Surface Reconstruction of Medical

Dicom Data
1
G Divya, 2Ramachandra A C ,3Prasanna Paga,4 Manish Sharma, 5Archita ,6Saambhavi
Barthwal
1
Assistant Professor, Professor & Head,3Associate Professor,456 Student, Dept of
2

E&CE, Nitte Meenakshi Institute of Technology, Yelahanka, Bengaluru-64

Corresponding author [email protected]

Abstract — Images of the chest ribs emission computed tomography (SPECT),

acquired using a Computed Tomography and others. As technology advances, 3D
(CT), Magnetic Resonance Imaging (MRI) images are produced from the 2D slices that
devices are a single channel image [2D CT and MRI machines store. The accuracy
Image]. The information obtained is very of 3D reconstruction has made it important
limited and it takes many images from in medical applications. The Digital Imaging
various sides to be able to determine the and Communications in Medicine (DICOM)
condition of the costs. This can slow down standard protocol is used in many healthcare
the diagnostic process and becomes very facilities for the management and
difficult for a non-medical student to transmission of medical pictures and related
understand. Our work aims to solve the data. The method for obtaining 3D
problems by developing a system of 3D reconstruction from 2D DICOM ribs CT
Reconstruction Algorithms. Multiple images is presented in this paper. Volume
Dicom 2D CT Images are collated from Rendering and Surface Rendering are the
different sides and then passed through the two categories into which the 3D
pre-processing pipeline to remove noise reconstruction techniques of medical pictures
and increase the clarity of it. The result of can be separated. The volume rendering
the pipeline is then passed to the Re- shows the object's overall effect, although it
Construction pipeline where the surface requires a lot of work. The surface rendering
rendering technique and Marching Cubes approach is still widely used in medical
(MC) algorithm are implemented. The image reconstruction as a result. The
algorithm used is Adaptive Marching Cube Marching Cube (MC) Algorithm is the most
Linear Interpolation Algorithm with 15 representative surface rendering algorithm.
unique combinations of cubes. W. Lorensen and Cline introduced the MC
Pretreatment is essential to obtain quality algorithm in 1987 [1]. In theory, it is
chest rib portions rendered from the straightforward, and in practice, it is simple.
original cardiac CT images.The rendered The fundamental concept behind it is to
3D object file can then be viewed using an separate an iso-surface from the Dicom
online 3D viewer or MeshLab. Image and combine those polygons into a 3D
data field. The use of visuals is used to
Keywords — Chest Ribs Dicom Images, 3D display the results.
Reconstruction Algorithm, Adaptive
Marching Cube Linear Interpolation
Algorithm, Online 3D viewer, MeshLab. II. RELATED WORK

In order to decrease the quantity of triangles,

I. INTRODUCTION C.Montani et al. suggested combining the
output triangles into sizable coplanar
Diagnoses, surgical diagnoses, and radiation polygons [2]. By adjusting the size of the
are all handled by medical imaging methods triangles, Renben Shu et al proposed
like Computed Tomography (CT), Magnetic Adaptive Marching Cubes lower the number
Resonance Imaging (MRI), single-photon of triangles [3]. Using 2D DICOM
information, V.S. Nguyen et al. demonstrate marching cube algorithm by considering the
an improved technique for reconstructing the processing time and the algorithms
shape of 3D objects. Processing DICOM considering different layers of data. The
data, which creates a 3D point cloud using studies proved that the reconstruction
digital image processing techniques, was the efficiency of the improved MC algorithm is
first focus. To create a smooth dataset for 20% more compared to the traditional
each 2D DICOM slice, this phase involves algorithms with the aid of Visualization tool
extracting the shape features' contour, kit. In [8], the authors Pratomo et.al applied
eliminating noisy data, and adding new the Marching cube algorithm for 3D Heart
pixels. Rebuild the 3D object using the 3D Image Reconstruction and Visualization.
point cloud's triangulation in the following Heart images taken using a CT-scan or MRI
step. An application for rendering items in is a two-dimensional format were passed
3D medical imaging was developed in the through a filter to remove the effect of noise
last stage. The proposed method is extremely and sharpen the image. In the proposed work
quick to build a 3D model from DICOM some 2D images were taken from different
images when compared to existing methods. sides and to reduce the processing time 64
The programme can help doctors, nurses, and images were used and the average error rate
other medical professionals with training, were below 1% by taking 15 different
diagnosis, and treatment [4]. For the purpose combinations of the cube for MC algorithm.
of 3D reconstruction of CT scans, Zhang In [9], the authors Qeethara et.al developed
Jing et al. offer the Marching Cube (MC) an efficient approach of 3D image
algorithm, which combines the golden reconstruction by incorporating four basic
section and isosurface direction smoothly. processing steps namely preprocessing,
Equivalence points image enhancement, image contour then
and normal vectors are computed using the image reconstruction and visualization. The
golden ratio approach, cutting down on the obtained results show a good reconstruction
amount of calculations from 4 to 1. In order of the tested image for X-ray medical image.
to smooth the direction of all triangle stains In this paper we have demonstrated that the
in the design of space, the isosurface proposed approach is a useful technique for
direction smooth algorithm determines the effective 3D visualization. In[10], the authors
average value of the normal support. Christos P Loizou demonstrated an
According to experimental findings, the integrated system for the register ion and 3D
Generative Adversarial Network [GAN] reconstruction of DICOM MRI images and
Image Broadcasting network blurring lesions of the brain acquired from multiple
algorithm performs better recovery than sclerosis (MS) subjects at two different time
other algorithms when comparing the results intervals (time 0 (T0) and time 1 (T1)). The
of the Shannon entropy ratio and peak signal- system facilitates the doctor to accurately
to-noise ratio for different blur angles and manage the follow up of the disease.A 6-
blur amplitudes. To decrease the retrieved stage analysis process consisting of
triangles, Tajian Du and Mike recommended preprocessing, lesion segmentation,
making topological configurations simpler. registration, 3D reconstruction, volume
Edge signs are used by Tianqi Jia et al. to estimation and method evaluation were used
avoid repeated counting [6] although this for the image reconstruction. The system was
results in a large number of triangular evaluated based on one MRI phantom and
patches. In [7], the authors Jin Zhao et.al a one DICOM MRI image of the brain. The
proposed an analysis and Optimization of 3D accuracy of the proposed registration and
Reconstruction CT by using Quadratic reconstruction (- / -) method were
median approximation technique. A 78.5%,97.2% and 95.4%/95.8% for the
comparative analysis was done between the phantom and the MRI images respectively.
traditional algorithms and the improved In[11], the authors Van Sinh NGUYEN et.al
proposed an improved method for were carried out using HSV method ,
reconstructing the shape of 3D objects from erosion, and dilation techniques to extract
2D DICOM datasets using digital image cancerous area from the lung image .The
processing to get a 3D point cloud. The extraction of information for reconstruction
processing techniques involved extracting process were carried out using Marching
features, boundary of the shape, removing Cubes (MC) algorithm. In[15],the authors
noisy data and inserting new pixels to obtain Zhang Jing et.al proposed a Novel 3D
a regular dataset on each 2D DICOM Reconstruction Algorithm for Motion-
slice.The 3D objects were reconstructed Blurred CT Image.The GAN deblurring
based on triangulation of 3D point clouds. algorithm were proposed to remove blurred
In [12], the authors Guanwu Jiang et.al results and subsequently Marching Cubes
proposed an improved Filtering algorithm (MC) algorithm based on the fusion of
based on 3 Dimensional Reconstruction of golden section and iso surface direction
CT Images.An improved adaptive weighted smooth (GI-MC) were used for 3D
median filter algorithm were used for reconstruction of CT images. The golden
reducing the processing time for CT images section algorithm is used to calculate the
and improve the process effect of the speckle equivalent points and normal vectors,
noise and environmental noise. Subsequently thereby reducing the number of vectors from
the automatic threshold segmentation four to one.The isosurface direction smooth
algorithm and the improved iterative algorithm computes the mean value of the
algorithm are used to realize the multi normal vector, so as to smooth the direction
threshold image segmentation. Finally used of all triangular patches in spatial
the marching cubes algorithm and equivalent arrangement. The experimental results
plane method were applied for the showed that for different blurred angle and
reconstruction of CT image through VTK blurred amplitude, the reconstruction
(Visualization Toolkit) to greatly improve accuracy of GI-MC algorithm were 9.9%,
the accuracy of medical diagnosis.In[13], 7.7%, and 3.9% higher than that of the
the authors Ignatius Luddy Indra Purnama traditional MC algorithm considering the
et.al proposed a 3D Image Reconstruction data set of liver
with Single-Slice CT using Improved In[15], For the 3D reconstruction scene of
Marching Cube Algorithm for the motion-blurred CT image, this paper consists
determination of the threshold in single-slice of two parts: firstly, a GAN image translation
Computerized Tomography (CT) for network deblurring algorithm is proposed to
interactive 3D image reconstruction by using remove blurred results. This algorithm
improved marching cube algorithm. The adopts the clear image to
medical image data set essentially consisted supervise the training process of the blurred
of skull bone and sternum-pelvis in Digital image, which creates solutions that are close
Imaging and Communications in Medicine to the clear image. Secondly, this paper
(DICOM) format. The threshold values were proposes an MC algorithm based on the
set at 200, except for the 3D image with fusion of the golden section and isosurface
quantity slice less than 10 .The difference in direction smooth (GI-MC) for 3D
the surface volume and surface area between reconstruction of CT images. The golden
the 3D image reconstruction output for skull section algo-
bone and sternum-pelvis were less than rithm is used to calculate the equivalent
0.5%. and the processing time to reconstruct points and normal vectors, which reduce the
the 3D image were five minutes.In[14] the calculation numbers from four to one.
authors K. A. Mandaliana et.al proposed a In the proposed study, an improved
Marching Cubes Algorithm for 3D algorithm has been suggested in response to
Visualization and Reconstruction of Lung the shortcomings of the conventional MC
Cancer Images .The segmentation process algorithm.
The 3D rendered result MeshLab was used improving the Interpretability or
to visualize the results, and Online 3D believability of information in photographs
Viewer was used to determine the for human viewers and providing better
ReConstruction Precision. The outcome input for various computers image
demonstrates that the enhanced MC processing techniques. There are numerous
approach may expedite isosurface extraction methods for improving a virtual image
and satisfy real-time user interaction without spoiling it. Each topic where the
requirements. images in Figure 2 should be studied and
analyzed implements image enhancement.
III. PROPOSED BLOCK DIAGRAM Before engaging in the process of 3D
reconstruction, a preprocessing stage is
The suggested 3D reconstruction Pipeline is utilized to improve and refine the image as
depicted in Figure 1. The PreProcessing and well as remove unnecessary elements of the
Equalizing pipeline receives the input Dicom image known as noise.
CT Scans and refines the image. These
A graphic interpretation of the brightness
images are then aligned using the instance
values of an image's pixels is called an
number found in Dicom images, and the data
image histogram. It can be thought of as a
is then transformed into a Numpy array. The
data structure that keeps track of the
MC Algorithm then receives these numpy
frequencies of every pixel's intensity level in
files and turns them into an object file that
an image. Histogram is used in the image
can be seen with any 3D viewer.
processing method known as "Histogram
Equalization" to modify the image contrast.
It is the expansion of the most prevalent
pixel brightness values to boost the image's
contrast or extend the brightness spectrum.
In this manner, sections of the image with
low contrast can obtain better contrast by
flattening the histogram.

Input Image Equalized Image

Histogram of Histogram of
Figure 1 : Block Diagram of_Proposed
input image. equalized image
Work
IV. PRE-PROCESSING Figure 2 : Comparison between Original
Image and Equalized Image
Image enhancement mainly involves
 IV. ADAPTIVE MARCHING binary data. Different algorithms and
CUBE LINEAR methods are required for such situations.
INTERPOLATION The volumetric data used in 3D-GIS spatial
ALGORITHM raster volumetric data can be easily
extracted as iso-surfaces using the MC
The basic goal of the MC Algorithm is to technique, but because of the discontinuity
classify the intersection of the threshold iso- in the volumetric data , it can also be used
plane and the Marching cube and rebuild the to reconstruct imperfect 3d Surfaces..
3-Dimensional surface based on the
intersection for each MC cube at a threshold
C consisting of 8 consecutive points of data.
The MC algorithm presumes continuous
linear fluctuation in the data fields at the
cube edge. Specifically, if the data of one
point on the MC cube's edge is larger than C
and the data of the other point is less than C,
only the data of the first point will be equal
to C edge of the cube has a point equal to the
value C. Regardless of whether C is large,
cubes can be categorized using the MC cube
point data to identify the reconstruction
mode of the MC cube isoplane. Each cube
point's data has two states, giving a total of
28 = 256 possible combinations of states.
These 256 states can be reduced to 15 kinds
using extra symmetry and rotational Figure 3 : Unique Marching Cube
symmetry, as shown in Figure 3. The black Configuration
dots here indicate that more data points in
the MC cube than C [Iso-Surface] were V.EXPERIMENT AND ANALYSIS
present. The location of the isosurface points
on the MC cube's edge is determined after Real medical imaging datasets were utilized
the reconstruction mode has been chosen. to gauge how well the proposed technique
The coordinates of the intersection point K performed. The system was configured with
along the edge points P1 and P2 as well as Graphic Card GeForce GTX 1050, i5-9th
the values of the edge points f (P1) and f GEN with 2.00 GHz, running 8Gb Ram, 512
(P2) are established under the assumption
SSD, 1TB HDD, Windows 10 64 bit
that the edge data field exhibits continuous
linear variation (P2). The equation for operating system. The output was visualized
intersection K is thus: in MeshLab. Figure 4 shows the Input
Dicom Images.
K is equal to P1 plus ((c f (P1)) / (f Figure 4 : Input Dicom Ribs Scans
(P2) f (P1)) (P2P1)), where c R 0
= c =1.

A nice feature for enhancing the voxel

display is the interpolation of edge-crossing
vertices. The interpolation, however,
assumes that there are already floating
points, therefore it is unable to smooth your
The proposed approach can significantly also be used by machine learning or deep
reduce the number of cubes visited and the learning algorithms to deliver useful
number of patches generated, thus requiring information. In this article, a technique for
less time investment. It is suitable for real- creating 3D images from DICOM files is
time interactive 3D reconstruction of suggested. The suggested approach may
medical images. resize the three-dimensional image with
greater quality than 2D resizing and create
VI. RESULTS AND OUTPUTS
3D images from slices of DICOM images.
The experimental findings show that the
In the proposed work Histogram
method, which uses the adaptive MC linear
Equalization was used for preprocessing and
interpolation algorithm, has a 94%
The MC Algorithm Technique as the main
reconstruction precision. The Difference
ReConstruction Algorithm. The MC Cubes
between the total SurfaceTriangles between
had 15 unique predefined cube
the Online 3D Viewer and MeshLab was
configurations. We used 361 Input DICOM
considered for validation
Images for the 3D reconstruction. The
amount was sufficient to obtain a reasonable
VIII. REFERENCES
result. The error is obtained by calculating
the difference in the number of grayscale
[1]. William E Lorensen, Harvey E Cline ,
pixels between the input image and the
output image. The 3D reconstruction result 1987, Marching Cubes: A High Resolution
is shown in Figure 5. 3D Surface Construction Algorithm
[Official Paper]

[2]. C. Montani, R. Scateni and R.

Scopigno, "Discretized Marching Cubes,"
Proceedings Visualization '94, 1994, pp.
281-287, doi:
10.1109/VISUAL.1994.346308.

[3] Shu, Renben & Zhou, Chen &

Kankanhalli, Mohan. (2003). Adaptive
Marching Cubes. The Visual Computer.
11. 10.1007/s003710050015.

[4].V. S. Nguyen, M. H. Tran and H. M.

Quang Vu, "A Research on 3D Model
Construction from 2D DICOM," 2016
International Conference on Advanced
Computing and Applications (ACOMP),
Figure 5 : Rendered 3D Output using MC 2016, pp. 158-163, doi:
Algorithm 10.1109/ACOMP.2016.031.
VII. CONCLUSIONS
[5].T. Jia, X. Feng and X. Guo,
Because 3D representations provide a more "Improvement of Marching Cubes
realistic viewpoint than 2D graphics, medical Algorithm Based on Sign Determination,"
images frequently use them. Such photos can 2010 Second International Workshop on
 Education Technology and Computer reconstruction of CT images based on
Science, 2010, improved filtering algorithm," 2018 IEEE
pp. 194-197, doi:10.1109/ETCS.2010.188. 9th International Conference on Mechanical
and Intelligent Manufacturing Technologies
[6].K. Sudharani, R.RavindraReddy,T. C. (ICMIMT), 2018,pp.207-212,doi:
Sarma, K.Satya Prasad, 2014,”3D DICOM 10.1109/ICMIMT.2018.8340450.
images reconstruction Using LabView.”
[13]. L. Indra Purnama, A. Edy Tontowi and
[7].Shuqin Liu, Jin H. Herianto, "3D Image Reconstruction with
Peng ,2015,”Optimization of Single-Slice CT using Improved Marching
Reconstruction of 2D Medical Images Cube Algorithm," 2019 International
Based on Computer 3D Biomedical Instrumentation and Technology
ReconstructionTechnology” Conference (IBITeC), 2019, pp. 84-87, doi:
10.1109/IBITeC46597.2019.9091687.
[8]. P. A. Nugroho, D. K. Basuki and R.
Sigit, "3D heart image reconstruction and [14] K. A. Mandaliana, T. Harsono and R.
visualization with marching cubes Sigit, "3D Visualization and Reconstruction
algorithm," 2016 International Conference of Lung Cancer Images using Marching
on Knowledge Creation and Intelligent Cubes Algorithm," 2019 International
Computing (KCIC), 2016, pp. 35-41, doi: Electronics Symposium (IES), 2019, pp.
10.1109/KCIC.2016.7883622. 143-147, doi:
10.1109/ELECSYM.2019.8901667.
[9].Qithara Kadhim Al-Shayea and [15] Jing, Zhang & Qiang, Guo & Fang,
MuzhirShaban Al- Ani, 2016,“An Efficient Han & Zhan-Li, Li & Hong-An, Li & Yu,
Approach to 3D Image Reconstruction.” Sun. (2020). A Novel 3D Reconstruction
Algorithm of Motion-Blurred CT Image.
Computational and Mathematical Methods in
[10]. C. P. Loizou et al., "Brain Image and
Medicine. 2020. 1-13.
Lesions Registration and 3D Reconstruction
10.1155/2020/9324689.
in Dicom MRI Images," 2017 IEEE 30th
International Symposium on Computer-
Based Medical Systems (CBMS), 2017, pp.
419-422, doi: 10.1109/CBMS.2017.53.

[11]. V. S. Nguyen, M. H. Tran and H. M.

Quang Vu, "An Improved Method for
Building A 3D Model from 2D DICOM,"
2018 International Conference on Advanced
Computing and Applications (ACOMP),
2018, pp. 125-131, doi:
10.1109/ACOMP.2018.

[12]. G. Jiang, M. Luo, S. Chen, G. Jiang and

K. Bai, "Research on 3 dimensional