technique for delivering intensity-modulated radiotherapy IMRT, Phys. Med. Biol. 51, L17-L21 2006.
34
L. Papiez and D. Rangaraj, DMLC leaf-pair optimal control for mobile,
deforming target, Med. Phys. 32, 275285 2005.
35
L. Papiez, D. Rangaraj, and P. Keall, Real-time DMLC IMRT delivery
for mobile and deforming targets, Med. Phys. 32, 30373048 2005.
36
D. Rangaraj and L. Papiez, Synchronized delivery of DMLC intensity
modulated radiation therapy for stationary and moving targets, Med.
Phys. 32, 18021817 2005.
37
R. McMahon, L. Papiez, and D. Rangaraj, Dynamic-MLC leaf control
utilizing on-flight intensity calculations: A robust method for real-time
IMRT delivery over moving rigid targets, Med. Phys. 34, 32113223
2007.
38
P. J. Keall, H. Cattell, D. Pokhrel, S. Dieterich, K. H. Wong, M. J. Murphy, S. S. Vedam, K. Wijesooriya, and R. Mohan, Geometric accuracy of
a real-time target tracking system with dynamic multileaf collimator
tracking system, Int. J. Radiat. Oncol., Biol., Phys. 65, 15791584
2006.
39
P. J. Keall, V. R. Kini, S. S. Vedam, and R. Mohan, Motion adaptive
x-ray therapy: A feasibility study, Phys. Med. Biol. 46, 110 2001.
40
T. Neicu, H. Shirato, Y. Seppenwoolde, and S. B. Jiang, Synchronized
moving aperture radiation therapy SMART: average tumour trajectory
for lung patients, Phys. Med. Biol. 48, 587598 2003.
41
A. Sawant, R. Venkat, V. Srivastava, D. Carlson, S. Povzner, H. Cattell,
and P. Keall, Management of three-dimensional intrafraction motion
through real-time DMLC tracking, Med. Phys. 35, 20502061 2008.
42
J. Zimmerman, S. Korreman, G. Persson, H. Cattell, M. Svatos, A.
Sawant, R. Venkat, D. Carlson, and P. Keall, DMLC motion tracking of
moving targets for intensity modulated arc therapy treatmenta feasibility study, Acta Oncol. 16 2008.
43
Y. Kamino, K. Takayama, M. Kokubo, Y. Narita, E. Hirai, N. Kawawda,
T. Mizowaki, Y. Nagata, T. Nishidai, and M. Hiraoka, Development of a
four-dimensional image-guided radiotherapy system with a gimbaled
X-ray head, Int. J. Radiat. Oncol., Biol., Phys. 66, 271278 2006.
44
C. Bert, N. Saito, A. Schmidt, N. Chaudhri, D. Schardt, and E. Rietzel,
Target motion tracking with a scanned particle beam, Med. Phys. 34,
47684771 2007.
45
S. O. Grozinger, E. Rietzel, Q. Li, C. Bert, T. Haberer, and G. Kraft,
Simulations to design an online motion compensation system for
scanned particle beams, Phys. Med. Biol. 51, 35173531 2006.
46
W. D. DSouza and T. J. McAvoy, An analysis of the treatment couch
and control system dynamics for respiration-induced motion compensation, Med. Phys. 33, 47014709 2006.
47
W. D. DSouza, S. A. Naqvi, and C. X. Yu, Real-time intra-fractionmotion tracking using the treatment couch: A feasibility study, Phys.
Med. Biol. 50, 40214033 2005.
48
P. Qiu, W. D. DSouza, T. J. McAvoy, and K. J. Ray Liu, Inferential
modeling and predictive feedback control in real-time motion compensation using the treatment couch during radiotherapy, Phys. Med. Biol. 52,
58315854 2007.
49
J. Wilbert, J. Meyer, K. Baier, M. Guckenberger, C. Herrmann, R. Hess,
C. Janka, L. Ma, T. Mersebach, A. Richter, M. Roth, K. Schilling, and M.
Flentje, Tumor tracking and motion compensation with an adaptive tumor tracking system ATTS: System description and prototype testing,
Med. Phys. 35, 39113921 2008.
50
Y. Seppenwoolde, H. Shirato, K. Kitamura, S. Shimizu, M. van Herk, J.
V. Lebesque, and K. Miyasaka, Precise and real-time measurement of
3D tumor motion in lung due to breathing and heartbeat, measured during
5692
radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 53, 822834 2002.

D. A. Low, P. J. Parikh, W. Lu, J. F. Dempsey, S. H. Wahab, J. P. Hubenschmidt, M. M. Nystrom, M. Handoko, and J. D. Bradley, Novel breathing motion model for radiotherapy, Int. J. Radiat. Oncol., Biol., Phys.
63, 921929 2005.
52
M. J. Murphy, J. Jalden, and M. Isaksson, Adaptive filtering to predict
lung tumor breathing motion during image-guided radiation therapy,
Proceedings of the 16th International Congress on Computer-assisted Radiology and Surgery 2002, pp. 539544.
53
S. S. Vedam, P. J. Keall, A. Docef, D. A. Todor, V. R. Kini, and R.
Mohan, Predicting respiratory motion for four-dimensional radiotherapy, Med. Phys. 31, 22742283 2004.
54
M. Isaksson, J. Jalden, and M. J. Murphy, On using an adaptive neural
network to predict lung tumor motion during respiration for radiotherapy
applications, Med. Phys. 32, 38013809 2005.
55
G. C. Sharp, S. B. Jiang, S. Shimizu, and H. Shirato, Prediction of
respiratory tumour motion for real-time image-guided radiotherapy,
Phys. Med. Biol. 49, 425440 2004.
56
A. Trofimov, E. Rietzel, H. M. Lu, B. Martin, S. Jiang, G. T. Chen, and T.
Bortfeld, Temporo-spatial IMRT optimization: Concepts, implementation and initial results, Phys. Med. Biol. 50, 27792798 2005.
57
W. A. Beckham, P. J. Keall, and J. V. Siebers, A fluence-convolution
method to calculate radiation therapy dose distributions that incorporate
random set-up error, Phys. Med. Biol. 47, 34653473 2002.
58
I. J. Chetty, M. Rosu, N. Tyagi, L. H. Marsh, D. L. McShan, J. M. Balter,
B. A. Fraass, and R. K. Ten Haken, A fluence convolution method to
account for respiratory motion in three-dimensional dose calculations of
the liver: A Monte Carlo study, Med. Phys. 30, 17761780 2003.
59
J. Leong, Implementation of random positioning error in computerised
radiation treatment planning systems as a result of fractionation, Phys.
Med. Biol. 32, 327334 1987.
60
A. E. Lujan, E. W. Larsen, J. M. Balter, and R. K. Ten Haken, A method
for incorporating organ motion due to breathing into 3D dose calculations, Med. Phys. 26, 715720 1999.
61
P. J. Keall, W. A. Beckham, J. T. Booth, S. F. Zavgorodni, and M. Oppelaar, A method to predict the effect of organ motion and set-up variations
on treatment plans, Australas. Phys. Eng. Sci. Med. 22, 4852 1999.
62
T. Craig, J. Battista, and J. Van Dyk, Limitations of a convolution
method for modeling geometric uncertainties in radiation therapy. II. The
effect of a finite number of fractions, Med. Phys. 30, 20122020 2003.
63
T. Craig, J. Battista, and J. Van Dyk, Limitations of a convolution
method for modeling geometric uncertainties in radiation therapy. I. The
effect of shift invariance, Med. Phys. 30, 20012011 2003.
64
G. T. Chen, J. H. Kung, and E. Rietzel, Four-dimensional imaging and
treatment planning of moving targets, Front. Radiat. Ther. Oncol. 40,
5971 2007.
65
S. Flampouri, S. B. Jiang, G. C. Sharp, J. Wolfgang, A. A. Patel, and N.
C. Choi, Estimation of the delivered patient dose in lung IMRT treatment
based on deformable registration of 4D-CT data and Monte Carlo simulations, Phys. Med. Biol. 51, 27632779 2006.
66
M. Guckenberger, J. Wilbert, T. Krieger, A. Richter, K. Baier, J. Meyer
and M. Flentje, Four-dimensional treatment planning for stereotactic
body radiotherapy, Int. J. Radiat. Oncol., Biol., Phys. 69, 276285
2007.
67
T. Guerrero, G. Zhang, W. Segars, T. C. Huang, S. Bilton, G. Ibbott, L.
Dong, K. Forster, and K. P. Lin, Elastic image mapping for 4-D dose
estimation in thoracic radiotherapy, Radiat. Prot. Dosim. 115, 497502
2005.
68
Y. Suh, E. Weiss, H. Zhong, M. Fatyga, J. V. Siebers, and P. J. Keall, A
deliverable four-dimensional intensity-modulated radiation therapyplanning method for dynamic multileaf collimator tumor tracking delivery, Int. J. Radiat. Oncol., Biol., Phys. 71, 15261536 2008.
69
A. Schlaefer, J. Fisseler, S. Dieterich, H. Shiomi, K. Cleary, and A. Schweikard, Feasibility of four-dimensional conformal planning for robotic
radiosurgery, Med. Phys. 32, 37863792 2005.
70
T. Yamamoto, U. W. Langner, and P. Keall, Retrospective analysis of
artifacts in the four-dimensional CT images of 50 patients, Int. J. Radiat.
Oncol., Biol., Phys. 72, 12501258 2008.
51
5693
71
P. Keall, 4-dimensional computed tomography imaging and treatment

planning, Semin. Radiat. Oncol. 14, 8190 2004.
72
P. J. Keall, S. Joshi, G. Tracton, V. Kini, S. Vedam, and R. Mohan,
4-Dimensional radiotherapy planning, Int. J. Radiat. Oncol., Biol.,
Phys. 57, S233 2003.
73
K. K. Brock, D. L. McShan, R. K. Ten Haken, S. J. Hollister, L. A.
Dawson, and J. M. Balter, Inclusion of organ deformation in dose calculations, Med. Phys. 30, 290295 2003.
74
M. Rosu, I. J. Chetty, J. M. Balter, M. L. Kessler, D. L. McShan, and R.
K. Ten Haken, Dose reconstruction in deforming lung anatomy: Dose
grid size effects and clinical implications, Med. Phys. 32, 24872495
2005.
75
E. Rietzel, G. T. Chen, N. C. Choi, and C. G. Willet, Four-dimensional
image-based treatment planning: Target volume segmentation and dose
calculation in the presence of respiratory motion, Int. J. Radiat. Oncol.,
Biol., Phys. 61, 15351550 2005.
76
E. Heath and J. Seuntjens, A direct voxel tracking method for fourdimensional Monte Carlo dose calculations in deforming anatomy, Med.
Phys. 33, 434445 2006.
77
P. J. Keall, S. Joshi, S. S. Vedam, J. V. Siebers, V. R. Kini, and R. Mohan,
Four-dimensional radiotherapy planning for DMLC-based respiratory
motion tracking, Med. Phys. 32, 942951 2005.
78
M. Engelsman, E. Rietzel, and H. M. Kooy, Four-dimensional proton
treatment planning for lung tumors, Int. J. Radiat. Oncol., Biol., Phys.
64, 15891595 2006.
79
Y. Kang, X. Zhang, J. Y. Chang, H. Wang, X. Wei, Z. Liao, R. Komaki, J.
D. Cox, P. A. Balter, H. Liu, X. R. Zhu, R. Mohan, and L. Dong, 4D
proton treatment planning strategy for mobile lung tumors, Int. J. Radiat.
Oncol., Biol., Phys. 67, 906914 2007.
80
R. Kashani, M. Hub, M. L. Kessler, and J. M. Balter, Technical note: A
physical phantom for assessment of accuracy of deformable alignment
algorithms, Med. Phys. 34, 27852788 2007.
81
R. Kashani, K. Lam, D. Litzenberg, and J. Balter, Technical note: A
deformable phantom for dynamic modeling in radiation therapy, Med.
Phys. 34, 199201 2007.
82
M. Serban, E. Heath, G. Stroian, D. L. Collins, and J. Seuntjens, A
deformable phantom for 4D radiotherapy verification: Design and image
registration evaluation, Med. Phys. 35, 10941102 2008.
83
E. Nioutsikou, N. S.-T. J. Richard, J. L. Bedford, and S. Webb, Quantifying the effect of respiratory motion on lung tumour dosimetry with the
aid of a breathing phantom with deforming lungs, Phys. Med. Biol. 51,
33593374 2006.
84
J. E. Cates, A. E. Lefohn, and R. T. Whitaker, GIST: An interactive,
GPU-based level set segmentation tool for 3D medical images, Med.
Image Anal. 8, 217231 2004.
85
A. E. Lefohn, J. M. Kniss, C. D. Hansen, and R. T. Whitaker, A streaming narrow-band algorithm: Interactive computation and visualization of
level sets, IEEE Trans. Vis. Comput. Graph. 10, 422433 2004.
86
K. O. Noe, B. D. De Senneville, U. V. Elstrom, K. Tanderup, and T. S.
Sorensen, Acceleration and validation of optical flow based deformable
registration for image-guided radiotherapy, Acta Oncol. 47, 12861293
2008.
87
K. O. Noe, K. Tanderup, J. C. Lindegaard, C. Grau, and T. S. Sorensen,
GPU accelerated viscous-fluid deformable registration for radiotherapy,
Stud. Health Technol. Inform. 132, 327332 2008.
88
G. C. Sharp, N. Kandasamy, H. Singh, and M. Folkert, GPU-based
streaming architectures for fast cone-beam CT image reconstruction and
demons deformable registration, Phys. Med. Biol. 52, 57715783
2007.
89
D. Ruan, J. A. Fessler, and J. M. Balter, Mean position tracking of
respiratory motion, Med. Phys. 35, 782792 2008.
90
R. George, Y. Suh, M. Murphy, J. Williamson, E. Weiss, and P. Keall,
On the accuracy of a moving average algorithm for target tracking during radiation therapy treatment delivery, Med. Phys. 35, 23562365
2008.
91
A. Trofimov, C. Vrancic, T. C. Chan, G. C. Sharp, and T. Bortfeld, Tumor trailing strategy for intensity-modulated radiation therapy of moving
targets, Med. Phys. 35, 17181733 2008.
92
M. van Herk, Different styles of image-guided radiotherapy, Semin.
Radiat. Oncol. 17, 258267 2007.
93
K. H. Wong, S. Dieterich, J. Tang, and K. Cleary, Quantitative measurement of CyberKnife robotic arm steering, Technol. Cancer Res. Treat. 6,
589594 2007.
94
5693
N. C. Choi and J. A. Doucette, Improved survival of patients with unresectable non-small-cell bronchogenic carcinoma by an innovated highdose en-bloc radiotherapeutic approach, Cancer 48, 101109 1981.
95
M. K. Martel, R. K. Ten Haken, M. B. Hazuka, M. L. Kessler, M. Strawderman, A. T. Turrisi, T. S. Lawrence, B. A. Fraass, and A. S. Lichter,
Estimation of tumor control probability model parameters from 3-D dose
distributions of non-small cell lung cancer patients, Lung Cancer 24,
3137 1999.
96
P. Okunieff, D. Morgan, A. Niemierko, and H. D. Suit, Radiation doseresponse of human tumors, Int. J. Radiat. Oncol., Biol., Phys. 32, 1227
1237 1995.
97
C. A. Perez, M. Bauer, S. Edelstein, B. W. Gillespie, and R. Birch, Impact of tumor control on survival in carcinoma of the lung treated with
irradiation, Int. J. Radiat. Oncol., Biol., Phys. 12, 539547 1986.
98
C. A. Perez, T. F. Pajak, P. Rubin, J. R. Simpson, M. Mohiuddin, L. W.
Brady, R. Perez-Tamayo, and M. Rotman, Long-term observations of the
patterns of failure in patients with unresectable non-oat cell carcinoma of
the lung treated with definitive radiotherapy. Report by the Radiation
Therapy Oncology Group, Cancer 59, 18741881 1987.
99
C. A. Perez, K. Stanley, P. Rubin, S. Kramer, L. Brady, R. Perez-Tamayo,
G. S. Brown, J. Concannon, M. Rotman, and H. G. Seydel, A prospective randomized study of various irradiation doses and fractionation
schedules in the treatment of inoperable non-oat-cell carcinoma of the
lung. Preliminary report by the Radiation Therapy Oncology Group,
Cancer 45, 27442753 1980.
100
J. Wulf, K. Baier, and M. P. Flentje, Dose escalation in radiothreapy of
lung tumors by stereotactic irradiation: Is there a dose-response relationship for local tumor control? Int. J. Radiat. Oncol., Biol., Phys. 63, S52
2005.
101
J. Bradley, J. O. Deasy, S. Bentzen, and I. El-Naqa, Dosimetric correlates for acute esophagitis in patients treated with radiotherapy for lung
carcinoma, Int. J. Radiat. Oncol., Biol., Phys. 58, 11061113 2004.
102
J. Bradley, M. V. Graham, K. Winter, J. A. Purdy, R. Komaki, W. H. Roa,
J. K. Ryu, W. Bosch, and B. Emami, Toxicity and outcome results of
RTOG 9311: A phase I-II dose-escalation study using three-dimensional
conformal radiotherapy in patients with inoperable non-small-cell lung
carcinoma, Int. J. Radiat. Oncol., Biol., Phys. 61, 318328 2005.
103
M. V. Graham, J. A. Purdy, B. Emami, W. Harms, W. Bosch, M. A.
Lockett, and C. A. Perez, Clinical dose-volume histogram analysis for
pneumonitis after 3D treatment for non-small cell lung cancer NSCLC,
104
M. L. Hernando, L. B. Marks, G. C. Bentel, S. M. Zhou, D. Hollis, S. K.
Das, M. Fan, M. T. Munley, T. D. Shafman, M. S. Anscher, and P. A.
Lind, Radiation-induced pulmonary toxicity: A dose-volume histogram
analysis in 201 patients with lung cancer, Int. J. Radiat. Oncol., Biol.,
Phys. 51, 650659 2001.
105
S. L. Kwa, J. V. Lebesque, J. C. Theuws, L. B. Marks, M. T. Munley, G.
Bentel, D. Oetzel, U. Spahn, M. V. Graham, R. E. Drzymala, J. A. Purdy,
A. S. Lichter, M. K. Martel, and R. K. Ten Haken, Radiation pneumonitis as a function of mean lung dose: An analysis of pooled data of 540
patients, Int. J. Radiat. Oncol., Biol., Phys. 42, 19 1998.
106
D. Oetzel, P. Schraube, F. Hensley, G. Sroka-Perez, M. Menke, and M.
Flentje, Estimation of pneumonitis risk in three-dimensional treatment
planning using dose-volume histogram analysis, Int. J. Radiat. Oncol.,
Biol., Phys. 33, 455460 1995.
107
Y. Seppenwoolde, J. V. Lebesque, K. de Jaeger, J. S. Belderbos, L. J.
Boersma, C. Schilstra, G. T. Henning, J. A. Hayman, M. K. Martel, and
R. K. Ten Haken, Comparing different NTCP models that predict the
incidence of radiation pneumonitis, Int. J. Radiat. Oncol., Biol., Phys.
55, 724735 2003.
108
E. D. Yorke, A. Jackson, K. E. Rosenzweig, S. A. Merrick, D. Gabrys, E.
S. Venkatraman, C. M. Burman, S. A. Leibel, and C. C. Ling, Dosevolume factors contributing to the incidence of radiation pneumonitis in
non-small-cell lung cancer patients treated with three-dimensional conformal radiation therapy, Int. J. Radiat. Oncol., Biol., Phys. 54, 329339
2002.
109
S. Hossain, P. Xia, C. Chuang, L. Verhey, A. Gottschalk, G. Mu, and L.
Ma, Simulated real time image guided intrafraction tracking-delivery for
hypofractionated prostate IMRT, Med. Phys. 35, 40414048 2008.
5694
110
A. Sawant, R. L. Smith, R. Venkat, L. Santaman, B. Cho, P. Poulsen, H.

Cattell, J. Newell, P. Parikh, and P. Keall, Geometric accuracy and latency of an integrated 4D IMRT delvery system using real-time internal
position monitoring and dynamic MLC tracking abstract, Int. J. Radiat.
111
R. L. Smith, A. Sawant, L. Santaman, R. Venkat, J. Newell, B. Cho, P.
Poulsen, H. Cattell, P. Keall, and P. Parikh, IMRT dosimetric measurements from a real-time internal position monitoring system coupled with
5694
a dynamic multileaf collimator tracking system abstract, Int. J. Radiat.

112
M. Uematsu, CT-guided focal high dose radiotherapy, presented at the 4th
S. Takahashi International Workshop on 3 Dimensional Conformal Radiotherapy, Nagoya, Japan, 2004.
113
Y. Suh, S. Dieterich, and P. J. Keall, Geometric uncertainty of 2D projection imaging in monitoring 3D tumor motion, Phys. Med. Biol. 52,
34393454 2007.
Prospects of photoacoustic tomography

Lihong V. Wanga
Optical Imaging Laboratory, Department of Biomedical Engineering, Washington University in St. Louis,
Campus Box 1097, One Brookings Drive, St. Louis, Missouri 63130-4899
Received 8 August 2008; revised 10 October 2008; accepted for publication 15 October 2008;
Commercially available high-resolution three-dimensional optical imaging modalitiesincluding
confocal microscopy, two-photon microscopy, and optical coherence tomographyhave fundamentally impacted biomedicine. Unfortunately, such tools cannot penetrate biological tissue deeper than
the optical transport mean free path 1 mm in the skin. Photoacoustic tomography, which combines strong optical contrast and high ultrasonic resolution in a single modality, has broken through
this fundamental depth limitation and achieved superdepth high-resolution optical imaging. In
parallel, radio frequency-or microwave-induced thermoacoustic tomography is being actively developed to combine radio frequency or microwave contrast with ultrasonic resolution. In this Vision
20/ 20 article, the prospects of photoacoustic tomography are envisaged in the following aspects:
1 photoacoustic microscopy of optical absorption emerging as a mainstream technology, 2
melanoma detection using photoacoustic microscopy, 3 photoacoustic endoscopy, 4 simultaneous functional and molecular photoacoustic tomography, 5 photoacoustic tomography of gene
expression, 6 Doppler photoacoustic tomography for flow measurement, 7 photoacoustic tomography of metabolic rate of oxygen, 8 photoacoustic mapping of sentinel lymph nodes, 9 multiscale photoacoustic imaging in vivo with common signal origins, 10 simultaneous photoacoustic
and thermoacoustic tomography of the breast, 11 photoacoustic and thermoacoustic tomography
of the brain, and 12 low-background thermoacoustic molecular imaging. 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.3013698
Key words: microscopy, melanoma detection, endoscopy, functional imaging, molecular imaging,
reporter gene imaging, Doppler effect, metabolic rate of oxygen, sentinel lymph nodes, multiscale
imaging, thermoacoustic tomography, breast imaging, brain imaging
I. INTRODUCTION
The field of photoacoustic tomography PAT has grown a
great deal in the past few years. PAT is cross-sectional or
three-dimensional imaging based on the photoacoustic effect.
Alexander Graham Bell first reported on the photoacoustic
effect in 1880. Only recently, however, was PAT developed
as an imaging technology.129 PAT combines high ultrasonic
resolution and strong optical contrast in a single modality,
capable of providing high-resolution structural, functional,19
and molecular30,31 imaging in vivo in optically scattering biological tissue at new depths.
In biological tissues, light transfer is dominated by scattering. The mean free path is on the order of 0.1 mm, and the
transport mean free path is on the order of 1 mm. While the
former measures the frequency of predominantly anisotropic
scattering, the latter assesses the frequency of equivalent isotropic scattering. As a result of scattering, photon propagation transitions from the ballistic regime into the diffusive
regime around one transport mean free path.4 Specifically,
the ballistic regime is within the mean free path, where a
significant number of photons have undergone no scattering.
The quasiballistic regime is from the mean free path to the
transport mean free path, where photons have sustained a
few scattering events but retain a strong memory of the original incidence direction. The quasidiffusive regime is from
the transport mean free path to ten times the transport mean
5758
free path, where photons have sustained many scattering

events and retain only a weak memory of the original incidence direction. The diffusive regime is beyond ten times the
transport mean free path, where photons have suffered many
scattering events that they have almost completely lost their
memory of the original incidence direction.4
Two important depth limits exist for optical imaging. One
is near the optical transport mean free path, representing the
depth of the quasiballistic regime in biological tissue. Ballistic light intensity attenuates exponentially with a decay constant equal to the mean free path. To reach one transport
mean free path deep, photons must undergo significant scattering, making focusing ineffective.32 We refer to this barrier
as the soft depth limit for high-resolution optical imaging.
Another depth limit is around 50 70 mm, which equals
roughly ten times the 1 / e optical penetration depth the reciprocal of the effective attenuation coefficient. To reach
this depth, light must experience 43 dB or 20 000 times oneway decay in intensity. Beyond this limit, even diffuse photons are too few for practical purposes. We refer to this limit
as the hard depth limit for optical imaging. Nevertheless, if
the tissue is illuminated from opposite sides, a thickness
greater than 10 cm can be potentially covered, which is adequate for many biomedical applications such as breast imaging.
0094-2405/2008/3512/5758/10/$23.00
5758
5759
Lihong V. Wang: Prospects of photoacoustic tomography
High-resolution optical imaging beyond the soft depth

limit, sometimes referred to as superdepth optical imaging,
remained a void until PAT was developed. None of the commercially available optical ballistic imaging modalities
including confocal microscopy, two-photon microscopy and
optical coherence tomographycan penetrate into scattering
biological tissue beyond the soft depth limit. By contrast,
diffuse optical tomographybased on multiple-scattered
photonscan provide rapid functional and molecular imaging beyond the soft depth limit; however, it has poor spatial
resolution. The motivation driving the development of PAT is
to overcome the poor spatial resolution of diffuse optical
tomography or the soft depth limit of existing highresolution optical imaging.
An approach to overcome the optical hard depth limit is to
adopt radiofrequency or microwaves for photoacoustic
excitation.6,7 In this case, the technology is referred to as
radio frequency- or microwave-induced acoustic tomography
or thermoacoustic tomography TAT. For simplicity here,
radio frequency refers to microwaves as well. In TAT, a radio
frequency generator instead of a laser is used. The radio
frequency generator transmits radio frequency pulses into the
tissue to be imaged. Radio frequency absorption produces
heat and subsequent ultrasonic waves. The ultrasound detection and image formation are similar to those in PAT.
Both PAT and TAT are designed to overcome the poor
spatial resolution of pure optical or radio frequency imaging
yet to retain the high electromagnetic contrasts. In terms of
spatial resolution, pure optical imaging beyond the soft depth
limit suffers from strong diffusion due to tissue scattering,
whereas pure radio frequency imaging suffers from strong
diffraction due to the long wavelength. Ultrasonic scattering
coefficient in tissue is 23 orders of magnitude less than the
optical counterpart, and the acoustic diffraction or wavelength is 23 orders of magnitude weaker or shorter than
the radio frequency counterpart. As a result, PAT and TAT
can provide high spatial resolution by detecting the induced
ultrasonic waves. Unlike ultrasonography or optical coherence tomography, PAT and TAT produce speckle-free
images.
The subsequent sections are organized as follows. First,
the basic principle of PAT is reviewed. Then, the future prospects of PAT and TAT are envisioned on twelve specific topics. Finally, a summary is provided.
II. BASICS OF PHOTOACOUSTIC TOMOGRAPHY
PAT involves optical irradiation, ultrasonic detection, and
image formation. The tissue is usually irradiated by a shortpulsed laser beam to produce thermal and acoustic impulse
responses. Locally absorbed light is converted into heat,
which is further converted to a pressure rise via thermoelastic expansion of the tissue. The initial pressure rise
determined by the local optical energy deposition also called
specific optical absorption in the unit of J / m3 and other
thermal and mechanical propertiespropagates in the tissue
as an ultrasonic wave, which is referred to as a photoacoustic
wave. The photoacoustic wave is detected by ultrasonic
5759
transducers placed outside the tissue, producing electric signals. The electric signals are then amplified, digitized, and
transferred to a computer, where an image is formed. PAT
depends on any absorbed photons, either unscattered or scattered, to produce photoacoustic signals as long as the photon
excitation is relaxed thermally. Readers are referred to recent
review and tutorial articles for more details.1,4
The formation of a photoacoustic image, mapping the
photoacoustic source according to the detected sound signals, can be exemplified with the simplest case of pinpointing a single sound source such as thunder in threedimensional 3D space. If the time delay between seeing
lightening and hearing thunder is recorded, one can infer that
the thunder took place on a spherical surface of a radius
determined by the product of the speed of sound and the time
delay. If three such measurements are taken at different locations, the sound source can be accurately triangulated. If
produced by the photoacoustic effect, the sound source can
be imaged by PAT. In most practical cases, the sound source
is spatially distributed rather than localized at a point.
PAT has been implemented in two major forms. One form
is based on a scanning focused ultrasonic transducer. Acoustic focusing in combination with time-resolved detection of
photoacoustic waves directly provides 3D spatial resolution.
Scanning over the tissue produces tomographic images.
Dark-field confocal photoacoustic microscopy belongs to
this category.3335 Dark field means that the light illumination
pattern on the tissue surface is donut shaped with a dark
core; confocal means that the ultrasonic detection shares the
same focus as the light illumination. The other form is based
on an array of unfocused ultrasonic transducers. This
method, also referred to as photoacoustic computed tomography, requires the use of an inverse algorithm to reconstruct
a tomographic image.36
The dominant contrast of interest is based on the optical
absorption in the photoacoustic excitation phase. By using
multiwavelength measurement, one can simultaneously
quantify concentrations of multiple chromophores of different colors, such as oxygenated and deoxygenated hemoglobin molecules in red blood cells. Such quantification of hemoglobin can provide functional imaging of the
concentration and oxygen saturation of hemoglobin. Both
parameters are related to hallmarks of cancer: the concentration of hemoglobin correlates with angiogenesis,27 whereas
the oxygen saturation of hemoglobin correlates with
hypermetabolism.31 Such parameters of hemoglobin can also
be used to image brain activity.19 In addition, extrinsic optical absorption contrast agents can be used to provide molecular imaging of biomarkers.30,31
PAT is highly sensitive to optical absorption. PAT works
on dark backgroundwhen no absorption exists, the background signal is zero. Such dark background enables sensitive detection. In the presence of absorption, the foreground
signal is proportional to the absorption coefficient. Any small
fractional change in optical absorption coefficient translates
into an equal amount of fractional change in PAT signal,
which means a relative sensitivity of 100%.2 When the frac-
5760
5760
tional change is large, light transport modeling can be used

to quantify the optical absorption coefficient.37
The spatial resolution is derived from ultrasonic detection
in the photoacoustic emission phase. Ultrasonic scattering is
much weaker than optical scattering. The wavelength of the
detected photoacoustic wave is sufficiently short. As a result,
photoacoustic waves provide better resolution than optical
waves beyond the soft depth limit. Among other parameters,
the center frequency and the bandwidth of the ultrasonic detection system predominantly determine the spatial resolution of PAT. The greater the center frequency and the wider
the bandwidth, the better the spatial resolution is.
III. PREDICTIONS
III.A. Photoacoustic microscopy of optical absorption
going mainstream
Prediction 1: Photoacoustic microscopy will join the suite

of mainstream optical microscopy technologies.
Photoacoustic microscopy PAM has shown the following capabilities. First, PAM has broken through the current
fundamental depth limitation of high-resolution optical imaging modalities.33,34 The commercially available microscopic optical imaging modalitiesincluding confocal microscopy, two-photon microscopy, and optical coherence
tomographycannot provide penetration into scattering biological tissue beyond the soft depth limit 1 mm in the
skin because they are based on ballistic and quasiballistic
photons. Taking advantage of photons that have been scattered an any number of times, PAM has filled the void of
penetration and reached superdepth optical imaging. As will
be shown in Sec. III A, imaging depths of 3 30 mm have
been experimentally demonstrated. Second, the depth-toresolution ratio maximum imaging depth divided by the
axial resolution can be approximately maintained at more
than 100 while the imaging depth and spatial resolution are
scaled. Third, PAM provides functional imaging based on
physiologically specific endogenous optical absorption contrasts. The red color in Fig. 1 shows a PAM image of blood
vasculature using endogenous contrast only, the contrast between the blood and the background is as high as 13:1.34 In
comparison, other imaging modalities such as x-ray computed tomography and magnetic resonance imaging usually
require the use of exogenous contrast agents to image blood
vessels clearly with high contrast. Further, magnetic resonance imaging cannot measure the concentration of hemoglobin and the absolute oxygenation of hemoglobin. Fourth,
PAM can provide molecular or genetic imaging based on
exogenous contrast mechanisms.30,31 Fifth, PAM can be constructed to image in real time. A B-scan frame rate of 50 Hz
has been achieved.38,39 Sixth, PAM images are free of
speckle artifacts. Finally, PAM is safe to human subjects and
ready for clinical applications.
The various high-resolution optical imaging modalities
are compared in Table I. While PAM is sensitive to optical
absorption, it is insensitive to optical scattering or
fluorescence.2 The various modalities complement each other
in both penetration and contrast; they will therefore coexist.
FIG. 1. a Photograph of a melanoma in a nude mouse; b image acquired

with the 50-MHz photoacoustic microscope operating at 584 and 764 nm.
Composite of the two maximum amplitude projection MAP images projected along the z axis. Six orders of vessel branching 16 can be observed. c 3D rendering of the melanoma image acquired at 764 nm. Two
MAP images at this wavelength projected along the x and y axes are shown
on the sidewalls. The composite image in panel b is redrawn at the bottom.
The top of the tumor is 0.32 mm below the skin surface, and the thickness of
the tumor is 0.3 mm. d Close-up two-dimensional image of the melanoma
in a cross-section parallel with the z-x plane at the dashed line in panel b. e
HE-stained histological section at the same marked location. M: melanoma.
Reproduced with permission Fig. 2 of Ref. 34.
In light of its appealing capabilities, PAM is expected to

become commercially available and widely used in biomedical research and clinical practice.
III.B. Melanoma detection using photoacoustic
microscopy
Prediction 2: Melanoma detection will become a high impact application of photoacoustic microscopy.
While melanoma is the most deadly form of skin cancer,
it is curable if detected early. The diagnosis of melanoma is
often based on its appearance. Abnormally appearing skin
lesions are resected for biopsy. Because visual inspection is
inaccurate and biopsy is invasive, an accurate noninvasive
method of melanoma detection is desirable.
PAM can acquire 1 structural images measuring tumor
burden and depth, 2 functional images of hemoglobin concentration measuring tumor angiogenesisa hallmark of
cancer, 3 functional images of hemoglobin oxygen saturation SO2 measuring tumor hypoxia or hypermetabolism
another hallmark of cancer, and 4 images of melanin concentration measuring tumor pigmentationa hallmark of
melanotic melanoma, consisting of 90% of melanomas
note that even amelanotic melanoma contains a low concentration of melanin. An in vivo PAM image of a melanoma in
a small animal is shown in Fig. 1. The depth of melanoma,
also an important prognostic parameter, was measured by
PAM. By contrast, other high-resolution optical microscopy
technologies cannot provide the required penetration. Because of biological variability in most biological problems, a
5761
5761
TABLE I. Comparison of high-resolution optical imaging modalities.

Modality
Confocal microscopy
Two-photon microscopy
Optical coherence tomography
Photoacoustic microscopy
Penetration
Primary contrast
0.5 mm
0.5 mm
1 mm
Up to 30 mm, scalable
Scattering, fluorescence
Fluorescence
Scattering, polarization
Absorption
single biomarker is unlikely able to provide accurate diagnosis. However, the combination of four complementary measures can significantly enhance the likelihood of high accuracy. In addition, as Doppler photoacoustic microscopy
matures, blood flow measurement can add more diagnostic
information. Therefore, it is expected that the combined rich
contrasts will provide the sensitivity and specificity needed
for the early detection of melanoma.
III.C. Photoacoustic endoscopy
Prediction 3: Photoacoustic endoscopy will play a unique

role in diagnosing gastrointestinal cancer.
Incidence of gastrointestinal cancer is on the rise yet often
diagnosed at an untreatable stage. Optical endoscopy is the
current practice in identifying suspicious lesions for subsequent biopsy. Operating in the mode of surface or ballistic
imaging, existing optical endoscopy has severely limited
penetration. Ultrasound endoscopy is being introduced into
the clinic as well. Yet the mechanical contrast in ultrasound
images often cannot provide the required sensitivity and
specificity. Other medical imaging modalities such as x-ray
computed tomography and magnetic resonance imaging have
not been implemented endoscopically because of diagnostic
insensitivity, technological difficulties or expenses.
Light and sound are well suited for endoscopic applications. Because both forms of energy have been used endoscopically, PAM can naturally be implemented endoscopically. As PAM can provide a multitude of endogenous
contrasts at new depths, it is likely to provide accurate diagnosis of superficial as well as deep lesions in the gastrointestinal tract. Similarly, intravascular photoacoustic imaging can
also be implemented.40 Of course, adding molecular imaging
capability to photoacoustic endoscopy can further enhance
the diagnostic accuracy albeit at the inconvenience of using
exogenous contrast agent.
ception of introducing exogenous contrast agent. By

combining imaging physics with molecular and cell biology,
molecular imaging can potentially lead to better methods for
preventing, diagnosing, and managing diseases. For example, tumor biology and angiogenesis can be characterized
with high specificity because molecular imaging probes can
be designed to target at specific molecular biomarkers expressed on neoendothelial-and tumor-cell surfaces. Such
characterization will accelerate the development of
molecular-targeted cancer therapeutics. Nevertheless, molecular imaging will not supplant functional imaging. The
former requires the use of foreign contrast agent, whereas the
latter depends on intrinsic contrast only. Therefore, the
former may involve some degree of toxicity and may not be
universally applicable. In addition, functional imaging can
measure some physiological parameters that cannot be easily
measured by molecular imaging.
Simultaneous functional and molecular PAT was demonstrated in live mice Fig. 2.31 One million human U87 glioblastoma tumor cells were implanted stereotactically into
young adult immunocompromised nude mice in the head
3 mm below the scalp surface. One week post inoculation,
20 nmol of molecular imaging probe IRDye800-cKRGDf
was administered through the tail vein while mannitol was
used to permeabilize the blood brain barrier. Approximately
20 h later, imaging with PAT was conducted. Four laser
wavelengths764, 784, 804, and 824 nmwere used to
measure the spectral distribution of optical absorption due to
oxyhemoglobin,
deoxyhemoglobin,
and
IRDye800cKRGDf. Figure 2a shows an in vivo pseudocolor functional PAT image of the oxygen saturation of hemoglobin. A
focal region pointed by the arrow shows lower oxygen satu-
III.D. Simultaneous functional and molecular

photoacoustic tomography
Prediction 4: Simultaneous functional and molecular photoacoustic tomography will provide unprecedented sensitivity and specificity to cancer diagnosis.
While functional imaging is based on endogenous contrast, molecular imaging is based on exogenous contrast
which essentially stains invisible biomarkers in vivo. The
former maps physiological activities at the tissue or organ
level, whereas the latter images biological and pathophysiological processes at the molecular level. Both can be accomplished quantitatively and noninvasivelywith the exMedical Physics, Vol. 35, No. 12, December 2008
FIG. 2. a In vivo functional PAT image of hemoglobin in a nude mouse

brain with a U87 glioblastoma xenograft. The arrow indicates the hypoxic
region. b In vivo molecular PAT image of tail-vein injected IRDye800cKRGDf in the same nude mouse brain. The circle indicates the region of
integrin overexpression. Reproduced with permission Figs. 2 and 3 of
Ref. 31.
5762
ration of hemoglobin than the surrounding area, indicating

hypoxia characteristic of tumor due to hypermetabolism.
Figure 2b shows an in vivo pseudocolor molecular PAT
image of the molecular imaging probe. Integrin v3 is
known to overexpress in U87 glioblastoma tumors while
IRDye800-cKRGDf is known to bind to v3 integrin. The
circled region indicates the overexpression of integrin in the
U87 glioblastoma tumor. The focal regions of both the functional and molecular contrasts were corroborated to be the
tumor region through postmortem fluorescence imaging and
histological examination.
PAT is competitive in sensitivity compared with other existing molecular imaging technologies. In comparison to the
nMpM sensitivity of positron emission tomography at mm
spatial resolution, PAT has poorer sensitivity M nM but
better spatial resolution. While positron emission tomography uses ionizing radiation, PAT uses nonionizing light and
ultrasound. Even at its current level, the sensitivity of PAT is
superior to that of magnetic resonance imaging mM. Furthermore, the sensitivity of PAT can be potentially improved
by increasing the local optical fluence, increasing the optical
absorption cross section of the contrast agent, increasing the
photoacoustic conversion efficiency, increasing signal averaging, and sacrificing the spatial resolution. In comparison to
pure ultrasound imaging, which measures mechanical contrast, PAT dominantly measures optical contrast and has
many choices of contrast agents. Ultrasound molecular imaging, based on microbubbles, is confined to intravascular
space only. It is worth mentioning that x-ray imaging has not
yet been demonstrated to provide molecular imaging because
of its low sensitivity.
In summary, PAT represents a new paradigm for functional and molecular imaging. Translation to the clinic is
expected to follow the availability of molecular contrast
agents for human use. By combining functional and molecular imaging in PAT, unprecedentedly high sensitivity and
specificity of cancer diagnosis are expected.
III.E. Photoacoustic tomography of gene expressions
Prediction 5: Deep penetrating high-resolution photoacoustic tomography will become a complementary tool to
conventional optical imaging of reporter genes.
Understanding the nature of genes during in vivo development and pathogenesis is of significant interest. Such understanding can be enabled by gene expression imaging, a
form of molecular imaging based on reporter genes. A reporter gene is a segment of DNA fused to another gene of
interest. Unlike the contrast-carrying molecular imaging
probe presented in the preceding section, reporter genes must
express into protein products to provide contrast. The two
fused genes, sharing the same promoter, are transcribed into
a single mRNA molecule, which is then translated into fused
proteins. The protein encoded by the reporter gene carries
contrast either directly or through enzymatic reactions. As a
result, a positive contrast in a gene expression image indicates up-regulations of both the reporter gene and the fused
gene. Alternatively, a reporter gene can be used without gene
5762
fusion in order to assay the activity of a particular promoter

in a cell. If the promoter is ubiquitous, a reporter gene can be
used to tag cells. The most widely used conventional optical
imaging of reporter genes include fluorescence imaging of
green fluorescence proteins or their mutants41 and bioluminescence imaging of luciferases.42
Molecular PAT of reporter gene lacZ was demonstrated in
living rats.30 The lacZ gene encodes enzyme -galactosidase.
The enzyme can cleave an optically transparent substrate
5-bromo-4-chloro-3-indolyl--D-galactoside x-gal, yielding a stable dark blue product. The blue product increases
optical absorption and serves as a contrast agent for PAT. So
far, x-gal was injected locally because systemic delivery
through the tail vein turned out to be inefficient.
We hope to collaborate with developers of reporter genes
to construct optical-absorption-based reporter gene systems
that are more suitable for PAT. Because of its high spatial
resolution at depths beyond the reach of pure high-resolution
optical imaging, PAT is expected to become a complementary tool to conventional optical imaging of reporter genes
in vivo.
III.F. Doppler photoacoustic tomography
Prediction 6: Doppler photoacoustic tomography will provide imaging of blood flow in vivo with high spatial and
velocity resolution.
The photoacoustic Doppler effect was recently discovered
experimentally.43,44 While the optical Doppler effect involves
light only, the acoustic Doppler effect involves sound only.
By contrast, the photoacoustic Doppler effect involves both
light and sound, resulting from the combination of the photoacoustic and Doppler effects. Here, the photoacoustic wave
undergoes a Doppler shift when the photoacoustic source
the part of the light absorbing medium generating the photoacoustic wavehas motion relative to the acoustic detector.
An important application is to image blood flow in vivo
by measuring the photoacoustic Doppler effect due to moving red blood cells. We coined photoacoustic Doppler flowmetry PADF for this imaging technology. PADF has potential advantages over conventional Doppler flowmetry. First,
PAT can image blood vessels at a depth of a few millimeters
with a spatial resolution of tens of micrometers, whereas
Doppler optical coherence tomography can penetrate only
one optical transport mean free path. Second, PADF tracks
light absorbing particles instead of scattering particles. Because red blood cells have much higher optical absorption
than the background outside blood vessels, PADF should
have low background noise and high sensitivity. By contrast,
acoustic flowmetry suffers from background scattering and
consequently has difficulty measuring slow flows. Third, the
Doppler shift in PADF is nearly independent of the excitation
light, enabling accurate measurement of both speed and direction of flow even at large depths inside an optically scattering medium. In comparison, laser Doppler flowmetry suffers from multiple light scattering, limiting the sensing depth
and obscuring the information of flow direction. Finally,
5763
PADF is free of speckle artifacts. Therefore, PADF is expected to break fresh ground for in vivo flow measurement
with high spatial resolution and high velocity sensitivity.
III.G. Photoacoustic tomography of metabolic rate of
oxygen
Prediction 7: Photoacoustic tomography will become the

first single modality that can independently image the metabolic rate of oxygen in vivo using endogenous contrast.
The metabolic rate of oxygen MRO2 quantifies metabolism, which can potentially be directly used to measure hypermetabolism, a hallmark of cancer. The MRO2 is defined
as the amount of oxygen consumed in a given tissue region
per unit time per 100 g of tissue or of the organ of interest.
Because in typical physiological conditions, hemoglobin is
the dominant carrier of oxygen the key measure of blood
oxygenation is oxygen saturation of hemoglobin SO2.
PAT can simultaneously image cross sections of blood
vessels, the concentration and oxygenation of hemoglobin, as
well as blood flow in vivo. These four parameters suffice to
quantify oxygen metabolism in tissues and organs. For a
single-vessel system, we have
out
in .
MRO2 SOin
2 SO2 CHb Ain
Here, Ain is the area of the incoming vessel, in is the mean

flow velocity of blood in the incoming vessel, CHb is the total
concentration of hemoglobin, SOin
2 is the SO2 of the incoming vessel to the region of interest, and SOout
2 is the SO2 of
the outgoing vessel. While Ain is measured by structural PAT,
CHb and SO2 are measured by multiwavelength functional
PAT; further, in is estimated using Doppler photoacoustic
flowmetry.
It is extremely fortuitous that all independent parameters
in Eq. 1 can be measured by PAT, a single hybrid imaging
modality. By contrast, magnetic resonance imaging detects
only temporal changes in deoxyhemoglobin concentration
instead of absolute values. Equation 1 can be extended to a
multivessel system. Although positron emission tomography
can image MRO2, exogenous radioactive tracers are
required.45,46 Diffuse optical tomography has also been used
to image MRO2 except that a theoretical model or another
technique diffuse correlation imaging is needed to estimate
the blood flow.47,48 Further, both positron emission tomography and diffuse optical tomography have low spatial resolution. Therefore, PAT will potentially become the first single
modality that can independently image the metabolic rate of
oxygen in vivo at high spatial resolution without the use of
exogenous contrast agent. It is expected to play a critical role
in metabolism-related biomedical research and clinical diagnosis.
III.H. Sentinel lymph node mapping
using photoacoustic tomography
Prediction 1: Sentinel lymph node mapping will become

one of the first niche applications of photoacoustic tomography.
5763
For patients with breast cancer but clinically negative axillae, biopsy of the sentinel lymph node SLNthe first
draining nodesfor axillary staging has become a standard
clinical practice. In this conventional method, both optical
dye and radioactive Tc-99 colloids are injected into the breast
tumor. Both contrast agents are accumulated by SLNs. Emission from radioactivity is first detected to locate approximately the SLNs. Then, the SLNs are located precisely
through an invasive surgical procedure by visually tracing
the accumulated optical dye such as methylene blue. The
identified SLNs are then resected for pathological examination. Unfortunately, the surgical procedure has associated
morbidity. A noninvasive method for accurately locating the
SLN in vivo is sought-after because noninvasive diagnostic
methods such as fine needle biopsy could be subsequently
utilized without surgery.
PAT can potentially map the SLN noninvasively with high
spatial resolution in real time, avoiding invasive surgery and
harmful radioactivity. Preliminary data demonstrated that
PAT has both the required penetration depth and spatial resolution for SLN mapping.49 PAT can identify whether a node
is sentinel by detecting the accumulated dye because the dye
has high optical absorption contrast. The peak absorption
wavelength of methylene blue dye is 690 nm, which lies near
the optimal optical penetration window. Since the optical
contrast agent has been used in clinical SLN biopsy already,
avoiding FDA approval for a new optical contrast agent can
significantly accelerate the translation of PAT into the
clinic.50 While PAT has broad potential applications, imaging
the SLN is a perfect niche clinical application of PAT with
anticipated high impact on breast axillary staging and management.
III.I. Multiscale photoacoustic imaging in vivo
with common signal origins
Prediction 9: Multiscale photoacoustic tomography of the

same signal origins in vivo will enable research in multiscale
biology.
Multiscale biology, recognized as the future of biomedical
sciences, needs multiscale imaging. To understand the workings of a whole biological system, components spanning
multiple spatial scales must be integrated. The components
may range from subcellular organelles sub-m scale,
through cells m, to organs cm. To facilitate research in
multiscale biology, we must provide multiscale imaging in
vivo with common signal origins. If the contrast mechanisms
were disconnected among the images of various length
scales, integrating discoveries across the scales would be
challenging.
PAT can provide multiscale imaging in vivo with the same
contrast origins Fig. 3. Our original 50-MHz photoacoustic
microscope can penetrate 3 mm at tens of m resolution.
The image resolution and the penetration limit are scalable
with the ultrasonic frequency within the reach of photons.
On the one hand, the penetration limit can be scaled up to
centimeters, i.e., the hard depth limit. In this case, photoacoustic microscopy is scaled to macroscopy. For example,
5764
5764
limited by the annihilation range of positrons, on the order of

1 mm. Although ultrasound can achieve microscopic imaging by increasing the frequency from the clinically used
MHz range to GHz range, the contrast reflects mechanical
rather than biochemical properties. Therefore, PAT is expected to be the first imaging technology, if not the only one
on the horizon, that can potentially span a length scale ranging from organelles to organs while providing consistent biochemical contrast.
FIG. 3. a Image acquired with the 5-MHz photoacoustic imaging system

showing the 30 mm penetration limit in chicken breast tissue. Reproduced
with permission Fig. 2 Ref. 51. b Photograph taken with transmission
optical microscopy showing the microvasculature in a nude mouse ear. c
In vivo photoacoustic image of the vasculature acquired with opticalresolution photoacoustic microscopy. CL: capillary; SG: sebaceous gland.
Reproduced with permission Fig. 3 of Ref. 52.
employing a 5-MHz ultrasonic transducer and an 804-nm

light source scaled the penetration to 3 cm as shown in Fig.
3a.51 On the other hand, the spatial resolution can be scaled
down to 10 m. Using a focused light beam resulted in
5 m lateral resolution, which is limited by the focal diameter of the excitation light beam. This resolution allows in
vivo imaging of capillariesthe smallest blood vesselsas
shown in Fig. 3c.52 Of course, optical focusing is effective
only within the soft depth limit. Theoretically, the ultimate
spatial resolution is optically diffraction limited, potentially
yielding sub-m resolution in PAT. Therefore, PAT can potentially image as small as organelles and as large as organs
with the same contrast mechanism.
The multiscale or multiresolution capability of PAT is
analogous to that of conventional optical microscopy, where
a single microscope possesses multiple attached objectives,
permitting users to select a magnification factor easily. Such
a capability allows one to image a large region of interest at
a relatively low resolution. Then, one may zoom into a
smaller region of interest with higher resolution to visualize
smaller features while the same contrast mechanism is maintained. This scalability is needed because of the tradeoff between the region of interest and the spatial resolution. Of
course, such scalability is not limited to a single imaging
system. On the contrary, multiple systems can potentially be
used to image at various resolution scales.
PAT is uniquely positioned for multiscale imaging in vivo.
Conventional optical microscopy in various forms is the tool
of choice for microscopic imaging of cells and tissues. It is
capable of offering a wealth of optical contrasts. PAT extends
the depth of conventional optical microscopy beyond the soft
depth limit and provides a continuum for multiscale imaging.
In comparison, conventional clinical imaging modalities
such as x-ray computed tomography, magnetic resonance imaging, positron emission tomography, and single photon
emission computed tomography have not been widely used
in microscopic form for various reasons. For example, the
resolution of positron emission tomography is fundamentally
III.J. Simultaneous photoacoustic and thermoacoustic

tomography of the breast
Prediction 10: Simultaneous photoacoustic and thermoacoustic tomography of the breast will supplement or supplant
x-ray mammography.
While breast cancer remains the leading cause of cancer
death among women, early diagnosis improves survival.
X-ray mammography, the only mass screening tool for detecting nonpalpable breast cancers, has been shown to reduce
breast cancer deaths. However, ionizing radiation is required,
and detecting dense breasts and early-stage tumors is met
with limited sensitivity. In addition, imaging tumors close to
the chest wall is difficult. Therefore, alternative nonionizing
radiation-based screening technologies are needed. So far,
ultrasound imaging has been used as an adjunct to x-ray
mammography, especially in differentiating cystic from solid
breast masses. Magnetic resonance imaging has high sensitivity but variable specificity. It is further limited by its high
cost and use of exogenous contrast agent.
PAT and TAT can be integrated into conventional ultrasound imaging to provide multiple complementary contrasts
simultaneously. Three types of images of the breast can be
acquired: 1 structural ultrasound pulse-echo images, 2
functional laser-induced PAT images of concentration and
oxygen saturation of hemoglobin, and 3 radio frequencyinduced TAT images of water and sodium concentrations.
Good optical contrast, related to hallmarks of cancer, between tumor and normal breast tissues has been observed.53
In TAT, up to 5:1 contrast was observed between tumor and
normal breast tissues.29 In comparison, x-ray contrast among
soft tissues is typically only a few percent. Again, a single
contrast is usually unable to provide accurate diagnosis because of biological variability. However, the combination of
multiple complementary measures will likely improve the
accuracy.
While compatible with ultrasound imaging, PAT and TAT
are also compatible with each other, sharing the same ultrasonic detection system and image reconstruction algorithm.
Integrating PAT with TAT adds significant contrast but does
not add significant costs. PAT and TAT can potentially be
distributed to physicians as an add-on feature to existing
clinical ultrasound imaging systems. While PAT and TAT are
emerging technologies, ultrasonography is a mature clinical
technology. Incorporating PAT and TAT into conventional
ultrasonography will facilitate physicians acceptance of the
new technologies.
5765
In summary, PAT and TAT use nonionizing radiation only

and can be implemented at relatively low cost. It can potentially provide high-resolution, multicontrast imaging of the
breast. Depending on the clinical trial outcome, PAT and TAT
could potentially serve as an adjunct or even replace x-ray
mammography if its sensitivity and specificity are superior.
III.K. Photoacoustic and thermoacoustic tomography
of the brain
Prediction 11: Photoacoustic and thermoacoustic tomography will provide low-cost, high-resolution, real-time, and
even bedside functional imaging of the brain.
Existing high-resolution human brain imaging
modalitiesincluding x-ray computed tomography, magnetic resonance imaging, and ultrasonographyhave limitations. Both x-ray computed tomography and magnetic resonance imaging are expensive and bulky. Further, x-ray
computed tomography uses ionizing radiation, and magnetic
resonance imaging requires a strong magnetic field. Therefore, it is advantageous to develop an affordable, nonionizing, high-resolution, and real-time imaging modality that can
be used in operating rooms or at the bedside. Such a system
can be used to monitor brain functions and conditions such
as strokes, head injuries, tumors, and brain infections. In certain cases, ultrasound brain imaging comes close to meeting
these objectives. Ultrasonography is an established pediatric
brain imaging modality when used before the fontanels are
closed. After the closure of the fontanels, the image quality
degrades significantly because the skull severely attenuates
and distorts ultrasonic waves.
Fortunately, unlike pulse-echo ultrasound imaging, PAT
and TAT involve only one-way ultrasound attenuation and
distortion due to the skull. As demonstrated previously,19 ultrasound penetration through small animal skulls is feasible,
and the associated distortions are relatively small. The authors group also demonstrated TAT through Rhesus monkey
skulls.54 The extension of PAT and TAT to human brain imaging is an area of research. The primary task is to correct
for the phase distortion due to the skull. The skull basically
functions as an irregular lens, whose aberration effect can be
compensated for if the phase distortion can be quantified. A
potential solution is to use geometric measurements of the
skull from x-ray computed tomography or magnetic resonance imaging to predict the phase distortion.
PAT and TAT can provide rich contrasts for highresolution structural and functional imaging. As demonstrated by diffuse optical tomography, near-infrared light can
penetrate through the skull to the cortex and has been used
for functional brain imaging at low spatial resolution. PAT
can potentially provide similar penetration but higher resolution. At an appropriate frequency, radio frequency can penetrate deeper than light. Therefore, TAT can potentially image deeper structures of the brain, while PAT will likely be
used to image cortical functions.
PAT and TAT are also relatively low-cost, real-time, and
portable. In comparison to x-ray computed tomography and
magnetic resonance imaging, ultrasonography is low cost.
5765
PAT and TAT essentially involve a laser or radio frequency

source in addition to the detection components of an ultrasound imaging system. Therefore, PAT and TAT can be
implemented at relatively low cost. When ultrasound arrays
are used as in ultrasonography, PAT and TAT can be performed in real time. A standard ultrasound imaging system is
portable and can be used at bedside. With the addition of a
laser or radio frequency source, the entire imaging system
can still be made relatively portable for bedside or operatingroom applications. Again, integrating PAT and TAT into existing ultrasound imaging systems can potentially accelerate
acceptance of the new technology by physicians.
In summary, the major challenge to PAT and TAT of the
human brain is the phase distortion due to the skull. Correcting the distortion is believed to be a tractable problem.
Therefore, PAT and TAT are expected to provide low-cost,
high-resolution, real-time, and bedside functional imaging of
the brain.
III.L. Low-background thermoacoustic molecular
imaging
Prediction 12: Radio frequency-induced thermoacoustic

tomography will provide high-resolution, high-sensitivity,
and deep-penetrating molecular imaging with the aid of targeted radio frequency contrast agents.
Molecular imaging based on TAT has immense potential.
PAT can provide molecular imaging within the hard depth
limit for optical imaging. By contrast, radio frequency waves
at sufficiently low frequency can penetrate beyond 10 cm as
demonstrated in magnetic resonance imaging. Ultrasound at
an appropriate frequency can travel a similar depth. Consequently, radio frequency-induced TAT can potentially provide penetration greater than 10 cm. Because low-frequency
radiofrequency waves have low absorption, endogenous thermoacoustic signals would be weak, providing a low background. Such a low background enables highly sensitive detection of any other radio frequency absorbers, which can be
targeted molecular contrast agents. While PAT boasts a
myriad of contrast agents, TAT has none available yet except
the recently tested carbon nanotubes.55 Molecular imaging
TAT is somewhat analogous to low-background positron
emission tomography except that TAT uses nonionizing radiation and provides high spatial resolution. The authors
group is actively seeking collaborations with chemists to
identify contrast agents for TAT. Once such contrast agents
become available, TAT is expected to avail a new paradigm
of high-resolution, low-background, high-sensitivity, and
deep-penetrating molecular imaging.
IV. SUMMARY
The author believes that PAT and TAT will find broad
applications in both basic research and clinical care. In general, it can potentially be applied in tumor studies, where
angiogenesis and hypermetabolism are highly correlated
with cancer aggressiveness; in neuroscience, where oxygen
saturation and concentration of hemoglobin are related to
neural activities; in radiotherapy and chemotherapy of can-
5766
cer, where hypoxia is responsible for resistance to therapy;

and in trauma evaluation, where optical absorption is associated with both hemorrhage and edema.
The aforementioned predictions represent the authors biased opinions of the most promising prospects of PAT and
TAT. The topics of the 12 predictions are recapitulated as
follows: 1 photoacoustic microscopy of optical absorption
emerging as a mainstream technology, 2 melanoma detection using photoacoustic microscopy, 3 photoacoustic endoscopy, 4 simultaneous functional and molecular photoacoustic tomography, 5 photoacoustic tomography of gene
expression, 6 Doppler photoacoustic tomography for flow
measurement, 7 photoacoustic tomography of metabolic
rate of oxygen, 8 photoacoustic mapping of sentinel lymph
nodes, 9 multiscale photoacoustic imaging in vivo with
common signal origins, 10 simultaneous photoacoustic and
thermoacoustic tomography of the breast, 11 photoacoustic
and thermoacoustic tomography of the brain, and 12 lowbackground thermoacoustic molecular imaging.
The following specific clinical applications are also worth
pursuing: Demarcation of cancer for Mohs surgery, treatment
planning, and monitoring for laser therapy of port-wine
stains, skin burn imaging to determine the need for skin
graft, intraoperative demarcation of brain tumors, and intraoperative imaging of brain functions. Undoubtedly, many
more applications will be identified.
While the photoacoustics research community continues
to discover new phenomena and invent new technologies,
several companies are actively commercializing photoacoustic tomography. We look forward to seeing the impact of PAT
and TAT on biomedicine.
ACKNOWLEDGMENTS
This work is sponsored in part by National Institutes of
Health Grant Nos. R01 EB000712 and R01 NS46214
Bioengineering Research Partnership. L.W. has a financial
interest in Endra, Inc., which, however, did not support this
work.
a

[email protected]
1
M. H. Xu and L. V. Wang, Photoacoustic imaging in biomedicine, Rev.
Sci. Instrum. 77, 041101 2006.
2
L. V. Wang, Tutorial on photoacoustic microscopy and computed tomography, IEEE J. Sel. Top. Quantum Electron. 14, 171179 2008.
3
A. A. Oraevsky and L. V. Wang, Photons Plus Ultrasound: Imaging and
Sensing SPIE, Bellingham, 2007, Vol. 643.
4
L. V. Wang and H. Wu, Biomedical Optics: Principles and Imaging
Wiley, Hoboken, NJ, 2007.
5
C. G. A. Hoelen, F. F. M. de Mul, R. Pongers, and A. Dekker, Threedimensional photoacoustic imaging of blood vessels in tissue, Opt. Lett.
23, 648650 1998.
6
L. V. Wang, X. M. Zhao, H. T. Sun, and G. Ku, Microwave-induced
acoustic imaging of biological tissues, Rev. Sci. Instrum. 70, 37443748
1999.
7
R. A. Kruger, D. R. Reinecke, and G. A. Kruger, Thermoacoustic computed tomographytechnical considerations, Med. Phys. 26, 18321837
1999.
8
A. A. Karabutov, E. V. Savateeva, N. B. Podymova, and A. A. Oraevsky,
Backward mode detection of laser-induced wide-band ultrasonic transients with optoacoustic transducer, J. Appl. Phys. 87, 20032014
2000.
5766
9
G. Ku and L. V. Wang, Scanning electromagnetic-induced thermoacoustic tomography: Signal, resolution, and contrast, Med. Phys. 28, 410
2001.
10
K. P. Kostli, D. Frauchiger, J. J. Niederhauser, G. Paltauf, H. P. Weber,
and M. Frenz, Optoacoustic imaging using a three-dimensional reconstruction algorithm, IEEE J. Sel. Top. Quantum Electron. 7, 918923
2001.
11
M. H. Xu, G. Ku, and L. V. Wang, Microwave-induced thermoacoustic
tomography using multi-sector scanning, Med. Phys. 28, 19581963
2001.
12
G. Paltauf, J. A. Viator, S. A. Prahl, and S. L. Jacques, Iterative reconstruction algorithm for optoacoustic imaging, J. Acoust. Soc. Am. 112,
15361544 2002.
13
M. H. Xu and L. V. Wang, Time-domain reconstruction for thermoacoustic tomography in a spherical geometry, IEEE Trans. Med. Imaging 21,
814822 2002.
14
Y. Xu, D. Z. Feng, and L. V. Wang, Exact frequency-domain reconstruction for thermoacoustic tomography-I: Planar geometry, IEEE Trans.
Med. Imaging 21, 823828 2002.
15
Y. Xu, M. H. Xu, and L. V. Wang, Exact frequency-domain reconstruction for thermoacoustic tomography-II: Cylindrical geometry, IEEE
16
V. G. Andreev, A. A. Karabutov, and A. A. Oraevsky, Detection of
ultrawide-band ultrasound pulses in optoacoustic tomography, IEEE
Trans. Ultrason. Ferroelectr. Freq. Control 50, 13831390 2003.
17
D. Finch, S. K. Patch, and Rakesh, Determining a function from its mean
values over a family of spheres, SIAM J. Math. Anal. 35, 12131240
2003.
18
K. P. Kostli and P. C. Beard, Two-dimensional photoacoustic imaging by
use of Fourier-transform image reconstruction and a detector with an
anisotropic response, Appl. Opt. 42, 18991908 2003.
19
X. D. Wang, Y. J. Pang, G. Ku, X. Y. Xie, G. Stoica, and L. V. Wang,
Noninvasive laser-induced photoacoustic tomography for structural and
functional in vivo imaging of the brain, Nat. Biotechnol. 21, 803806
2003.
20
X. D. Wang, Y. J. Pang, G. Ku, G. Stoica, and L. V. Wang, Threedimensional laser-induced photoacoustic tomography of mouse brain with
the skin and skull intact, Opt. Lett. 28, 17391741 2003.
21
M. H. Xu and L. V. Wang, Analytic explanation of spatial resolution
related to bandwidth and detector aperture size in thermoacoustic or photoacoustic reconstruction, Phys. Rev. E 67, 056605 2003.
22
M. Haltmeier, O. Scherzer, P. Burgholzer, and G. Paltauf, Thermoacoustic computed tomography with large planar receivers, Inverse Probl. 20,
16631673 2004.
23
G. Ku, X. D. Wang, G. Stoica, and L. V. Wang, Multiple-bandwidth
photoacoustic tomography, Phys. Med. Biol. 49, 13291338 2004.
24
Y. Xu and L. V. Wang, Time reversal and its application to tomography
with diffracting sources, Phys. Rev. Lett. 92, 033902 2004.
25
B. T. Cox and P. C. Beard, Fast calculation of pulsed photoacoustic fields
in fluids using k-space methods, J. Acoust. Soc. Am. 117, 36163627
2005.
26
G. Ku and L. V. Wang, Deeply penetrating photoacoustic tomography in
biological tissues enhanced with an optical contrast agent, Opt. Lett. 30,
507509 2005.
27
G. Ku, X. D. Wang, X. Y. Xie, G. Stoica, and L. V. Wang, Imaging of
tumor angiogenesis in rat brains in vivo by photoacoustic tomography,
Appl. Opt. 44, 770775 2005.
28
J. Zhang, M. A. Anastasio, X. C. Pan, and L. V. Wang, Weighted expectation maximization reconstruction algorithms for thermoacoustic tomography, IEEE Trans. Med. Imaging 24, 817820 2005.
29
G. Ku, B. D. Fornage, X. Jin, M. H. Xu, K. K. Hunt, and L. V. Wang,
Thermoacoustic and photoacoustic tomography of thick biological tissues toward breast imaging, Technol. Cancer Res. Treat. 4, 559565
2005.
30
L. Li, R. J. Zemp, G. Lungu, G. Stoica, and L. V. Wang, Photoacoustic
imaging of lacZ gene expression in vivo, J. Biomed. Opt. 12, 02050413 2007.
31
M. Li, J. Oh, X. Xie, G. Ku, W. Wang, C. Li, G. Lungu, G. Stoica, and L.
V. Wang, Simultaneous molecular and hypoxia imaging of brain tumors
in vivo using spectroscopic photoacoustic tomography, Proc. IEEE 96,
481489 2008.
32
L. V. Wang and G. Liang, Absorption distribution of an optical beam
focused into a turbid medium, Appl. Opt. 38, 49514958 1999.
5767
33
K. Maslov, G. Stoica, and L. V. Wang, In vivo dark-field reflection-mode

photoacoustic microscopy, Opt. Lett. 30, 625627 2005.
34
H. F. Zhang, K. Maslov, G. Stoica, and L. V. Wang, Functional photoacoustic microscopy for high-resolution and noninvasive in vivo imaging,
Nat. Biotechnol. 24, 848851 2006.
35
H. F. Zhang, K. Maslov, and L. V. Wang, In vivo imaging of subcutaneous structures using functional photoacoustic microscopy, Nature Protocols 2, 797804 2007.
36
M. H. Xu and L. V. Wang, Universal back-projection algorithm for photoacoustic computed tomography, Phys. Rev. E 71, 016706 2005.
37
B. T. Cox, S. R. Arridge, K. P. Kostli, and P. C. Beard, Two-dimensional
quantitative photoacoustic image reconstruction of absorption distributions in scattering media by use of a simple iterative method, Appl. Opt.
45, 18661875 2006.
38
R. J. Zemp, R. Bitton, M. L. Li, K. K. Shung, G. Stoica, and L. V. Wang,
Photoacoustic imaging of the microvasculature with a high-frequency
ultrasound array transducer, J. Biomed. Opt. 12, 010501-13 2007.
39
R. J. Zemp, L. Song, R. Bitton, K. K. Shung, and L. V. Wang, Realtime
photoacoustic microscopy in vivo with a 30-MHz ultrasound array transducer, Opt. Express 16, 79157928 2008.
40
S. Sethuraman, J. H. Amirian, S. H. Litovsky, R. W. Smalling, and S. Y.
Emelianov, Ex vivo characterization of atherosclerosis using intravascular photoacoustic imaging, Opt. Express 15, 1665716666 2007.
41
R. Y. Tsien, The green fluorescent protein, Annu. Rev. Biochem. 67,
509544 1998.
42
C. H. Contag and M. H. Bachmann, Advances in vivo bioluminescence
imaging of gene expression, Annu. Rev. Biomed. Eng. 4, 235260
2002.
43
H. Fang, K. Maslov, and L. V. Wang, Photoacoustic Doppler effect from
flowing small light-absorbing particles, Phys. Rev. Lett. 99, 18450114 2007.
44
H. Fang, K. Maslov, and L. V. Wang, Photoacoustic Doppler flow measurement in optically scattering media, Appl. Phys. Lett. 91, 264103
264101 2007.
45
N. Kudomi, H. Watabe, T. Hayashi, and H. Iida, Separation of input
function for rapid measurement of quantitative CMRO2 and CBF in a
single PET scan with a dual tracer administration method, Phys. Med.
Biol. 52, 18931908 2007.
5767
46
S. H. Yee, K. Lee, P. A. Jerabek, and P. T. Fox, Quantitative measurement of oxygen metabolic rate in the rat brain using microPET imaging of
briefly inhaled O-15-labelled oxygen gas, Nucl. Med. Commun. 27,
573581 2006.
47
D. A. Boas, G. Strangman, J. P. Culver, R. D. Hoge, G. Jasdzewski, R. A.
Poldrack, B. R. Rosen, and J. B. Mandeville, Can the cerebral metabolic
rate of oxygen be estimated with near-infrared spectroscopy?, Phys.
Med. Biol. 48, 24052418 2003.
48
J. P. Culver, T. Durduran, T. Furuya, C. Cheung, J. H. Greenberg, and A.
G. Yodh, Diffuse optical tomography of cerebral blood flow, oxygenation, and metabolism in rat during focal ischemia, Biopolymers 23,
911924 2003.
49
K. H. Song, E. W. Stein, J. A. Margenthaler, and L. V. Wang, Noninvasive photoacoustic identification of sentinel lymph nodes containing methylene blue in vivo in a rat model, J. Biomed. Opt. 13, 054033-16
2008.
50
E. M. Sevick-Muraca, R. Sharma, J. C. Rasmussen, M. V. Marshall, J. A.
Wendt, H. Q. Pham, E. Bonefas, J. P. Houston, L. Sampath, K. E. Adams,
D. K. Blanchard, R. E. Fisher, S. B. Chiang, R. Elledge, and M. E.
Mawad, Imaging of lymph flow in breast cancer patients after microdose
administration of a near-infrared flurophore: Feasibility study, Radiology
246, 734741 2008.
51
K. Song and L. V. Wang, Deep reflection-mode photoacoustic imaging
of biological tissue, J. Biomed. Opt. 12, 060503 2007.
52
K. Maslov, H. F. Zhang, S. Hu, and L. V. Wang, Optical-resolution
photoacoustic microscopy for in vivo imaging of single capillaries, Opt.
Lett. 33, 929931 2008.
53
Q. Zhu, C. Xu, P. Y. Guo, A. Aquirre, B. H. Yuan, F. Huang, D. Castilo,
J. Gamelin, S. Tannenbaum, M. Kane, P. Hedge, and S. Kurtzman, Optimal probing of optical contrast of breast lesions of different size located
at different depths by US localization, Technol. Cancer Res. Treat. 5,
365380 2006.
54
Y. Xu and L. V. Wang, Rhesus monkey brain imaging through intact
skull with thermoacoustic tomography, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 53, 542548 2006.
55
M. Pramanik, M. Swierczewska, D. Green, B. Sitharaman, and L. V.
Wang, Carbon nanotubes as a multimodalthermoacoustic and
photoacousticcontrast agent unpublished.
Vision 20/ 20: Proton therapy

Alfred R. Smitha
Department of Radiation Oncology, The University of Texas M. D. Anderson Cancer Center,
1515 Holcombe Boulevard, Houston, Texas 77030
Received 6 August 2008; revised 5 December 2008; accepted for publication 8 December 2008;
published 26 January 2009
The first patients were treated with proton beams in 1955 at the Lawrence Berkeley Laboratory in
California. In 1970, proton beams began to be used in research facilities to treat cancer patients
using fractionated treatment regimens. It was not until 1990 that proton treatments were carried out
in hospital-based facilities using technology and techniques that were comparable to those for
modern photon therapy. Clinical data strongly support the conclusion that proton therapy is superior
to conventional radiation therapy in a number of disease sites. Treatment planning studies have
shown that proton dose distributions are superior to those for photons in a wide range of disease
sites indicating that additional clinical gains can be achieved if these treatment plans can be reliably
delivered to patients. Optimum proton dose distributions can be achieved with intensity modulated
protons IMPT, but very few patients have received this advanced form of treatment. It is anticipated widespread implementation of IMPT would provide additional improvements in clinical
outcomes. Advances in the last decade have led to an increased interest in proton therapy. Currently,
proton therapy is undergoing transitions that will move it into the mainstream of cancer treatment.
For example, proton therapy is now reimbursed, there has been rapid development in proton therapy
technology, and many new options are available for equipment, facility configuration, and financing. During the next decade, new developments will increase the efficiency and accuracy of proton
therapy and enhance our ability to verify treatment planning calculations and perform quality
assurance for proton therapy delivery. With the implementation of new multi-institution clinical
studies and the routine availability of IMPT, it may be possible, within the next decade, to quantify
the clinical gains obtained from optimized proton therapy. During this same period several new
proton therapy facilities will be built and the cost of proton therapy is expected to decrease, making
proton therapy routinely available to a larger population of cancer patients. 2009 American
Key words: proton therapy, particle therapy, hadron therapy, radiation oncology
I. ABBREVIATED HISTORY OF PROTON THERAPY
E. Rutherford proposed the existence of protons in 1919;1
however, it was not until 1955 that the first patients were
treated with proton beams at the Lawrence Berkeley Laboratory LBL in California.2 In the years between 1919 and
1955, two events occurred that had an important impact on
proton therapy. First, in 1930, E. O. Lawrence built the first
cyclotron, paving the way for future particle accelerators
with energies high enough for cancer treatment applications.
In 1946, Robert Wilson then proposed that because of their
physical properties, beams of protons would be advantageous for the treatment of deep-seated cancers stating, It
will be easy to produce well collimated narrow beams of fast
protons, and since the range of the beam is easily controllable, precision exposure of well defined small volumes
within the body will soon be feasible.3 Wilson also described the use of a rotating wheel of variable thickness, i.e.,
a range modulation wheel RMW, interposed in a proton
beam as a method of adding together several Bragg peaks of
variable energies and weights in order to spread the proton
stopping region in depth to deliver a uniform dose to an
extended target volume. Figure 1 illustrates the concepts proposed by Wilson.
556
Med. Phys. 36 2, February 2009
Further progress in proton therapy was rather slow in the

35 years following the first patient treatments at LBL. During this time, patients were treated in a few research facilities, notably in the United States, Sweden, and Russia, where
the methods and techniques for proton therapy were further
refined and new technology was developed. The clinical results obtained in these research facilities demonstrated the
feasibility and efficacy of proton therapy. In 1990, the modern era of particle therapy then began when the first hospitalbased proton therapy facility was opened at the Loma Linda
University Medical Center LLUMC in California.4 Now
there are approximately 25 facilities around the world treating cancer patients with proton beams with more than 55,000
patients treated.5
II. PROTON BEAM DELIVERY TECHNIQUES
In order for the reader to understand the material in the
following sections, it is necessary to understand the various
methods used to deliver proton radiation therapy, which fall
into two general categories: passive scattering and spot scanning also called pencil beam scanning techniques. A basic
overview of the techniques is given here; for further reading,
good general descriptions can be found in the literature.6,7
0094-2405/2009/362/556/13/$25.00
556
557
Alfred R. Smith: Proton Therapy
Photons
557
15 MV Photons
vs.
Protons
Ideal dose
distribution
Protons
The passive scattering technique uses scattering devices

in the treatment delivery nozzle usually double scatterers for
large fields and single scatterers for small e.g., radiosurgery
fields to spread the beam laterally and an RMW or ridge
filter7 to create a spread-out-Bragg peak SOBP in the target
volume. Between the nozzle exit and the patient surface, a
treatment field-specific collimator is used to shape the field
laterally to conform to the maximum beams-eye-view extent
of the target volume and a range compensator is used to
correct for patient surface irregularities, density heterogeneities in the beam path, and changes in the shape of the distal
target volume surface. The size of the SOBP is chosen to
cover the greatest extent in depth of the target volume. The
SOBP size is constant over the entire target volume; therefore, in general, there is some pull back of the high dose
region into normal tissues proximal to the target volume
Fig. 2. Because several treatment fields are usually used,
FIG. 1. Comparison between the depth dose curves for

15 MV photons and a proton spread-out-Bragg peak
SOBP. A target volume is shown in red. Shown also
in red lines is an ideal dose distribution for the target
volume, which provides uniform, maximum dose to the
target volume and zero dose outside the target volume.
The proton dose distribution approaches the ideal case
to a much greater extent than does the photon dose
distribution. Notably, the proton dose stops abruptly
distal to the target volume and delivers less dose to the
region proximal to the target volume.
each directed from a different angle, this high dose pull back
into normal tissue is not additive over all fields. In principle,
the use of an RMW to obtain an SOBP in passive scattering
techniques modulates both energy and beam intensity. However, this should not be confused with intensity modulated
proton treatment IMPT achieved with beam scanning techniques as described below.
The second type of beam delivery utilizes spot scanning
techniques also often called pencil beam scanning. Figure 3
illustrates how a target can be scanned by placing Bragg
peaks throughout the volume by the use of scanning magnets
and energy changes. Energy changes in scanning techniques
can be carried out with various methods including: 1 energy changes in the accelerator when a synchrotron is used;
2 energy changes made with an energy selection system
when a cyclotron is used; or 3 either of the above methods
plus energy absorbers in the treatment nozzle.
FIG. 2. Diagram of a typical passive

scattering system. The range modulation wheel or ridge filter creates the
SOBP; the double scattering system
spreads the beam laterally; and the
collimator and range compensator
shape the beam to the lateral extent
and distal surface of the target volume.
The size of the SOBP is chosen to
cover the greatest extent of the target
volume and, where the target volume
is smaller, the high dose is pulled back
into normal tissues proximal to the target volume.
Medical Physics, Vol. 36, No. 2, February 2009
558
558
High-density
structure
Scanning Magnets
Critical
Structure
Scanning Magnets
Proton Pencil
Beam
FIG. 3. Diagram of a typical pencil beam scanning system. Two sets of scanning magnets scan the beam in a
2D pattern: the beam range is adjusted by changing the
beam energy entering the nozzle. In the usual case, all
spots for the deepest range are scanned, the energy is
changed, and all spots with the new range are scanned,
etc., until the entire target volume has been scanned. In
pencil beam scanning, the high dose region can be confined to the target volume and no high dose spills over
into normal tissue.
Body
Surface
Currently, no standard definitions exist for the various

techniques used to deliver scanned beam proton treatments.
For this paper, we will use the definitions given below.
Scanning techniques can be used to spread the beam laterally to generate large uniform treatment fields instead of
using passive scattering techniques; this is designated as
uniform field scanning. This scanning technique is often
called wobbling; however, this name was originally applied to a different technique developed at the Lawrence Berkeley Laboratory7 and it is misleading to apply the term here.
In uniform field scanning, collimators and range compensators are still required to shape the treatment beam laterally
and distally to the target volume. The primary advantages of
uniform field scanning are that large field sizes up to about
40 cm 40 cm or so are feasible and, because of the absence of a double scattering system, protons are used more
efficiently and fewer secondary neutrons are produced. An
additional advantage is that the deliverable proton range is
greater because there is no range loss in the scattering system: the additional range is typically 1 3 gm/ cm2 depending on the beam energy and field size. However, the disadvantages of uniform field scanning are that the resultant dose
distributions in patients are essentially the same as those
achieved with passive scattering techniques and a higher sensitivity is introduced to target motion errors than is present in
passive scattering techniques.
Additional techniques used in scanned beam treatment delivery are those that use the results of optimization techniques in treatment planning to achieve a desired dose distribution usually uniform in the target volume. Objective
functions can be used to optimize spot patterns for single
treatment fields. This technique would appropriately be
called single optimized uniform field but it has become
customary to define this as single field uniform dose, or
SFUD, and this designation will be used throughout. An
SFUD treatment plan consists of one or more individually
optimized fields.
Another technique for delivering scanned beam proton
therapy is IMPT, which is analogous to intensity modulated
Target
Volume
photon treatments that have been called IMRT but, which we

will call IMXT. IMPT is defined as the simultaneous optimization of all Bragg peaks from all fields with or without
additional dose constraints to organs at risk OAR, such
that, when all fields are delivered to the patient, their combination results in a desired dose distribution in the target
volume and OARs. Figure 4 shows an IMPT treatment plan
having four nonuniform fields optimized to deliver a uniform
target dose when all fields are delivered. The desired dose
distribution in the target volume may be uniform or nonuniform, depending upon the treatment goalsfor example, the
treatment goal may be to deliver a simultaneous boost dose
in which case the resulting dose distribution would be nonuniform. Even when the treatment goal is to deliver a uniform dose distribution, there are usually low-level hot and/or
cold spots in the dose distribution. It should be noted that
IMPT is characterized by variations in intensity and energy
Intensity Modulated Proton Therapy (IMPT)

Composite dose from all fields
FIG. 4. Illustration of intensity modulated proton therapy. Four nonuniform

fields have been optimized to deliver a uniform dose to the target volume.
The dose scale applies only to the central figure showing the composite dose
courtesy of A. Trofimov and A. Chan, Massachusetts General Hospital,
Department of Radiation Oncology.
559
rather than variations in intensity alone as is the case for

IMXT. IMPT permits realization of the full potential of particle beams to provide highly localized dose distributions for
cancer treatment; however, as stated elsewhere in this paper,
more research and development is required in order to reduce
treatment errors in IMPT, especially those related to uncertainties in treatment planning calculations of proton ranges in
tissues and errors arising from target motion during treatment
delivery.
Spot scanning is usually achieved by one of two methods:
discrete spot scanning or dynamic raster spot scanning,
which are described later in this paper.
To be clear, passive scattering, uniform field scanning,
and SFUD techniques, even though they may have components of energy and/or intensity modulation, do not comply
with the definition for IMPT used in this paper.
III. REASONS FOR INTEREST IN PROTON

THERAPY
Robert Wilsons observations on the potential advantages
of proton therapy for cancer treatment created interest among
clinical scientists and led to the hypothesis that highly localized dose distributions achieved by proton beams would result in higher probabilities for local control and disease-free
survival, and lower probabilities for normal tissue damage.
This hypothesis remains the motivating force for clinical
studies in proton therapy. It is important to note that the
advantages of a proton dose distribution hold true for every
beam used in a treatment plan and, for treatments of equal
complexity in terms of such factors as technique, number of
beams, and beam directions, etc., protons will almost always
have superior dose distributions to photons. An exception is
when skin sparing is clinically important as photons offer
more skin-sparing benefits than protons. It is also important
to note that, in some clinical situations, the advantage of
protons may be reduced by errors caused by proton range
uncertainties in the treatment plan or by target motion during
treatment delivery.8,9
Several developments occurred in the decade or so before
the early 2000s that generated increased interest in proton
therapy. These events took place in a number of areas including medical physics, clinical results, reimbursement, technology, and changes in the proton therapy equipment marketplace.
III.A. Medical physics developments
Between the mid-1980s and mid-1990s, medical physicists developed methods to fully utilize medical imaging and
to calculate three-dimensional 3D treatment plans for external beams of photons, electrons, protons, and light ions. In
addition, inverse planning and optimization techniques were
developed that made it possible to calculate treatment plans
for intensity modulated radiation therapy. These developments led to a large number of treatment planning comparisons including those for:
559
Conformal photons vs. conformal, passively scattered

protons,
IMXT vs. conformal, passively scattered protons, and
IMXT vs. IMPT.
Many treatment planning studies have been published1014
all of which had the general conclusion that, when photon
and proton treatment plans with comparable complexity are
compared, protons provided superior dose distributions.
These results were obtained from the analysis of dose volume histograms DVHs for both target volumes and normal
tissues. In particular, the integral dose in the proton treatment
plans was two to three times less than the integral dose in the
photon plans.14 These results obtained from treatment planning comparisons provided strong evidence of the potential
for proton beams to deliver dose distributions that could improve clinical results if those dose distributions could be realized in actual cancer treatments.
III.B. Clinical results
Several published reports have confirmed that the use of

proton beams results in improvements in both local control
and normal tissue complications over those achieved with
photon beams; however, these comparisons were predominantly made against historical photon controls.1517 It should
be noted that many of these results were obtained from proton treatments carried out in research-based facilities with
horizontal, passively scattered beams, limited energies, and
restricted flexibility in the complexity and types of treatments that could be carried out. Further, the number of patients treated in a given disease site was relatively small and,
thus, the analysis of these studies often did not produce statistically significant results. It should also be noted that during the mid to late 1990s, no patients had been treated with
IMPT techniques; therefore, the best achievable dose localization for proton beams had not been realized.
Based on these preliminary results, it was reasonable to
conclude that, should patients be treated with proton beams
in modern, hospital-based facilities with improved treatment
technology and improved treatment techniques, additional
clinical gains could be achieved.
III.C. Reimbursement
Until the late 1990s, proton treatment delivery had not

been assigned procedure codes by the American Medical Association. Billing was carried out using special procedure
codes, which were evaluated by Medicare and third-party
insurance carriers on a case-by-case basis. Therefore, reimbursement for proton therapy could be, and often was, denied. With no assurance of reimbursement, there was no financial incentive in the private sector for building new
facilities. Both LLUMC and Massachusetts General Hospital
MGH proton treatment facilities were financed with institutional and government funding.
In the late 1990s, MGH and LLUMC successfully applied
to the American Medical Association AMA for protonspecific treatment delivery procedure codes. Medicare and
560
private insurance carriers then assigned payment rates to the

new procedure codes. Moreover, when Medicare set rates for
proton treatment delivery, they stated that proton treatments
were not considered to be investigational. The achievement
of reimbursement provided financial incentives for the private sector to finance, build, and operate proton therapy facilities, with the expectation that they could realize a reasonable return on investments. Current reimbursement rates for
proton therapy delivery remain attractive; however, one
should not expect that these rates will persist in the long
term.
III.D. Advances in technology
Research laboratories where proton therapy was being

carried out have continued to improve proton therapy techniques and develop new technologies for the acceleration,
transport, and delivery of proton therapy. In the private sector, manufacturers have developed isocentric gantries, compact cyclotrons, medical synchrotrons, and advanced treatment delivery systems.18 Superconducting technology has
been used to develop smaller and higher energy cyclotrons19
and the efficiency of cyclotrons in terms of beam extraction
and power consumption has been substantially improved.
Robotics have been introduced into proton therapy systems
resulting in improved efficiency in the management of patient transport, patient treatment positioning, treatment appliances, and imaging systems.20,21 In addition, proton capabilities have been developed for commercial treatment planning
and data management systems.
Vendors have made major improvements in treatment delivery techniques during the past decade. New technologies
include: 1 beam scanning techniques; 2 universal
nozzles that have the ability to deliver both passive scattering
and spot-scanning treatments; 3 RMWs installed permanently in the nozzle that produce variable SOBPs by a combination of beam gating and beam intensity variation;
4 advanced treatment control and safety systems; 5 improved imaging systems; and 6 multileaf collimators.
A noteworthy technological achievement was made at the
Paul Sherrer Institute in Switzerland where techniques for
proton pencil beam scanning were developed and implemented. This technology was first used for SFUD treatments
and then extended to IMPT treatments.22,23 Several vendors
now offer advanced proton therapy systems that have U.S.
Food and Drug Administration FDA 510k clearance.
These achievements have matured proton therapy into fully
integrated, modern radiation therapy systems that provide
advanced cancer treatments in state-of-the-art hospital settings.
III.E. Increased vendor options
Until 1990, all proton therapy took place in researchbased facilities with equipment developed by the clinical researchers and engineers who worked in those facilities. The
lack of a sizable market was a disincentive for manufacturers
to spend large sums of money to develop proton therapy
equipment. The confluence of factors during the 1990s and
560
early 2000s discussed above created a more favorable climate for vendors and there are currently several companies
marketing proton therapy systems and offering systems that
provide both protons and carbon ions for cancer therapy.
New proton/carbon systems based on superconducting technology have been developed that will make it possible to
build more compact isocentric gantries and accelerators for
hospital-based systems. In addition, companies have been
formed to develop and operate proton therapy facilities and
others offer various levels of consulting e.g., feasibility
studies, marketing studies, business plans, technology evaluation, staffing and operating plans, etc.. The success of new
hospital-based proton therapy projects like those at the University of Florida/Shands Cancer Center in Jacksonville,
Florida and The University of Texas M. D. Anderson Cancer
Center in Houston, Texas has demonstrated that these complex, expensive facilities can be built on time and on budget,
and can rapidly reach targeted patient treatment capacity in
order to satisfy ambitious clinical programs and business
plans.
III.F. Summary
All the factors discussed above have contributed to

greatly increased interest in proton therapy in the United
States and worldwide. In fact, there has been such an explosion of interest in proton therapy during the past 3 years that
one could argue that the current enthusiasm is overheated.
There is an indication that some institutions seeking proton
therapy facilities are doing so without critical evaluation of
the clinical role of proton therapy and the relatively large
financial, clinical, and personnel commitments required for
developing and operating these facilities. Moreover, there is
an acute shortage of trained proton therapy staff and few
training programs. Therefore, institutions that are building
new proton therapy facilities may have a difficult time recruiting trained, proficient staff for clinical operations.
IV. FUTURE PROSPECTS
Based on the current trends, the events that may occur in
particle therapy during the next decade are listed below:
IV.A. Clinical studies
MGH and LLUMC have carried out dose escalation studies for prostate cancer.24 Recently, the National Cancer Institute approved funding for a P01 grant application submitted
by MGH and M. D. Anderson Cancer Center MCACC to
conduct clinical studies in proton therapy with the objective
of applying advanced proton radiation planning and delivery
techniques to improve outcomes for patients with nonsmallcell lung cancer proton dose escalation and proton vs. photon randomized trials, liver tumors, pediatric medulloblastoma and rhabdomyosarcoma, spine/skull base sarcomas,
and paranasal sinus malignancies. It is expected that the University of Florida and University of Pennsylvania facilities
will also be participating in multi-institutional clinical
studies.
561
There is an ongoing discussion in the radiation oncology

community regarding the appropriateness of prospective,
randomized clinical trails comparing protons with photons.
The basic argument is that the dose distributions of protons
will always be superior to those for photons when both modalities are optimized and one cannot justify randomized
clinical trials of protons vs. photons because there would not
be equipoise between the two arms of the trial.25,26 Whether
or not one supports the concept of randomized clinical trials
for protons, it is difficult to argue against the need for additional clinical data; it is not feasible to draw statistically
meaningful comparisons between photon and proton treatment techniques without data acquired through carefully
controlled clinical studies in an expanded range of disease
sites.
Research programs will also be directed toward the reduction of treatment errors in proton therapy treatment planning
and treatment delivery, especially those related to uncertainties in the calculation of proton ranges in inhomogeneous
tissues and errors arising from target motion during treatment
delivery.
IV.B. Lower costs for proton therapy
Currently, proton therapy is more expensive than photon

therapy. In 2003, Goitein and Jermann estimated the relative
costs of proton and photon therapy, concluding that, with
some foreseeable improvements, the ratio of costs protons/
photons was likely to be about 1.7.27 However, these estimates are probably outdated. Reimbursement rates currently
make it possible to develop and operate proton therapy facilities with a reasonable profit margin. In the future, it is
likely that reimbursement rates for proton therapy treatment
delivery will decrease as capital costs of current facilities
will be spread among more patients as these facilities reach
full-capacity operations.
In addition, the overall costs of proton therapy will be
reduced as proton therapy procedures become more efficient
and patient throughput increases. Increased patient throughput will make it possible to spread the capital costs of proton
therapy among a larger number of patients and the cost per
patient will decrease. Improved efficiencies can be realized
in the following ways:
Using treatment setup rooms outside the treatment room
will increase patient throughput, especially for pediatric
patients who require anesthesia, which currently requires these patients to spend more time in the treatment room than patients who do not require anesthesia;
Using faster, automated imaging techniques for patient
positioning both outside and inside the treatment room;
Using robotics both outside and inside the treatment
room for transferring and positioning patients, for moving imaging devices, and for handling treatment appliances;
Using improved accelerator, beam transport, and treatment delivery technologies, which will reduce the time
currently required to change energies and to switch the
beam from one room to another;
561
Using IMPT techniques, which have the potential to

decrease treatment times as fewer patient appliances
field-shaping collimators and range compensators will
need to be manually inserted and removed for each
treatment field. It will also be possible after the treatment setup procedure for the first treatment field to treat
all subsequent fields in a treatment session containing
coplanar fields without re-entering the treatment room;
Potentially using larger and fewer fractions hypofractionation in proton therapy than are currently used in
photon therapy. Proton therapy spares normal tissues to
a much greater extent than is possible with photon
beams, making hypofractionation potentially effective.
With reduced numbers of treatment fractions, the cost to
patients for a course of therapy will decrease; and
As the demand for proton therapy increases, more facilities will be built and more equipment manufacturers
will enter the market. Competition between manufacturers should drive down the cost of particle therapy systems. Further, rapid growth of proton therapy facilities,
spurred by strong market acceptance, will allow manufacturers to sell more systems i.e., high business startup and technology development costs for proton
therapy systems can be spread among larger numbers of
systems.
IV.C. Advances in technology
New developments, many already underway, will have an

impact on proton therapy.
IV.C.1. Single-room proton therapy systems
There is an increasing interest in smaller proton therapy
facilities that are better suited for community hospitals and
large private practices. Several vendors are developing the
technologies necessary for single-room solutions for proton therapy systems. Single-room proton therapy systems
have been stated to have the following advantages:
They provide a lower-cost option for implementing proton therapy.
All or most components accelerator, beam transport,
energy selection, and beam delivery can be mounted
on a rotating gantry or very near the gantry, thus, reducing the size of the complete proton therapy system
enough that it can be installed in its entirety or nearly
so in a single treatment room.
There is no competition with other treatment rooms for
proton beams because each room has its own accelerator.
If multiple rooms are used, the entire proton treatment
capability is not lost if the accelerator goes down, as is
the case when one large accelerator generates beams for
several treatment rooms.
Single-room systems may have the following disadvantages, however:
The costs of these systems may not necessarily be
cheaper per treatment room as compared to large systems using single accelerators to feed several treatment
rooms.
Additional rooms may have to be installed if demand
for proton therapy increases in a location with a singleroom system. If several treatment rooms are eventually
needed, the cost effectiveness of the single-room concept may be lost; for example, an additional accelerator
will be needed for each treatment room. Maintaining
multiple accelerators will also increase maintenance
costs.
Because of the pulse structure of some accelerators
e.g., synchrocyclotrons chosen for the proposed systems, it may be difficult to use spot scanning techniques
to deliver IMPT treatments.28 IMPT could be delivered
using a multileaf collimator; however, this method may
have problems such as poor efficiency in the use of
protons and increased neutron production.
These systems are new, and, thus, their operability, reliability, and maintainability are not known.
IV.C.2. Accelerator development
Advances in accelerator technology are also expected.
Several developments already underway show considerable
potential to provide a new generation of more compact, more
efficient, and less expensive accelerators for proton therapy.
Some developments will improve upon current technologies,
while others will provide entirely new types of accelerators,
some of which are highlighted here. Flanz29 has provided an
excellent overview of new accelerator technologies.
IV.C.2.a. Superconducting cyclotrons and synchrocyclotrons. Cyclotron and synchrocyclotron development has resulted in more compact systems with increased energy. These
gains have been achieved by the use of new accelerator designs and the application of superconducting technology.
Figure 5 illustrates how the weight of cyclotrons/
synchrocyclotrons has deceased since 1995. Notably, the accelerator energy has increased while the weight/size of the
accelerators has decreased. At the current reduced weights of
approximately 20 tons, accelerators can be installed on isocentric gantries and rotated around patients, eliminating large
and expensive beam transport systems.19,30 There is at least
one vendor marketing a one-room proton therapy system
based on this concept. Such systems will be particularly attractive to small community hospitals and large radiation oncology private practices.
IV.C.2.b. Fixed field alternating gradient (FFAG) accel
erators. To date, most particle therapy facilities have used
isochronous cyclotrons, and synchrotrons for particle acceleration. However, FFAG accelerators may be able to replace
both cyclotrons and synchrotrons for hadron therapy. FFAG
accelerators combine many of the positive features of cyclotrons and synchrotrons with fixed magnetic fields as in cyclotrons and pulsed acceleration as in synchrotrons. FFAG
accelerators have the following potential advantages:
FFAG accelerators can be cycled faster than synchrotrons, limited only by the rate of the radiofrequency
562
250
IBA 230
200
150
Tons
562
100
ACCEL K250
50
New
0
1995
IBA 230
1999
2003
ACCEL
K250
2006
New
2010
Date
FIG. 5. This graph indicates how the weight and, therefore, the size of
clinical cyclotrons has decreased since 1995. The IBA 230 room temperature cyclotron was the first cyclotron to be installed in a hospital-based
proton therapy facility. More recently, the ACCEL K250 now the Varian/
ACCEL K250 has been installed in two European facilities. New superconducting cyclotrons will weigh about 10% of the IBA 230.
RF modulation. Higher duty factors will permit higher

average beam currents and high repetition rates for spot
scanning. Fixed fields require simpler and cheaper
power supplies and are more easily operated than synchrotrons.
With respect to fixed field cyclotrons, nonscaling FFAG
accelerators allow strong focusing and, hence, smaller
aperture requirements, which leads to low beam losses
and better control over the beam.
FFAG accelerators like synchrotrons have a magnetic
ring that allows beam extraction at variable energies
rather than just a single energy, as in a cyclotron.
Their superconducting magnets and compact size make
FFAG accelerators attractive for proton therapy applications.
Because of the possibility of changing energy and location with each spot and having a repetition rate of about
100 Hz, spot scanning with FFAG proton beams can be
carefully controlled in three dimensions.
The FFAG concept was developed in the 1950s in Japan,

Russia, and the United States.31 The first electron FFAG accelerator as developed in the late 1950s and several others
were built in the early 1960s. In the 1980s, proposals for
FFAG-based neutron spallation sources were unsuccessful
due to the perceived complexity of the magnets. In scalingtype FFAG accelerators, the magnetic field has to be nonlinear with zero chromatic beam optics. Because the magnets
are complex and expensive to manufacture, the resulting cost
has limited their use in medicine. However, new magnet designs have been made possible by 3D magnetic field simulation codes and large-scale computers. In addition, for a proton FFAG, a broad band and high gradient RF cavity is
required. A new type of RF cavity developed at the High
563
Energy Accelerator Research Organization KEK in Japan32

made it possible to overcome these problems, and in 1999,
the first proton FFAG, was built at KEK in Japan.33 This
machine was the proof of principle POP for proton FFAG
accelerators and was named POP-FFAG. After the success of
POP-FFAG, a 150 MeV FFAG accelerator was developed at
KEK. To date, several FFAG accelerators have been built in
the energy range 150 250 MeV and others have been
proposed.
The nonscaling FFAG accelerator was invented in 1999.34
This accelerators magnet design provides a variation of orbit
length with energy, which can be arranged to greatly compress the range of the orbit radii and, thus, the magnet aperture, while maintaining linear magnetic field dependence. In
addition, the small apertures and linear fields allow simplification and cost reduction compared with scaling FFAG accelerators. Keil et al. have proposed a hadron cancer therapy
system using a nonscaling FFAG accelerator and gantry
composed of nonscaling FFAG cells.35 This accelerator will
accelerate carbon ions up to 400 MeV/ and protons up to
250 MeV.
IV.C.2.c. Laser acceleration. Proton laser acceleration is
achieved by focusing a high-power approximately 1015 W
laser on a thin target such as a 5 m-thick titanium foil, or
multilayer targets, such as 100 nm-thick aluminum hydrogen. To achieve proton energies of 200 250 MeV with adequate intensities, focused short laser pulses to intensities of
1022 W / cm2 or higher are required.36 The short
30 80 1015 sec laser pulse width produces a high
peak power intensity that causes massive ionization in the
target, expelling a large number of relativistic electrons. The
sudden loss of electrons gives the target a high positive
charge and this transient positive field accelerates protons to
high energies. The resultant proton beams have a broad energy spectrum; therefore, a magnetic spectrometer must be
used to select a narrow energy band for patient treatment.
This selection process throws away about 99.5% of the
beam, and, thus, the energy selection system must be heavily
shielded against neutrons resulting from proton losses in the
spectrometer.
Achieving the energy and beam intensity required for proton therapy is a major challenge. Groups in several countries
are working on proton laser acceleration;3638 however, because of the technical difficulties, it is not expected that the
current development efforts will be able to achieve the required beams even for treatment of ocular melanomas
70 MeV at a dose rate of approximately 10 12 Gy/ min
over the next 5 7 years.
IV.C.2.d. Dielectric wall accelerators. Conventional accelerator cavities have an accelerating field only in their
gaps, which occupy only a small fraction of the cavity
length, and have an accelerating gradient of approximately
1 2 MeV/ m. In contrast, dielectric wall accelerators
DWAs have the potential of producing gradients of approximately 100 MeV/ m. In a DWA, the beam pipe is replaced by an insulating wall so that protons can be accelerated uniformly over the entire length of the accelerator,
563
yielding a much higher accelerating gradient. The DWA uses

fast-switched high-voltage transmission lines to generate
pulsed electric fields on the inside of a high gradient insulating HGI acceleration tube. High electric field gradients are
achieved by the use of alternating insulators and conductors
and short pulse times. An electric field propagates down the
bore of the accelerator, pushing the proton packet in front
of it. The system will produce individual pulses that can be
varied in intensity, energy, and spot width and, therefore,
should be suitable for IMPT applications.
The enabling technologies for DWA are the high gradient
insulator, fast SiC switching, and new dielectric materials.
The DWA technology is being developed at the Lawrence
Livermore Laboratory,39 with a full scale prototype proton
therapy system expected to be operational in approximately
4 or 5 years. The private sector partner for this development
is TomoTherapy40 and the first prototype is expected to be
installed at the University of California Davis Cancer Center.
IV.C.3. Pencil beam spot scanning techniques
Most patients who have been treated with proton beams
have been treated with passive scattering techniques. Modern
passive scattering systems are much more efficient and versatile than the systems used before hospital-based facilities
were developed, and additional advances have continued to
the present time. Recently, there has been an effort to develop multileaf collimators MLCs for passive scattering applications; however, to date, proton MLCs have been less
than optimal e.g., they have small field sizes, and, for some,
their size does not permit positioning the MLC close to the
patient surface, resulting in large air gaps that degrade the
lateral penumbra of the treatment beam.
To date, only a handful of facilities are using beam scanning techniques for particle therapy. During the next decade,
the development and refinement of beam scanning systems
will continue. Importantly, pencil beam scanning makes it
possible to deliver IMPT treatments, which provide a substantial dosimetric improvement over passive-scattering
techniques. Analogous to IMXT, IMPT enables further optimization of dose distributions for proton therapy.
Scientists at the PSI in Switzerland developed a spotscanning system where an energy selection system, one set
of scanning magnets for 1D scanning, and couch shifts are
used together to deliver pencil beam Bragg peaks to a target
volume.22 The magnet and energy selection system are used
to scan a section of the target volume, the couch is shifted,
then another section is scanned. In this way, the target volume can be scanned in three dimensions. This scanning system has been used to deliver both SFUD and IMPT treatments.
More recent scanning systems use two sets of scanning
magnets to scan the beam in a plane perpendicular to the
nozzle axis for 2D scanning. The target volume is divided
into energy layers highest to lowest energy, and each layer
is sequentially scanned with pencil beams.
The treatment is carried out by moving the beam spot
throughout each of several energy layers in a target volume.
564
The spot can be moved dynamically i.e., continuously in a

raster or line pattern or by a discrete spot scanning method,
where the beam is cut off between spots and the dose at each
spot is varied to achieve the prescribed dose pattern. In dynamic raster scanning, either the scan speed or beam intensity, or both, are varied to produce the intensity distribution
prescribed by the treatment plan.
A fundamental challenge for proton therapy using scanning techniques will be to develop methods to reduce or
eliminate treatment uncertainties caused by range errors or
target motions. Errors due to target motion during the delivery of scanned treatment fields are a problem for several
treatment situations, particularly, for the treatment of lung
tumors. Currently there are two primary ways to reduce motion errors: 1 beam gating, where the beam is turned off
when the target has moved out of the target position for
which the treatment plan was calculated;41 or 2 repainting
spots, repainting layers, or repainting the entire treatment
volume in this context, repainting means to scan a particular
spot position, layer, or volume two or more times. Errors in
target motion can be reduced in proportion to the number of
repaintings. In addition, repainting of spots and layers can be
done in a straightforward manner using treatment planning
and beam delivery methods that have already been developed. However, rapid repainting of the entire volume will
require a considerable amount of new technology to be developed for beam delivery, primarily in the area of very rapid
scanning and energy changes. When combined with repainting, beam gating increases the complexity of treatment delivery. Furukawa et al. have proposed a design for a heavyion radiotherapy system that addresses this problem.42 Bert
et al. have proposed a system in which pencil beam positions, as well as the beam energy, are changed according to
the tracked target motion to reduce motion errors.43 A considerable amount of development effort is currently being
directed toward decreasing the errors due to range uncertainties and target motion errors when spot scanning techniques
are used and one can expect that good solutions will be
found to reduce these errors during the next decade.
IV.C.4. Treatment planning calculations
IV.C.4.a. Monte Carlo treatment planning. Conventional
proton treatment planning using pencil beam algorithms has
been the mainstay of treatment planning using passivescattering techniques. However, calculations by pencil beam
algorithms can have appreciable errors in some situations,
notably when there are tissue interfaces or inhomogeneities
in the beam path or when metallic implants are present adjacent to or in the target volume. Monte Carlo MC calculations have the potential to increase the accuracy of dose
calculations in those instances where pencil beam algorithms
have difficulty. MC treatment planning has been developed
for proton therapy applications.4446 However, MC calculations are time intensive and are not used routinely, although
they can be useful in benchmarking pencil beam calculations. Within the next few years, MC calculations may become faster and more efficient and, therefore, can be used
more routinely in treatment planning. MC calculations have
564
also been quite useful in calculating data used in commissioning proton treatment planning systems47,48 MC calculations can decrease the time required to commission a treatment planning system by eliminating a large number of
measurements, especially in the case of spot beam scanning
where approximately 100 energies need to be commissioned.
It is reasonable to expect that, with future developments,
particularly in the area of increasing the speed of MC calculations, MC will play an increasing role in proton treatment
planning.
IV.C.4.b. Optimization for IMPT treatment planning. Protons are much more sensitive to tissue inhomogeneities in the
beam path than photons and current optimization algorithms
do not accurately account for the errors resulting from range
uncertainties and changes in scattering conditions caused by
such inhomogeneities. For passive-scattering techniques,
such uncertainties are partially offset by expanding the margins around the clinical target volume or by using internal
target volumes determined from 4D CT treatment planning
scans. In IMPT, the dose distribution is sensitive to the uncertainties in each individual beam of the potentially several
thousand Bragg peaks used in the treatment plan, and uncertainties in the spot placements can result in unacceptable hot
and cold regions in the target volume. One approach to improving IMPT treatment planning is to include errors in the
optimization calculations in such a way as to ensure that the
treatment plan predicts actual dose distributions more reliably. Work in this area is underway and these approaches
have the potential to improve the accuracy and robustness of
IMPT treatment plans.49
IV.C.5. Positron emission tomography PET
for treatment verification
Ideally, one would have a method available to measure
proton dose distributions delivered to patients either online
during treatment or offline immediately or soon after treatment. Such treatment verification would make it possible to
correct the treatment plan after the first treatment session or
to make corrections during the course of treatment. Such
adjustments may be necessitated by changes in the patients
weight or changes in the tumors size, shape, or position
during treatment. Positron emission tomography PET is a
potentially useful tool for validating the fidelity of the treatment plan delivery.
PET can verify treatment delivery by detecting annihilation -rays arising from the decay of small amounts of +
emitters such as 11C, 15O, and 10C produced by nuclear fragmentation reactions between the primary charged particle
and the target nuclei in the irradiated tissue. Particle treatment delivery verification can be achieved by comparing the
measured + activity distribution with an expected pattern
calculated by treatment planning or by MC calculations of
the expected + activity. In particular, PET techniques offer
the potential to detect and quantify range uncertainties in
treatment planning calculations.
Physicists at MGH and elsewhere5053 have been exploring the use of PET imaging, both offline and online, for
565
treatment verification. The preliminary results look promising and it is expected that with further development, this
technique will become an important tool for verifying particle treatment plan delivery.
IV.C.6. Proton CT
The idea of using proton beams for imaging has been
around for some time. In his seminal papers on the application of photon absorption in radiology, Cormack pointed out
the possibility of using heavy charged particles in CT
applications.54,55 In 1968, Koehler demonstrated that
160 MeV protons could produce images on radiographic
films that had much better contrast than those produced with
photons.56 Cormack and Koehler published a study on proton
CT in 1976.57 In the late 1970s and early 1980s, Hanson et
al. published a series of papers on proton CT including one
of the earliest studies of human specimens.58 In addition,
Schneider et al. conducted studies on the use of proton radiography as a quality assurance tool for proton treatment
planning.59,60 These studies, including measurements taken
on an animal patient, concluded that inaccuracies in treatment planning as a result of incomplete modeling of multiple
Coulomb scattering effects and errors in proton stopping
powers derived from photon CT scans could result in sizeable deviations in predicted versus measured proton dose
distributions. Zygmanski et al. demonstrated the feasibility
of using proton cone-beam CT to determine proton stopping
powers for materials of various densities.61 In 2004, Schulte
et al. at the LLUMC proton therapy center reported on a
conceptual design for a proton CT system for applications in
proton radiation therapy.62
Currently, most proton treatment planning dose calculations use a pencil beam algorithm that accounts for tissue
inhomogeneities by means of pathlength water equivalent
depth scaling based on relative stopping powers. Hounsfield
numbers obtained from x-ray treatment planning CT scans
are converted to proton stopping powers using measured relationships between Hounsfield numbers and the relative
stopping powers for materials closely matched to human tissues such as soft tissue, fat, lung tissue, soft bone, and hard
bone.63 These conversions are frequently inexact, which can
lead to errors in the calculation of proton ranges in patients
of approximately 3% or more of the proton range. These
errors could be substantially reduced by the direct measurement of relative proton stopping powers using proton CT
techniques.
There are, however, also errors in proton CT data. For
instance, limitations in the spatial and density resolution of
proton CT arise from various sources: multiple Coulomb
scattering in the patient leads to spatial uncertainties in the
CT image and energy loss straggling due to momentum
spread of the beam entering the patient and proton range
straggling in the patient can cause errors in density resolution. In addition, secondary protons arising from nonelastic
nuclear interactions can lead to noise in the proton CT images. Recent studies have demonstrated that density resolutions of 1%2% are possible with proton CT techniques,
565
which can be achieved with proton doses of 10 mGy or less,

comparable to or better than doses stated for cone beam CT
with kilovoltage x-ray beams.64
With the widespread use of isocentric gantries in proton
treatment facilities, further development and study of proton
CT and cone-beam CT are anticipated. Development of new
high-bandwidth silicon strip detector systems will also make
proton-by-proton data collection feasible. Individual proton
tracking techniques will make it possible to measure the position of each particles track and each particles energy
deposition,65,66 which will greatly increase the spatial and
energy resolution of proton CT measurements.
IV.C.7. QA systems for scanned proton beams
Implementing scanned proton beams for cancer treatment
has generated the need for a new class of quality assurance
QA systems that permit rapid and accurate 2D and 3D data
acquisition. QA systems for scanned proton beams must also
be adaptable to the timing structure and energy stacking of
pencil beam scanning beam treatment delivery. In contrast to
dosimetric measurements for passive scattering methods,
when pencil beam scanning techniques are employed, the
dose measured at a point in the phantom can be obtained
only by collecting the detector signal during the entire scan
sequence. In many cases, it is also necessary to characterize
a single pencil beam, which places additional requirements
on the measuring system.
In 2000, Jkel et al. described QA procedures and methods for a treatment planning system used for scanned ion
beam therapy.67 In 2005, Pedroni et al. reported on aspects of
dosimetric methods used to characterize the scanned proton
beams at PSI in Switzerland68 including ionization chambers,
films, and a charged-coupled-device CCD camera coupled
with a scintillating screen. In addition, Pedroni et al. have
used a linear array of 26 ionization chambers and the CCD
system to verify scanned dose distributions. In 1999, Karger
et al. described a 24 ionization chamber planar array used for
the QA of heavy ion beams.69 More recently, Nichiporov and
Solberg reported on two ionization chamber array systems: a
128-chamber planar array designed to measure beam profiles
in fields up to 38 cm in diam and an array of 122 smallvolume multilayer ionization chambers used for depth-dose
measurements in clinical proton beams.70 Users of these ionization chamber array systems have agreed that these systems provide considerable time savings in dosimetry measurements. Unfortunately, the ionization chamber arrays
described above were developed in house at research facilities and are not commercially available. As more proton
treatment facilities implement scanning techniques and the
demand for appropriate dosimetry systems increases, vendors may be encouraged to develop similar commercial devices.
A dosimetry system utilizing a CCD camera plus scintillating screen71 has been shown to be useful for measuring
proton fluences in air, and by placing absorbers in front of
the screen, relative dose delivered at depth. Pedroni et al.
stated that the CCD system works well for conducting rela-
566
tive dosimetry measurements to compare treatment planning

calculations with the dose distributions delivered by the proton scanning system.68 A similar CCD system has also been
used at the M. D. Anderson Proton Therapy Center to measure fluence patterns in air at the isocenter for single spots,
uniform spot patterns, and for spot patterns calculated for a
SFUD prostate treatment delivery. However, CCD camerabased proton dosimetry systems are not readily available, but
they can be made in house without undue difficulty.
Other dosimetric systems such as radiochromic films,72,73
alanine detectors,73 and polymer gels74 can be used for QA
measurements. However, the use of these systems is limited
because of saturation effects in the proton stopping region
and because they are not real-time instruments. These systems continue to be investigated, and it is expected that they
will, in time, become more widely used.
IV.C.8. Carbon ions

There is a growing interest in using carbon ions for cancer
therapy. The rationale for using carbon ions is that in addition to their highly localized physical dose distributions,
which are quite similar to those for proton beams, carbon ion
beams are high linear energy transfer LET in their interactions in tissues. High LET provides increased relative biological effectiveness RBE and a low oxygen enhancement
ratio OER; therefore, carbon ions may be more effective
than protons in treatment situations where cancer cells are
anoxic or hypoxic.7577 Carbon ion facilities in Japan
HIMAC-Chiba and HIMC-Hyogo and Germany GSI,
Darmstadt have treated approximately 5,000 patients to
date5 and a hospital-based facility is being constructed at the
DKFZ/German Cancer Research Center/University of
Heidelberg, which will commence patient treatments in
2009. New facilities are being built in Japan, Italy, and Germany and are in an advanced stage of planning in Austria
and France. Most of these carbon ion facilities will also have
the ability to treat patients with proton beams. There are at
least three institutions in the United States that are contemplating proton/carbon ion facilitiesat this time they are all
planning to start with protons only and, at a later date, add
carbon ion capabilities, but there are several impediments to
the development of carbon ion facilities in the United States:
1 carbon ion systems do not have FDA clearance; 2 there
are no procedure codes or reimbursement rates for carbon
ion treatments; and 3 carbon ion facilities are much more
expensive than proton facilities. The current financial climate, high costs of construction, and unfavorable foreign
exchange rates are also strong financial disincentives for
building carbon ion facilities.
In addition, the clinical value of carbon ions has not been
sufficiently demonstrated; however, clinical results are
promising.15 It is certain that investigations of both protons
and carbon ions, and clinical comparisons between the two
particles will be carried out during the next decade as more
facilities are opened that have the capability of treating patients with both types of particles.
566
V. CONCLUSIONS
Proton therapy is in the midst of a remarkable transition.
During the next decade, important advances in proton
therapy treatment delivery, treatment planning, and quality
assurance systems will be made that will improve proton
therapys efficiency, robustness, and accuracy. It is also expected that the cost to patients will decrease, making proton
therapy more financially competitive with photon therapy.
IMPT treatments will become available for routine use in
an increasing number of facilities, and methods will be developed to improve the accuracy and precision of this treatment modality. With the availability of advanced IMPT techniques, it may be possible to quantify the clinical gains
obtained from optimized proton therapy and to compare the
clinical outcomes of photons and protons using the best dose
distributions for both modalities.
It will be necessary to develop methods to make proton
treatment plan calculations more accurate and to verify that
treatment plans are actually realized in the delivery of proton
treatments. It is not meaningful to compare protons with
photons unless we can demonstrate that the superior dose
distributions of proton treatment plans can actually be delivered to patients.78
More hospital-based, state-of-the-art proton therapy facilities will be built and increasing numbers of patients will be
treated. As stated by Schultz-Ertner and Tsujii, The potential of particle therapy can only be exploited if a full integration of particle therapy into clinical environments and interdisciplinary treatment strategies is sought and if new medical
and technologic advances are properly incorporated into the
total treatment process.15
a
Author to whom all correspondence should be addressed. Electronic

mail: [email protected]
E. Rutherford, Collisions of alpha particles with light atoms. III. Nitrogen and oxygen atoms, Philos. Mag. 37, 571580 1919.
2
C. A. Tobias, J. H. Lawrence, J. L. Born, R. K. McCombs, J. E. Roberts,
H. O. Anger, V. V. A. Low-Beer, and C. B. Huggins, Pituitary irradiation
with high-energy proton beams: a preliminary report, Cancer Res. 18,
121134 1958.
3
R. R. Wilson, Radiological use of fast protons, Radiology 47, 487491
1946.
4
J. M. Slater, J. O. Archambeau, D. W. Miller, M. I. Notarus, W. Preston,
and J. D. Slater, The proton treatment center at Loma Linda University
Medical Center, Int. J. Radiat. Oncol., Biol., Phys. 22, 383389 1991.
5
Particles newsletter. Available from: http://ptcog.web.psi.ch/
ptcentres.html
6
B. Gottschalk and E. Pedroni, Treatment delivery systems, Proton and
Charged Particle Radiotherapy, T. F. DeLaney and H. M. Kooy, eds.
Lippincott Williams and Wilkins, Philadelphia, 2008, pp. 3349.
7
W. T. Chu, B. A. Ludewigt, and T. R. Renner, Instrumentation for treatment of cancer using proton and light-ion beams, Rev. Sci. Instrum. 64,
20552122 1993.
8
A. J. Lomax, Intensity modulated proton therapy and its sensitivity to
treatment uncertainties. 1. The potential effects of calculational uncertainties, Phys. Med. Biol. 53, 10271042 2008.
9
A. J. Lomax, Intensity modulated proton therapy and its sensitivity to
treatment uncertainties. 2. The potential effects of inter-fraction and interfield motions, Phys. Med. Biol. 53, 10431056 2008.
10
B. Glimelius, U. Isacsson, E. Blomquist, E. Grusell, B. Jung, and A.
Montelius, Potential gains using high-energy protons for therapy of malignant tumors, Acta Oncol. 38, 137145 1999.
11
B. Glimelius et al., Number of patients potentially eligible for proton
therapy, Acta Oncol. 44, 836849 2005.
1
567
12
R. Flynn, D. Barbee, T. Mackie, and R. Jeraj, Comparison of intensity

modulated x-ray therapy and intensity modulated proton therapy for selective subvolume boosting: A phantom study, Phys. Med. Biol. 52,
60736091 2007.
13
L. Haisen, H. Romeijn, H. Fox, J. Palta, and J. Dempsey, A computational implementation and comparison of several intensity modulated proton therapy treatment planning algorithms, Med. Phys. 35, 11031112
2008.
14
A. J. Lomax et al., A treatment planning inter-comparison of protons and
intensity-modulated photon therapy, Radiother. Oncol. 51, 257271
1999.
15
D. Schultz-Ertner and H. Tsujii, Particle radiation therapy using proton
and heavier ion beams, J. Clin. Oncol. 25, 953964 2007.
16
H. Tsujii, et al. Clinical results of fractionated proton therapy, Int. J.
17
H. Suit et al., Proton beams to replace photon beams in radical dose
treatments, Acta Oncol. 42, 800808 2003.
18
A. R. Smith, Innovations and technical developments for particle therapy
in the United States, Proceedings, The 21st International Symposium:
Modern Radiation Oncology: Innovative Technologies and Translational
Research, 2008.
19
H. Klein, C. Baumgarten, A. Geisler, J. Heese, A. Hobl, D. Krischel, M.
Schillo, S. Schmidt, and J. Timmer, New superconducting cyclotron
driven scanning proton therapy systems, Nucl. Instrum. Methods Phys.
Res. B 241, 721726 2005.
20
A. Mazal et al., Robots in high precision patient positioning for conformal radiotherapy, Proceedings of World Congress on Medical Physics
and Biomedical Engineering, Medical and Biological Engineering and
Computing, 1997.
21
C. E. Allgower, A. N. Schreuder, J. B. Farr, and A. E. Mascia, Experiences with an application of industrial robotics for accurate patient positioning in proton radiotherapy, Int. J. Med. Robotics Comput. Assist.
Surg. 3, 7281 2007.
22
E. Pedroni et al., The 200 MeV proton therapy project at the Paul Scherrer Institute: Conceptual design and practical realization, Med. Phys. 22,
3753 1995.
23
A. Lomax et al., Treatment planning and verification of proton therapy
using spot scanning: Initial experience, Med. Phys. 31, 31503157
2004.
24
A. L. Zeitman et al., Comparison of conventional-dose vs. high-dose
conformal radiation therapy in clinically localized adenocarcinoma of the
prostate: A randomized controlled trial, JAMA, J. Am. Med. Assoc. 294,
12331239 2005.
25
M. Goitein and J. Cox, Should randomized clinical trial be required for
proton radiotherapy? J. Clin. Oncol. 26, 175176 2008.
26
H. Suit et al. Should positive phase III clinical trial data be required
before proton beam therapy is more widely adopted? No, Radiother.
Oncol. 86, 148153 2008.
27
M. Goitein and M. Jermann, The relative costs of proton and x-ray
radiation therapy, Clin. Oncol. 15, S37S50 2003.
28
Y. Jongen, Cyclotrons and synchrocyclotrons, PTCOG Educational
Workshop, Jacksonville, Fl, May 2008, http://ptcog.web.psi.ch/
ptcog47_talks.html accessed November 1, 2008.
29
J. Flanz, Particle therapy: Technical approaches, PTCOG Educational
Workshop, Jacksonville, Fl, May 2008, http://ptcog.web.psi.ch/
ptcog47_talks.html accessed November 1, 2008.
30
R. L. Maughan and W. E. Powers, A superconducting cyclotron for
neutron radiation therapy, Med. Phys. 21, 779785 1994.
31
M. D. Craddock, New concepts in FFAG design for secondary beam
facilities and other applications, Proceedings of EPAC 2005, 2005, pp.
261263.
32
Y. Mori et al., A new type of rf cavity for high intensity proton synchrotron using high permeability magnetic alloy, Proceedings of EPAC 1998,
pp. 299301.
33
M. Aiba et al., Development of a FFAG proton synchrotron, Proceedings of EPAC 2000, pp. 581583.
34
C. Johnstone et al., Fixed field circular accelerator designs, Proceedings of EPAC 1999, pp. 30683070.
35
E. Keil, A. M. Sessler, and D. Trbojevic, Hadron cancer therapy complex using nonscaling fixed field alternating gradient accelerator and gantry design, Phys. Rev. ST Accel. Beams 10, 054701 2007.
36
C.-M. Ma et al., Development of a laser-driven proton accelerator for
cancer therapy, Laser Phys., 16, 18 2006.
567
37
E. Fourkal, B. Shahine, M. Ding, J.-S. Li, T. Tajima, and C.-M. Ma,

Particle in cell simulation of laser-accelerated proton beams for radiation
therapy, Med. Phys. 29, 27882798 2002.
38
S. S. Bulanov et al., Accelerating protons to therapeutic energies with
ultraintense, ultraclean, and ultrashort laser pulses, Med. Phys. 35,
17701776 2008.
39
G. J. Caporaso et al., A compact linac for intensity modulated proton
therapy based on a dielectric wall accelerator, Phys. Medica 24, 98101
2008.
40
T. R. Mackie et al., A proposal for a novel compact intensity modulated
proton therapy system using a dielectric wall accelerator, Med. Phys. 34,
2628 2007.
41
H. M. Lu, R. Brett, G. Shapr, S. Safai, S. Jiang, J. Flanz, and H. Kooy, A
respiratory-gated treatment system for proton therapy, Med. Phys. 34,
32733278 2007.
42
T. Furukawa, T. Inaniwa, S. Sato, T. Tomitani, S. Iinohara, K. Noda, and
T. Kanai, Design study of a raster scanning system for moving target
irradiation in heavy-ion therapy, Med. Phys. 34, 10851097 2007.
43
C. Bert, N. Laito, A. Schmidt, N. Chaudhri, D. Schardt, and E. Rietzel,
Target motion tracking with scanned particle beam, Med. Phys. 34,
47684771 2007.
44
H. Paganetti, H. Jiang, K. Parodi, R. Slopsema, and M. Engelsman,
Clinical implementation of full Monte Carlo dose calculation in proton
beam therapy, Phys. Med. Biol. 53, 48254853 2008.
45
H. Paganetti, H. Jiang, and A. Trofimov, 4D Monte Carlo simulation of
proton beam scanning: Modeling of variations in time and space to study
the interplay between scanning pattern and time-dependent patient geometry, Phys. Med. Biol. 50, 983990 2005.
46
R. D. Ilie, V. Spasic-Jokic, P. Belicev, and M. Dragovic, The Monte
Carlo SRNA-VOX code for 3D proton dose distribution in voxelized
geometry using CT data, Phys. Med. Biol. 50, 10111017 2005.
47
W. Newhauser et al., Monte Carlo simulations of a nozzle for the treatment of ocular tumors with high-energy proton beams, Phys. Med. Biol.
50, 52295249 2005.
48
W. Newhauser et al., Monte Carlo simulations for configuring and testing an analytical proton dose-calculation algorithm, Phys. Med. Biol. 52,
45694584 2007.
49
D. Pflugfelder, J. J. Wilkens, and U. Oelfke, Worst case optimization: A
method to account for uncertainties in the optimization of intensity modulated proton therapy, Phys. Med. Biol. 53, 16891700 2008.
50
K. Parodi et al., Patient study of in vivo verification of beam delivery
and range, using positron emission tomography and computed tomography imaging after proton therapy, Int. J. Radiat. Oncol., Biol., Phys. 68,
920934 2007.
51
K. Parodi, T. Bortfeld, and T. Haberer, Comparison between in-beam
and offline positron emission tomography imaging of proton and carbon
ion therapeutic irradiation at synchrotron- and cyclotron-based facilities,
52
T. Nishio and T. Ogino, Dose-volume delivery guided proton therapy
using beam on-line PET system, Med. Phys. 33, 41904197 2006.
53
A. Knopf, K. Parodi, H. Paganetti, E. Cascio, A. Bonab, and T. Bortfeld,
Quantitative assessment of the physical potential of proton beam range
verification with PET/CT, Phys. Med. Biol. 53, 41374151 2008.
54
A. M. Cormack, Representation of a function by its line integrals with
some radiological applications, J. Appl. Phys. 34, 27222727 1963.
55
A. M. Cormack, Representation of a function by its line integrals with
some radiological applications II, J. Appl. Phys. 35, 29082913 1964.
56
A. M. Koehler, Proton radiography, Science 160, 303304 1968.
57
A. M. Cormack and A. M. Koehler, Quantitative proton tomography:
Preliminary experiments, Phys. Med. Biol. 21, 560569 1976.
58
K. M. Hanson, J. N. Bradbury, T. M. Cannon, R. I. Hutson, D. B.
Laubacher, R. Macek, M. A. Paciotti, S. A. Sandford, and V. W. Steward,
Proton computed tomography of human specimens, Phys. Med. Biol.
26, 965983 1981.
59
U. Schneider and E. Pedroni, Proton radiography as a tool for quality
control in proton therapy, Med. Phys. 22, 353363 1995.
60
U. Schneider, J. Besserer, P. Pemler, M. Dellert, M. Moosburger, E.
Pedroni, and B. Kaser-Holz, First proton radiography of an animal patient, Med. Phys. 31, 10461051 2004.
61
P. Zygmanski, K. P. Gall, M. S. Z. Rabin, and S. J. Rosenthal, The
measurement of proton stopping powers using proton-cone-beam computed tomography, Phys. Med. Biol. 45, 511528 2000.
62
R. Schulte et al., Conceptual design of a proton computed tomography
568
system for applications in proton radiation therapy, IEEE Trans. Nucl.

Sci. 51, 866872 2004.
63
U. Schneider, E. Pedroni, and A. Lomax, The calibration of CT Hounsfield units for radiotherapy treatment planning, Phys. Med. Biol. 41,
111124, 1996.
64
R. W. Schulte, V. Bashkirov, M. C. L. Klock, T. Li, A. J. Wroe, I. Evessv,
D. C. Williams, and T. Satogata, Density resolution of proton computed
65
J. T. de Assis et al., Proton computed tomography as a tool for proton
therapy planning: Preliminary computer simulations and comparisons
with x-ray CT basics, X-Ray Spectrom. 34, 481492 2005.
66
H. F.-W. Sadrozinski et al., Toward proton computed tomography,
IEEE Trans. Nucl. Sci. 51, 39 2004.
67
O. Jkel, G. H. Hartmann, C. P. Karger, P. Heeg, and J. Rassow, Quality
assurance for a treatment planning system in scanned ion beam therapy,
Med. Phys. 27, 15881600 2000.
68
E. Pedroni, S. Scheib, T. Bhringer, A. Coray, M. Grossmann, S. Lin, and
A. Lomax, Experimental characterization and physical modelling of the
dose distribution of scanned proton pencil beams, Phys. Med. Biol. 50,
541561 2005.
69
C. P. Karger, O. Jkel, and G. H. Hartmann, A system for threedimensional dosimetric verification of treatment plans in intensitymodulated radiotherapy with heavy ions, Med. Phys. 26, 21252132
1999.
568
70
D. Nichiporov and K. Solberg, Multichannel detectors for profile measurements in clinical proton fields, Med. Phys. 34, 26832690 2007.
71
S. Boon, P. van Luijk, T. Bhringer, A. Coray, A. Lomax, E. Pedroni, B.
Schaffner, and J. M. Schippers, Performance of a fluorescent screen and
CCD camera as a two dimensional system for dynamic treatment techniques, Med. Phys. 27, 21982208 2000.
72
S. Vatnitsky, Radiochromic film dosimetry for clinical proton beams,
Appl. Radiat. Isot. 48, 643651 1997.
73
D. Nichiporov et al., Investigation of applicability of alanine and radiochromic detectors to dosimetry of proton clinical beams, Appl. Radiat.
Isot. 46, 13551362 1995.
74
W. B. Warren, Evaluation of Bang polymer gel dosimeters in proton
beams, M.S. thesis, The University of Texas Health Science Center at
Houston, Graduate School of Biomedical Sciences, 2007.
75
G. Kraft, The radiobiological and physical basis for radiotherapy with
protons and heavier ions, Strahlenther. Onkol. 166, 1013 1990.
76
T. Kanai et al., Biophysical characteristics of HIMAC clinical irradiation
system for heavy-ion radiation therapy, Int. J. Radiat. Oncol., Biol.,
Phys. 64, 201210 1999.
77
A. Brahme, Recent advances in light ion radiation therapy, Int. J. Radiat. Oncol., Biol., Phys. 58, 603616 2004.
78
R. Mohan, Achieving what-you-see-is-what-you-get in proton therapy,
PTCOG Scientific Meeting, Jacksonville, Fl, May 2008, http://
ptcog.web.psi.ch/ptcog47_talks.html accessed November 1, 2008.
Tomosynthesis imaging: At a translational crossroads

James T. Dobbins IIIa
Carl E. Ravin Advanced Imaging Laboratories, Department of Radiology, Department of Biomedical Engineering,
and Medical Physics Graduate Program, Duke University Medical Center, Durham, North Carolina 27705
Received 23 July 2008; revised 2 February 2009; accepted for publication 27 March 2009;
Tomosynthesis is a decades-old technique for section imaging that has seen a recent upsurge in
interest due to its promise to provide three-dimensional information at lower dose and potentially
lower cost than CT in certain clinical imaging situations. This renewed interest in tomosynthesis
began in the late 1990s as a new generation of flat-panel detectors became available; these detectors
were the one missing piece of the picture that had kept tomosynthesis from enjoying significant
utilization earlier. In the past decade, tomosynthesis imaging has been investigated in a variety of
clinical imaging situations, but the two most prominent have been in breast and chest imaging.
Tomosynthesis has the potential to substantially change the way in which breast cancer and pulmonary nodules are detected and managed. Commercial tomosynthesis devices are now available or
on the horizon. Many of the remaining research activities with tomosynthesis will be translational
in nature and will involve physicist and clinician alike. This overview article provides a forwardlooking assessment of the translational questions facing tomosynthesis imaging and anticipates
some of the likely research and clinical activities in the next five years. 2009 American Association of Physicists in Medicine. DOI: 10.1118/1.3120285
Key words: tomosynthesis, breast imaging, chest imaging, tomography
I. OVERVIEW
Digital tomosynthesis is a simple and relatively inexpensive
method of producing section images using conventional digital x-ray equipment. It is a form of limited angle tomography
that produces section, or slice, images from a series of
projection images acquired as the x-ray tube moves over a
prescribed path. The total angular range of movement is often less than 40. Because the projection images are not acquired over a full 360 rotation about the patient, the resolution in the z direction i.e., in the depth direction
perpendicular to the x-y plane of the projection images is
limited, and thus tomosynthesis does not produce the isotropic spatial resolution achievable with computed tomography
CT. However, the resolution of images in the x-y plane of
the reconstructed slices is often superior to CT, and the ease
of use in conjunction with conventional radiography makes
tomosynthesis a potentially quite useful imaging modality.
There has been a high degree of research interest in tomosynthesis imaging in the past decade, and at least two
commercial products have recently been approved by the
Food and Drug Administration FDA and released on the
market. It is expected that other approved devices will soon
follow. As such, tomosynthesis imaging is in a period of a
high rate of change, with an increasing number of investigators and manufacturers nearing completion of projects involving both the physics and clinical aspects of the technique. If one were to compare the current state of
tomosynthesis imaging to a metaphorical calendar year, the
field is firmly in the middle of spring. It has moved beyond
the winter of initial research investigation but is not yet in
the summer as a mature and accepted clinical modality.
Just as the rate of change in daylight hours is at its greatest at
1956
the equinoxes and slows down near the solstices, so the rate
of progress in a new imaging modality is greatest in the
middle of its developmental pathway. Tomosynthesis imaging is clearly in March of its allegorical calendar year of
development. Change is happening at a rapid pace, which is
exciting for investigators, but it also makes the long-term
future difficult to predict with certainty.
Although tomosynthesis has not yet passed its pivotal
clinical evaluation, it can, to borrow another metaphor, be
said to have passed the light box test. A clinical colleague
has often remarked that he can gauge which new imaging
modalities will be clinically successful from an intuitive appraisal of how substantially the new technique appears to
improve upon the modality it is vying to replace by viewing
the images on a light box across the room. In the opinion
of this colleague, if improvements from a new modality are
only demonstrable by comparing small changes in Az values
following an extensive ROC study, then the improvement of
the new modality may be statistically significant but the level
of its clinical impact may be hard to predict in advance. On
the other hand, if the new modality produces images that are
so substantially superior to the predecessor technique that the
difference is visible on a light box viewed across the room,
then the clinical impact is likely to be substantial. After having viewed several thousand tomosynthesis images in early
clinical studies, most observers would agree that there is a
subjective but substantial improvement in the ability to appreciate abnormal anatomy or disease in tomosynthesis images relative to conventional radiography. At the current time
there are only a limited number of completed clinical studies
to quantify this improvement and to determine how tomosynthesis affects specificity as well as sensitivity. Nonethe-
0094-2405/2009/366/1956/12/$25.00
1956
1957
James T. Dobbins III: Tomosynthesis imaging 2020
less, based purely on the gestalt impression of the images,

most would agree that tomosynthesis passes the light box
test with what appears to be a substantial visual improvement over conventional radiographic imaging.
Notwithstanding the improved visibility of anatomy in tomosynthesis images, the true test of the impact of a new
imaging modality is how it affects management of patients
and clinical outcomes. Issues of cost, radiation dose, flexibility of use, and clinical workflow must be considered in addition to measures of sensitivity and specificity when determining how important a new modality will be. It is precisely
these questions of clinical translation that will be the focus of
much of the research effort in tomosynthesis in the next five
years.
As with weather forecasting, it is difficult to predict with
any certainty the long-range forecast, but one can make some
reasonable estimates of near-term effects. So it is with tomosynthesis. In this article, the short-term future of tomosynthesis research and clinical utilization in the next 5 years will
be assessed. A brief history of tomosynthesis will be provided so that the reader can have a context for current work
in the field. A summary of the current state of the art will be
given, and then many of the current questions of clinical
translation will be explored in more detail.
II. A BRIEF HISTORY OF TOMOSYNTHESIS

Tomosynthesis is one of a number of imaging modalities
whose conception preceded its feasible implementation by
several decades. The basic theoretical framework for limited
angle tomography was provided by Ziedses des Plantes1 in
the 1930s. However, with the development of digital detectors being decades in the future, the implementation of tomosynthesis was clearly not feasible in those early days.
Grant2 coined the term tomosynthesis in a landmark paper
in 1972 that described the method of simple tomosynthesis
reconstruction. A number of variants of tomosynthesis imaging were developed in the 1970s and 1980s, including ectomography by Edholm et al.,3 and flashing tomosynthesis4
that provided rapid imaging for vascular applications.
One of the difficulties encountered with these early techniques was the residual blur from objects outside of the plane
of interest. As anyone familiar with conventional screen-film
based geometrical tomography knows, when the x-ray tube
and sometimes the detector moves during image acquisition, objects in the fulcrum plane of motion remain in sharp
focus but objects outside the fulcrum plane are blurred in
relation to their distance from the fulcrum plane. In the case
of intravenous pyelograms IVPs, for example, a zone of
relatively sharp focus is centered about the opacified kidneys, ureters, and bladder, and the anatomy outside of this
zone is relatively blurry. This type of geometrical tomography works well for imaging high contrast opacified structures, as with IVPs, but is not very successful at imaging
unopacified soft-tissue anatomy because the residual blur
from above and below the plane of interest masks the lowcontrast soft-tissue anatomy of interest.
1957
In the mid-1980s, several investigators explored methods

to reduce the blur artifacts associated with tomosynthesis imaging. Ghosh Roy et al.,5 Chakraborty et al.,6 Ruttimann et
al.,7 Dobbins et al.,8 and others were successful in eliminating all or part of the residual blur from overlying anatomy by
solving for the blurring function in different reconstructed
tomosynthesis planes. These deblurring algorithms made tomosynthesis suitable for consideration in a wider range of
clinical applications
Another more serious challenge to the development of
tomosynthesis imaging was the lack of a suitable digital detector for acquisition of the projection images. At the time
that much of the work on deblurring algorithms was being
done in the 1980s, a suitable large-area, self-scanned digital
detector was not available. The earliest evaluation of the matrix inversion tomosynthesis MITS algorithm in our laboratory, for example, was tested using a conventional geometric tomography table and screen-film imaging; a series of
screen-film projection images was acquired over a range of
tube movement and then digitized.8 The digitized images
were then reconstructed using the MITS algorithm, and a
small phantom comprised of geometric shapes was visible in
several reconstructed planes. Needless to say, this approach
was extremely time consuming and certainly not clinically
feasible. Computed radiography CR imaging plates became
commercially available in the mid-1980s, and while a substantial improvement over digitizing film, were also not suitable because of the logistical impracticality of acquiring and
scanning multiple CR plates during tube movement.
A few investigators used image intensifiers to demonstrate
the potential of tomosynthesis imaging. Image intensifiers
allowed rapid acquisition of images, thereby resolving the
issue of how to acquire multiple images in a clinically realistic time frame. These images demonstrated that the concept
of using tomosynthesis imaging could be useful in appreciating disease in the chest and were a substantial
breakthrough.9 However, image intensifiers have notable
drawbacks, including pincushion distortion and variable geometric uniformity when moving in the earths gravitational
field, which make them less than ideal for tomosynthesis
imaging. A number of corrections are necessary to adequately use image intensifiers for this purpose.
The advent of spiral CT in the late 1980s, coupled with
the lack of a suitable digital radiographic detector, caused
much of the research in digital tomosynthesis to come to a
halt for about a decade. This author, and others, felt that
spiral CT would become the way that all volumetric x-ray
imaging would be done in the future and relegated tomosynthesis to the dustheap of research ideas that looked promising
at first but just did not seem to pan out. The situation
changed substantially in the late 1990s, however, when flatpanel radiographic detectors were introduced. With these
new detectors there was finally a high-DQE, stable, lownoise, self-scanned imaging device without geometric distortion that could image at the speeds needed for reasonable use
in tomosynthesis. Several investigators, including this author,
retrieved tomosynthesis from the shelf of research relics,
dusted it off, and picked up again in earnest with tomosyn-
1958
1958
FIG. 1. Images of pulmonary nodules in chest tomosynthesis images of a human subject. a Coned view of digital PA radiograph shows one clearly visible
right lung nodule arrow. b Tomosynthesis image shows the same nodule vertical arrow as seen on the PA radiograph in a. A second nodule horizontal
arrow is also visible that was not seen in the PA radiograph in a. c Tomosynthesis image at a more posterior level shows an additional left lung nodule
arrow not seen in the PA radiograph in a. d CT image lung window confirms left lower lobe nodule seen in c. Reprinted with permission from
Medical Physics, Ref. 16, Copyright 2008, American Association of Physicists in Medicine AAPM.
thesis research at that time. At last, there was a detector that

made tomosynthesis feasible, and the rate of research exploration in tomosynthesis accelerated considerably.
One of the early applications of tomosynthesis with flatpanel detectors was conducted in our laboratory and was
related to chest imaging. Starting in 1998, we developed,
optimized, and evaluated tomosynthesis with flat-panel detectors for pulmonary nodule imaging.1016 Figure 1 illustrates the advantage of using tomosynthesis for improving
the visibility of subtle pulmonary nodules. At about the same
time, tomosynthesis with flat-panel detectors was also applied to breast imaging.1719
Tomosynthesis has been applied to a variety of clinical
applications over the years, including dental imaging, angiography, and imaging of the chest, breast, and bones see
the review article in Ref. 20 for a more detailed list of citations to these applications. The two that have garnered the
most attention since the late 1990s have been breast imaging
and pulmonary nodule imaging, which will be described in
more detail below.
III. CURRENT STATE OF THE ART

III.A. Geometry of motion
The first item to consider when describing tomosynthesis

imaging is the geometry of motion of the tube and/or detector. There are three basic motion geometries see Fig. 2:
parallel path where the tube moves in a plane parallel to the
detector plane; the detector may also move within its plane,
full isocentric motion where the tube and detector are fixed
rigidly with respect to each other and move in tandem in a
circular path around the patient, and partial isocentric motion the detector remains stationary, and the x-ray tube
moves in an arc above the detector. See Ref. 20 for a review article containing more details about these motions.
All three of these motion geometries are used in contemporary tomosynthesis imaging. The parallel-path motion is typically used in chest and abdominal tomosynthesis in an upright configuration for chest tomosynthesis and a tabletop
configuration for IVP and other abdominal applications.
Partial isocentric motion is used in virtually all current breast
1959
1959
FIG. 2. Geometries of motion for tomosynthesis image acquisition: a

Parallel-path motion, b partial isocentric motion, and c full isocentric
motion. Parallel-path motion in a illustrates how objects in two planes
circle and triangle are projected onto
different locations in the image plane
due to parallax as the tube moves. Reprinted by permission from Physics in
Medicine and Biology, Ref. 20, Copyright 2003, Institute of Physics
IOP Publishing Ltd.
tomosynthesis devices because of the ease of constructing a

compact rotational gantry for the x-ray tube; for simplicity,
the detector remains fixed beneath the breast. Complete isocentric motion is used in cone-beam CT which, for reasons
of nomenclature, is not included in the family of tomosynthesis techniques because cone-beam CT gives a fully isotropic 3D reconstruction. Full isocentric motion is also being
investigated in tomosynthesis applications using C-arm imagers and in radiation oncology applications where the x-ray
source and detector are mounted on the linac gantry as it
rotates through a limited angle of rotation in order to verify
positioning of the patient. Limited-angle tomosynthesis acquisition can be reconstructed using a Feldkamp approach as
described by Godfrey et al.21 or through image manipulation
described by Kolitsi et al.22
Of the three geometries of motion, the parallel-path motion enables the simplest reconstruction algorithm and also
maintains uniform magnification at each tube position. The
partial isocentric motion leads to variable magnification at
different tube orientations, and thus can distort small structures unless care is taken in the backprojection process.23
The full isocentric motion can provide excellent reconstructions but with a more complicated algorithm than the
parallel-path motion.
III.B. Reconstruction algorithms
The most frequently used reconstruction algorithm for tomosynthesis is commonly referred to as shift and add SAA.
In the case of parallel-path geometry of motion of the tube

and/or detector, SAA involves shifting each of the projection
images by a given amount and then adding them together. By
selecting the shift amount correctly, objects in a given plane
can be brought into sharp focus.
When performing SAA, it is important to align the shifted
anatomical information correctly. With most x-ray tube gantries or overhead cranes, there is sufficient mechanical stability such that SAA can be performed adequately based
solely on the known positions of the x-ray tube as it travels.
However, patient motion must also be taken into account if
the desired structures are going to align properly after the
SAA image shifting. In the case of breast tomosynthesis,
motion is mitigated by the light compression of the breast
during image acquisition. In other applications, patient motion can be accounted for by placing fiducial markers on the
patient in order to register the final images after taking into
account motion of the patient. Webber et al.24 introduced a
formalism for incorporating fiducial marker information into
the image reconstruction in a method that is a variant of
SAA; their method, called tuned aperture computed tomography TACT, allows images to be acquired at random
angles and orientations and then reconstructed in arbitrary
planes using the projected locations of the fiducial markers.
In our own research laboratory we have adopted a simplified
version of the fiducial marker approach of Webber et al. in
conjunction with parallel-path geometry to register projection images for chest tomosynthesis. In a research study of
1960
chest tomosynthesis, we found that most human subjects

shifted vertically or horizontally by less than 2 mm during
acquisition of the set of projection images; however, one
subject shifted by as much as 8 mm. This motion did not
result in blur within individual projection images due to the
very short exposure times used. Rather, the subject motion
meant that the sharp structures in individual projection images did not align precisely during shift-and-add reconstruction. Despite this motion, most anatomical structures in the
subjects were reconstructed quite well with no apparent blur;
however, some low-contrast or small objects were reconstructed more poorly unless correction for patient motion
was used. As a result, we have developed an automated algorithm to locate the position of a fiducial marker placed on
the patients back, and we recommend the use of fiducial
markers for improved chest tomosynthesis. Such fiducial
marker registration is not yet available in commercial tomosynthesis devices.
Simple SAA reconstruction is basically the same as unfiltered backprojection. SAA forms the basis of most tomosynthesis algorithms today because of its simplicity. It is, however, insufficient to use SAA alone for high-quality
tomosynthesis reconstructions due to the overlapping blurry
anatomy from outside of the plane of interest. As mentioned
earlier, deblurring algorithms are used to correct for the outof-plane blur and are almost universally adopted in contemporary tomosynthesis imaging. The two deblurring algorithms that have received the most attention in recent years
are MITS and filtered backprojection FBP. MITS, which
was developed in our laboratory, solves for the out-of-plane
blur using the known blurring functions of all other planes
when a given plane is reconstructed.1113,20 Describing the
SAA tomosynthesis reconstructions as a sum of blurry components from all planes, the unblurred structures can be obtained by using matrix algebra to solve the set of coupled
equations in frequency space. This method is quite fast computationally, and given an object composed of a finite number of planes, can render an exact solution in the absence of
noise. Real patients are not merely a set of planes but are
three-dimensional structures, so there is some sharing of
anatomy from plane to plane with MITS; MITS functions in
a well-behaved way for the structures between planes, however.
A technique that is similar to MITS is the iterative restoration approach by Ruttimann et al.7 This method solves for
the blur in each of the planes using known blurring functions, but unlike MITS, it solves the equations iteratively in
the forward direction rather than by matrix algebra as with
MITS. The advantage of image restoration is that it can include all elements of the imaging system, including truncation of structures at the edge of the detector at wide angles. It
is, however, an iterative process, and thus is computationally
slower than MITS.
The deblurring algorithm that is used today by many tomosynthesis investigators, and by most manufacturers, is filtered backprojection. FBP is well known from decades of
work in CT and, like MITS, is a computationally fast algorithm. FBP takes the known sampling density in frequency
1960
space to generate an approximate point spread function that

is used to correct for the blur of out-of-plane anatomy.25 FBP
requires that the Fourier transform of projection images be
multiplied by a ramp function to correct for the point spread
function of the SAA algorithm. This ramp function tends to
produce exaggerated noise in the high frequencies; therefore,
a roll-off filter is typically used to suppress high-frequency
noise amplification. The main difference between FBP and
MITS is in the noise spectrum; FBP has better noise properties at low frequencies than MITS, but MITS has better noise
properties at high frequencies no roll-off filter is required
with MITS. Effort is underway in our laboratory to combine
FBP and MITS to get the optimum noise properties of each
at different parts of the frequency spectrum.
There are also several iterative algorithms that are currently being investigated for tomosynthesis reconstruction.
Wu et al.19 published breast tomosynthesis reconstructions
using maximum likelihood expectation maximization MLEM. An advantage of this approach is that all the components of the imaging system can be modeled, but the downside is that the technique is iterative and quite
computationally intensive. In one study, ML was found to be
superior to FBP for masses and small calcifications.26 Other
investigators have also recently reported on a simultaneous
algebraic reconstruction technique SART that was found to
give results comparable to ML methods but requiring fewer
iterations.27,28
Regardless of the algorithm used to do the basic tomosynthesis reconstruction, it is important when using partial isocentric motion to account for the arc-shaped path of objects
projected onto the detector plane. It has been shown by Chen
et al.23 that such arc-shaped projections can lead to distortion
of the shape of microcalcifications in breast tomosynthesis. It
is important for this effect to be included in the geometry of
motion in the reconstruction algorithm.
Recent work by Fahrig et al.29 and Pineda et al.30 investigated a novel inverse geometry in tomosynthesis imaging.
This inverse geometry uses a scanned source with a small
detector element, just the opposite of the geometry of conventional x-ray imaging. Advantages of this device include a
reduced level of patient skin exposure, which is potentially
advantageous in high-exposure areas such as cardiac
imaging.31 The inverse geometry also inherently provides tomosynthesis imaging capabilities.
III.C. Acquisition parameters
The optimum image acquisition parameters have been investigated in several laboratories and depend on both the
clinical application as well as the reconstruction algorithm
used. In the case of chest imaging, research in our own laboratory has determined that 71 projection images an odd
number is preferable acquired over 20 of tube movement is
best for detection of pulmonary nodules with the MITS
algorithm.15,16 A commercial device currently on the market
uses about 60 images acquired over 40 for chest imaging.
While the number of projection images is fairly consistent
between these two implementations, the angle obviously dif-
1961
fers substantially. The MITS algorithm used in our laboratory has an impulse response that performs better at
midrange angles15 and FBP used with the commercial devices performs better at wider angles,32 thus explaining this
difference. In application to breast imaging, anywhere from
11 to 49 projection images have been used,33,34 with a total
angle of tube movement in the range of 1550.33,35,36
The spacing of the reconstructed planes also varies with
the clinical application and the reconstruction algorithm. In
breast imaging, it is common to use 1 mm plane spacing,19,33
but in chest imaging, 5 mm is more typical.16
III.D. Commercial implementation
Commercial manufacturers have recently introduced tomosynthesis products that can be used for chest, abdominal,
and musculoskeletal applications. Two such devices are currently on the market and are FDA approved. Devices for
breast tomosynthesis are being developed by several manufacturers, and at least one manufacturer has a submission to
the FDA for breast imaging applications.37 It is likely that
within 1 year there will be an FDA-approved device on the
market for breast imaging in the U.S.
IV. TOWARD CLINICAL TRANSLATION
Tomosynthesis research over the past two decades has
addressed many of the fundamental physics questions, and
the field has reached a level of maturity reminiscent of the
first or second generation CT or MR scanners. With the introduction of commercial products, many more investigators
and clinicians will have access to tomosynthesis imaging,
and a host of new questions will arise regarding its appropriate clinical utilization. Indeed, an increased interest in tomosynthesis has already been seen, as reflected in the exponentially expanding number of papers and presentations on this
topic at recent scientific meetings. Many of the questions
ahead will largely be translational in nature, which is not to
say that physicists will not be actively involved; rather, research and applications in tomosynthesis will increasingly
involve the combined efforts of both physicists and clinicians. In this section, some of the likely translational activities will be explored, and a synthesis of where the field is
likely to go next will be given.
IV.A. Continued physics optimization
IV.A.1. Reconstruction algorithms

Reconstruction algorithms in tomosynthesis are relatively
mature at this point, but continued work on optimizing the
deblurring algorithms will occur for the next few years. For
example, Chen et al.38 are investigating the combination of
MITS and FBP in a technique termed Gaussian frequency
blending GFB. This method combines the mid- and highfrequency components of MITS with the low-frequency
components of FBP. Doing so utilizes the excellent highfrequency noise properties of MITS with the excellent lowfrequency noise properties of FBP. The combination should
enable an improved noise power spectrum and slice profile
1961
response across the frequency spectrum. Other investigators

are looking at alternative reconstruction algorithms, including SART.27 Various efforts are underway to compare these
algorithms to one another26,27,3941 and to determine the optimum set of acquisition parameters.15,4245
There is also ongoing work to determine the best visual
presentation of tomosynthesis images. Subjectively, images
processed and reconstructed with deblurring algorithms can
have a high-pass filtered appearance and may lose some of
the overall grayscale range typical of conventional chest or
breast imaging. Although it would be ideal to have the
equivalent appearance of a sagittal or coronal CT reconstruction, tomosynthesis has a very broad slice profile at low frequencies that can appear somewhat unnatural if the image
is presented with too much of a high-pass filtered look. For
this reason, with MITS imaging, for example, a fraction of
the very lowest frequency bins of a conventional tomosynthesis image is added back to the deblurred image. This process produces more of the grayscale range of a conventional
chest radiograph, which may make it easier for radiologists
to acclimate to the new modality. Likewise, with FBP it may
be desirable to add back in a fraction of the lowest frequency
grayscale variation to avoid the situation of the raw beam
outside the patient having the same average grayscale as the
interior anatomy. Again, such change is largely cosmetic but
should improve the confidence of radiologists as the technique is gaining acceptance.
Clinical evaluation studies should be performed to determine if these modifications in visual presentation improve
radiologists performance or make no difference. These
projects will involve the input of both physicists and radiologists, and will likely not be completed until tomosynthesis
has been commercially available for a while and radiologists
have developed some experience with it.
IV.A.2. Dose optimization
The question of optimum dose for tomosynthesis has not
been resolved. The doses reported for a single-view breast
tomosynthesis exam are about one to two times that of a
single-view full-field digital mammogram FFDM33,46,47
doses of 2 mGy Ref. 47 and 4 mGy Ref. 33 have been
reported; many publications do not state absolute dose values, rather merely citing dose relative to that of single-view
FFDM. Early clinical experience in chest tomosynthesis indicates that exposure equivalent to that of one to three lateral
chest exams is about right for reasonable imaging14,16,48 in
one study, total tomosynthesis exposures of 68135 mR were
reported,16 and in another study, an effective dose of 0.12
mSv was reported for chest tomosynthesis as compared with
0.04 mSv for a PA/lateral chest exam48. In both of these
clinical applications, however, the issue has not been fully
resolved regarding whether to target the dose level to produce an overall acceptable image appearance or to address
a specific clinical task. For example, in chest imaging, the
dose level required solely to identify pulmonary nodules is
likely to be perhaps one-half or less of the dose that produces
an image suitable for general diagnostic use.49 The reason for
1962
this dose disparity is that the clinical task of nodule detection

involves looking for larger objects than other general diagnostic tasks in chest imaging. In the case of breast tomosynthesis, the appropriate dose level depends on whether one is
looking for both masses and calcifications or just for masses.
One component of dose optimization will be determining
the optimum x-ray spectrum to yield the highest signal-tonoise ratio for the diagnostic tasks at hand. Currently, chest
tomosynthesis has been performed with spectra equivalent to
conventional radiography 125 kVp in one report48 and 120
kVp with 0.2 mm Cu filtration in another16, but further
study of whether these spectra are optimum is needed. The
spectra reported in breast tomosynthesis have been typical of
those used in conventional digital mammography, but the
optimum spectrum may not be finally decided upon until it is
determined whether breast tomosynthesis will be used for
both masses and calcifications or just for masses.
Some initial effort has gone into determining generally
acceptable dose levels in tomosynthesis using limited numbers of human subjects or computer simulation and modeling
studies.49,50 Further studies will be needed to first determine
the range of diagnostic tasks for which tomosynthesis is useful and then the appropriate dose level for each. These studies will likely take several years as radiologists gain experience with tomosynthesis and develop a sense for its best
clinical utilization.
IV.B. Breast imaging
Breast imaging has seen the greatest number of studies to

date in tomosynthesis. Because there is currently no FDAapproved breast tomosynthesis device, all of the experience
in breast tomosynthesis has been gained in research studies
with human subjects. The total number of human subjects
imaged with breast tomosynthesis is currently over 4000
worldwide. Despite the large number of human subjects participating in clinical studies, there is scant published data to
date indicating sensitivity and specificity results. One study
with 98 subjects found that breast tomosynthesis had image
quality lesion conspicuity and feature analysis rated as
equivalent or superior to diagnostic mammography in 89%
of cases.33 Observer studies of computer simulated lesions in
a structured 3D breast model found Az scores of 0.93 for
tomosynthesis
compared
with
0.76
for
digital
mammography.46 Thus, completion of the various ongoing
clinical trials will be necessary before definitive sensitivity
and specificity data are known in human subjects.
Several important translational questions in breast tomosynthesis need study. First, it will be important to determine
whether breast tomosynthesis will be most useful in screening, diagnostic use, or both. Some investigators feel that tomosynthesis will eventually be useful in both screening and
diagnostic use, but the case for screening seems stronger at
present. Because tomosynthesis may improve sensitivity of
detection, it is likely to be most useful in screening where
most breast cancer is initially discovered. However, because
more potential cancers may be visualized with tomosynthesis, it will be important to carefully evaluate specificity in
1962
large clinical trials. It will also be necessary to evaluate recall rate and biopsy rate, as well as whether the positive
predictive value for biopsy recommendation improves with
tomosynthesis.36 The higher sensitivity of tomosynthesis
may mean that more objects are seen that look suspicious,
and thus the number of recalls may initially increase; however, there is some evidence that tomosynthesis may allow a
better look at potential pathologies, thereby ruling out some
false positives, and thus ultimately the number of cases that
would be recalled may be reduced compared with conventional mammography.33 The very important translational
question of how tomosynthesis affects diagnostic decision
making and patient outcomes will need larger clinical studies
to resolve.47,51
A second important question is how many views to take
with breast tomosynthesis. Currently, some investigators feel
that tomosynthesis should be performed in both mediolateral
oblique MLO and craniocaudal CC views,52 although that
opinion is not universally held. There also remains a question
about whether or not a static view should also be obtained.
The answer to this question depends largely on how effective
tomosynthesis is at imaging microcalcifications. There are
several challenges with imaging microcalcifications with tomosynthesis, including the reduced resolution of the method
compared with conventional single-image mammography
and also the fact that the pattern of distribution of microcalcifications is more difficult to appreciate in section imaging
unless a relatively thick section is used. The reduced resolution with tomosynthesis relative to a single-view FFDM
image is due to the need to shift images by fractional pixel
amounts prior to adding during reconstruction. This reduced
resolution may impact the ability to appreciate the morphology of small calcifications, the importance of which is open
to debate among clinical observers. If it is decided that a
static view is needed to adequately evaluate calcifications,
then a scenario may develop where an MLO static view
mammogram is acquired along with MLO and CC view tomosynthesis exams. Nishikawa et al.53 have proposed a hybrid scheme where a regular MLO or CC view digital mammogram at normal dose is evaluated for microcalcifications,
and a reduced dose and reduced resolution tomosynthesis
data set is reviewed for masses. Clinical trials will be needed
to determine which of these is the most appropriate utilization scheme.
A third translational question for breast tomosynthesis is
how tomosynthesis used for screening will impact diagnostic
workup procedures. Tomosynthesis used for screening will
likely result in fewer diagnostic workups, and traditional
methods such as ultrasound and magnification view mammography can still be used to evaluate suspicious opacities
noted on screening tomosynthesis. Will tomosynthesis have
sufficient specificity that certain patients can go directly to
biopsy? And how will the very small number of cancers
e.g., circumscribed that do not display well on tomosynthesis be handled in screening? These practical questions of
utilization will likely be resolved as radiologists gain clinical
experience with tomosynthesis in the first few years after
FDA-approved devices are available.
1963
There are other new breast imaging approaches under investigation that may impact how tomosynthesis is used. One
example is the development of novel x-ray sources using
carbon nanotube field-emission cathodes.54 If these devices
are successful and reasonably priced, they may enable very
rapid collection of tomosynthesis projection images that
would eliminate the need for lengthy compression. A second
example is the use of iodinated contrast in breast imaging. If
injection of contrast is found useful in breast cancer imaging,
will it also be advantageous in tomosynthesis imaging? Initial studies indicate that it may be.55 If contrast-enhanced
tomosynthesis is found to be advantageous, it may be less
costly and more readily available than alternatives such as
contrast-enhanced magnetic resonance imaging. A third and
larger question is how breast tomosynthesis and cone-beam
CT of the breast will compare. CT is advantageous in that it
has better depth resolution than tomosynthesis, but it also
requires a larger room with more specialized equipment, requires the patient to lie prone, and has difficulty imaging
near the chest wall.56 Computer simulation studies indicated
that both tomosynthesis and breast CT outperformed digital
mammography for detection of masses, but tomosynthesis
and CT performed roughly comparably to each other.46 The
jury is out on this comparison, and studies comparing these
two modalities will be necessary to decide ultimately
whether CT or tomosynthesis will prove superior.
IV.C. Chest imaging
Although chest imaging was one of the first applications

evaluated with tomosynthesis, there have not been as many
studies in the chest as with breast tomosynthesis. We have
investigated detection of pulmonary nodules with tomosynthesis for many years in our laboratory, and there are other
investigators entering this area of research now that commercial chest tomosynthesis devices are available.
Imaging of the chest is very complex due to the wide
range of disease with thoracic manifestations. Partly due to
the wide range of disease in the chest, investigators have
traditionally selected pulmonary nodules as one of the principal areas of research in chest imaging. Pulmonary nodules
are often subtle and difficult to appreciate with conventional
radiography due to the overlying anatomy that reduces their
conspicuity. Tomosynthesis is ideally suited for improving
detection of pulmonary nodules because it eliminates the visibility of overlying structures while still producing an image
that is reminiscent of a posteroanterior PA chest radiograph
Fig. 1. Tomosynthesis can be performed with minor modifications to a conventional digital chest imaging room, making it possible to acquire tomosynthesis images at the same
time as a conventional chest exam. Thus, although tomosynthesis does not have the depth resolution of CT, it has certain
practical advantages and a price and radiation dose likely
less than those of CT.
Early studies have demonstrated that tomosynthesis substantially increases detection sensitivity for pulmonary nodules over conventional PA radiography. In a study of 175
nodules in 21 subjects, we demonstrated that nodule detecMedical Physics, Vol. 36, No. 6, June 2009
1963
tion improved from 22% with PA radiography to 70% with

tomosynthesis; substantial and significant p 0.004 improvement was noted with tomosynthesis over a range of
nodule sizes 35, 510, and 10 mm.16 In a recent study
at the University of Gothenburg, the lesion localization fraction LLF increased from 0.16 in PA radiography to 0.56
with tomosynthesis; however, the nonlesion localization fraction i.e., false positives was approximately 50% higher
with tomosynthesis than with PA radiography.48 In both of
these studies, the sensitivity of detection of pulmonary nodules tripled with tomosynthesis relative to conventional PA
radiography, indicating the potential for improved clinical
detection of lung cancer. However, the increase in false positives may reflect the need for additional clinical training with
the new technique. Investigating reasons for the increase in
false positives will no doubt be the subject of future work.
The primary translational issue for chest tomosynthesis is
determining the best clinical utilization strategy. Tomosynthesis is partway between conventional chest radiography
CXR and CT in its imaging performance. In application to
detection of pulmonary nodules, it is much more appropriate
to see tomosynthesis not as a replacement for CT but as a
vast improvement over CXR. With that perspective, how can
tomosynthesis fit into the clinical management of patients
along with CXR and CT?
There are four likely scenarios in which tomosynthesis
may have clinical utility for pulmonary nodule detection.
First, tomosynthesis may function as a problem-solving tool
for suspicious opacities noted at CXR; if a nodulelike opacity is noted on conventional radiography, tomosynthesis may
enable nodule mimics to be ruled out without the expense of
sending the patient to CT. Second, it may be possible to
perform tomosynthesis on every patient scheduled for CXR
who is at elevated risk for pulmonary malignancies e.g.,
smokers over age 50. By acquiring a tomosynthesis exam
on all high-risk patients scheduled for CXR, it is likely that
some asymptomatic nodules will be detected earlier. Third, it
may be possible to follow patients with known nodules using
tomosynthesis rather than CT. Such an approach would
likely result in monetary savings, but it would need to be
demonstrated in clinical trials that such an approach did not
miss new small nodules that would be actionable. Fourth, if
large-scale screening for lung cancer were adopted in the
future, then tomosynthesis might be a lower-dose and lowercost alternative to CT in such screening.
All four of the above scenarios require additional clinical
studies to determine if they are feasible and effective. The
larger issue with pulmonary nodule detection is more difficult to solve, and that is whether any of the above procedures
results in better patient outcomes. Unlike breast imaging,
which has demonstrated reduced mortality from screening
mammography, there is still considerable controversy over
the role of chest imaging for screening for lung cancer. This
controversy is irrespective of the type of chest imaging i.e.,
CXR, tomosynthesis, or CT. The problem lies in the biology
of lung cancer. Studies have yet to demonstrate that detecting
lung cancer earlier through imaging-based screening results
1964
in reduced mortality. One recent study found that screening

with low-dose CT resulted in lengthened survival,57 while
another study found that low-dose CT screening did not
change mortality from lung cancer.58 These two prominent
papers came to differing conclusions as to whether imagingbased screening is beneficial. It is likely that this controversy
will not be resolved until results from the National Lung
Screening Trial NLST Ref. 59 are available some time in
the future. Tomosynthesis appears to be a promising technique for improving pulmonary nodule detection, but the ultimate benefit to patient outcomes will need to await the
conclusion of the NLST. In the meantime, additional clinical
studies are needed to address the best way to use tomosynthesis as a problem-solving tool in chest imaging and to determine whether it makes sense to image high-risk patients
scheduled for CXR with an additional tomosynthesis exam.
One additional translational question to address is
whether or not the lateral chest image can be eliminated
when using tomosynthesis. The lateral image is not as important as the PA radiograph for most chest imaging cases
but accounts for a majority of the radiation exposure in a
conventional PA/lateral chest exam. The lateral is most commonly used to confirm presence of an anomaly noted in the
PA radiograph or to triangulate the location of a noted object.
Tomosynthesis is likely to be better at both those tasks, and
thus, it raises the question of whether the lateral can be eliminated when using tomosynthesis; doing so would save
enough radiation exposure that a composite exam composed
of a PA radiograph and tomosynthesis may have comparable
dose to a conventional chest exam. Eliminating the lateral
would make the clinical acceptance of chest tomosynthesis
much easier due to dose concerns, but a clinical study will
be required to satisfactorily answer this question.
IV.D. Radiation oncology applications
Tomosynthesis also has generated considerable interest

recently for application in determining patient positioning in
radiation oncology.21,6066 On-board imaging has become an
important part of verification of patient positioning, especially with intensity-modulated radiation therapy IMRT.
Tomosynthesis offers a quicker approach to validating positioning than a full-circle cone-beam computed tomography
CBCT acquisition, although it does not have the full isotropic volumetric information obtainable from CBCT. Tomosynthesis has shown improved localization based upon bony
anatomy when compared with traditional orthogonal radiographs e.g., for head and neck cancers,61 but it has been
more difficult to use with soft-tissue localization, although it
has been shown subjectively to be good for visualizing thoracic and abdominal soft tissues in breath-hold treatment
techniques. Tomosynthesis has also been used for the localization of brachytherapy seeds.67 There continues to be interest from vendors in evaluating tomosynthesis in radiation
oncology applications. Further physics development work
and clinical studies will be needed to confirm its role in these
applications.
1964
IV.E. Other clinical applications
An additional area of exciting prospect for tomosynthesis

is in orthopedic imaging. Studies of joint and bone disease
can be difficult with CT due to the limited resolution in sagittal or coronal planes and with conventional radiography
because of the inability to distinguish three-dimensional
structures. Tomosynthesis naturally addresses both of these
concerns. Flynn et al.68 showed approximately threefold improvement in spatial resolution in tomosynthesis images of
bone relative to that in CT, and yet with dose considerably
below that of CT. Flynns group is investigating tomosynthesis in application to knee, femur, shoulder, arm, leg, ankle,
and wrist and has found in early results that tomosynthesis is
good at demonstrating subtle fractures and at imaging metallic implants that would be difficult to image with CT. Webbers TACT method has also been used in the evaluation of
metallic implants in joint arthroplasty.69
Another application of tomosynthesis to orthopedic use is
in evaluation of arthritis. Duryea et al.70 showed that tomosynthesis could demonstrate the joint space in the fingers
with better clarity than on plain radiographic imaging. They
demonstrated the ability to measure the joint space accurately and speculated that tomosynthesis would also have the
potential to evaluate erosions of the joint space; both of these
measures can be indicative of arthritic changes, lending support to the idea that tomosynthesis could become part of the
screening for and monitoring of arthritic disease. Others are
investigating tomosynthesis joint imaging with a C-arm
device.71
Tomosynthesis has also been used in dental imaging for
indications such as evaluation of implants, and there is a
commercial dental tomosynthesis imaging unit on the market. However, the trend today seems to be more toward
CBCT for a broader range of dental uses, and so the future of
tomosynthesis imaging in dental applications is uncertain.
Tomosynthesis is also being evaluated for use
intraoperatively72 or in emergency settings where standard
CT imaging would be impractical or too time consuming.
IV.F. Computer-aided detection and diagnosis
Computer-aided detection CAD and computer-aided detection diagnosis CADx schemes are often mentioned in
connection with tomosynthesis73 and for two primary reasons. First, the reduction of overlying anatomy offers the
potential for improved sensitivity and specificity for CAD
algorithms. One of the factors that limits the success of some
CAD algorithms is the large number of false-positive findings. Using tomosynthesis to eliminate much of the confusing background should theoretically reduce the number of
false positives. However, in some cases such as with chest
tomosynthesis, this reduction in false positives from reduction of overlying soft-tissue might be limited by increased
false positives from circular cross sections of imaged vessels.
Thus, exactly how much tomosynthesis will benefit CAD
algorithms remains to be seen. There is ongoing research in
several laboratories, including ours, into these issues.
1965
A second reason that tomosynthesis and CAD may make

a good marriage is that CAD may reduce some of the overhead of viewing the large number of images in a set of tomosynthesis slices. One concern about tomosynthesis that is
sometimes expressed by clinical colleagues is that tomosynthesis produces many more images than conventional projection imaging, and therefore, may slow down workflow. It is
the opinion of this author that such reduction in workflow
will not scale with the number of images and will ultimately
not be too burdensome on the clinical staff. The reasoning
behind that opinion is that the stack of images may be
viewed at a PACS station dynamically by paging through the
set quickly. The entire set of images viewed in such a dynamic fashion will take more time than reading a single conventional image, to be sure, but will be much faster than
viewing each of the slice images statically as was the case
long ago with CT images printed side by side on hardcopy
film. Furthermore, the human eye is very sensitive to dynamic changes, and subjectively when anomalous objects in
tomosynthesis slices appear in a dynamically viewed stack,
they are quite apparent. That being said, if CAD algorithms
could be developed to review tomosynthesis image stacks,
they might be able to speed up the process with which radiologists could review the data. Clearly, ongoing research in
this area is needed.
Despite the potential advantages of combining CAD and
tomosynthesis outlined above, there are considerations that
might argue that tomosynthesis would benefit less from CAD
than either conventional radiography or CT. Because objects
are much more easily detected in tomosynthesis images than
in conventional radiography, the need for CAD may well be
less in tomosynthesis than in conventional imaging. On the
other hand, chest CT, for example, can demonstrate such an
abundance of small nodules that are below the actionability
threshold of 5 mm based on the recommendation of the
Fleischner Society74 that reader fatigue may make it difficult
to find all nodules in a chest CT data set. Tomosynthesis of
the chest, on the other hand, does not display as many small
nodules, and therefore, may not be as substantially helped by
CAD as CT would be.
It is clear that substantial research into CAD algorithms
for tomosynthesis using both slice-by-slice evaluation as
well as full three-dimensional evaluation will be an important area of research in the future. The impact of using CAD,
either as a prereader or a secondary reader, on user workflow
will be important things to measure before tomosynthesis
makes its full translational move from the bench to the bedside.
1965
sensitivity and specificity of tomosynthesis in cancer imaging. Further studies are needed to address its full clinical
performance.
Tomosynthesis is in somewhat of a chicken and egg situation at the present time with regard to its clinical acceptance. Before tomosynthesis will be widely adopted in clinical use, its benefit to patient outcomes must be demonstrated
with further studies and wider clinical experience. However,
in order to complete such studies and gain such wider clinical experience, the technique must be used on more patients.
This type of impasse is often seen early in the adoption of
any new modality, and it is certain that gaining reimbursement for the technique will be an important component of
accelerating its clinical utilization.
Tomosynthesis is fundamentally at a translational crossroads at the present time, with many of the remaining questions not ones of physics optimization or even clinical performance but rather pertaining to practical matters. How will
tomosynthesis impact workflow, and what will be the consequences of that effect? How will tomosynthesis fit into the
paradigm of clinical management as a tool halfway between
traditional radiography and CT? Will patients that receive
tomosynthesis be sent on to CT or other standard diagnostic
workup afterwards in most cases anyway, thus ultimately
not saving much money for the healthcare enterprise? Or will
tomosynthesis find a niche that enables better utilization of
dose and imaging dollars than the existing clinical paradigm?
This author believes the latter will ultimately be the case,
although further trials are needed to demonstrate it. Tomosynthesis is somewhat unique as a new imaging modality in
that it is not likely to produce a revolution as CT did but
rather an evolution of better clinical management. We are
currently in the exciting and uncertain times when the tipping point will likely be seen in the next 13 years that will
determine the future of this interesting imaging modality.
ACKNOWLEDGMENTS
The author gratefully acknowledges the input and suggestions of Joseph Y. Lo, Ph.D., Jay A. Baker, MD, Devon J.
Godfrey, Ph.D., Michael J. Flynn, Ph.D., and Donald A. Tyndall, Ph.D., DDS in the preparation of this manuscript. Some
of the research results from the authors laboratory were supported by grants from NIH Grant No. R01 CA80490 and
GE Healthcare. GE and Duke jointly hold a patent related to
certain aspects of tube movement in tomosynthesis imaging.
Collaborative work on breast tomosynthesis at the authors
institution was supported in part by a grant from Siemens
Medical Solutions.
a
V. CONCLUSION AND SUMMARY

It seems clear from early studies that tomosynthesis offers
the potential for superior performance to conventional radiography for several important clinical applications, and the
prospect exists that it may become a cost-effective and lowdose imaging strategy to improve earlier detection of disease.
Important clinical studies are underway to quantify both the

[email protected]
1
B. G. Ziedses des Plantes, Eine neue methode zur differenzierung in der
roentgenographie planigraphie, Acta Radiol. 13, 182192 1932.
2
D. G. Grant, Tomosynthesis: A three-dimensional radiographic imaging
technique, IEEE Trans. Biomed. Eng. BME-19, 2028 1972.
3
P. Edholm, G. Granlund, H. Knutsson, and C. Petersson, Ectomography:
A new radiographic method for reproducing a selected slice of varying
thickness, Acta Radiol. 21, 433442 1980.
4
H. Becher, M. Schluter, D. G. Mathey, W. Bleifeld, E. Klotz, P. Haaker,
R. Linde, and H. Weiss, Coronary angiography with flashing tomosyn-
1966
thesis, Eur. Heart J. 6, 399408 1985.

D. N. Ghosh Roy, R. A. Kruger, B. Yih, and P. Del Rio, Selective plane
removal in limited angle tomographic imaging, Med. Phys. 12, 6570
1985.
6
D. P. Chakraborty, M. V. Yester, G. T. Barnes, and A. V. Lakshminarayanan, Self-masking subtraction tomosynthesis, Radiology 150, 225229
1984.
7
U. E. Ruttimann, R. A. J. Groenhuis, and R. L. Webber, Restoration of
digital multiplane tomosynthesis by a constrained iteration method,
IEEE Trans. Med. Imaging 3, 141148 1984.
8
J. T. Dobbins III, A. O. Powell, and Y. K. Weaver, Matrix inversion
tomosynthesis: Initial image reconstruction, in RSNA 73rd Scientific
Assembly, Chicago, IL, 1987 unpublished.
9
S. Sone, T. Kasuga, F. Sakai, K. Oguchi, A. Itoh, F. Li, Y. Maruyama, K.
Kubo, T. Honda, M. Haniuda, and K. Takemura, Digital tomosynthesis
imaging of the lung, Radiat. Med. 14, 5363 1996.
10
J. T. Dobbins III, R. L. Webber, and S. M. Hames, Tomosynthesis for
improved pulmonary nodule detection, in RSNA 84th Scientific Assembly, Chicago, IL, 1998 unpublished.
11
R. J. Warp, D. J. Godfrey, and J. T. Dobbins III, Applications of matrix
inverse tomosynthesis, Proc. SPIE 3977, 376383 2000.
12
D. J. Godfrey, R. J. Warp, and J. T. Dobbins III, Optimization of matrix
inverse tomosynthesis, Proc. SPIE 4320, 696704 2001.
13
D. J. Godfrey, A. Rader, and J. T. Dobbins III, Practical strategies for the
clinical implementation of matrix inversion tomosynthesis MITS, Proc.
SPIE 5030, 379390 2003.
14
J. T. Dobbins III, D. J. Godfrey, and H. P. McAdams, in Advances in
Digital Radiography: RSNA Categorical Course in Digital Radiography,
edited by E. Samei and M. J. Flynn Radiological Society of North
America, Oak Brook, 2003.
15
D. J. Godfrey, H. P. McAdams, and J. T. Dobbins III, Optimization of
the matrix inversion tomosynthesis MITS impulse response and modulation transfer function characteristics for chest imaging, Med. Phys. 33,
655667 2006.
16
J. T. Dobbins III, H. P. McAdams, J.-W. Song, C. M. Li, D. J. Godfrey, D.
M. DeLong, S.-H. Paik, and S. Martinez-Jimenez, Digital tomosynthesis
of the chest for lung nodule detection: Interim sensitivity results from an
ongoing NIH-sponsored trial, Med. Phys. 35, 25542557 2008.
17
L. T. Niklason, B. T. Christian, L. E. Niklason, D. B. Kopans, D. E.
Castleberry, B. H. Opsahl-Ong, C. E. Landberg, P. J. Slanetz, A. A. Giardino, R. Moore, D. Albagli, M. C. DeJule, P. F. Fitzgerald, D. F. Fobare,
B. W. Giambattista, R. F. Kwasnick, J. Liu, S. J. Lubowski, G. E. Possin,
J. F. Richotte, C.-Y. Wei, and R. F. Wirth,, Digital tomosynthesis in
breast imaging, Radiology 205, 399406 1997.
18
S. Suryanarayanan, A. Karellas, S. Vedantham, S. J. Glick, C. J. DOrsi,
S. P. Baker, and R. L. Webber, Comparison of tomosynthesis methods
used with digital mammography, Acad. Radiol. 7, 10851097 2000.
19
T. Wu, A. Stewart, M. Stanton, T. McCauley, W. Phillips, D. B. Kopans,
R. H. Moore, J. W. Eberhard, B. Opsahl-Ong, L. Niklason, and M. B.
Williams, Tomographic mammography using a limited number of lowdose cone-beam projection images, Med. Phys. 30, 365380 2003.
20
J. T. Dobbins III and D. J. Godfrey, Digital x-ray tomosynthesis: Current
state of the art and clinical potential, Phys. Med. Biol. 48, R65R106
2003.
21
D. J. Godfrey, F.-F. Yin, M. Oldham, S. Yoo, and C. Willett, Digital
tomosynthesis with an on-board kilovoltage imaging device, Int. J. Radiat. Oncol., Biol., Phys. 65, 815 2006.
22
Z. Kolitsi, V. Anastassopoulos, A. Scodras, and N. Pallikarakis, A multiple projection method for digital tomosynthesis, Med. Phys. 19, 1045
1050 1992.
23
Y. Chen, J. Y. Lo, and J. T. Dobbins III, Importance of point-by-point
back projection correction for isocentric motion in digital breast tomosynthesis: Relevance to morphology of structures such as microcalcifications, Med. Phys. 34, 38853892 2007.
24
R. L. Webber, R. A. Horton, D. A. Tyndall, and J. B. Ludlow, Tunedaperture computed tomography TACT-TM: Theory and application for
three-dimensional dento-alveolar imaging, Dentomaxillofac Radiol. 26,
5362 1997.
25
G. Lauritsch and W. H. Harer, A theoretical framework for filtered backprojection in tomosynthesis, Proc. SPIE 3338, 11271137 1998.
26
T. Wu, R. H. Moore, E. A. Rafferty, and D. B. Kopans, A comparison of
reconstruction algorithms for breast tomosynthesis, Med. Phys. 31,
26362647 2004.
5
1966
27
Y. Zhang, H. P. Chan, B. Sahiner, J. Wei, M. M. Goodsitt, L. M. Hadjiiski, J. Ge, and C. Zhou, A comparative study of limited-angle conebeam reconstruction methods for breast tomosynthesis, Med. Phys. 33,
37813795 2006.
28
Y. Zhang, H. P. Chan, B. Sahiner, Y.-T. Wu, C. Zhou, J. Ge, J. Wei, and
L. M. Hadjiiski, Application of boundary detection information in breast
tomosynthesis reconstruction, Med. Phys. 34, 36033613 2007.
29
R. Fahrig, A. R. Pineda, E. G. Solomon, A. N. Leung, and N. J. Pelc,
Fast tomosynthesis for lung cancer detection using the SBDX geometry, Proc. SPIE 5030, 371378 2003.
30
A. R. Pineda, S. Yoon, D. S. Paik, and R. Fahrig, Optimization of a
tomosynthesis system for the detection of lung nodules, Med. Phys. 33,
13721379 2006.
31
E. G. Solomon, M. S. Van Lysel, R. E. Melen, J. W. Moorman, and B.
Skillicorn, Low-exposure scanning-beam x-ray fluoroscopy system,
Proc. SPIE 2708, 140149 1996.
32
B. Li, G. B. Avinash, J. W. Eberhard, and B. E. H. Claus, Optimization
of slice sensitivity profile for radiographic tomosynthesis, Med. Phys.
34, 29072916 2007.
33
S. P. Poplack, T. D. Tosteson, C. A. Kogel, and H. M. Nagy, Digital
breast tomosynthesis: Initial experience in 98 women with abnormal digital screening mammography, AJR Am. J. Roentgenol. 189, 616623
2007.
34
W. Zhao, B. Zhao, P. R. Fisher, P. Warmoes, T. Mertelmeier, and J.
Orman, Optimization of detector operation and imaging geometry for
breast tomosynthesis, Proc. SPIE 6510, M1M12 2007.
35
T. Wu, R. H. Moore, and D. B. Kopans, Voting strategy for artifact
reduction in digital breast tomosynthesis, Med. Phys. 33, 24612471
2006.
36
J. M. Park, E. A. J. Franken, M. Garg, L. L. Fajardo, and L. T. Niklason,
Breast tomosynthesis: Present considerations and future applications,
Radiographics 27, S231S240 2007.
37
T. Wallack, New dimension in detection, Boston Globe, June 6, 2008.
38
Y. Chen, J. Y. Lo, J. A. Baker, and J. T. Dobbins III, Gaussian frequency
blending algorithm with matrix inversion tomosynthesis MITS and filtered back projection FBP for better digital breast tomosynthesis reconstruction, Proc. SPIE 6142, 0E10E9 2006.
39
Y. Chen, J. Y. Lo, and J. T. Dobbins III, Impulse response analysis for
several digital tomosynthesis mammography reconstruction algorithms,
Proc. SPIE 5745, 541549 2005.
40
J. T. Rakowski and M. J. Dennis, A comparison of reconstruction algorithms for C-arm mammography tomosynthesis, Med. Phys. 33, 3018
3032 2006.
41
J. Zhou, B. Zhao, and W. Zhao, A computer simulation platform for the
optimization of a breast tomosynthesis system, Med. Phys. 34, 1098
1109 2007.
42
H. C. Gifford, C. S. Didier, M. Das, and S. J. Glick, Optimizing breasttomosynthesis acquisition parameters with scanning model observers,
Proc. SPIE 6917, 0S10S9 2008.
43
T. Deller, K. N. Jabri, J. M. Sabol, X. Ni, G. Avinash, R. Saunders, and R.
UppaluriEffect of acquisition parameters on image quality in digital tomosynthesis, Proc. SPIE 6510, 1L11L11 2007.
44
I. Reiser, B. A. Lau, and R. M. Nishikawa, Effect of scan angle and
reconstruction algorithm on model observer performance in tomosynthesis, in IWDM 2008, edited by E. A. Krupinski Springer-Verlag, Berlin,
2008, Vol. LNCS 5116, pp. 606611.
45
H. P. Chan, J. Wei, Y. Zhang, B. Sahiner, L. M. Hadjiiski, and M. A.
Helvie, Detection of massses in digital breast tomosynthesis mammography: Effects of the number of projection views and dose, in IWDM
2008, edited by E. A. Krupinski Springer-Verlag, Berlin, 2008, Vol.
LNSC 5116, pp. 279285.
46
X. Gong, S. J. Glick, B. Liu, A. A. Vedula, and S. Thacker, A computer
simulation study comparing lesion detection accuracy with digital mammography, breast tomosynthesis, and cone-beam CT breast imaging,
Med. Phys. 33, 10411052 2006.
47
W. F. Good, G. S. Abrams, V. J. Catullo, D. M. Chough, M. A. Ganott, C.
M. Hakim, and D. Gur, Digital breast tomosynthesis: A pilot observer
study, AJR Am. J. Roentgenol. 190, 865869 2008.
48
J. Vikgren, S. Zachrisson, A. Svalkvist, . A. Johnsson, M. Boijsen, A.
Flinck, S. Kheddache, and M. Bth, Comparison of chest tomosynthesis
and chest radiography for detection of pulmonary nodules: Human observer study of clinical cases, Radiology 249, 10341041 2008.
49
C. M. Li and J. T. Dobbins III, Methodology for determining dose re-
1967
duction for chest tomosynthesis, Proc. SPIE 6510, 2D12D10 2007.

P. Timberg, M. Bth, I. Andersson, T. Svahn, M. Ruschin, B. Hemdal, S.
Mattsson, and A. Tingberg, Impact of dose on observer performance in
breast tomosynthesis using breast specimens, Proc. SPIE 6913, 4J1
4J10 2008.
51
D. Gur, Tomosynthesis: Potential clinical role in breast imaging, AJR
Am. J. Roentgenol. 189, 614615 2007.
52
E. A. Rafferty, L. T. Niklason, and L. A. Jameson-Meehan, Breast tomosynthesis: One view or two?, in RSNA 92nd Scientific Assembly,
Chicago, IL, 2006 unpublished.
53
R. M. Nishikawa, I. Reiser, and P. Seifi, A new approach to digital breast
tomosynthesis for breast cancer screening, Proc. SPIE 6510, 3C13C8
2007.
54
G. Yang, R. Rajaram, G. Cao, S. Sultana, Z. Liu, D. Lalush, J. Lu, and O.
Zhou, Stationary digital breast tomosynthesis system with a multi-beam
field emission x-ray source array, Proc. SPIE 6913, 1A11A10 2008.
55
S. C. Chen, A.-K. Carton, M. Albert, E. F. Conant, M. D. Schnall, and A.
D. A. Maidment, Initial clinical experience with contrast-enhanced digital breast tomosynthesis, Acad. Radiol. 14, 229238 2007.
56
A. Karellas, J. Y. Lo, and C. G. Orton, Cone beam x-ray CT will be
superior to digital x-ray tomosynthesis in imaging the breast and delineating cancer, Med. Phys. 35, 409411 2008.
57
The International Early Lung Cancer Action Program Investigators, Survival of patients with stage I lung cancer detected on CT screening, N.
Engl. J. Med. 355, 17631771 2006.
58
P. B. Bach, J. R. Jett, U. Pastorino, M. S. Tockman, S. J. Swensen, and C.
B. Begg, Computed tomography screening and lung cancer outcomes,
JAMA, J. Am. Med. Assoc. 297, 953961 2007.
59
D. R. Aberle, C. Chiles, C. Gatsonis, B. J. Hillman, C. D. Johnson, B. L.
McClennan, D. G. Mitchell, E. D. Pisano, M. D. Schnall, and A. G.
Sorensen, Imaging and cancer: Research strategy of the American College of Radiology Imaging Network, Radiology 235, 741751 2005.
60
D. J. Godfrey, L. Ren, H. Yan, Q. J. Wu, S. Yoo, M. Oldham, and F.-F.
Yin, Evaluation of three types of reference image data for external beam
radiotherapy target localization using digital tomosynthesis DTS, Med.
Phys. 34, 33743384 2007.
61
Q. J. Wu, D. J. Godfrey, Z. Wang, J. Zhang, S. Zhou, S. Yoo, D. M.
Brizel, and F.-F. Yin, On-board patient positioning for head-and-neck
IMRT: Comparing digital tomosynthesis to kilovoltage radiography and
cone-beam computed tomography, Int. J. Radiat. Oncol., Biol., Phys. 69,
598606 2007.
50
1967
62
H. Yan, L. Ren, D. J. Godfrey, and F.-F. Yin, Accelerating reconstruction

of reference digital tomosynthesis using graphics hardware, Med. Phys.
34, 37683776 2007.
63
L. Ren, D. J. Godfrey, H. Yan, Q. J. Wu, and F.-F. Yin, Automatic
registration between reference and on-board digital tomosynthesis images
for positioning verification, Med. Phys. 35, 664672 2008.
64
H. Yan, D. J. Godfrey, and F.-F. Yin, Fast reconstruction of digital tomosynthesis using on-board images, Med. Phys. 35, 21622169 2008.
65
G. Pang, A. Bani-Hashemi, P. Au, P. F. OBrien, J. A. Rowlands, G.
Morton, T. Lim, P. Cheung, and A. Loblaw, Megavoltage cone beam
digital tomosynthesis MV-CBDT for image-guided radiotherapy: A
clinical investigational system, Phys. Med. Biol. 53, 9991013 2008.
66
M. Descovich, O. Morin, J. F. Aubry, M. Aubin, J. Chen, A. BaniHashemi, and J. Pouliot, Characteristics of megavoltage cone-beam digital tomosynthesis, Med. Phys. 35, 13101316 2008.
67
I. B. Tutar, R. Managuli, V. Shamdasani, P. S. Cho, S. D. Pathak, and Y.
Kim, Tomosynthesis-based localization of radioactive seeds in prostate
brachytherapy, Med. Phys. 30, 31353142 2003.
68
M. J. Flynn, R. McGee, and J. Blechinger, Spatial resolution of x-ray
tomosynthesis in relation to computed tomography for coronal/sagittal
images of the knee, Proc. SPIE 6510, 0D10D9 2007.
69
F. H. Fahey, R. L. Webber, F. S.-K. Chew, and B. A. Dickerson, Application of TACT to the evaluation of total joint arthroplasty, Med. Phys.
30, 454460 2003.
70
J. Duryea, J. T. Dobbins III, and J. A. Lynch, Digital tomosynthesis of
hand joints for arthritis assessment, Med. Phys. 30, 325333 2003.
71
S. Li and H. Jiang, A practical method for three-dimensional reconstruction of joints using a C-arm system and shift-and-add algorithm, Med.
Phys. 32, 14911499 2005.
72
G. Bachar, J. H. Siewerdsen, M. J. Daly, D. A. Jaffray, and J. C. Irish,
Image quality and localization accuracy in C-arm tomosynthesis-guided
head and neck surgery, Med. Phys. 34, 46644677 2007.
73
H. P. Chan, J. Wei, B. Sahiner, E. A. Rafferty, T. Wu, M. A. Roubidoux,
R. H. Moore, D. B. Kopans, L. M. Hadjiiski, and M. A. Helvie,
Computer-aided detection system for breast masses on digital tomosynthesis mammograms: Preliminary experience, Radiology 237, 1075
1080 2005.
74
H. MacMahon, J. H. M. Austin, G. Gamsu, C. J. Herold, J. R. Jett, D. P.
Naidich, E. F. Patz, Jr., and S. J. Swensen, Guidelines for management of
small pulmonary nodules detected on CT scans: A statement from the
Fleischner Society, Radiology 237, 395400 2005.
The evolution of brachytherapy treatment planning

Mark J. Rivarda
Department of Radiation Oncology, Tufts University School of Medicine, Boston, Massachusetts 02111
Jack L. M. Venselaar
Department of Medical Physics, Instituut Verbeeten, P.O. Box 90120, 5000 LA Tilburg, The Netherlands
Luc Beaulieu
Dpartement de Radio-Oncologie et Centre de Recherche en Cancrologie de lUniversit Laval,
Centre Hospitalier Universitaire de Qubec, 11 Cte du Palais, Qubec, Qubec G1R 2J6, Canada
and Dpartement de Physique, de Gnie Physique et dOptique, Universit Laval, Qubec,
Qubec G1K 7P4, Canada
Received 2 February 2009; revised 5 April 2009; accepted for publication 6 April 2009;
Brachytherapy is a mature treatment modality that has benefited from technological advances.
Treatment planning has advanced from simple lookup tables to complex, computer-based dosecalculation algorithms. The current approach is based on the AAPM TG-43 formalism with recent
advances in acquiring single-source dose distributions. However, this formalism has clinically
relevant limitations for calculating patient dose. Dose-calculation algorithms are being developed
based on Monte Carlo methods, collapsed cone, and solving the linear Boltzmann transport equation. In addition to improved dose-calculation tools, planning systems and brachytherapy treatment
planning will account for material heterogeneities, scatter conditions, radiobiology, and image
guidance. The AAPM, ESTRO, and other professional societies are working to coordinate clinical
integration of these advancements. This Vision 20/20 article provides insight into these
endeavors. 2009 American Association of Physicists in Medicine. DOI: 10.1118/1.3125136
Key words: brachytherapy, treatment planning
I. HISTORY OF BRACHYTHERAPY DOSIMETRY
Many of our technological discoveries are distinctly marked
in history. Independent of each other, two of these discoveries launched the start of the radiation oncology era. Wilhelm
Rentgen discovered x rays in November 1895, and shortly
afterward, Henri Becquerel accidentally exposed a photographic plate to uranium in 1896, identifying the phenomenon of emitted radiation.1 The work of Pierre and Marie
Curie in 1898 was needed to extract radium from pitchblende
ore to determine the origin of this penetrating radiation. Very
soon after these first steps, new pathways were explored to
apply these radiations in the first treatments of patients. Becquerel himself experienced the effects of radiation exposure
by carrying a tube containing decigrams of radium chloride
in his vest pocket. He recorded the progression of his skin
reaction.2 The first medical experiences belong to Danlos and
Bloch3 in 1901 of Paris and Abb4 in 1904 of New York. The
Radium Biological Laboratory was created in Paris in 1906,
and Finze in London started treatments with radium in 1909.
A book on radium therapy, which is now known as brachytherapy, was published in 1909 by Wickham and Degrais.5
These early 20th century achievements established the medical application of radiation within one decade of its discovery. The first investigators were by necessity led by clinical
observations and developed from their experiences the rules
to avoid errors in clinical applications. In medical science,
such processes soon lead to systems or schools. For brachytherapy, 226Ra was the only radionuclide for these systems.6
The basic principles of its use in brachytherapy were estab2136
lished after World War I in medical institutes such as the

Radium Hemmet in Stockholm, the Memorial Hospital in
New York, and the Radium Institute in Paris. Arrangement of
the radioactive sources in certain geometric patterns, with
definitions of the strength, distance, and treatment time, developed into a set of rules for patient treatments. For intracavitary treatments, the Stockholm and Paris methods were
described in 1914 and 1919, respectively. Paterson and
Parker set the basis of the 1930 rules for the Manchester
system,7 which was extensively described in a book by
Meredith.8
Two other steps are important to date the development of
brachytherapy. First, the discovery of artificial radioactivity
in 1934, allowing the use of manufactured radioactive materials in radiotherapy. Second, in the 50s and 60s of the 20th
Century, the development of remote afterloading devices allowed improved personnel radiation protection and gave
more flexibility to the applications. New radionuclides like
60
Co, 137Cs, 182Ta, and 198Au were applied with designs initially similar to 226Ra sources. Henschke9 was the first to
explore the use of 192Ir. This radionuclide is currently the
most widely applied radioactive source in the field of brachytherapy. The rules for interstitial therapy using low-dose rate
LDR 192Ir wire sources, which could be applied in various
lengths and strengths, were developed in a very precise and
consistent manner in Paris.10,11 Hollow tubes, in the form of
either needles or plastic catheters, were surgically implanted,
equidistant and parallel, to allow placement of radioactive
wires or seed ribbons with an afterloading procedure. Using
0094-2405/2009/366/2136/18/$25.00
2136
2137
Rivard, Venselaar, and Beaulieu: The evolution of brachytherapy treatment planning
conventional x-ray and dummy sources, imaging for implant

localization was possible to completely eliminate operating
room staff exposure.
The first afterloaders were developed simply to minimize
exposure from the radioactive sources. Using a cable attachment, they just pushed the sources mechanically into the preinserted applicators.12,13 Only later its function changed toward multiple-programmable source trains and, eventually,
to miniature stepping source technology. High-activity
sources, for high-dose rate HDR and pulsed dose rate
PDR applications with the latter type mostly used in Europe, have largely replaced the use of LDR afterloadingthe
exception is permanent implant technology for LDR prostate
brachytherapy. Miniaturized 192Ir sources with a typical outer
diameter of 1 mm replaced the 137Cs tube and pellet applications. Optimization, modulating the dose distribution by
varying the dwell times, has become a standard feature in
brachytherapy systems, providing the user of the afterloader
with greater flexibility. Further development and history of
brachytherapy technology and medical physics practice during the past 50 years, covering the period in which the remotely controlled afterloaders were introduced and refined,
has been described in a review article by Williamson in
Physics in Medicine and Biology.14 Ibbott et al.15 described
50 years of AAPM involvement in radiation dosimetry, including dosimetric issues in brachytherapy, in Medical Physics. Issues and trends in brachytherapy physics were further
discussed in the Medical Physics Anniversary Paper by Thomadsen et al.16 In this invited Vision 20/20 article, we focus
on the current trends and challenges in brachytherapy, specifically on the medical physics practice of brachytherapy
treatment planning and the research activities for advancement.
I.A. Brachytherapy dosimetry systems
The term dosimetry can have a very narrow interpretation

but may just as well have a much broader meaning. It may
refer only to the measurements in a radiation field of radioactive sources or of radiation beams, it may refer to the
methodology for calculation of a dose or dose rate value at a
specific point in a given medium, or it may also reflect the
full chain from imaging and reconstruction to dose delivery
and then also includes the prescription of dose. In this article,
dosimetry is used mainly in the context of the calculation
process. A dosimetry system essentially consists of three
parts:
i
a set of rules to distribute the sources inside a defined

volume to achieve a clinically acceptable dose distribution,
ii a method to calculate patient dose, and
iii a system for dose prescription.
These steps were already acknowledged and defined in an
era when computers were not available and imaging tools
were not integrated into brachytherapy procedures. A standard system relies essentially on two steps: i implant source
reconstruction, originally and for decades based on orthogo-
2137
nal x rays, and ii summation of dose patterns of the individually reconstructed sources. The conventional systems
were applicator based, not anatomy based. They did not utilize the anatomical information, so the relation between dose
deposition and the tumor volume as well as the volume of
healthy tissue and organs at risk OARs was underdeveloped. Implementation of modern resources with accurate calculation techniques and imaging procedures allows us to increase the efficacy of brachytherapy in terms of clinical
outcome by demonstrating whether or not a chosen isodose
covers the clinical target volume CTV and spares OARs.
Patient-specific treatment planning is presently considered
useful in our treatments and is essential for achieving the
treatment planning standards of modern radiation oncology
in more complicated cases.
I.B. Prescription dose
Historically, brachytherapy prescriptions were stated in

terms of exposure and exposure rate. In intracavitary gynecologic brachytherapy, the use of mg h radium or radiumequivalent dosimetry was the standard for decades. Interstitial brachytherapy was often performed using the rules of the
Quimby or Manchester implant systems.17,18 These systems
specified the geometric arrangement of radium needles and
their radioactive contents relative to the target volume. The
basis for a treatment time calculation was the implanted area
or volume to obtain exposure or mg h estimates from dosespecification criteria for the idealized needle arrangements
described by the system. Reconstructions of the real implant
were made from planar x-ray images. This method could
reconstruct the needles, but fundamentally had limitations
recognizing the clinical target. Using manual tools, such as
the tables and nomograms published for these systems, the
reference dose rate was estimated and the treatment time
calculated. Real implants frequently deviated significantly
from the idealized source arrangements used as examples for
the system rules.18 Nevertheless, the dose calculation itself
was generally accurate as demonstrated in comparisons with
published data tables. For example, see the work of Shalek
and Stovall who reviewed the calculation models for 226Ra
tubes and needles.19
For 192Ir wire interstitial implants with continuously
loaded and parallel placed sources, the Paris system was described in conjunction with an escargot snail diagram Fig.
1 to ease the calculation process.11,20 Such a diagram contains information on a variety of dose rates for a standardized
linear source strength of 1 mCi/cm for a series of source
lengths. Using such a diagram, one can derive the contribution of dose to a point in terms of dose rates from each
individual source. Basal dose points are used in the Paris
system and are found as the points at the center of gravity
between the neighboring sources at the central plane of the
implant. By adding the contributions and successively repeating the procedure for each basal dose point, the basal
dose or the average dose in all basal dose points can be
found. The reference prescription isodose line/surface is then
taken as 85% of the basal dose in this system, assuming to
2138
2138
FIG. 2. Coordinate system for the AAPM TG-43 brachytherapy source dosimetry formalism from Rivard et al. Ref. 24.
FIG. 1. The escargot diagram in the Paris system is a tool for manual
calculation of the dose rate at points at distances in the plane transversal to
the source. The curves represent the dose rate in cGy h1 for sources of 1
up to 7 cm length and a linear strength of 1 mR h1 m2 cm1
8.7 Gy h1 m2 cm1. From Dutreix et al. Ref. 20.
cover the planned target volume PTV. As with the older

Quimby and Manchester rules, the methods were manual and
based on tabulated or nomogram-derived datasets.
I.C. Classical form of calculation algorithm
for point sources
Classically, the dose rate for a source is derived from its

contents, i.e., the amount of radioactivity contained within
the source. The activity was first expressed as the weight of
the contents, for 226Ra in mg. With the unit Ci defined as the
activity of 1 g of 226Ra, an important step was made. The
dose rate in Gy/h at distance r from a point source was described as
r = A f as,Wr Fm ,
D
W
r2
in which A is the effective activity in Ci, deduced from dose

rate measurements in air, is the exposure rate per Ci of the
radionuclide in R/h at 1 m, f as,Wr is the correction for absorption and scatter effects of the photons at distance r in
water when compared to the same point in vacuo, and Fm is
the medium conversion factor from R to Gy.
Inspection of this equation reveals a number of flaws and
possibilities for errors. First, there is the confusion between
the real and the effective, assumed, or apparent activity. Second, we have, at least quantitatively, a poorly defined medium that is related to the value of the Fm factor. It differs for
water, tissue, and materials with strongly deviating composiMedical Physics, Vol. 36, No. 6, June 2009
tion such as bone and lung. Then, these values must be taken
from textbooks or other resources where it is easy to find
great 25% disparities in the tabulated values.21 This is
apparent when trying, for example, to list values from the
literature for a radionuclide with a complex photon spectrum
such as 192Ir. These flaws may lead to errors, for example,
when the effective activity is measured from dose rate in air
by the vendor while the user of a given calculation system
has included a different value. The fact that this concern
is not imaginary but very real is demonstrated in the IAEA
survey on misadministrations in medical radiation applications, often occurring in coordination with a lack of
training and supervision of the personnel involved in the
procedures.22
As the quantities for A, , and Fm in Eq. 1 are not in
accordance with the SI system, and therefore considered obsolete, Eq. 1 needs to be rephrased for the dose rate to
r. A very detailed step-by-step description is
water D
W
given in the recent book by Baltas et al.23 resulting in Eqs.
2 and 3,
1 g en
r = K
D
W
R
r = S 1 g en
D
W
K
r0
r
1
r
f as,Wr,
f as,Wr.
and air-kerma strength S

The reference air-kerma rate K
R
K
are mutually related through the inverse-square law to the
reference distance r0 as KR = SK1 / r02. g is the fraction of
electron energy liberated by photons in air that is lost due to
radiative processes such as bremsstrahlung and fluorescence.
The term en / W is the ratio of the mass-energy absorption
coefficient of water to that of air.
I.D. Calculations with encapsulated and elongated
sources
A real source such as the one depicted in Fig. 2 has finite

dimensions compared to an idealized point source. The most
common shape of a brachytherapy source is the cylinder. The
design of the source, its shape, and the choice of the active
material influence the resulting dose distribution. Brems-
2139
strahlung photons generated in the capsule by beta particles

may contribute to the dose, and the capsule may filter the
radiation emitted by the bare source material. The photon
energy spectrum of a realistic source can therefore deviate
substantially from that of a bare source, with especially significant deviations along the source long axis. The calculation process according to the Sievert-integration procedure
considers the dose at a point near the source as the summation of point-source contributions. An elongated source is
then divided into a series of infinitesimal source elements.
The path-length attenuation through the source and the capsule wall using the effective-attenuation coefficients for the
small source elements takes into account most of the
effects.25 This method is accurate to within 2%8% for 137Cs
sources but is less accurate at low- and intermediate-photon
energies e.g., less than 0.4 MeV.26 The evolution of dosecalculation methods has been discussed and reviewed elsewhere for 137Cs and low-energy photon-emitting sources.27,28
Computer algorithms have previously been based on the
use of tabulated data in Cartesian or polar coordinates, or
made use of directly calculated data from the Sievertintegration procedure. Both the manual and the classical approaches in dose-calculation algorithms used in computerized systems are rather simplistic. Generally, the influence of
tissue heterogeneity, organ or body outline, or applicator heterogeneity is ignored, excluding an occasional onedimensional 1D correction to the total dose to account for a
predefined ovoid shield.29 Interseed attenuation ISA or interapplicator shielding is not addressed. Other effects are difficult to quantify since they depend on many patient-specific
circumstances. These effects will be discussed in Sec. III.
II. CURRENT STATUS OF BRACHYTHERAPY
TREATMENT PLANNING
As of 2009, the current approach for brachytherapy dose
calculation is based on the AAPM TG-43 dosimetry
formalism,24,30 which relies on superposition of single-source
dose distributions obtained in a liquid water phantom with a
fixed volume for radiation scattering. This approach was developed through computerized treatment planning systems
TPSs to replace the antiquated approaches such as the
Manchester and Paris systems.3133
II.A. Dosimetry formalism
Based on the assumption of cylindrically symmetric

brachytherapy source dose distributions, the TG-43 brachytherapy dosimetry formalism utilizes a polar coordinate system along the source long axis z axis with the coordinate
system origin located at the center of radioactivity, as shown
in Fig. 2. Dose distributions at point Pr , are obtained in
the vicinity of the source with radial distance r and polar
angle expressed relative to the origin and source long axis,
respectively. A special reference point Pr0 , 0 is positioned
at r0 = 1 cm and 0 = 90 on the transverse plane of the
source. For Fig. 2, = 2 1, and t is the capsule thickness
in the transverse-plane direction.
2139
Either the 1D or 2D equation used to calculate dose rate

are used in modern TPS, as shown in Eqs. 4 and 5, respectively, with dosimetry parameters defined below,
r = S r0
D
K
r
g Pr anr,
r, = S GLr, g r Fr, .
D
K
L
GLr0, 0
Identical to Eqs. 2 and 3, SK is the air-kerma strength of

the brachytherapy source and is determined for specific
sources either by the clinical medical physicist or provided
by the manufacturer.34 SK has units U = cGy cm2 h1 and its
measured value should be traceable to a primary standards
dosimetry laboratory PSDL like the National Institute of
Standards and Technology NIST.35 The dose rate constant
r , / S since
has units of cGy h1 U1 such that = D
0 0
K
the other dosimetry parameters from Eq. 5 take values of
unity at Pr0 , 0. Dose rate variations in the source transverse plane are accounted for by the geometry function
and radial dose function. The geometry function takes the
simple form of point-source or inverse-square approximation
r0 / r2 for the 1D formalism as is most commonly used in
brachytherapy TPS for LDR low-energy photon-emitting
brachytherapy using sources such as 125I. However, the
2D formalism approximates the distribution of radiation
emissions as a line-segment with length L and is most commonly used for HDR high-energy photon-emitting brachytherapy sources such as 192Ir. Here, GLr , = / L r sin.
The
point-source
radial
dose
function
g Pr
2
= Dr / Dr0 r / r0 , whereas the line-source radial dose

r , / D
r , G r , / G r , . For
function gLr = D
0
0 0
L 0 0
L
0
calculating dose rate off the transverse-plane where dose anisotropy from brachytherapy sources comes into play, the 1D
anisotropy function anr or 2D anisotropy function Fr ,
is used. All these dosimetry parameters are explained in great
detail in the 2004 AAPM TG-43U1 report,24 along with reporting criteria and good practice recommendations for dosimetry investigators obtaining these parameters using measurements and/or Monte Carlo MC methods.
II.B. Dosimetry parameter datasets
Often there are several dosimetry publications on a given

brachytherapy source model, and thus there is potential for
variable administration of dose with clinically relevant ramifications. To subjugate this problem, the AAPM prepares
consensus dosimetry datasets for use in brachytherapy TPS.
As of 2009, 16 consensus datasets for LDR 125I and 103Pd
sources have been issued by the AAPM TG-43U1 and TG43U1S1 reports.24,36 These datasets are incorporated by TPS
software vendors as machine data, analogous to external
beam treatment planning data, and are also used in practice
for treatment planning of patients on cooperative group clinical trials. Incorrect application of data could cause variability
in clinical practice and unnecessary variations in administered dose or even serious dose-calculation errors. A brief
2140
2140
description is provided of the main publicly accessible archives or databases of dosimetry datasets to provide guidance on dosimetry parameter choice. Toward providing guidance on dataset choice, recommendations to medical
physicists by Rivard et al.37 are presented below:
a
b
c
AAPM-recommended consensus data from the TG43U1 and TG-43U1S1 reports;24,36

data from the joint AAPM/Radiological Physics Center
RPC Source Registry using data in the original publications to check for agreement; and
if a given source model is not posted on the joint
AAPM/RPC Source Registry, this suggests that it may
not have met even the most basic dosimetric practice
standards. The burden then falls upon the user to perform an analysis to ensure that seed calibrations are
NIST traceable with an acceptable level of accuracy
and that available published and unpublished dosimetry
data exhibit an acceptable level of rigor and redundancy to safely treat patients. Early adopters of a new
seed product should consider closely working with the
vendor, the vendors dosimetry consultant, and the
treatment planning software vendor to ensure that the
conditions outlined in the relevant AAPM recommendations are met.24,38,39 Coordinated action by the medical community using a particular source model, rather
than isolated and possibly inconsistent individual
analyses, are preferred to maximize uniformity of dosimetric practice. For a product with a track record of
use for which the vendor has not successfully submitted an AAPM/RPC Registry application e.g., a source
widely used in Europe but new to North America,
clinical users should consider following GEC-ESTRO
recommendations. If available, datasets on the ESTRO
or Carleton University websites may be useful.
In all instances the medical physicist should document the

source of the brachytherapy dosimetry parameter data and
provide rationale for why a given dataset and/or website
tabulation was chosen.
III. DOSIMETRY FORMALISM LIMITATIONS
While dataset standardization is a key element for acquisition of high-quality clinical results, especially from clinical
trials, dosimetry parameter datasets used in the AAPM
TG-43 dosimetry formalism are obtained in a liquid water
phantom with a fixed volume for radiation scattering. Consequently, the formalism does not readily account for several
important aspects that undermine clinical application of the
TG-43 formalism with these datasets.
FIG. 3. Effect of phantom medium on absorbed dose and attenuation at

r = 1 cm as a function of photon energy.
III.A.1. Differences between absorbed dose in

water and tissue
Under charged-particle equilibrium, the mass-energy absorption coefficient as a function of photon energy E and
atomic number Z, notated as en / E,Z, may be used to account for differences in absorbed dose between water and
tissue as Dtissueen / E,tissue = Dwateren / E,water. Using the
soft tissue characterization from ICRU 44,40 with water and
tissue mass densities of 1.00 and 1.06 g / cm3, Fig. 3 presents
water
as a functhe ratio of absorbed dose in water to tissue Dtissue
tion of photon energy. Over the range of photon energies
examined, 0.0110 MeV, the absorbed dose in tissue is
nearly equal to that in liquid water as shown by the solid
curve. This ratio dips to 0.96 near 0.05 MeV due to slightly
higher photoelectric effect cross sections of tissue compared
to water.41,42
III.A.2. Differences between radiation attenuation
in water and tissue
Similarly, the mass-attenuation coefficient as a function of
photon energy and atomic number / E,Z may approximate
differences in radiation attenuation between water and tissue
I = I0eZ / E,Zt or I = I0eE,Zt.43 Again using the water and
tissue mass densities, but with a simple path length approach
using t = 1 cm, the ratio of radiation attenuation between the
two media I / I0water / I / I0tissue = eE,water1 cm / eE,tissue1 cm
is presented in Fig. 3 as the dashed curve. Unlike the absorbed dose ratios, the larger mass density of tissue contributes to the increased attenuation. Further, increased attenuation by tissue below 0.05 MeV is caused by the larger
photoelectric effect cross section of tissue compared to water.
III.A. Dosimetric circumstances highlighting

formalism limitations
III.A.3. Source shielding or applicator radiation

interactions
Due to the simplicity of brachytherapy dosimetry in comparison to external beam dosimetry, a basic understanding of
radiological physics interactions can generally explain five
limitations of the AAPM TG-43 dosimetry formalism.
For multisource clinical applications, ISA may be significant. Applicator:radiation interactions and applicator shielding may detract from the accuracy of conventional TG-43
based dose calculations.44 For radiation shielding between
2141
FIG. 4. Photon transmission ratios through water or high-Z attenuators as a

function of photon energy. The Ag K edge 0.02 551 MeV is indicated in
the dashed curve.
sources or by an applicator, the attenuation of water is replaced with a high-Z material. For LDR low-energy photonemitting sources, this high-Z material may for example be a
silver Z = 47 radio-opaque marker used for identification
during postimplant CT localization. For HDR high-energy
photon-emitting sources, the material may be a stainless steel
applicator Z = 26. Using Ag = 10.5 g / cm3 and Fe = 7.8 g /
cm3 with a minimal thicknesses of 0.1 mm and negligible
water attenuation, the high-Z photon attenuations are given
in Fig. 4. As in the prior two examples, the photoelectric
effect is largely responsible for the differences in comparison
to water.
III.A.4. Differences between radiation scattering for
dataset acquisition and patient treatment
The effect of phantom dimensions and volume in MC
dosimetry has been reported in literature as early as
1991.43,4549 For high-energy photon-emitting brachytherapy
sources such as 192Ir, dose differences greater than 5% are
possible within 10 cm of the source, as shown in Fig. 5.
However, studying varying phantom dimensions quantifies
only the effect of excess material or missing backscatter
close to the phantom boundary as source position is fixed
and phantom radius decreases. Therefore, as argued and
shown in Pantelis et al.50 and later by Lymperopoulou et
al.,51 corresponding findings cannot be readily translated to
the effect expected in actual clinical practice where a source
dwell position is programmed close to the patient contour as
scattering conditions are altered in total, including lateral
scatter.
Further, the proportion of dose due to primary and scattered radiation significantly changes as a function of distance
from the source, as shown in Fig. 6, which is also for 192Ir.52
2141
FIG. 5. Comparison of 192Ir gr values calculated for a given spherical

water phantom radius R compared to R = 50 cm, which is assumed to provide full scatter conditions. Data include curves for R = 5, 10, 15, 20, 25, 30,
35, 40, and 45 cm. Reproduced from Fig. 1c of Ref. 43 with permission
from C. S. Melhus and M. J. Rivard, Med. Phys. 33, 1729 2006. Copyright
2006, American Association of Physicists in Medicine.
disequilibrium in homogeneous media due to brachytherapy

inverse-square increases in photon fluence upon moving toward a brachytherapy source. Concern for dose:kerma
equivalence is not commonly observed in brachytherapy
measurements or MC calculations of conventional highenergy sources since the required distance is 5 mm for
greater than a 5% effect.54 For LDR 137Cs sources, this distance is adjacent to the source capsule. For HDR 192Ir or
60
Co sources, dose/kerma of 1.05 occurs at distances below 2 and 7 mm from the source center, respectively. Breakdown of CPE can also occur at material boundaries such as
tissue:high-Z interfaces. Another aspect of electrons in
brachytherapy dosimetry ignored with current TPS is the
dose contribution from betas emitted upon radionuclide
disintegration.55
III.A.5. Dose, kerma, and electrons

Subtleties associated with dose calculation include the assumption of equivalence of absorbed dose and kerma.
Roesch53 first considered the possibility for charged-particle
FIG. 6. Components of 192Ir dose as a function of radial distance. Reproduced with permission from R. E. Taylor and D. W. Rogers, Med. Phys. 35,
4933 2008. Copyright 2008, American Association of Physicists in
Medicine.
2142
2142
TABLE I. Sensitivity of commonly treated anatomic sites to dosimetric limitations of the current brachytherapy
dose calculation formalism. Items flagged as Y indicate the authors opinion that significant differences
between administered and delivered dose are possible due to the highlighted dosimetric limitation.
Anatomic site
Prostate
Breast
GYN
Skin
Lung
Penis
Eye
Source energy
Absorbed dose
Attenuation
Shielding
Scattering
Beta/kerma dose
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
N
N
Y
N
N
N
Y
N
Y
N
Y
Y
N
Y
Y
N
N
N
N
Y
Y
N
N
Y
N
N
N
Y
Y
Y
Y
Y
Y
Y
Y
N
N
N
N
N
N
N
N
Y
N
N
N
Y
N
III.A.6. Summary of dosimetric concerns

Absorbed dose in water is about 4% and +2% compared
to tissue for low- and high-energy photons, respectively. Per
centimeter, the attenuation of high-energy photons is about
the same between water and tissue. However, there are significant differences in attenuation for low-energy photons,
increasing as photon energy decreases. The presence of
high-Z materials can substantially alter dose distributions for
low-energy photons. Dose differences of 5% are possible
for high-energy sources within 5 cm of the skin. Equivalence
of dose and kerma within 5% does not hold true within a few
millimeters of high-energy photon-emitting sources. Dosimetric contributions from beta emissions at these distances are
also ignored in the current TG-43 formalism.
III.B. Clinical circumstances highlighting formalism
limitations
In addition to dosimetric limitations, there are limitations

to applying the AAPM TG-43 dosimetry formalism to certain
clinical circumstances. For instance, the formalism does not
readily permit calculation of dose distributions for curved
brachytherapy sources such as LDR 192Ir wire. This is also
r L / 2 , is needed, but is
true for long sources where D
0
challenging to implement on TPS with the 2D formalism
since high Fr , gradients near the source long axis require
high-resolution dosimetry parameters. Beyond these TPS
limitations, the scope of anatomic sites commonly considered for application of brachytherapy are practically subject
to the limitations of the AAPM TG-43 dosimetry formalism
as described in the previous dosimetric section. A breakdown
is presented in Table I of the relative sensitivity of these
anatomic sites to the five dosimetric limitations.
In our opinion, Table I highlights the sensitivity of each
anatomic site to the aforementioned dosimetric limitations.
Prostate implants using high-energy sources such as HDR
192
Ir are deep seated and not as sensitive to radiation scattering conditions as are other sites. Further, a single source is
translated throughout the implanted plastic catheters and thus

none of the other effects are prominent. For low-energy
photon-emitting sources in the prostate, the D90 is generally
within 1 cm of the gland and water:tissue attenuation differences are minimal. However, absorbed dose can differ by
several percent, and ISA can exceed 10% along the needle
direction. LDR prostate implants typically utilize the largest
number of implanted sources among all anatomic sites, and
may be subject to dosimetric effects of calcifications.61
Breast implants using HDR 192Ir share similar circumstances
as HDR 192Ir prostate implants. However, the radiation scattering conditions differ and generally overestimate dose
deposition in clinical implantsespecially near the skin surface. For breast implants using low-energy photon emitting
sources such as LDR 103Pd or the Xoft Axxent 50 kVp
source, scattering differences are negligible between the
treatment plan using superposition of source locations in liquid water and the clinical environment. However, absorbed
dose in breast at these energies is generally underestimated
by several percent, source-to-source shielding can be significant for 103Pd sources, and differences between prescribed
and administered doses can also be significant due to differences in radiation attenuation between water and tissue
especially at large distances. Gynecological implants often
use high-Z colpostats for shielding the bladder and rectum
particularly for treating the uterus. For treating the vagina/
cuff, treatment through a plastic applicator is de rigueur and
contributes to differences in radiation attenuation. Absorbed
dose is underestimated by several percent for low-energy
sources. However, radiation scattering is not of concern due
to the deep-seated position within the body. Brachytherapy
of the skin is most sensitive to scattering conditions. Use of
bolus material will provide better agreement between
planned dose and delivered dose, but skin dose outside the
treatment field can be minimized without bolus. Further,
most high- and low-energy brachytherapy treatments of the
skin use some form of shielding to protect the unintended
2143
tissues, and conventional TPS cannot account for this shielding. Again, absorbed dose is underestimated by several percent for low-energy brachytherapy sources. Lung implants
typically include tissues with mass densities three to four
times less than conventional tissues. Consequently, scattering
conditions for both high- and low-energy brachytherapy
sources are not congruent between the clinical implant and
the planned approach. Absorbed dose and attenuation are
also significantly different for low-energy sources, and dose
calculations close to high-energy sources such as HDR 192Ir
can be in error by several percent within a few millimeters of
the source. Brachytherapy of the penis is performed using
either HDR high-energy sources or LDR low-energy sources.
Both applications overestimate administered dose due to radiation scattering conditions. Further, low-energy sources underestimate dose due to differences in tissue composition between water and tissue. Eye plaque brachytherapy has been
conducted for several decades but is sensitive to all of the
aforementioned dosimetric limitations.
In summary, it is clear that most applications of brachytherapy are subject to limitations of the current dosecalculation formalism. Thus, it is imperative that advances be
made to resolve these discrepancies and provide clinical users a realistic depiction of individualized brachytherapy dose
administration.
IV. BRACHYTHERAPY TREATMENT PLANNING
RESEARCH
IV.A. Introduction
In recent years, advanced dose-calculation methods have

been considered to perform explorative and retrospective
clinical studies. These studies identified limitations of the
current protocol for clinical practice. Others have looked at
dose comparison in clinical geometries such as a shielded
vaginal cylinder51,57 or shielded rectal applicator.56 In these
cases, the dose differences were dramatic. For seed implants,
ISA was thought to be a negligible effect due to the physical
size of the seed.58 Meigooni et al.59 first showed that the
minimum peripheral dose was decreased by 6% due to
ISA; Mobit and Badragan60 calculated a 10% decrease depending on the seed configuration. Clinical seed configurations were studied by Chibani et al.61 and Carrier et al.62 A
decrease of 2%5% can be expected on the clinical parameter, D90.63 The decrease was dependent on the radionuclide
choice,61 seed density,62 and seed design.64 Furthermore, tissue heterogeneity was also shown to be of importance for
low-energy seeds, contributing to further decrease in D90.62,65
This effect was more pronounced for 103Pd than 125I.66
However, for 192Ir, tissue heterogeneity was negligible for
r 10 cm,43 covering most of the clinical situations. Finally,
eye plaque dosimetric effects were also studied in detail.67,68
While most of these studies were conducted using
general-purpose MC codes MCNP69,70 and GEANT471,72 or
thoroughly tested radiation therapy-specific MC codes e.g.,
73
74
75,76
EGS4, EGSnrc, PTRAN
, new MC codes are being developed specifically for brachytherapy such as MCPI Ref. 77
and BRACHYDOSE.78 Other avenues are also explored such as
2143
the collapsed cone CC,79,80 and attempts to directly solve

the transport equation via discrete ordinates DO using
Attila/Acuros are underway.81,82 We also refer the readers
to reviews on this subject.83,84
IV.B. Analytical models primary/scatter
Separation of primary and scattered photon contributions

to the total dose has been successful in the Comptondominated conditions of external beam conditions,85,86 ultimately leading to various dose-calculation algorithms such
as the pencil beam,87 CC,79 and superposition convolution.88,89 Advantages of CC and superposition-convolution
implementations are their availability as TPS dosecalculation algorithms. Furthermore, they can accurately account for material inhomogeneities in external beam radiation therapy through clinically validated algorithms.
For convolution methods, the energy deposition is generally composed of two terms. The first one describes the interaction of the primary photons with the medium, specifically the total energy release per unit mass or terma. The
second component addresses the details of how the energy
release is deposited at various distances from the interaction
site. This is accomplished by energy deposition kernels. Usually, the kernels are separated in two or three components:
primary dose, the first-scatter dose, and the multiple-scatter
dose the last two components may be combined into a
single kernel. When inhomogeneities are present, the kernels can be scaled according to the average mass or electron
density, depending on the implementation. It is the superposition or convolution of the various kernels, weighted by
the terma, that generates the dose distribution.
In the CC approach, the kernels are polyenergetic and
represented by an analytical function of the nonisotropic energy deposition around the primary interaction site in a
spherical coordinate system.79 The interaction site is assumed to be located at the origin of a set of lines, each being
the central axis of a cone. The energy deposited in this cone
is further described, or collapsed, on the central line. Implementation for dose calculation along Cartesian coordinate
volumes requires the number of cone axes or transport lines
to adequately cover the voxels in the geometry of interest.
Photon energies are lower in brachytherapy than for external beam. As such, the radiation path length is shorter and
the contribution of scattered radiation to the dose, at relevant
distances, can be of the same order of magnitude as the primary dose.90 Furthermore as photon energy decreases, the
amount of energy released per interaction decreases and multiscatter quickly becomes the dominant contribution of the
total scatter component.91 This is illustrated in Fig. 6, which
shows that for 192Ir, the multiple-scatter component has a
higher contribution to the dose than the single scatter at 6 cm
from the source and constitute the highest contributor to the
total dose after 10 cm.52 For most modern brachytherapy
radionuclides with photon energies of 192Ir or less, the secondary electron range is small and kerma well approximates
dose. In this context, the calculation of primary dose can be
performed quickly and with high accuracy. However, accu-
2144
rate calculations of the multiple-scatter components necessitate the full 3D geometry, especially in the presence of material inhomogeneities for which the interaction cross
sections can vary greatly.
An early implementation of a convolution method for
brachytherapy has been demonstrated by Williamson et al.92
In this approach, the multiple-scatter component is provided
by precomputed MC dose spread array. For 125I and 137Cs
point-source dose calculations, the results were shown to be
within 3% of the MC calculations with a gain in speed of
factors of 2050 at that time. In this approach, heterogeneity
corrections were accounted for by rescaling the dose spread
of the once-scattered photons. Another approach proposed by
Williamson et al.93 used a scatter-subtraction method based
on 1D scatter integration. While being about 1000 times
faster than MC, it was limited to water-equivalent heterogeneities for cylindrical geometries. This method was further
extended to a 2D scatter integration,94 allowing arbitrary geometry. However, its heterogeneity correction was limited by
analytical correction factors for a wide range of atomic numbers and material densities.95 The latter incarnation was
shown to be 50 000 100 000 times faster than MC while
maintaining an accuracy of 10% except for low-energy photons and high-atomic number heterogeneity materials. However, in this approach multiple scattering is neglected and the
heterogeneity geometry was limited to a thin slab.
The application of a CC algorithm in brachytherapy has
been pursued for many years by the Helax/Nucletron group
in Uppsala, Sweden. With this algorithm, a successivescattering superposition method has been developed in which
the first-scatter dose distribution is used to derive the higher
order multiple-scatter distributions. In this formalism, the
multiple-scatter kernel can be isotropic for low-energy
photons96 or when oriented along the primary photon.80 A
scaling method was also developed for the successivescattering superposition in the presence of high-Z and highdensity materials.97 A source description formalism was also
proposed in the context of primary and scatter dose separation PSS formalism that is backward compatible with the
current TG-43 dose-calculation formalism and is used as input into the CC algorithm.98 CC performance was recently
studied for relevant brachytherapy geometries and for 28, 60,
and 350 keV sources. A total of 31 800 voxels/ s 0.2 mm3
voxel, 52 directions were calculated for permanent prostate
seed geometry.90 The authors further tested implementation
on parallel hardware, specifically GPU cards, and showed it
to be 100 times faster for the same geometry.90
Finally, Wang and Sloboda99,100 also proposed a formalism of primary, once-scatter and multiple-scatter separation.
The novelty is that the once-scatter is evaluated using a microbeam ray-tracing algorithm, which includes basic interaction physics such as Compton, Rayleigh, and photoelectric
effect. The advantage of this approach is that the once-scatter
contribution does not have to be scaled to account for material heterogeneities.99 The multiple-scatter dose is obtained
from a nonlinear curve fitting of MC multiple-scatter dose in
various heterogeneous media.100 Accuracy was within 10%
for all cases studied.
2144
IV.C. Deterministic solution to the transport equation
Monte Carlo uses stochastic approaches random numbers to sample the probability density functions describing
the phenomena underlying the transport of particles through
matter for simulations. With sufficient statistics or particle
histories, MC obtains precise solutions to a variety of problems in radiation therapy. Another calculation approach consists of directly solving the linear Boltzmann transport equation LBTE through deterministic means. Two general
approaches with application to brachytherapy can be seen in
the literature: solving the differential LBTE81,101 or the integral formulation.102 The LBTE is commonly solved by discretization of the parameters phase space. The most commonly seen method in scholarly articles is the angular
domain or the use of DO. Discretization is also performed in
space finite difference or finite element and in energy with
appropriate multigroup cross sections. The system can be
coupled to handle neutral, charged, and coupled photonelectron-positron transport for iterative solutions. The discretization of the space dimension is well suited to problems
in radiation therapy where voxel-based geometries of patients are constructed from tomographic images.101
It is generally understood that the accuracy of deterministic approaches is directly related to discretization,101,103
with fine steps leading to accurate solutions at the price of a
larger system of equations to be solved. Thus, the technique
is numerically intensive. As noted by Daskalov et al.,101 the
deterministic concept of accuracy is not the same as for MC
dose calculation and is related to a systematic difference between the solution and the truth. With deterministic tools,
uncertainty does not have a stochastic component, while MC
uncertainties have stochastic and systematic components. An
important issue for brachytherapy is the ray effect, which
results from the nonphysical fluence buildup due to the limited number of angles for near pointlike sources in lowdensity media. To diminish this artifact, stochastic or semianalytic methods are generally used to determine the oncescattered dose distribution within the discretized phase
space.81,101
The first application of a deterministic approach to HDR
brachytherapy used a finite-difference multigroup-DO solver
from Los Alamos National Laboratory LANL called
DANTSYS.104 Though limited to 2D, the results were within
2 of EGS4 MC calculations.101 This approach was further
extended, namely, by reducing the energy group cross sections, to calculation of the absorbed dose rate around a 125I
brachytherapy seed for a 2D cylindrical geometry. The deterministic method was shown to be 100-fold more efficient
than the EGS4 MC code.105
Zhou and Inanc102 introduced another approach to solve
the integral representation of the LBTE and studied ISA in
125
I seed implants. The algorithms were extended to account
for fluorescence x rays, and parallelization was introduced.
Single seeds and more complex configurations such as the
Quality Assurance Review Center www.qarc.org geometry
for clinical protocol accreditation were studied.106 The calcu-
2145
lation took 2.5 h on a 64 processor computer cluster.

However, the multiseed configuration results were not compared to reference MC calculations.
More recently, the LANL developments were transferred
to a commercial spin-off, Transpire Inc. Gig Harbor, WA.
Their Attila product is a multidimensional finite-element
multigroup-DO solver. A mathematical description of the
various components can be found in the 2006 paper of Gifford et al.81 Born from nuclear science, it handles electron
and photon beams and brachytherapy sources. A specific
package enables 3D geometry construction from various
CAD formats. Interestingly, the results of solving the transport equation are energy-dependent fluence maps, not dose.81
The final dose distribution is obtained by multiplying the
fluence by the appropriate / E,Z value. Gifford et al. conducted a comparative study between Attila, MCNPX, and
EGS4 Ref. 81 and a detailed application to 3D dose calculation around a HDR 192Ir source. In the latter case, calculation times were about 15 min and the deterministic method
was over 100 times more efficient than MCNPX.81
IV.D. Monte Carlo
Direct MC simulation is based on a random sampling of

particle histories to estimate the quantity of interest, absorbed dose in the patient. A comprehensive overview of the
possibilities of MC methodology is given in the doctoral
thesis of Carlsson Tedgren, since interaction details can be
implemented with no other limits than time, this is potentially the most accurate method available for dose
calculations.107 MC already plays an important role in several aspects of brachytherapy dose calculation. For example,
MC methods revealed theoretically that dose rates from 125I
were lower by 10%14% due to air-kerma contributions
from titanium characteristic x rays in the NIST SK,N85 calibration standard.26,108 MC is an obligatory part of the confirmatory process and is equal to measurements for AAPM
brachytherapy dosimetry parameter datasets.24 It is also used
to characterize radiation sources in terms of their spatial distribution of primary and scattered radiation to allow simple
input data for modeling clinical sources. Further, it can be
used to derive transmission data through materials and thus
contributes to radiation protection data as shown recently.109
As discussed in detail in Sec. IV A, MC methods are a key
tool to characterize shielding effects, ISA, and other relevant
factors to clinical brachytherapy dose calculation. A thorough discussion of dose-calculation algorithms was prepared
by Carlsson Tedgren and others.80,84,98
However, the MC calculation time for direct brachytherapy applications is affected by the inverse-square law fall
off in the primary photon fluence with distance, as it leads to
a low density of primary photons at a distance. Another problem is the nearly elastic scattering at relatively low energies,
as it requires many interactions until the initial energy is lost.
Such reasons have obstructed the use of MC algorithms in
direct calculation of dose distributions around implants.
Methods have been developed to increase MC calculation
2145
speed by looking into variance reduction techniques such as

correlated sampling.110 Chibani and Williamson77 developed
MCPI, an MC code dedicated to low-energy permanent seed
implants. MCPI uses source phase-space files to avoid tracking particles within the seeds but allows full calculation of
ISA, arbitrary seed orientation and material heterogeneities.
A similar dedicated route was taken with EGSnrc that led
to BRACHYDOSE.78
Others have pursued MC integration by implementing
DICOM-RT support to generate voxel-based geometry from
clinical implants exported directly from the TPS and imported into general-purpose codes such as GEANT471 or
111
PTRAN.
This has enabled retrospective dose comparison
studies, essential when considering new dose-calculation
methods. One such study has argued strongly in favor of MC
dose calculation as a clinical standard for postimplant dose
evaluation, since time constraints are less important for
follow-up procedures and will lead to better dose-outcome
studies.65 Similarly, in HDR brachytherapy, retrospective
dose comparisons for a shielded rectal applicator have shown
large differences in delivered doses for a cohort of 40
patients.56 The success of subminute MC codes for seed implants has yet to be repeated for the higher energy 192Ir; the
longer mean free path makes the calculation time for similar
statistical uncertainty much longer. To include some of the
benefits of MC in the presence of high-Z materials but keep
calculation time clinically acceptable, a hybrid approach has
been proposed in which MC is used to calculate modified
TG-43 parameters that includes the source and the applicators skin applicators, breast brachytherapy applicators, and
COMS-based eye plaque.112 In the above examples, the importance of material heterogeneities has been demonstrated
for a wide range of brachytherapy procedures. While advanced MC software may soon be available for dose recalculation, the question of advanced methods for dose optimization has also been put forward. A method has recently been
shown using a MC dose-calculation technique for plan optimization in afterloading brachytherapy. This method involves a full MC precalculation in the treatment geometry
including heterogeneities for all possible dwell positions,
which are then used in an inverse planning software.113
Even though MC has been an integral part of many different aspect of brachytherapy, commercial implementation
into a brachytherapy TPS is currently unavailable. A number
of open questions remain to be addressed for efficient clinical integration. The first task would be to address source
geometry definition. It would be a great loss of resources for
each institution to characterize its commonly available
sources. A standard reference dataset must be made available, leaving the medical physicist to validate test cases
against published data. These will probably need to also include a standard format for applicator definitions. Similarly,
the issue of organ or tissue segmentation in terms of material
composition can be critical for low-energy seeds and electronic brachytherapy sources. While CT is the imaging standard, it only provides electron density, not average elemental
composition or mass density. Dual-energy CT could be a
2146
potential solution.114116 Still, some guidelines must be provided to the end user, otherwise large variations would result
when using MC methods.
V. PREDICTIONS AND RATIONALE FOR
BRACHYTHERAPY TPS TRENDS
The previous sections describe the past, present, and future of brachytherapy dosimetry in a narrow interpretation of
the term dosimetry as a dose-calculation procedure. We followed the history from the systems of the precomputer era,
from the table or nomogram based manual systems, via the
classical Sievert-based dose calculation Sec. I D, to the
present-day TG-43 based system Sec. II, toward the complex algorithms for primary-scatter separation superposition/
convolution models and the MC approach Sec. IV D. We
now summarize a number of activities by committees within
the AAPM and jointly with other professional societies to
help shape the future of brachytherapy dosimetry; here used
in the wider interpretation of the term. A detailed uncertainty
analysis will help identify areas where the largest benefits
can be achieved from research activities. Easily accessible
traceable source strength calibration methods, preferably in
the quantity dose to water are helpful and reassuring for inhouse check procedures of source deliveries. We must strive
for a situation where dose is interpreted unequivocally
among institutions, for the same starting points whichever
system is used. Easily accessible consensus datasets for
source characterization, both in TG-43 format and in the
more complex algorithms, are then required for commissioning TPS and for quality control. Maybe the most important
evolution in brachytherapy dosimetry is the consequence of
modern imaging technologies. Section V presents our views
on a number of these issues.
V.A. Working groups and joint international
coordination
The aim of physicists work in radiation oncology, and

thus in brachytherapy dosimetry, is to provide a safe and
reliable technical environment for patient treatments. As discussed in previous sections and elsewhere in several recent
publications,1416 there are distinct areas where significant
improvements can be made, either to improve the accuracy
of given steps in the dosimetry chain or to speed up or improve the procedures for certain techniques to benefit patients or improve cost effectiveness. This is essentially the
work of individual researchers or the aim of joint professional groups. The results of such efforts are usually published in national or international reports. Tables II and III
show the activities of the current working groups and task
groups and their chairs under the Brachytherapy Subcommittee BTSC of the AAPM Therapy Physics Committee
TPC, as well as a list of the reports that were published on
brachytherapy by the AAPM. As is clear from this table, a
tremendous effort was demonstrated especially from 1995
onward, leading to production of about one report per year.
Furthermore, other professional societies have presented
work in the same area usually in the form of national reports
2146
TABLE II. Overview of active AAPM Brachytherapy Subcommittee working

groups and task groups.
AAPM BTSC Active Working Groups
Brachytherapy Source Registry Geoffrey Ibbott
Special Brachytherapy Modalities Bruce Thomadsen
High-Energy Brachytherapy Dosimetry Jos Prez-Calatayud
Low-Energy Brachytherapy Dosimetry Michael Mitch
Robotic Brachytherapy Bruce Thomadsen and Yan Yu
AAPM BTSC Active Task Groups
TG-129 Eye Plaque Dosimetry Sou-Tung Chiu-Tsao
TG-137 Prostate Dose Prescription and Reporting Methods
Ravinder Nath
TG-138 Brachytherapy Dosimetric Uncertainties Larry DeWerd
TG-143 Dosimetric Evaluation of Elongated Brachytherapy Sources
Ali Meigooni
TG-144 Dosimetry for 90Y microsphere Brachytherapy for Liver Cancer
Andy Dezarn
TG-167 Novel and Developmental Brachytherapy Sources
Ravinder Nath
TG-186 Advanced Brachytherapy Dosimetry Luc Beaulieu
and recommendations. It is not our intention to provide an

extensive overview of these committees and reports, but examples are easily found in the report series of the German
DIN, French SFMP, Dutch/Belgian NCS, IPEM in the U.K.,
and those of ESTRO GEC-ESTRO, Braphyqs and IAEA.
Many of these reports can be found on and downloaded from
the Internet.
It is worth mentioning the current status of groups within
this framework. The high-energy brachytherapy dosimetry
HEBD working group aims to prepare and issue AAPM
recommendations on brachytherapy dosimetry parameters
and dosimetry algorithms for high-energy photon-emitting
sources. At the same time, this group works on solutions for
a number of practical problems associated with differences
compared to low-energy photon-emitting brachytherapy
sources. The low-energy brachytherapy dosimetry LEBD
working group is developing consensus datasets for new
low-energy photon-emitting brachytherapy sources such as
131
Cs and will include expansion of acceptable and benchmarked measurement and simulation techniques considering
advances in dosimetry research. LEBD works closely with
the Calibration Laboratory Accreditation Subcommittee under TPC, NIST, and the RPC to oversee dosimetric and calibration aspects. Under LEBD are TG-137, TG-138, and TG167. The report of the TG-137 on Prostate dose prescription
and reporting methods is in the final AAPM review stage,
with plans to publish in Medical Physics. The charge of TG138 is Photon brachytherapy source dosimetric uncertainty
analysis, and the report is nearing completion. This report
focuses on measurement system uncertainty components,
MC simulation uncertainty components, uncertainty in the
TG-43 dosimetry formalism parameters, and the uncertainties associated with the transfer to the clinics. If this topic
was further extended to include the uncertainties of the more
clinically related steps in the dosimetry chain, including im-
2147
TABLE III. Overview of AAPM reports on brachytherapy with report number

and year of publication.
21 1987 Radiation Therapy Committee Task Group No. 32. Specification
of brachytherapy source strength
41 1993 Remote Afterloading Technology Task Group No. 41. Remote
afterloading technology
51 1995 Radiation Therapy Committee Task Group No. 43. Dosimetry
of interstitial brachytherapy sources
59 1997 Radiation Therapy Committee Task Group No. 56. Code
of practice for brachytherapy physics
61 1998 Radiation Therapy Committee Task Group No. 59. High-dose
rate brachytherapy treatment delivery
66 1999 Radiation Therapy Committee Task Group No. 60. Intravascular
brachytherapy physics.
68 1999 Radiation Therapy Committee Task Group No. 64. Permanent
prostate seed implant brachytherapy
69 2000 Recommendations of the American Association of Physicists
in Medicine on 103Pd interstitial source calibration and dosimetry:
Implications for dose specification and prescription
84 2004 Brachytherapy Subcommittee Workgroup on low-energy
brachytherapy source dosimetry. Update of AAPM Task Group No. 43
Report: A revised AAPM protocol for brachytherapy dose calculations
84S 2007 Brachytherapy Subcommittee Workgroup on low-energy
brachytherapy source dosimetry. Supplement to the 2004 update
of the AAPM Task Group No. 43 Report
89 2005 Therapy Physics Committee Subcommittee on photon-emitting
brachytherapy dosimetry. Recommendations of the AAPM regarding
the impact of implementing the 2004 Task Group 43 report on dose
specification for 103Pd and 125I interstitial brachytherapy
98 2008 Brachytherapy Subcommittee Workgroup on low energy
brachytherapy source calibration. Report of the AAPM Low-Energy
Brachytherapy Source Calibration Working Group: Third-party
brachytherapy source calibrations and physicist responsibilities
aging and reconstruction procedures and registration of

images,117,118 the contouring of target organs, and tissues at
risk and the analysis of dosimetric parameters, this would
lead to a rather complete overview of all uncertainties in
brachytherapy dosimetry. In this context, work is ongoing by
the GEC-ESTRO Group Europen de Curiethrapie Braphyqs physics network on quality of imaging procedures,119121 on the accuracy of present day dose volume
histogram calculation algorithms,122 and the initiation of a
new project on uncertainties in seed definition and contouring due to interobserver variability in prostate postplanning
by De Brabandere. These topics are complementary to the
TG-138 work. Thus, the time is ready for a joint effort to
delineate high-priority topics for further basic or clinical research. The joint AAPM/RPC Brachytherapy Source Registry is overseen by the Brachytherapy Source Registry working group, and endeavors to expand the current Registry to
include high-energy photon-emitting sources. The Robotic
Brachytherapy working group was established in late 2008
and is underway to gather data and recommend robotic
brachytherapy standards. The Special Brachytherapy Modalities working group oversees TG-129, TG-143, and TG-144.
Its TG-149 on intravascular brachytherapy was published by
Medical Physics in 2007. In 2009, another task group TG186 was established directly under BTSC to build on the
expertise of international scientists for a report on ModelMedical Physics, Vol. 36, No. 6, June 2009
2147
based dose-calculation techniques in brachytherapy: Status

and clinical requirements for implementation beyond AAPM
TG-43. The aim is to review the potential next generation
brachytherapy dose-calculation algorithms, and to provide
guidance for early adopters in using model-based dosecalculation algorithms in brachytherapy, especially for commissioning, handling of patient data, and prescription definition.
V.B. Dose-calculation engines for clinical routine
To be useful in the clinic, MC TPS may come with predefined CAD based applicator geometries and physical properties. We expect that source modeling and its associated
phase-space file will be mandatory as per a new AAPM prerequisite. These might be specific to the type of MC code
implemented by the commercial vendor, but would still have
to meet stringent validation, akin to the AAPM brachytherapy dosimetry prerequisites. As we expect most of the
applicators, catheters, and needles to become MRI/CT compatible, artifacts will become less of an issue. However, image processing will be implemented to minimize streaking of
permanent seed implants.
It is expected that the next generation algorithms will be
able to deal with a number of the problems related to the
accuracy of dose calculation discussed in Sec. III. The influence of tissue density and tissue composition effects on the
resulting dose distribution was discussed in detail in the recent AAPM Anniversary Paper on Brachytherapy Physics by
Thomadsen et al.16 While the reader is referred to that document, we summarize that a direct MC calculation approach
can address the substantial dose effects of several percent
possible with tissue composition changes such as soft tissue
calcifications,61 adipose, and fatlike tissues123 depending on
the photon energy. The challenge remains in the design of
imaging technology to allow advanced brachytherapy TPS to
account for the relevant spatial and density information. Applicator and catheter shielding effects have similar influence
on the dose deposition.124,125 Further, those involved with
TPS algorithm development would benefit if vendors provide
a full description of their applicators. This is one step beyond
the tools currently available for full applicator geometry description for reconstruction, viewing, and manipulation in
3D. An open database to share the information on material
composition in addition to a series of systematic MC research studies on the influence of a wide spectrum of photon
radiation on dose is welcomed. Brachytherapy TPS development should be less vendor specific and instead provide compatibility with afterloading equipment from other vendors.
Standardization efforts for brachytherapy DICOM-BT by
the cross-societal Integrating the Healthcare Environment for
Radiation Oncology IHE-RO taskforce are underway.
There are more problems that can be addressed in the
coming years. One is a systematic and comprehensive study
on the influence of secondary photon radiation on dose from
beta components e.g., bremsstrahlung of radionuclides used
or proposed for brachytherapy. Often the beta particles themselves or low-energy photons are absorbed in the source ma-
2148
terial or source capsule or have a penetration of 12 mm.

However, if high-energy electrons are emitted and sources
have high-Z capsule materials, penetration of the secondary
radiation may be significant. MC studies are ideal for such
work. Discrepancies between the outcome of the codes used
for MC calculations can sometimes be attributed to the use
of different cross-section libraries.126 Differences in the use
of certain databases for energy spectra of the radionuclide
under investigation might be a cause for deviation.42 As mentioned in Sec. II B, source characterization databases are useful for further development of direct MC or MC-based algorithms. Similar to the consensus datasets, clear
recommendations on which data to be used for such work
would be useful. The professional organizations might take
this task to implement a web-based resource for sharing information.
Unmentioned in the previous sections, brachytherapy
dose optimization is a critical tool. This is due to the flexibility of single HDR source stepping technology. It is often
not recognized by the user that for a set of restrictions or
requirements, multiple solutions are possible. Inverse planning has been shown to be superior to any form of manual
planning in term of simultaneous fulfillment of CTV coverage, organs at risk, and dose sparing while providing better
dose conformity and homogeneity. More importantly, it
handles increased complexity without effort and the reproducibility from patient to patient becomes user independent.
Just like for IMRT, dose optimization will become the standard for brachytherapy treatment planning for many anatomic sites. Still, there is an adagium that tells us that optimization cannot make a good dose distribution out of a bad
implant. The larger contiguous high-dose volumes e.g.,
V200 that should be avoided are evident after optimization of
bad implants,127 but these are often not used for optimization
restriction. Furthermore, there is a wide variety of parameters
proposed to help evaluate implantsgenerally based on the
outcome of DVH analysis. Still, their value for predicting
clinical outcomes has not been validated, and clinical studies
to demonstrate efficacy of certain approaches need to be addressed in the next decade.
Use of radiobiology in external beam TPS is only recently
being considered, and brachytherapy would also benefit from
planning beyond physical dose. However, many basic questions and reference data still need to be addressed. Unlike for
external beam today, radiobiology is faced with an interesting challenge in brachytherapy. For the prostate example, the
duration of the dose delivery varies markedly from very lowdose rates as with 125I seed implants to hypofractionated
HDR monotherapy schedules. A better understanding of normal tissue and target tissue response to these dose rates is
required. Essentially this process, tissue recognition and
characterization in terms of / ratios, is completely analogous to external beam planning and there is no reason to
assume that it will not be implemented for brachytherapy.
However, patient-specific radiobiological information is
lacking in a general sense, and further investigation is
needed.
2148
All modern TPSs are built in a modular fashion. Software

components are developed at specialized centers, often university groups, and are combined into the main system. The
TPS vendor is then responsible for component integration
and the user interface. This process will continue as a form
of outsourcing research and development. Consequently, the
splitting of external beam and brachytherapy TPS components over the past two decades may be undone by some of
the vendors, making possible dual-modality dose summation
within a 3D patient matrix.
V.C. Dose-to-water calibration
A dose-to-water calibration standard for source strength

determination appears to be the next goal of the PSDLs and
secondary standard dosimetry laboratories SSDLs in their
support to the brachytherapy community. Just as with external beam therapy, the brachytherapy dose calculation system
is specific to a single point.128,129 For example, if the refer r , NIST as determined at
ence dose rate to liquid water D
0 0 water
NIST was to replace SK in Eq. 5 for source strength as
determined at an accredited dosimetry calibration laboratory
r , NIST / M C C A , then
ADCL or SSDL ND,W = D
0 0 water
el
el
ion
the reference dose rate to liquid water as determined in a
r , clinic for use in a new 2D formalism could
clinic D
0 0 water
follow,
r , clinic = N
D
0 0 water
D,W M Cel CTP Pion
with the dose rate at any point determined as

r , clinic GLr, g r Fr, ,
r, = D
D
0 0 water
L
GLr0, 0
where ND,W is the ionization chamber calibration as used

currently, M is the chamber reading, Cel and CTP are the
electrometer and temperature/pressure correction factors, respectively, and Aion and Pion are the chamber collection efficiency and recombination factors.
At the European laboratories, there is noticeably a renewed interest in these questions for brachytherapy and it is
accepted as an explicit part of a new research activity in the
EURAMET network of national measurement institutes
NMIs. The iMERA joint research project on brachytherapy
aims at developing a primary standard of absorbed dose to
water for 192Ir to be available at three NMIs, and similarly
for 125I available at four NMIs in about 3 years. While the
National Physical Laboratory NPL in the United Kingdom
offers 125I calibrations for only a specific source model
6711 and in an ad hoc manner, the Physikalisch-Technische
Bundesanstalt PTB in Germany provides a similar calibration service which is not limited to only one source type and
accepted by EURAMET. Implicitly this means that the previous lack of European PSDLs offering calibration of lowand intermediate-energy photon-emitting sources appears to
be solved. PTB has performed ad hoc intercomparisons of
125
I and 103Pd seeds with the University of Wisconsin
ADCL.130 Regardless of how this advance is further implemented, there is still much work to be done before direct
2149
chamber calibrations for liquid water dose rate are available

to clinical users. We believe it is the responsibility of the
international PSDL community to coordinate efforts to uniformly transition to develop dose-to-water standards worldwide and to disseminate these standards to the clinical source
users. Coordination for this transition from reference air
kerma to dose to water across the entire field of brachytherapy is key. A distinct requirement from the clinical user
is that such a new approach should at least give the same, but
preferably, lower source calibration uncertainties. Further cooperation with NIST and other PSDLs is essential for mutual
validation and to harmonize methodologies. Within the next
few years, PSDL intercomparisons will be organized by the
BIPM and will include 192Ir source calibrations. There is no
plan yet for a similar intercomparison for 125I seeds. The
success of such collaboration between laboratories is clearly
demonstrated in the review paper of Williamson14 in which
development of modern quantitative approaches to brachytherapy dosimetry at NIST is linked with clinical utilization
of low-energy seeds.
It is expected that the AAPM will continue to play a major role in the further development of both the calibration
standards and the recommendations of the dose-calculation
algorithm datasets. Research groups working with or in contact with the vendors and the standards laboratories should
be included in the working groups and task groups in order
to avoid any disparities to occur either from new consensus
algorithm types or the dose-to-water concept for calibration
that might otherwise distract and confuse individual users.
The importance of a clinical implementation plan cannot be
underestimated. It will require an international organizational
leadership to bring together the treatment planning vendors,
PSDLs, source manufacturers, and medical physics societies,
possibly on an annual basis.
V.D. Developments in brachytherapy imaging
Use of x-ray CT and other 3D imaging modalities MR,

US, and more recently PET marks a distinct improvement in
brachytherapy planning, to be considered as a departure from
the surgical practice of brachytherapy. As with many other
developments, the ideas are not new as demonstrated by
early publications of, for example by Herskovic et al.,131
Pierquin and Fayos,132 Ling et al.,133 and Schoeppel et al.134
The step forward is made primarily due to direct availability
of the radiological equipment for interventional procedures.
Low-cost ultrasound forms the basis of prostate implant
technology, but the far more expensive other modalities are
nowadays available in every modest-sized hospital, whereas
in many radiotherapy departments 3D imaging has replaced
the conventional RT simulator for patient setup. CTV localization and dose-limiting normal tissues allows the absorbed
dose from the implanted sources to be specified according to
anatomy-based parameters, such as the coverage index. In
the classical systems, dosimetry quantities for analysis of
implant quality were difficult to correlate with clinical outcome as they lacked 3D volumetric information of the relevant structures. As discussed by Feuvret et al.,135 this led to
2149
proposed indices for which the evidence of the predictive

value is often poor or lacking. TRUS-guided permanent136,137
and HDR temporary138,139 implantation have become standard practice for prostate brachytherapy in North America
and Europe. For other body sites, especially gynecology, 3D
imaging is now or will be soon included in the newest recommendations as a standard guidance for volume definition
and dose prescription,140,141 and will lead to new clinical
studies for a systematic evaluation of the responses.142 As a
spin-off of these developments, the interaction between radiation oncologists, medical physicists, and the equipment
vendors may lead to a new generation of applicators in
which a combination of intracavitary and interstitial techniques is pursued to increase the flexibility for source position optimization in the implant geometry. As a consequence,
ICRU reports such as Report 38 from 1985 on gynecology
and Report 58 from 1997 on interstitial implants will need to
be updated.143,144
Although MRI is slowly entering the field, it is widely
recognized that this technique readily adds to the quality of
the treatments for its ability to discriminate between different
tissue types. Contouring will increasingly be based on this
imaging technology. Functional MRI depends on its ability
to demonstrate active parts of the tissue under consideration,
allowing suborgan identification. If tissue contouring can be
performed in this way, this would allow a spatially modulated dose deposition dose painting) procedure, in which the
full organ e.g., a prostate is treated to the required dose but
subparts are administered a much higher dose.145,146 There is
a similar expectation for the use of PET and PET-CT imaging for identifying active tumor areas.147,148
One of the challenges in brachytherapy imaging is the
further development of registration for different modalities
and image sets made at different points in time. A key to this
challenge is the need to address deformable image registration. In brachytherapy, not only organ movements or the increase or decrease of tumor volume, but also the required
transportation of a patient from one location to the other
e.g., operating area to radiology department may influence
the relative position of the applicators and the organs. Not
only the tools themselves but also QA of these imaging procedures adapted to the needs of radiation oncology. New QA
protocols are required specific to the clinical applications in
brachytherapy.
VI. CONCLUSION
It is a great challenge to introduce these new techniques
into medical practice and in cooperation with physicians to
help design prospective clinical trials to demonstrate their
values and efficacy. It is expected that as the new-generation
algorithms find their way to early adopters, retrospective dosimetric studies will be conducted to revisit prescription
level, target dose, OAR doses, and their relationship to outcomes, either survival or toxicities. Improved accuracy of
brachytherapy dose calculations, and flexible coregistration
of external beam and brachytherapy dose distributions, will
enable true dose summation of external beam with LDR
2150
and/or HDR brachytherapy e.g., a salvage prostate HDR

brachytherapy after a seed implant. These advances will
also allow researchers to explore implants containing multiple radionuclides based on a solid foundation that will take
into account varying RBE. Joint databases for MC-based algorithms will form the backbone of the new generation of
commercial treatment planning systems, in combination with
a new validation prerequisite for those new algorithms. A
dose-to-water standard for brachytherapy source calibration
will become available with a better or at least similar uncertainty level. It is expected that above all, if anything will
make a change in clinical outcome to the benefit of our patients, it is the wide introduction of modern imaging tools
into the clinical practice.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the careful and detailed comments from the referees. The comments and suggestions of Tony Aalbers, Todd Wareing, Marco Zaider,
Michael Mitch, Larry DeWerd, Hans-Joachim Selbach,
Christopher Melhus, Ali Meigooni, and Panos Papagiannis
are also greatly appreciated. The authors have not received
any financial support in relation to the preparation of this
article.
a

J. Dutreix, M. Tubiana, and B. Pierquin, The hazy dawn of brachytherapy, Radiother. Oncol. 49, 223232 1998.
2
H. Becquerel and P. Curie, Action physiologique des rayons du radium,
Acad. Sci., Paris, C.R. 132, 12891291 1901.
3
H. Danlos and P. Bloch, Note sur le traitement du lupus rythmateux
par des applications de radium, Ann. Dermatol. Syphil. Paris 2, 986
988 1901.
4
R. Abb, Notes on the physiologic and therapeutic action of radium,
Wash. Med. Ann. 2, 363377 1904.
5
L. Wickham and P. Degrais, Radiumthrapie Baillire, Paris, 1909.
6
R. F. Mould, A Century of X-Rays and Radioactivity in Medicine Institute of Physics, Bristol, 1993.
7
R. Paterson and H. M. Parker, A dosage system for interstitial radium
therapy, Br. J. Radiol. 1, 252340 1938.
8
W. J. Meredith, Radium Dosage: The Manchester System Livingston,
Edinburgh, 1967.
9
U. Henschke, B. S. Hilaris, and G. D. Mahan, Afterloading in interstitial
and intracavitary radiation therapy, Am. J. Roentgenol., Radium Ther.
Nucl. Med. 90, 386395 1963.
10
D. Chassagne, M. Raynal, and B. Pierquin, Technique of endocurietherapy by iridium 192 with plastic tubes in the breast tumors, J. Radiol., Electrol., Med. Nucl. 44, 269271 1963.
11
B. Pierquin, D. J. Chassagne, C. M. Chahbazian, and J. F. Wilson,
Brachytherapy W.H. Green, St Louis, MO, 1978.
12
R. Walstam, Remotely-controlled afterloading radiotherapy apparatus,
Phys. Med. Biol. 7, 225228 1962.
13
U. K. Henschke, B. S. Hilaris, and G. D. Mahan, Intracavitary radiation
therapy of cancer of the uterine cervix by remote afterloading with cycling sources, Am. J. Roentgenol., Radium Ther. Nucl. Med. 96, 4551
1966.
14
J. F. Williamson, Brachytherapy technology and physics practice since
1950: A half-century of progress, Phys. Med. Biol. 51, R303R325
2006.
15
G. S. Ibbott, C. M. Ma, D. W. O. Rogers, S. M. Seltzer, and J. F. Williamson, Anniversary paper: Fifty years of AAPM involvement in radiation dosimetry, Med. Phys. 35, 14181427 2008.
16
B. R. Thomadsen, J. F. Williamson, M. J. Rivard, and A. S. Meigooni,
Anniversary paper: Past and current issues, and trends in brachytherapy
physics, Med. Phys. 35, 47084723 2008.
17
E. H. Quimby and V. Castro, The calculation of dosage in interstitial
1
2150
radium therapy, Am. J. Roentgenol., Radium Ther. Nucl. Med. 70, 739
740 1953.
18
W. J. Meredith, Radiotherapeutic physics: V. Dosage control in interstitial gamma-ray therapy, Br. J. Radiol. 24, 380384 1951.
19
R. J. Shalek and M. A. Stovall, in Radiation Dosimetry, edited by F. H.
Attix and E. Tochlin Academic, New York, 1969, Vol. 3, pp. 776798.
20
A. Dutreix, G. Marinello, and A. Wambersie, Dosimtrie en Curiethrapie Masson, Paris, 1982.
21
S. Jayaraman, L. H. Lanzl, and S. K. Agarwal, An overview of errors in
line source dosimetry for gamma-ray brachytherapy, Med. Phys. 10,
871875 1983.
22
IAEA Report No. 17 IAEA, Vienna, 2000.
23
D. Baltas, L. Sakelliou, and N. Zamboglou, The Physics of Modern
Brachytherapy for Oncology Taylor & Francis, New York, 2007.
24
M. J. Rivard, B. M. Coursey, L. A. DeWerd, W. F. Hanson, M. S. Huq, G.
S. Ibbott, M. G. Mitch, R. Nath, and J. F. Williamson, Update of AAPM
Task Group No. 43 Report: A revised AAPM protocol for brachytherapy
dose calculations, Med. Phys. 31, 633674 2004.
25
J. F. Williamson, Recent developments in brachytherapy physics, Proceedings of the ABS 1994 Summer School Medical Physics, Madison,
WI, 1995.
26
J. F. Williamson, Monte Carlo and analytic calculation of absorbed dose
near 137Cs intracavitary sources, Int. J. Radiat. Oncol. Biol. Phys. 15,
227237 1988.
27
J. F. Williamson, The Sievert integral revisited: evaluation and extension
to 125I, 169Yb, and 192Ir brachytherapy sources, Int. J. Radiat. Oncol.,
Biol., Phys. 36, 12391250 1996.
28
J. F. Williamson, Monte Carlo-based dose-rate tables for the Amersham
CDCS.J and 3M model 6500 137Cs tubes, Int. J. Radiat. Oncol. Biol.
Phys. 41, 959970 1998.
29
R. van der Laarse, presented at Proceedings of the Seventh Intern Conference on the Use of Computers in Radiation Therapy, Japan, 1980 unpublished.
30
R. Nath, L. L. Anderson, G. Luxton, K. A. Weaver, J. F. Williamson, and
A. S. Meigooni, Dosimetry of interstitial brachytherapy sources: Recommendations of the AAPM Radiation Therapy Committee Task Group No.
43. American Association of Physicists in Medicine, Med. Phys. 22,
209234 1995.
31
R. Paterson, A dosage system for gamma ray therapy. Part I. Clinical
aspects, Br. J. Radiol. 7, 592612 1943.
32
H. Parker, A dosage system for gamma ray therapy. Part II. Physical
aspects, Br. J. Radiol. 7, 612632 1943.
33
B. Pierquin and G. Marinello, A Practical Manual of Brachytherapy
Medical Physics, Madison, WI, 1997.
34
W. M. Butler, W. S. Bice, Jr., L. A. DeWerd, J. M. Hevezi, M. S. Huq, G.
S. Ibbott, J. R. Palta, M. J. Rivard, J. P. Seuntjens, and B. R. Thomadsen,
Third-party brachytherapy source calibrations and physicist responsibilities: Report of the AAPM Low Energy Brachytherapy Source Calibration
Working Group, Med. Phys. 35, 38603865 2008.
35
S. M. Seltzer, P. J. Lamperti, R. Loevinger, M. G. Mitch, J. T. Weaver,
and B. M. Coursey, New national air-kerma-strength standards for 125I
and 103Pd brachytherapy seeds, J. Res. Natl. Inst. Stand. Technol. 108,
337358 2003.
36
M. J. Rivard, W. M. Butler, L. A. DeWerd, M. S. Huq, G. S. Ibbott, A. S.
Meigooni, C. S. Melhus, M. G. Mitch, R. Nath, and J. F. Williamson,
Supplement to the 2004 update of the AAPM Task Group No. 43 Report, Med. Phys. 34, 21872205 2007.
37
M. J. Rivard, C. S. Melhus, and J. F. Williamson, Brachytherapy dose
calculation formalism, dataset evaluation, and treatment planning system
implementation, in Clinical Dosimetry for Radiotherapy: AAPM Summer School, edited by D. W. O. Rogers and J. E. Cygler Medical Physics,
Madison, WI, 2009.
38
L. A. DeWerd, M. S. Huq, I. J. Das, G. S. Ibbott, W. F. Hanson, T. W.
Slowey, J. F. Williamson, and B. M. Coursey, Procedures for establishing and maintaining consistent air-kerma strength standards for lowenergy, photon-emitting brachytherapy sources: Recommendations of the
Calibration Laboratory Accreditation Subcommittee of the American Association of Physicists in Medicine, Med. Phys. 31, 675681 2004.
39
Z. Li, R. K. Das, L. A. DeWerd, G. S. Ibbott, R. W. Kline, A. S. Meigooni, J. Perez-Calatayud, M. J. Rivard, R. S. Sloboda, and J. F. Williamson, Dosimetric prerequisites for routine clinical use of photon emitting
brachytherapy sources with average energy higher than 50 keV, Med.
Phys. 34, 3740 2007.
2151
40
ICRU Report No. 44, 1989.

D. E. Cullen, J. H. Hubbell, and L. Kissel, Lawrence Livermore National
Laboratory Report No. UCRL-50400, Vol. 6, Rev. 5, 1997.
42
National Nuclear Data Center, NUDAT 2.5, Brookhaven National
Laboratory, 2009; http://www.nndc.bnl.gov/nudat2/reCenter.jsp?z77&n
115 last accessed 3 May 2009.
43
C. S. Melhus and M. J. Rivard, Approaches to calculating AAPM TG-43
brachytherapy dosimetry parameters for 137Cs , 125I, 192Ir, 103Pd, and
169
Yb sources, Med. Phys. 33, 17291737 2006.
44
J. Markman, J. F. Williamson, J. F. Dempsey, and D. A. Low, On the
validity of the superposition principle in dose calculations for intracavitary implants with shielded vaginal colpostats, Med. Phys. 28, 147155
2001.
45
J. Perez-Calatayud, D. Granero, and F. Ballester, Phantom size in
brachytherapy source dosimetric studies, Med. Phys. 31, 20752081
2004.
46
D. Granero, J. Perez-Calatayud, M. C. Pujades-Claumarchirant, F. Ballester, C. S. Melhus, and M. J. Rivard, Equivalent phantom sizes and
shapes for brachytherapy dosimetric studies of 192Ir and 137Cs, Med.
Phys. 35, 48724877 2008.
47
T. Mauceri, P. Biggs, J. Beatty, and D. Flynn, Scatter and attenuation
measurements at distances greater than 10 cm from an 192Ir source, Med.
Phys. 26, 9799 1999.
48
J. F. Williamson, Comparison of measured and calculated dose rates in
water near I-125 and Ir-192 seeds, Med. Phys. 18, 776786 1991.
49
L. Sakelliou, K. Sakellariou, K. Sarigiannis, A. Angelopoulos, A. Perris,
and G. Zarris, Dose rate distributions around 60Co , 137Cs, 198Au, 192Ir,
241
Am, 125I models 6702 and 6711 brachytherapy sources and the nuclide 99mTc, Phys. Med. Biol. 37, 18591872 1992.
50
E. Pantelis, P. Papagiannis, P. Karaiskos, A. Angelopoulos, G. Anagnostopoulos, D. Baltas, N. Zamboglou, and L. Sakelliou, The effect of finite
patient dimensions and tissue inhomogeneities on dosimetry planning of
192
Ir HDR breast brachytherapy: A Monte Carlo dose verification study,
51
G. Lymperopoulou, E. Pantelis, P. Papagiannis, H. Rozaki-Mavrouli, L.
Sakelliou, D. Baltas, and P. Karaiskos, A Monte Carlo dosimetry study
of vaginal 192Ir brachytherapy applications with a shielded cylindrical
applicator set, Med. Phys. 31, 30803086 2004.
52
R. E. Taylor and D. W. Rogers, EGSnrc Monte Carlo calculated dosimetry parameters for 192Ir and 169Yb brachytherapy sources, Med. Phys.
35, 49334944 2008.
53
W. C. Roesch, Evaluation of high-energy brachytherapy source electronic disequilibrium and dose from emitted electrons, Radiat. Res. 9,
399410 1958.
54
D. Baltas, P. Karaiskos, P. Papagiannis, L. Sakelliou, E. Loeffler, and N.
Zamboglou, Beta versus gamma dosimetry close to Ir-192 brachytherapy
sources, Med. Phys. 28, 18751882 2001.
55
F. Ballester, D. Granero, J. Perez-Calatayud, C. S. Melhus, and M. J.
Rivard, Evaluation of electronic equilibrium and electron contribution to
dose near brachytherapy sources, Med. Phys. submitted.
56
E. Poon, J. F. Williamson, T. Vuong, and F. Verhaegen, Patient-specific
Monte Carlo dose calculations for high-dose-rate endorectal brachytherapy with shielded intracavitary applicator, Int. J. Radiat. Oncol.,
Biol., Phys. 72, 12591266 2008.
57
C. S. Sureka, P. Aruna, S. Ganesan, C. S. Sunny, and K. V. Subbaiah,
Computation of relative dose distribution and effective transmission
around a shielded vaginal cylinder with 192Ir HDR source using
MCNP4B, Med. Phys. 33, 15521561 2006.
58
Y. Yu, L. L. Anderson, Z. Li, D. E. Mellenberg, R. Nath, M. C. Schell, F.
M. Waterman, A. Wu, and J. C. Blasko, Permanent prostate seed implant
brachytherapy: Report of the American Association of Physicists in Medicine Task Group No. 64, Med. Phys. 26, 20542076 1999.
59
A. S. Meigooni, J. A. Meli, and R. Nath, Interseed effects on dose for
125
I brachytherapy implants, Med. Phys. 19, 385390 1992.
60
P. Mobit and I. Badragan, Dose perturbation effects in prostate seed
implant brachytherapy with I-125, Phys. Med. Biol. 49, 31713178
2004.
61
O. Chibani, J. F. Williamson, and D. Todor, Dosimetric effects of seed
anisotropy and interseed attenuation for 103Pd and 125I prostate implants,
Med. Phys. 32, 25572566 2005.
62
J. F. Carrier, L. Beaulieu, F. Therriault-Proulx, and R. Roy, Impact of
interseed attenuation and tissue composition for permanent prostate implants, Med. Phys. 33, 595604 2006.
41
63
2151
R. G. Stock, N. N. Stone, A. Tabert, C. Iannuzzi, and J. K. DeWyngaert,

A dose-response study for I-125 prostate implants, Int. J. Radiat. Oncol., Biol., Phys. 41, 101108 1998.
64
H. Afsharpour, M. DAmours, B. Cote, J. F. Carrier, F. Verhaegen, and L.
Beaulieu, A Monte Carlo study on the effect of seed design on the
interseed attenuation in permanent prostate implants, Med. Phys. 35,
36713681 2008.
65
J. F. Carrier, M. DAmours, F. Verhaegen, B. Reniers, A. G. Martin, E.
Vigneault, and L. Beaulieu, Postimplant dosimetry using a Monte Carlo
dose calculation engine: a new clinical standard, Int. J. Radiat. Oncol.,
Biol., Phys. 68, 11901198 2007.
66
B. Reniers, F. Verhaegen, and S. Vynckier, The radial dose function of
low-energy brachytherapy seeds in different solid phantoms: Comparison
between calculations with the EGSnrc and MCNP4C Monte Carlo codes
and measurements, Phys. Med. Biol. 49, 15691582 2004.
67
C. S. Melhus and M. J. Rivard, COMS eye plaque brachytherapy dosimetry simulations for 103Pd, 125I, and 131Cs, Med. Phys. 35, 33643371
2008.
68
R. M. Thomson, R. E. P. Taylor, and D. W. O. Rogers, Monte Carlo
dosimetry for 125I and 103Pd eye plaque brachytherapy, Med. Phys. 35,
55305543 2008.
69
X-5 Monte Carlo Team, MCNPA General Monte Carlo N-Particle
Transport Code, Version 5 Los Alamos National Laboratory, Los Alamos, 2003.
70
L. S. Waters, G. W. McKinney, J. W. Durkee, M. L. Fensin, J. S. Hendricks, M. R. James, R. C. Johns, and D. B. Pelowitz, The MCNPX
Monte Carlo radiation transport code, AIP Conf. Proc., 896, 8190
2007.
71
S. Agostinelli, J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. Arce,
M. Asai, D. Axen, S. Banerjee, G. Barrand, F. Behner, L. Bellagamba, J.
Boudreau, L. Broglia, A. Brunengo, H. Burkhardt, S. Chauvie, J. Chuma,
R. Chytracek, G. Cooperman, G. Cosmo, P. Degtyarenko, A. DellAcqua,
G. Depaola, D. Dietrich, R. Enami, A. Feliciello, C. Ferguson, H. Fesefeldt, G. Folger, F. Foppiano, A. Forti, S. Garelli, S. Giani, R. Giannitrapani, D. Gibin, J. J. Gomez Cadenas, I. Gonzalez, G. Gracia Abril, G.
Greeniaus, W. Greiner, V. Grichine, A. Grossheim, S. Guatelli, P. Gumplinger, R. Hamatsu, K. Hashimoto, H. Hasui, A. Heikkinen, A. Howard,
V. Ivanchenko, A. Johnson, F. W. Jones, J. Kallenbach, N. Kanaya, M.
Kawabata, Y. Kawabata, M. Kawaguti, S. Kelner, P. Kent, A. Kimura, T.
Kodama, R. Kokoulin, M. Kossov, H. Kurashige, E. Lamanna, T. Lampon, V. Lara, V. Lefebure, F. Lei, M. Liendl, W. Lockman, F. Longo, S.
Magni, M. Maire, E. Medernach, K. Minamimoto, P. Mora de Freitas, Y.
Morita, K. Murakami, M. Nagamatu, R. Nartallo, P. Nieminen, T. Nishimura, K. Ohtsubo, M. Okamura, S. ONeale, Y. Oohata, K. Paech, J.
Perl, A. Pfeiffer, M. G. Pia, F. Ranjard, A. Rybin, S. Sadilov, E. Di Salvo,
G. Santin, T. Sasaki, N. Savvas, Y. Sawada, S. Scherer, S. Sei, V. Sirotenko, D. Smith, N. Starkov, H. Stoecker, J. Sulkimo, M. Takahata, S.
Tanaka, E. Tcherniaev, E. Safai Tehrani, M. Tropeano, P. Truscott, H.
Uno, L. Urban, P. Urban, M. Verderi, A. Walkden, W. Wander, H. Weber,
J. P. Wellisch, T. Wenaus, D. C. Williams, D. Wright, T. Yamada, H.
Yoshida, and D. Zschiesche, G4a simulation toolkit, Nucl. Instrum.
Methods Phys. Res. A 506, 250303 2003.
72
J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. A. Dubois, M. Asai,
G. Barrand, R. Capra, S. Chauvie, R. Chytracek, G. A. P. Cirrone, G.
Cooperman, G. Cosmo, G. Cuttone, G. G. Daquino, M. Donszelmann, M.
Dressel, G. Folger, F. Foppiano, J. Generowicz, V. Grichine, S. Guatelli,
P. Gumplinger, A. Heikkinen, I. Hrivnacova, A. Howard, S. Incerti, V.
Ivanchenko, T. Johnson, F. Jones, T. Koi, R. Kokoulin, M. Kossov, H.
Kurashige, V. Lara, S. Larsson, F. Lei, O. Link, F. Longo, M. Maire, A.
Mantero, B. Mascialino, I. McLaren, P. M. Lorenzo, K. Minamimoto, K.
Murakami, P. Nieminen, L. Pandola, S. Parlati, L. Peralta, J. Perl, A.
Pfeiffer, M. G. Pia, A. Ribon, P. Rodrigues, G. Russo, S. Sadilov, G.
Santin, T. Sasaki, D. Smith, N. Starkov, S. Tanaka, E. Tcherniaev, B.
Tome, A. Trindade, P. Truscott, L. Urban, M. Verderi, A. Walkden, J. P.
Wellisch, D. C. Williams, D. Wright, and H. Yoshida, Geant4 developments and applications, IEEE Trans. Nucl. Sci. 53, 270278 2006.
73
W. R. Nelson, H. Hirayama, and D. W. O. Rogers, Stanford Linear Accelerator Report No. SLAC-265, 1985.
74
I. Kawrakow and D. W. O. Rogers, NRCC Report No. PIRS-701, 2006.
75
J. F. Williamson, Monte Carlo evaluation of kerma at a point for photon
transport problems, Med. Phys. 14, 567576 1987.
76
Z. Li and J. F. Williamson, Volume-based geometric modeling for radiation transport calculations, Med. Phys. 19, 667677 1992.
2152
77
O. Chibani and J. F. Williamson, MCPI: A sub-minute Monte Carlo dose

calculation engine for prostate implants, Med. Phys. 32, 36883698
2005.
78
R. E. Taylor, G. Yegin, and D. W. Rogers, Benchmarking brachydose:
Voxel based EGSnrc Monte Carlo calculations of TG-43 dosimetry parameters, Med. Phys. 34, 445457 2007.
79
A. Ahnesjo, Collapsed cone convolution of radiant energy for photon
dose calculation in heterogeneous media, Med. Phys. 16, 577592
1989.
80
A. K. Carlsson and A. Ahnesjo, The collapsed cone superposition algorithm applied to scatter dose calculations in brachytherapy, Med. Phys.
27, 23202332 2000.
81
K. A. Gifford, J. L. Horton, Jr., T. A. Wareing, G. Failla, and F. Mourtada,
Comparison of a finite-element multigroup discrete-ordinates code with
Monte Carlo for radiotherapy calculations, Phys. Med. Biol. 51, 2253
2265 2006.
82
O. N. Vassiliev, T. A. Wareing, I. M. Davis, J. McGhee, D. Barnett, J. L.
Horton, K. Gifford, G. Failla, U. Titt, and F. Mourtada, Feasibility of a
multigroup deterministic solution method for three-dimensional radiotherapy dose calculations, Int. J. Radiat. Oncol., Biol., Phys. 72, 220
227 2008.
83
A. Ahnesjo and M. M. Aspradakis, Dose calculations for external photon
beams in radiotherapy, Phys. Med. Biol. 44, R99R155 1999.
84
D. W. O. Rogers, Fifty years of Monte Carlo simulations for medical
physics, Phys. Med. Biol. 51, R287R301 2006.
85
H. E. Johns and J. R. Cunningham, The Physics of Radiology, 4th ed.
Charles C Thomas, Springfield, IL, 1983.
86
F. M. Khan, The Physics of Radiation Therapy, 3rd ed. Lippincott Williams & Wilkins, 2003.
87
R. Mohan, C. Chui, and L. Lidofsky, Differential pencil beam dose
computation model for photons, Med. Phys. 13, 6473 1986.
88
T. R. Mackie, J. W. Scrimger, and J. J. Battista, A convolution method of
calculating dose for 15-MV x rays, Med. Phys. 12, 188196 1985.
89
A. Boyer and E. Mok, A photon dose distribution model employing
convolution calculations, Med. Phys. 12, 169177 1985.
90
A. Carlsson Tedgren and A. Ahnesjo, Optimization of the computational
efficiency of a 3D, collapsed cone dose calculation algorithm for brachytherapy, Med. Phys. 35, 16111618 2008.
91
K. R. Russell and A. Ahnesjo, Dose calculation in brachytherapy for a
192
Ir source using a primary and scatter dose separation technique, Phys.
Med. Biol. 41, 10071024 1996.
92
J. F. Williamson, R. S. Baker, and Z. F. Li, A convolution algorithm for
brachytherapy dose computations in heterogeneous geometries, Med.
Phys. 18, 12561265 1991.
93
J. F. Williamson, Z. Li, and J. W. Wong, One-dimensional scattersubtraction method for brachytherapy dose calculation near bounded heterogeneities, Med. Phys. 20, 233244 1993.
94
A. S. Kirov and J. F. Williamson, Two-dimensional scatter integration
method for brachytherapy dose calculations in 3D geometry, Phys. Med.
Biol. 42, 21192135 1997.
95
G. M. Daskalov, A. S. Kirov, and J. F. Williamson, Analytical approach
to heterogeneity correction factor calculation for brachytherapy, Med.
Phys. 25, 722735 1998.
96
A. K. Carlsson and A. Ahnesjo, Point kernels and superposition methods
for scatter dose calculations in brachytherapy, Phys. Med. Biol. 45, 357
382 2000.
97
A. K. Carlsson Tedgren and A. Ahnesjo, Accounting for high Z shields
in brachytherapy using collapsed cone superposition for scatter dose calculation, Med. Phys. 30, 22062217 2003.
98
K. R. Russell, A. K. Carlsson Tedgren, and A. Ahnesjo, Brachytherapy
source characterization for improved dose calculations using primary and
scatter dose separation, Med. Phys. 32, 27392752 2005.
99
R. Wang and R. S. Sloboda, Brachytherapy scatter dose calculation in
heterogeneous media. I. A microbeam ray-tracing method for the singlescatter contribution, Phys. Med. Biol. 52, 56195636 2007.
100
R. Wang and R. S. Sloboda, Brachytherapy scatter dose calculation in
heterogeneous media. II. Empirical formulation for the multiple-scatter
contribution, Phys. Med. Biol. 52, 56375654 2007.
101
G. M. Daskalov, R. S. Baker, D. W. Rogers, and J. F. Williamson, Dosimetric modeling of the microselectron high-dose rate 192Ir source by the
multigroup discrete ordinates method, Med. Phys. 27, 23072319
2000.
102
C. Zhou and F. Inanc, Integral-transport-based deterministic brachyMedical Physics, Vol. 36, No. 6, June 2009
2152
therapy dose calculations, Phys. Med. Biol. 48, 7393 2003.

C. Borgers, Complexity of Monte Carlo and deterministic dosecalculation methods, Phys. Med. Biol. 43, 517528 1998.
104
R. E. Alcouffe, R. S. Baker, F. W. Brinkley, D. R. Marr, R. D. ODell,
and W. F. Walters, Los Alamos National Laboratory Report No. LA12969-M, 1995.
105
G. M. Daskalov, R. S. Baker, D. W. Rogers, and J. F. Williamson, Multigroup discrete ordinates modeling of 125I 6702 seed dose distributions
using a broad energy-group cross section representation, Med. Phys. 29,
113124 2002.
106
C. Zhou, F. Inanc, and J. M. Modrick, Distortions induced by radioactive
seeds into interstitial brachytherapy dose distributions, Med. Phys. 31,
33933405 2004.
107
A. Carlsson Tedgren, Ph.D. thesis, Stockholm University, 2003.
108
H. Kubo, Comparison of two independent exposure measurement techniques for clinical I-125 seeds, Med. Phys. 12, 221224 1985.
109
P. Papagiannis, D. Baltas, D. Granero, J. Perez-Calatayud, J. Gimeno, F.
Ballester, and J. L. Venselaar, Radiation transmission data for radionuclides and materials relevant to brachytherapy facility shielding, Med.
Phys. 35, 48984906 2008.
110
H. Hedtjarn, G. A. Carlsson, and J. F. Williamson, Accelerated Monte
Carlo based dose calculations for brachytherapy planning using correlated
sampling, Phys. Med. Biol. 47, 351376 2002.
111
E. Poon, Y. Le, J. F. Williamson, and F. Verhaegen, BrachyGUI: An
adjunct to an accelerated Monte Carlo photon transport code for patientspecific brachytherapy dose calculations and analysis, J. Phys.: Conf.
Ser. 102, 012018 2008.
112
M. J. Rivard, C. S. Melhus, D. Granero, J. Perez-Calatayud, and F. Ballester, An approach to using conventional brachytherapy software for
clinical treatment planning of complex, Monte Carlo-based brachytherapy
dose distributions, Med. Phys. 36, 19681975 2009.
113
M. DAmours, J. Carrier, E. Lessard, J. Pouliot, F. Verhaegen, and L.
Beaulieu, A new approach for afterloading brachytherapy inverse
planned dose optimization based on the accurate Monte Carlo method,
Med. Phys. 35, 2969 2008.
114
J. F. Williamson, S. Li, S. Devic, B. R. Whiting, and F. A. Lerma, On
two-parameter models of photon cross sections: Application to dualenergy CT imaging, Med. Phys. 33, 41154129 2006.
115
M. Bazalova, J. F. Carrier, L. Beaulieu, and F. Verhaegen, Tissue segmentation in Monte Carlo treatment planning: A simulation study using
dual-energy CT images, Radiother. Oncol. 86, 9398 2008.
116
M. Bazalova, J. F. Carrier, L. Beaulieu, and F. Verhaegen, Dual-energy
CT-based material extraction for tissue segmentation in Monte Carlo dose
calculations, Phys. Med. Biol. 53, 24392456 2008.
117
P. Petric, J. Dimopoulos, C. Kirisits, D. Berger, R. Hudej, and R. Potter,
Inter- and intraobserver variation in HR-CTV contouring: Intercomparison of transverse and paratransverse image orientation in 3D-MRI assisted cervix cancer brachytherapy, Radiother. Oncol. 89, 164171
2008.
118
K. Tanderup, T. P. Hellebust, S. Lang, J. Granfeldt, R. Potter, J. C. Lindegaard, and C. Kirisits, Consequences of random and systematic reconstruction uncertainties in 3D image based brachytherapy in cervical cancer, Radiother. Oncol. 89, 156163 2008.
119
B. Al-Qaisieh, D. W. Smith, E. Brearley, S. St Clair, and P. Bownes,
Comprehensive I-125 multi-seed comparison for prostate brachytherapy:
Dosimetry and visibility analysis, Radiother. Oncol. 84, 140147 2007.
120
F. A. Siebert, M. De Brabandere, C. Kirisits, G. Kovacs, and J. Venselaar,
Phantom investigations on CT seed imaging for interstitial brachytherapy, Radiother. Oncol. 85, 316323 2007.
121
F. A. Siebert, A. Srivastav, L. Kliemann, H. Fohlin, and G. Kovacs,
Three-dimensional reconstruction of seed implants by randomized
rounding and visual evaluation, Med. Phys. 34, 967975 2007.
122
C. Kirisits, F. A. Siebert, D. Baltas, M. De Brabandere, T. P. Hellebust, D.
Berger, and J. Venselaar, Accuracy of volume and DVH parameters
determined with different brachytherapy treatment planning systems,
123
A. S. Meigooni and R. Nath, Tissue inhomogeneity correction for
brachytherapy sources in a heterogeneous phantom with cylindrical symmetry, Med. Phys. 19, 401407 1992.
124
J. F. Williamson, Dose calculations about shielded gynecological colpostats, Int. J. Radiat. Oncol., Biol., Phys. 19, 167178 1990.
125
G. Lymperopoulou, P. Papagiannis, A. Angelopoulos, P. Karaiskos, E.
Georgiou, and D. Baltas, A dosimetric comparison of 169Yb and 192Ir for
103
2153
HDR brachytherapy of the breast, accounting for the effect of finite patient dimensions and tissue inhomogeneities, Med. Phys. 33, 45834589
2006.
126
J. J. DeMarco, R. E. Wallace, and K. Boedeker, An analysis of MCNP
cross-sections and tally methods for low-energy photon emitters, Phys.
Med. Biol. 47, 13211332 2002.
127
A. Gerbaulet, R. Ptter, J.-J. Mazeron, H. Meertens, and E. Van Limbergen, The GEC ESTRO Handbook of Brachytherapy ESTRO, Brussels,
2002.
128
R. Nath, L. Anderson, D. Jones, C. Ling, R. Loevinger, J. Williamson,
and W. Hanson, AAPM Report No. 21, 1987.
129
NCS Report No. 4, 1991.
130
H.-J. Selbach, H.-M. Kramer, and W. S. Culberson, Realization of reference air-kerma rate for low-energy photon sources, Metrologia 45,
422428 2008.
131
A. Herskovic, S. Lee, and T. Padikal, Utilization of the computed tomography scanner in interstitial dosimetry, Radiology 135, 781782
1980.
132
B. Pierquin and J. V. Fayos, Dosimetry by tomography in interstitial
curietherapy: Point technique, Am. J. Roentgenol., Radium Ther. Nucl.
Med. 87, 585592 1962.
133
C. C. Ling, M. C. Schell, K. R. Working, K. Jentzsch, L. Harisiadis, S.
Carabell, and C. C. Rogers, CT-assisted assessment of bladder and rectum dose in gynecological implants, Int. J. Radiat. Oncol., Biol., Phys.
13, 15771582 1987.
134
S. L. Schoeppel, M. L. LaVigne, M. K. Martel, D. L. McShan, B. A.
Fraass, and J. A. Roberts, Three-dimensional treatment planning of intracavitary gynecologic implants: Analysis of ten cases and implications
for dose specification, Int. J. Radiat. Oncol., Biol., Phys. 28, 277283
1994.
135
L. Feuvret, G. Noel, J. J. Mazeron, and P. Bey, Conformity index: A
review, Int. J. Radiat. Oncol., Biol., Phys. 64, 333342 2006.
136
H. H. Holm, N. Juul, J. F. Pedersen, H. Hansen, and I. Stroyer, Transperineal 125iodine seed implantation in prostatic cancer guided by transrectal ultrasonography, J. Urol. Paris 130, 283286 1983.
137
J. C. Blasko, P. D. Grimm, and H. Ragde, Brachytherapy and organ
preservation in the management of carcinoma of the prostate, Semin.
Radiat. Oncol. 3, 240249 1993.
138
A. A. Martinez, I. Pataki, G. Edmundson, E. Sebastian, D. Brabbins, and
G. Gustafson, Phase II prospective study of the use of conformal highdose-rate brachytherapy as monotherapy for the treatment of favorable
2153
stage prostate cancer: A feasibility report, Int. J. Radiat. Oncol., Biol.,

Phys. 49, 6169 2001.
139
G. Kovacs, R. Potter, T. Loch, J. Hammer, I. K. Kolkman-Deurloo, J. J.
de la Rosette, and H. Bertermann, GEC/ESTRO-EAU recommendations
on temporary brachytherapy using stepping sources for localised prostate
cancer, Radiother. Oncol. 74, 137148 2005.
140
C. Haie-Meder, R. Potter, E. Van Limbergen, E. Briot, M. De Brabandere,
J. Dimopoulos, I. Dumas, T. P. Hellebust, C. Kirisits, S. Lang, S. Muschitz, J. Nevinson, A. Nulens, P. Petrow, and N. Wachter-Gerstner, Recommendations from Gynaecological GYN GEC-ESTRO Working
Group I: Concepts and terms in 3D image based 3D treatment planning
in cervix cancer brachytherapy with emphasis on MRI assessment of
GTV and CTV, Radiother. Oncol. 74, 235245 2005.
141
R. Potter, C. Haie-Meder, E. Van Limbergen, I. Barillot, M. De Brabandere, J. Dimopoulos, I. Dumas, B. Erickson, S. Lang, A. Nulens, P.
Petrow, J. Rownd, and C. Kirisits, Recommendations from gynaecological GYN GEC ESTRO working group II: Concepts and terms in 3D
image-based treatment planning in cervix cancer brachytherapy-3D dose
volume parameters and aspects of 3D image-based anatomy, radiation
physics, radiobiology, Radiother. Oncol. 78, 6777 2006.
142
R. Potter, J. Dimopoulos, P. Georg, S. Lang, C. Waldhausl, N. WachterGerstner, H. Weitmann, A. Reinthaller, T. H. Knocke, S. Wachter, and C.
Kirisits, Clinical impact of MRI assisted dose volume adaptation and
dose escalation in brachytherapy of locally advanced cervix cancer, Radiother. Oncol. 83, 148155 2007.
143
144
145
J. Pouliot, Y. Kim, E. Lessard, I. C. Hsu, D. B. Vigneron, and J.
Kurhanewicz, Inverse planning for HDR prostate brachytherapy used to
boost dominant intraprostatic lesions defined by magnetic resonance spectroscopy imaging, Int. J. Radiat. Oncol., Biol., Phys. 59, 11961207
2004.
146
K. Tanderup, D. R. Olsen, and C. Grau, Dose painting: Art or science?,
147
R. S. Malyapa, S. Mutic, D. A. Low, I. Zoberi, W. R. Bosch, R. Laforest,
T. R. Miller, and P. W. Grigsby, Physiologic FDG-PET threedimensional brachytherapy treatment planning for cervical cancer, Int. J.
148
U. Nestle, W. Weber, M. Hentschel, and A. L. Grosu, Biological imaging
in radiation therapy: Role of positron emission tomography, Phys. Med.
Biol. 54, R1R25 2009.

Vision 2020 Collection

Uploaded by

Copyright:

Available Formats

Vision 2020 Collection

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vision 2020 Collection

Uploaded by

Copyright:

Available Formats

Quality in radiation oncology

Todd Pawlickia and Arno J. Mundt

Med. Phys. 34 5, May 2007

In early 1942, Deming was invited to the Stanford University

2007 Am. Assoc. Phys. Med.

T. Pawlicki and A. J. Mundt: Quality in radiation oncology

Identify many possible causes of a problem

Also known as the fishbone diagram or Ishikawa diagram.

opportunities to improve patient care. In the next section, we

The focus of research on quality in radiation oncology is

Organizes a large number of ideas into

Sometimes replaced with the prioritization matrix.

Medical Physics, Vol. 34, No. 5, May 2007

T. Pawlicki and A. J. Mundt: Quality in radiation oncology

the main mode of operation to improve quality. An example

T. Pawlicki and A. J. Mundt: Quality in radiation oncology

All technology research in radiation oncology is geared

the physician care of patients.49 All of these aspects argue for

Cultural change and strong management are necessary to

T. Pawlicki and A. J. Mundt: Quality in radiation oncology

A closer and more formal collaboration with vendors

Collaboration with other professional societies that have

F. Adopt a patient view of quality

Electronic mail: [email protected]

T. Pawlicki and A. J. Mundt: Quality in radiation oncology

Medical Physics, Vol. 34, No. 5, May 2007

The future of PACS

Med. Phys. 34 7, July 2007

The fundamental predictor often cited in the computer

2007 Am. Assoc. Phys. Med.

Paul G. Nagy: The future of PACS

days standard model dual-core 64-bit processors with more

The high-density online market hosts a growing industry

FIG. 1. Growth in number of images per study and number of CT studies

heat and consume less power so that more can be aggregated

Paul G. Nagy: The future of PACS

Prediction 3. The commoditization of the server

years. The number of images per study in the department is

A blade farm is a server processing farm in which the

Paul G. Nagy: The future of PACS

processing power these systems can provide. Consider, as an

Prediction 4. The next generation of PACS will rely

To better understand this prediction, it is important to look

radiology. The second innovation was the use of dynamic

The client application of newer generations of PACS will

Paul G. Nagy: The future of PACS

known as thick clients, have traditionally been written in

creasingly untenable for radiologists to sacrifice productivity

Paul G. Nagy: The future of PACS

William Gibson, www.wikipedia.com, http://wn.wikiquote.org/wiki/

Paul G. Nagy: The future of PACS

461/661, Spring 1997, http://www.eecs.tulane.edu/Terry/OpenGL/

Medical Physics, Vol. 34, No. 7, July 2007

G. C. Nikiforidis, G. C. Kagadis, and C. G. Orton, It is important that

BGRT: Biologically guided radiation therapyThe future is fast

Med. Phys. 34 10, October 2007

strategy to integrate biological information into treatment

2007 Am. Assoc. Phys. Med.

R. D. Stewart and X. A. Li: Biologically guided radiation therapy

icity. A method to identify isoeffective prescription doses is