AI IN TELEMEDICINE: AN APPRAISAL ON

DEEP LEARNING-BASED APPROACHES TO

VIRTUAL DIAGNOSTIC SOLUTIONS (VDS)
Ozioma Collins Oguine and Kanyifeechukwu Jane Oguine

Department of Computer Science, University of Abuja, Nigeria

ABSTRACT
Advancements in Telemedicine as an approach to healthcare delivery have heralded a new
dawn in modern Medicine. Its fast-paced development in our contemporary society is credence
to the advances in Artificial Intelligence and Information Technology. This paper carries out a
descriptive study to broadly explore AI's implementations in healthcare delivery with a more
holistic view of the usability of various Telemedical Innovations in enhancing Virtual
Diagnostic Solutions (VDS). This research further explores notable developments in Deep
Learning model optimizations for Virtual Diagnostic Solutions. A further research review on the
prospects of Virtual Diagnostic Solutions (VDS) and foreseeable challenges was also
highlighted. Conclusively, this research gives a general overview of Artificial Intelligence in
Telemedicine with a central focus on Deep Learning-based approaches to Virtual Diagnostic
Solutions.

KEYWORDS
Biomedical imaging, Telemedicine, Smart Healthcare, Medical Imaging, AI, Virtual Diagnostic
Solutions.

1. INTRODUCTION
Healthcare and Medicine are areas of modern society which has gained quite an outstanding level
of research attention given current antecedents of virus outbreaks and spikes in anomalies
regarding human health. Over the years, advancement in Artificial Intelligence and its resonating
research areas such as Telecommunication and information technology has stirred up questions
and advanced solutions regarding Human health. Affirmatively, we can infer that these
improvements have notably impacted the medical and healthcare delivery scale and quality.
However, healthcare access and delivery have struggled extensively to meet anticipated
simultaneous prospects, as is the situation in many parts of the world, predominantly in
underdeveloped and developing nations. A significant reason for this decline is the ever-
increasing number of healthcare users and medical patients leading to the overutilization of
medical resources such as healthcare providers, medical staff, and access to medical
infrastructures. Hence, a more suitable and sustainable healthcare delivery approach was
necessary to eliminate the issues arising from overpopulation and access to healthcare
infrastructures.

A notable convergence of Machine Learning, Robotics, Telecom, computational neuroscience,

and cloud computing has established a new infrastructure for global healthcare delivery known as
Telemedicine. According to Khemapech et al., “Telemedicine is the delivery of health care
services, with significant consideration of distance in service delivery and accessibility by all
David C. Wyld et al. (Eds): SIPP, NLPCL, BIGML, SOEN, AISC, NCWMC, CCSIT - 2022
pp. 229-243, 2022. CS & IT - CSCP 2022 DOI: 10.5121/csit.2022.121320
230 Computer Science & Information Technology (CS & IT)
stakeholders as key variables, using information and communications technologies” [1]. An
inferred purpose of this innovation is to exchange valid information for diagnosing, treating, and
preventing disease and injuries. Also, for research and evaluation, continuing education of health
care providers and users, all aimed at advancing the healthcare systems of individuals and their
communities. Although research on Telemedicine has been ongoing for decades, the emergence
of the COVID-19 pandemic has reinvented its usage instantaneously, resulting in its scalability in
fully implementing essential health and safety protocols in service delivery. This research area is
an innovation that transcended from being a convenient alternative for technically savvy patients
to the mainstay for healthcare delivery today across nations of the world, a reality that may
continue long after the COVID pandemic. Telemedicine is the most suitable and sustainable
approach to delivering healthcare services to patients by incorporating technological
advancements in affordable and low-cost implementations. It also eliminates factors that
adversely affect privileges to healthcare by strategically expanding virtual solutions to
accommodate the growing populace and integrating smart tech to facilitate healthcare access and
efficiency. The telemedicine framework has enabled collective improvements in Virtual
Diagnostic Solutions, as shown in Fig 1.

Fig 1. Telemedicine Infrastructure and Components

Deep Learning: Deep learning can be referred to as a modern state-of-the-art mechanism for
computational processes applied to solving contemporary problems. Its potential for drawing
patterns and insights from a massive amount of data boasts its wide adoption.

Virtual Diagnostic Solutions (VDS): This is a systematic approach to undertaking medical

prognosis and administering or proffering medical solutions with little or no supervision from
medical Professionals.

2. STAKEHOLDERS IN TELEMEDICINE
Telemedicine as an infrastructure for healthcare improvement entails the participation of four
significant stakeholders, namely:
 Computer Science & Information Technology (CS & IT) 231
Patients (Key Stakeholders): These are the receivers or users of healthcare services. They are
perhaps the most critical stakeholder to consider when thinking of approaches to Telemedicine.

Medical Professionals: These are the next most essential stakeholders in Telemedicine. Their
knowledge area and research lay the foundation for building a telemedical Infrastructure.

Developers/ Tech Experts: These stakeholders are necessary because they constantly research
innovative ways to build solutions to medical problems (Middlemen between Patients and
Medical Professionals).

Policy Makers: These are folks who formulate and regulate laws that guide the implementation
and utilization of Telemedicine

Telemedicine is currently gaining legislative and regulatory support in most developed countries;
there have been no nationally representative estimates on its implementation by physicians and
health practitioners across all medical specialties [2]. “To tackle this information gap, the
American Medical Association (AMA) conducted a study in 2016 that surveyed 3,500 physicians
who provided needed data to help assess potential barriers and create strategies to promote
telemedicine adoption. The data analysis report from the AMA’s 2016 Physician Practice
Benchmark Survey clearly described the Physician-to-Patient ratio of Virtual Diagnostic
Solutions (VDS) in Telemedicine, as shown in Fig. 2. Consequently, this paradigm also proved
efficient in maintaining and sharing patients’ medical records among hospitals and medical
institutions.

Fig 2. AMA 2016 Data Analysis on Patient to Healthcare Professional ratio in Telemedicine

The advancement of Artificial Intelligence and its research areas, such as the Internet of Things
(IoT), Machine Learning, Image processing, and Deep Learning, has highly improved the level
and quality of services provided by Telemedical software and applications. Researchers have and
are currently up scaling on services and innovative solutions such as medical prognosis,
Analytical Medicine, Robotic Surgery, DNA (genome) Sequence Analysis (DSA), Drug research
and discovery, Medical Data Security, Clinical Trials, Medical Risk Prediction, emergency
services, and Medical Image Analysis (brain monitoring, computer tomography, and radiology)
offered by this remarkable means of healthcare delivery. Amid all the solution-oriented
approaches put forward to dress health challenges, medical imaging is one area that has seen
more promising results and prospects. X-Ray, Ultrasound (US), MRI, and CT Scanners have
been interfaced with computers to transfer medical images to the remote center for careful
analysis and early detection of medical abnormalities [3].
232 Computer Science & Information Technology (CS & IT)
This paper explicitly discusses the implementation of artificial intelligence in Telemedicine with
a more holistic view of Annotation-Efficient deep learning models for Medical Imaging in
Virtual Diagnostic Solutions (VDS). It also elaborates on how data (input images) from medical
imaging equipment are passed to sophisticated algorithms for accurate Diagnosis and treatment
of Imaging-related ailments. Furthermore, application scenarios of medical Imaging models for
VDS are subsequently highlighted and reviewed to understand its challenges and perceived
prospects.

3. ARTIFICIAL INTELLIGENCE IN TELEMEDICINE

Artificial intelligence is a broad research field that facilitates machine simulation of human
intelligence and behaviors ranging from learning to problem-solving. Its rapid growth and
development in the past decade are due to its vast implementation in all human endeavors. Given
the recent upsurge in big data generation, powerful computing coupled with refined
computational models and algorithms, developments in AI have accelerated exponentially [4].
This trend has flagged the emergence of subfields such as Machine Learning (ML), Natural
Language Processing (NLP), AI voice technology, Medical Imaging, AI assistants, Computer
Vision, and robotics.

Telemedicine is a resurging innovation that has gained wide adoption in most developed and
developing countries to Fastrack the accessibility to sustainable healthcare. It has achieved quite
a laudable level of attention and research over the last two decades as ongoing studies are
exploring ways to improve the existing infrastructure to a State-of-the-Art (SOTA). For efficient
healthcare provision in contemporary societies, humans and machines have complimented each
other in effectively delivering healthcare services through Virtual Diagnostic Solutions (VDS),
thus providing a platform that bridges the gap between communication and accessibility to
medical services. AI in Telemedicine has seen rapid adoption primarily based on the volume of
medical data (Big data) recurrently generated. According to Ozioma et al., traditional data-
handling techniques have proved ineffective in utilizing these data to obtain viable medical
insights or solutions, given the gargantuan nature of the data [5]. Hence, sophisticated
demonstrations of Artificial Intelligence approaches and models in Medical Diagnostics are
becoming relatively popular. The adaptability and flexibility of these AI-based approaches and
models have also driven the necessity for their implementation.

Andressa et al., amongst other scholars (see table 1), proposed an architecture that relies on
fingerprinting and FLIPER framework to Fastrack the versatility and interconnectivity of
healthcare applications. They also anticipated impact of their research was to enable quick,
customizable resources that meet the level of reliability required for AI in Smart-health
applications [6].

Table 1. Comparative Table describing Researches implementation of AI in Telemedicine [7]

Research Article Year Trend Category Methodology

A Predictive Model for 2014 Information Analysis KNN and other data mining
Assistive Technology Adoption and Collaboration algorithms were utilized to analyze
for People with Dementia [8] the behavior of patients with
dementia and their adaptation to
technology.
 Computer Science & Information Technology (CS & IT) 233
A Telerehabilitation 2014 Healthcare Issues of Telemedicine under low-
Application with Pre-defined Information bandwidth network conditions were
Consultation Classes [9] Technology addressed using customized
consultation classes demonstrating
rehabilitation practices with preset
parameters and a bandwidth adaption
algorithm
An application of fuzzy systems 2014 Intelligent Assistance “This paper explores principled
was used to identify the best Diagnosis machine learning approaches to
course of action for a given interpreting large quantities of
situation to Monitor Mobile continuously acquired, multivariate
Patients by Combining Clinical physiological data. Early warning of
Observations with Data from serious physiological determination
Wearable Sensors [10] was done using wearable patient
monitors, such that a degree of
predictive care may be provided.”

Ankle Rehabilitation System 2015 Patient Monitoring This study uses a smartphone
with Feedback from a application, a
Smartphone Wireless wireless gyroscope platform,
Gyroscope Platform and machine learning, and 3D printing to
Machine Learning record usage and effects of therapy
Classification [11] on an ankle and measure the
strategy's efficacy.
Intelligent decision systems in 2015 Intelligent Assistance The study presents “a short review of
Medicine -a short survey on Diagnosis some current Machine Learning
medical Diagnosis and patient algorithms (neural networks, genetic
management [12] algorithms, support vector machines,
Bayesian decision, k-nearest
neighbor, etc.) used for automated
Diagnosis of different major
diseases, such as breast, pancreatic,
and lung cancer, heart attacks,
Diabetes.”
Smartphone-Based Recognition 2015 Intelligent Assistance This paper proposes a system of
of States and Changes in Diagnosis using a
Bipolar Disorder Patients [13] smartphone-sensing wearable device
to evaluate the behavior and
recognize depressive and manic
states of patients with bipolar
disorder.
An Effective Telemedicine 2016 Information This paper proposes an algorithm
Security Using Wavelet-Based Technology that embeds and reads digital wavelet
Watermarking [14] watermarks on medical images to
secure confidentiality.

Mobile Cyber-Physical 2016 Patient Monitoring This paper proposes the use of a
Systems for Health Care: monitoring system embedded in
Functions, Ambient Ontology wearable devices for the doctor or
and e-Diagnostics [15] family members to receive updates
on the patient's status.
234 Computer Science & Information Technology (CS & IT)
Detection of Fetal 2016 Patient Monitoring This study uses a neuro-fuzzy logic
Electrocardiogram through system to monitor and detect the
OFDM, Neuro-Fuzzy logic, and exact
Wavelets Systems for electrocardiogram and other signals
Telemetry [16] of a fetus inside an abdomen.
Using CART for Advanced 2016 Intelligent Assistance This study created an algorithm with
Prediction of Asthma Attacks Diagnosis data from a home-based
Based on Telemonitoring Data telemonitoring system to predict
[17] asthma exacerbation.

A Wireless Continuous 2017 Patient Monitoring The paper presents “a wireless

Patient Monitoring System for monitoring system for patients who
Dengue: Wi-Mon [18] need continuous monitoring, using
the Wireless Body Area Network
(WBAN) concept.”

This research also aims to describe and evaluate the impact of Artificial Intelligence in
Telemedicine. Some documented implementations of this powerful paradigm are discussed
below:

Medical Imaging: This generally entails training AI models with images of medical scans that
have been scientifically collected and stored in data repositories and databases. AI has
significantly reduced the cost and time involved in analyzing scans through advanced deep
learning models to diagnose health disorders. Hence, potentially allowing more scans to be taken
and improving prognosis proficiency and accuracy [19].A vast amount of resources and research
has been expended in developing this field, as will be accentuated later in this paper. These
researches have paved the way for State-of-the-art detection methodologies of medical ailments
such as Brain tumors, skin and breast cancer, Pneumonia, and eye diseases.

Fig 3. Covid-19 Detection in Lungs using Multibox SSD Model [20]

Echocardiography: Heartbeat patterns and coronary heart disease diagnosis and detection
utilizethe Ultromics system, an AI framework trialed at John Radcliffe Hospital in Oxford, to
analyze echocardiography scans [21].

Fig 4. Classification of Echocardiograms using Deep Learning [22]

 Computer Science & Information Technology (CS & IT) 235
Screening for Neurological Conditions: Several AI models are being developed and
employed in speech patterns analysis to predict psychotic episodes, Schizophrenia, recognize and
monitor symptoms of neurological disorders such as Parkinson’s disease [23].

Fig 5. fMRI Study in Individuals with First-Episode Psychosis [24]

Emotion Recognition for Psychological Prognosis: Artificial Intelligence Deep learning

models have proven efficient in recent times in observing and predicting emotions through
individual reactions and psychological states at a given time. Ozioma et al. opined in their
research on Facial Expression Recognition that Emotions are fundamental in human
communication, driven by the erratic nature of the human mind and the perception of relayed
information from the environment [25]. They proposed a hybrid deep learning model that makes
real-time predictions of a person’s emotional state, categorizing the individual’s emotion into one
of seven classes. The need for novelties in advancing this research field stems from the alarming
rate of Emotion disorders, Suicide, Post-traumatic stress disorders (PTSD), and
Psychological/Mental breakdown suffered by people in our contemporary society.

Fig 6. Facial Emotion Recognition using Hybrid Deep Learning Model [25]

Teleradiology: “Telecommunication is employed to transmit digital radiological images, like X-

rays, Computed Tomograms (CTs), and Magnetic Resonance Images (MRI’s) across
geographical locations for interpretation and consultation” [26]. Reducing financial costs is one
of the primary benefits of Teleradiology as it significantly reduces constraints in accessing
radiological images, reports, and feedback between health professionals and patients.

Teleradiology is a crucial means for optimizing radiology workflow by sending the images to the
radiologist rather than traditionally going to the radiology facility. By its very nature,
Teleradiology is an efficient and high-quality manner by which patients’ images can be
interpreted and diagnosed by qualified specialists. Cloud services are primarily employed in this
Telemedical service, where health stakeholders can utilize numerous privileges from the cloud,
thus, upscaling the quality of radiology services. AI in Teleradiology will also enable the sharing
of clinical information, medical imaging studies, and patient diagnostics [27] between patients
and healthcare professionals.
236 Computer Science & Information Technology (CS & IT)
Teledermatology: This service implements Telecommunication to transmit medical information
concerning skin conditions (e.g., tumors) for interpretation and consultation. According to Eedy
and Wotton, “Teledermatology model has received extensive advocate as a healthcare delivery
model that may reduce inequalities encountered in the utilization of overstretched dermatological
services, with implementation concentration ranging from remote to isolated communities” [28].
Landow et al., in their research, highlighted four factors that have stirred a relative increase in
face-to-face appointments which teledermatology strategies have tackled: (1) effective pre-
selection of patients for teleconsultation, (2) high-quality photographic images, (3) dermoscopy if
pigmented lesions are evaluated, and (4) adequate infrastructure and culture in place to
implement teleconsultation recommendations [29].

4. APPLICATIONS OF DEEP LEARNING-BASED DIAGNOSTIC SOLUTIONS IN

TELEMEDICINE
State-of-the-art Deep Learning models have seen advancements in methodologies over various
medical problems such as object detection, recognition and segmentation in computer vision,
voice recognition, and genotype/phenotype prediction. Telemedicine employs deep learning
models in several Virtual Diagnostic Solutions today. While early studies focused on 2D medical
images, such as chest X-rays, mammograms, and histopathological images, recent studies are
looking toward applying sophisticated deep learning models to volumetric medical images.

CNNs form the basis for most State-of-the-art medical imaging DL models, which have gained
wide prominence since achieving impressive results at the ImageNet [30] competition in 2012.
Akkus et al. opined that CNNs remain a popular choice of DL approaches to image processing
given their laudable tendency to weight sharing across convolutional layers or feature maps, in
contrast, to fully connected ANNs [22]. And a rational reason for this was that, for 2D/3D image
processing, ANNs utilize heavy computational processing that consumes a relatively high amount
of GPU memory.

Several scholars have conducted evaluation studies and proposed several methodologies for
different Virtual Diagnostic Solutions employing medical Imaging to solve health-related issues
ranging from collecting image data to evaluating and diagnosing medical ailments. Qin et al.
utilized Computer-aided Detection (CAD) in chest radiography based on contrast enhancement
and segmentation in diagnosing various lung diseases such as early lung cancer, Pneumonia,
Tuberculosis, and, more recently, lung inflammation levels caused by Covid-19 [31]. Numerous
well-known DCNN architectures tested by Shin et al. emphasized the efficiency of transfer
learning approaches in CT patch-based thoracoabdominal lymph node detection and ILD
classification [32]. A Recurrent full Convolutional Neural Network (RFCNN) proposed by
Poudel et al. was used to segment the left ventricle from cardiac MR images [33]. Hosseini-Asl et
al. proposed a 3D CNN model to determine the progression of Alzheimer’s disease from
structural brain MR images [34]. In the method, they employed a transfer learning approach that
utilized pre-trained weights of features from a Computer-Aided Engineering (CAE) with a small
number of source domain images to fine-tune the target domain data to train the actual
classification model. Yu et al. proposed a 3D volumetric CNN for prostate segmentation on MR
images through U-net expansion, initially used for 2D biomedical image segmentation. They also
added residual connections to combine multiple-scale information [35, 36]. An alternative study
proposed a slice-level classification model to detect Interstitial Lung Diseases (ILD) from chest
CT scans [37]. Jamaludin et al. detected several diseases simultaneously from spinal MR images
through a trained multi-task learning model. They visualized salient regions in the image for
corresponding predictions as ‘evidence hotspots’ as seen in Fig.7 [38].
 Computer Science & Information Technology (CS & IT) 237

Fig 7. “Evidence hotspot” Visualization in Spinal MRI [38]

5. PROSPECTS OF AI-BASED VIRTUAL DIAGNOSTIC SOLUTIONS (VDS) IN

TELEMEDICINE
Early Diagnosis and treatment of diseases have been a primary focus of modern-day healthcare
infrastructure. In previous sections above, this research has significantly highlighted some
outstanding advancements and implementation of the Deep Learning approach of Artificial
Intelligence to improve Virtual Diagnostic Solutions. This review paper also seeks to forecast
based on the current state-of-the-art prospects of this frequently evolving paradigm. A crucial
benefit of this paper is to buttress the promises of AI in future healthcare development.

Early Diagnosis and Treatment: Diagnosis and treatment have ensured the advancement of
research conducted in the field of Medicine today. With the current pace in development, a
significant prospect of AI in Telemedicine is seen to be geared towards establishing and
improving early diagnosis and treatment mechanisms of numerous medical conditions.
Jacobsmeyer showed the effectiveness of AI in the early detection and management of infectious
diseases and epidemics such as Water contamination, worldwide virus outbreaks, etc. [39].
Research Labs and tech firms are currently pushing the limits on a proactive scale in ensuring
this, as not a week goes by without the introduction of new approaches to medical solutions using
AI or Big data with improved accuracy and precision.

AI systems will also become more advanced in engaging in broader medical problem-solving
tasks with little human supervision or control. Hence, advanced ethical learning as imposed by
state-of-the-art deep learning models will play a crucial role in the effective implementation
strategy and adoption of Artificial Intelligence approaches in Virtual Diagnostic Solutions.

Independence: A major gap identified in traditional Medicine is the inability of patients to

access healthcare when, where, and how they want it. Deep learning approaches to Telemedicine
have gone far beyond filling this gap by empowering patients to systematically, thematically, and
objectively evaluate symptoms and proffering possible solutions through virtual diagnostic
solutions. Another significant advantage is the noteworthy reduction in the strain on medical
resources and healthcare professionals. While concerns have been raised regarding the negative
impact of AI solutions on Telemedicine [19], several notable research has been observed to
acknowledge its importance and necessity in delivering cost-effective, high quality and accessible
healthcare services [40, 41].
238 Computer Science & Information Technology (CS & IT)
Improved Clinical and Therapeutic Coordination: Not only have the efficiency and
advancements of AI ensured quality in the level of medical potential, but they have also shown a
remarkable promise in the level of service delivery mechanisms. As more implementations of
Virtual Diagnostic Solutions are utilized, huge amounts of data are also generated from
healthcare professionals and patients. Over time, this aggregate amount of data will create an
abundance of information on stakeholders involved in Telemedicine utilization. A foreseeable
impact of this will be the development of systematic and coordinated clinical therapeutic
processes and services that ensure that VDS prognosis are subjected to second opinion
evaluations to establish accurate Diagnosis, the improved continuity in medical care through
research and training mechanisms, simplification of medical procedures and the analysis of
medical inconsistencies and patterns in patient health to foster the development of new patient-
centered models.

Scalability: Global healthcare witnessed new dawn with the discovery of Telemedicine which
ensured that medical solutions could be provided outside hospital buildings. Virtual workflows
and technologies have been set up over time to create optimal infrastructures to implement this
process. The scalability of Virtual Diagnostics Solutions used in Telemedicine has been a
significant metric of development in our contemporary society. Evolutionary standards in AI
have ultimately ensured the continuous upgrade of Telecommunications, software, and digital
technologies used to implement virtual and real-time diagnostic solutions. A relative expansion in
the number of telemedical services provided by Virtual Diagnostic Solutions has spiraled the
frequent adoption of this paradigm in tackling global healthcare challenges such as epidemics,
pandemics, and critical health conditions.

Another reason for improvements in scalability is the ever-changing needs of healthcare

stakeholders to meet state-of-the-art next-generation service level requirements. Scalability could
entail efficient resource allocation, enhanced data integrated hospital-grade wearable devices,
expansion of patient management software systems, robust databases, cloud services, and data-
sharing infrastructures. As we advance, a perceived rise in the adoption of the telemedical model
of healthcare delivery will no doubt facilitate accessibility and efficiency with the sole purpose of
bringing a paradigm shift to Global health solutions.

6. CHALLENGES OF AI-BASED VIRTUAL DIAGNOSTIC SOLUTIONS (VDS)

IN TELEMEDICINE

Despite the numerous benefits promised by the integration of AI in Virtual Diagnostics Solutions
(VDS) to provide geographically accessible, affordable, acceptable, and quality healthcare, there
have also been challenges mitigating its full-scale adoption and development potential. As stated
earlier, this paper will highlight some key barriers facing the advancement of Virtual Diagnostic
Solutions in Telemedicine.

High-Cost: Although Virtual Diagnostic Solutions have been leveraged to reduce the high cost
of healthcare accessibility, this is only valid from a patient's perspective. While this issue might
seem improbable in our contemporary society, it is pertinent to note that ‘Cost’ sums up the
general resources required to develop, implement, maintain and advance Virtual Diagnostic
Solutions in Telemedicine. These resources include but are not limited to human, technical,
financial, and academic resources that have become relatively expensive to acquire and employ
over time. Cost implications for efficient initiation and delivery of telemedical research and
projects are seen as economic excess and, as such, given less consideration. This effect has led to
the relatively reduced advocacy for adopting telemedical solutions in most developing and
underdeveloped countries. From an AI perspective, huge amounts of technical resources,
 Computer Science & Information Technology (CS & IT) 239
skillsets, and training are required to build robust Telemedical deep learning models for Virtual
Diagnostic Solutions. Financial resources are expended on the acquisition, installation,
utilization, and maintenance of telemedical equipment. On a more holistic scale, this factor is the
most critical, responsible for the decline and a foreseeable reduction in VDS development and
adoption in Telemedicine.

Unavailability and Underdevelopment of Technical Infrastructures: Technology,

Telecommunication, and Artificial Intelligence advancements are the backbone for functioning
state-of-the-art Telemedical solutions, as have been elucidated in earlier sections of this paper.
Hence, the robust nature of Telemedicine has necessitated the requirement of sophisticated
infrastructures to ensure efficient and effective development and deployment. For instance, Tele-
ophthalmology, Teleradiography, and real-time emergency consultation, amongst others, are
some of the telemedical services which require heavy computing technologies and fast internet
connectivity. Systematically excellent and diverse medical data are also needed for training
Machine Learning (ML) and deep learning models to enhance effective generalization.
Technology literacy levels also play a part in ensuring telemedical services' smooth
implementation and sustainability by abstracting the process workflow to patients and medical
professionals. The unavailability of these requirements is predicted to pose a significant gap in
harnessing the potential of Telemedicine. On a more holistic scale, a lack of technical
infrastructures will more likely hinder the development of Telemedicine, thereby causing a
decline in the progress achieved so far. Despite the encouragement by the WHO encouraging the
adoption of Telemedical innovations by member states, a major drawback emanating from this
challenge has been observed in underdeveloped and developing countries.

Reliability Issues: A significant concern regarding the adoption and implementation of Virtual
Diagnostics Solution is its reliability and generalization ability. Given this issue's validity, several
Virtual Diagnostics solutions have come under severe scrutiny, raising questions and sentiments
concerning their utilization. Reliability is a dominant parameter in determining the rate of
adoption and research in Telemedical solutions. Hence, at any slightest detection of
untrustworthiness or inconsistency, Telemedicine could lose the attention and participation of key
stakeholders. Error-prone VDS solutions have stereotypically discouraged total reliability in the
potentials of Telemedicine. While a significant level of this challenge is due to lapses in
technicalities and operational models, several other factors can also contribute to this issue, such
as stakeholders’ resistance to accepting change, Digital illiteracy leading to poor awareness of
modern tech, cultural perceptions, and malpractice liabilities.

Ethical Violations, Confidentiality, and Privacy issues: Ever since the revolutionization of the
Internet, privacy and ethical policy stability have been a critical challenge among internet users.
Just like Medical professionals, patients (VDS users) need orientation and training on data access,
privacy and protection measures when utilizing services on Virtual Diagnostic Solutions. For
efficiency of Telemedicine, data privacy such as Doctor-Patient confidentiality, training, and
licensing of personnel is expected to be strictly adhered to. However, stability in providing robust
security infrastructures and interoperability features to tackle these challenges, coupled with the
ever-rising trend of cyber-crimes, has hindered the trustworthy adoption of Telemedicine as an
effective approach to healthcare access. Another propelling reason for this challenge is the result
of insufficient legal policies, guidelines, and Standard Operating Procedures (SOPs). In addition
to the absence of defined policies and regulatory procedures, a lack of international regulatory
uniformity has stirred several controversies regarding Telemedical services and solutions. While
AI’s propensity for good has been established in earlier sections of this paper, Malpractice
liability is another factor to consider. This issue has necessitated holistic reflections on AI's dual
potentials by governments, Researchers, and Engineers developing Telemedical solutions.
240 Computer Science & Information Technology (CS & IT)
Continuity/Sustainability: The recent global pandemic has necessitated the sustainability of
advanced medical technologies. As more medical issues arise, so should the pervasiveness of
healthcare approaches employed for treatment purposes. A major drawback, especially in
underdeveloped and developing countries, is their incapability to adequately sustain and
encourage advancements in Telemedicine either through research or system analysis. Several
factors can be attributed to this enthusiastic acceptance of the already existing Virtual Diagnostic
Solutions, formulation of ethical and privacy policies, standardization of technological
equipment, skill set and infrastructures, cost-effectiveness and coordination, etc. In light of this,
the sustainability and continuity of this ever-growing trend hinge on the improvements of all
participating stakeholders in creating stable infrastructures and approaches to tackle existing
challenges and foster the growth and application of AI in Virtual Diagnostic Solutions (VDS).

7. CONCLUSION
Artificial Intelligence has no small impact on Global health in our contemporary society. In most
developed and some developing countries, Telemedicine has gained popularity for its benefits in
improving healthcare access, reducing healthcare costs, and enhancing the quality of healthcare
services. These benefits necessitated the rapid paradigm shift from the usual traditional healthcare
(provider-centric) infrastructures to more robust (patient-centric) methodologies and
infrastructures, given the tremendous pressure on healthcare providers to provide affordable,
accessible, and quality healthcare services. This paper holistically discussed the growth and
revolution of Telemedicine as a modern research field; it also introduced a broad insight towards
implementing Artificial Intelligence in Telemedicine with specific inclinations toward services
provided by Virtual Diagnostic Solutions (VDS). A review of works of literature from
researchers citing several Deep Learning Approaches employed in detecting and treating several
medical ailments was also discussed in this paper, recommending the importance of deep
learning in the advancement of AI services. Several Applications of Artificial Intelligence in
sustainable healthcare provision and access were also reviewed and described.

The overall significance of this paper is to throw more light on the importance of DEEP
LEARNING-BASED AI methodologies in advancing state-of-the-art Virtual Diagnostic
Solutions in Telemedicine. Consequently, this paper enumerated and discussed crucial prospects
and challenges associated with incorporating, implementing, adopting, and advancing Artificial
Intelligence in Telemedicine. Further holistic research on advanced Telemedical approaches is
encouraged in tandem with these views.

ABBREVIATIONS
IoT - Internet of Things CAD - Computer-Aided Detection
DSA - DNA Sequence Analysis AI - Artificial Intelligence
VDS - Virtual Diagnostic Solutions MRI - Magnetic Resonance Imaging
ML - Machine Learning US - Ultrasound
CT - Computed Tomography AMA - American Medical Association
NLP - Natural Language Processing SOTA - State-of-the-Art
CAE - Computer-Aided Engineering RFCNN - Recurrent Fully Convolutional Neural
Network
ILD - Interstitial Lung Diseases SOPs - Standard Operating Procedures

DISCLOSURE
The authors declare that they have no competing interests with anyone in publishing this paper.
 Computer Science & Information Technology (CS & IT) 241
AUTHOR CONTRIBUTIONS

All authors made substantial contributions in conscripting the paper and revising it critically for
important intellectual content; agreed to submit it to the current journal; and final approval of the
version.

REFERENCES

[1] T. Khemapech, W. Sansrimahachai, and M. Toahchoodee, "Telemedicine–Meaning, Challenges, and

Opportunities," Siriraj Medical Journal, 2022. [Online]. Available:
http://dx.doi.org/10.33192/Smj.2019.38. [Accessed: 06- Jul- 2022].
[2] C. Kane, “AMA offers first national estimate of telemedicine use by physicians,” American Medical
Association, 2018. [Online]. Available: https://www.ama-assn.org/press-center/press-releases/ama-
offers-first-national-estimate-telemedicine-use-physicians. [Accessed: 12- Jun- 2022].
[3] Andr’e Pereira Alves, Tiago Marques Godinho, and Carlos Costa, “Assessing the relational database
model for optimization of content discovery services in medical imaging repositories,” 2016 IEEE
18th International Conference on e-Health Networking, Applications and Services (Healthcom), pp.
1-6, 2016.
[4] M. Chen and M. Decary, “Artificial intelligence in healthcare: An essential guide for health leaders,”
Healthcare Management Forum, vol. 33, no. 1, pp. 10-18, 2019. Available:
https://doi.org/10.1177/0840470419873123. [Accessed 2 June 2022].
[5] O. Oguine, K. Oguine, and H. Bisallah, “Big Data and Analytics Implementation in Tertiary
Institutions to Predict Students Performance in Nigeria,” Science Open, 2021. Available:
10.14293/s2199-1006.1.sor-.ppfhsfb.v1 [Accessed 21 June 2022].
[6] A. Banerjee, C. Chakraborty, A. Kumar, and D. Biswas, Emerging trends in IoT and big data
analytics for biomedical and health care technologies. Elsevier Inc., 2019.
[7] D. M. Mitch, E. D. Subido, and N. T. Bugtai. "Trends in telemedicine utilizing artificial intelligence"
AIP Conference Proceedings 1933, 040009 (2018); https://doi.org/10.1063/1.5023979 Published
Online: 13 February 2018.
[8] Shuai Zhang, S. I. McClean, C. D. Nugent, M. P. Donnelly, L. Galway, B. W. Scotney, and I.
Cleland, IEEE J. Biomed.Heal. Informatics 18(1), pp. 375–383 (2014).
[9] T. K. Kiong and A. S. Narayanan, “A telerehabilitation application with pre-defined consultation
classes,” in 2014International Conference on Circuits, Power and Computing Technologies
[ICCPCT-2014], (2014, pp. 1238–1244.
[10] L. Clifton, D. A. Clifton, M. A. F. Pimentel, P. J. Watkinson, and L. Tarassenko, IEEE J. Biomed.
Heal. Informatics,18(3), pp. 722–730(2014).
[11] R. LeMoyne, T. Mastroianni, A. Hessel, and K. Nishikawa, “Ankle Rehabilitation System with
Feedback from aSmartphone Wireless Gyroscope Platform and Machine Learning Classification,” in
2015 IEEE 14th InternationalConference on Machine Learning and Applications (ICMLA), (IEEE,
Miami, FL, 2015), pp. 406–409.
[12] F. Gorunescu, “Intelligent decision systems in Medicine? A short survey on medical diagnosis and
patientmanagement,” in 2015 E-Health and Bioengineering Conference (EHB), (IEEE, Iasi, 2015),
pp. 1–9.
[13] A. Grunerbl, A. Muaremi, V. Osmani, G. Bahle, S. Öhler, G. Tröster, O. Mayora, C. Haring, and P.
Lukowicz, IEEE J.Biomed. Heal. Informatics 19(1), pp. 140–148 (2015).
[14] J. Singh and A. K. Patel, “An effective telemedicine security using wavelet-based watermarking,” in
2016 IEEE (ICCIC), (IEEE, Chennai, 2016), pp. 1–6.
[15] A. Costanzo, A. Faro, D. Giordano, and C. Pino, “Mobile cyber-physical systems for health care:
Functions, ambientontology, and e-diagnostics,” in 2016 13th IEEE Annual Consumer
Communications & Networking Conference (CCNC), (LasVegas, NV, 2016) pp. 972–975.
[16] P. Kumar, S. K. Sharma, and S. Prasad, “Detection of fetal electrocardiogram through OFDM, neuro-
fuzzy logic andwavelets systems for telemetry,” in 2016 10th International Conference on Intelligent
Systems and Control (ISCO),(IEEE, Coimbatore, 2016), pp. 1–4.
[17] J. Finkelstein and I. C. Jeong, “Using CART for advanced prediction of asthma attacks based on
telemonitoring data,” in 2016 IEEE 7th Annual Ubiquitous Computing, Electronics & Mobile
Communication Conference (UEMCON), (IEEE, New York, NY, 2016), pp. 1–5
242 Computer Science & Information Technology (CS & IT)
[18] S. Nubenthan and C. Shalomy, “A wireless continuous patient monitoring system for dengue: Wi-
Mon,” in 2017 6thNational Conference on Technology and Management (NCTM), (IEEE, Nagercoil,
2017), pp. 23–27.
[19] House of Lords Select Committee on Artificial Intelligence (2018) AI in the UK: ready, willing, and
able?
[20] F. Saiz, and I. Barandiaran, "COVID-19 Detection in Chest X-ray Images using a Deep Learning
Approach". International Journal of Interactive Multimedia and Artificial Intelligence. In Press. 1.
10.9781/ijimai.2020.04.003, 2020.
[21] See http://www.ultromics.com/
[22] Z. Akkus et al., "Artificial Intelligence (AI)-Empowered Echocardiography Interpretation: A State-of-
the-Art Review." J. Clin. Med. 2021, 10, 1391. https://doi.org/ 10.3390/jcm10071391
[23] G. Bedi et al., “Automated analysis of free speech predicts psychosis onset in high-risk youths, NPJ
Schizophrenia, 1: 15030; IBM Research (5 January 2017) IBM 5 in 5”: with AI, our words will be a
window into our mental health.
[24] C. Bartholomeusz et al., "An fMRI study of the theory of mind in individuals with first-episode
psychosis," Psychiatry Research: Neuroimaging, vol. 281, pp. 1-11, 2018. Available:
10.1016/j.pscychresns.2018.08.011 [Accessed 26 June 2022].
[25] O. C. Oguine, K. A. Kinfu, K. J. Oguine, H. I. Bisallah and D. Ofuani, “Hybrid Facial Expression
Recognition (FER2013) Model for Real-Time Emotion Classification and Prediction”, arXiv.org,
2022. [Online]. Available: https://doi.org/10.48550/arXiv.2206.09509. [Accessed: 22- Jun- 2022].
[26] M. Fatehi, R. Safdari, M. Ghazisaeidi, M. Jebraeily, and M. Koolaee, “Data Standard for in Tele-
radiology.”, Acta Inform. Med. 23(3): 165- 168, 2015.
[27] E. J. Monteiro, C. Costa, and J. L. Oliveira, “A Cloud Architecture forTeleradiology-as-a-Service”.
Methods. Inf. Med. 55(3):203-14, 2016.
[28] D. J. Eedy, and R. Wootton,“Teledermatology: a review.” Br. J.Dermatol. 144:696-707, 2001.
[29] S. M. Landow, A. Mateus, K. Korgavkar, D. Nightingale, and M. A. Weinstock, “Teledermatology:
key factors associated with reducing face-to-face dermatology visits.” J. Am. Acad. Dermatol.
71(3):570-6, 2014.
[30] O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla,
M.Bernstein, et al.ImageNet Large Scale Visual Recognition Challenge. Int. J. Comput. Vis. 2015,
115, 211–252. [CrossRef].
[31] C. Qin, D. Yao, Y. Shi, and Z. Song, “Computer-aided detection in chest radiography based on
artificial intelligence: a survey.” Biomed Eng Online 17(1):113, 2018.
[32] H. C. Shin, H. R. Roth, M. Gao, “Deep Convolutional Neural Networks for Computer-Aided
Detection: CNN Architectures, Dataset Characteristics, and Transfer Learning.” IEEE Trans Med
Imaging 2016; 35:1285–98.
[33] R. P. K. Poudel, P. Lamata, and G. Montana,“Recurrent fully convolutional neural networks for
multi-slice MRI cardiac segmentation.” Vol. 10129 of Lecture Notes Comput Sci 2017:83–94
[34] E. Hosseini-Asl, R. Keynton, and A. El-Baz, “Alzheimer’s disease diagnostics by adaptation of 3D
convolutional network”, Arxiv.org, 2016. [Online]. Available: https://arxiv.org/pdf/1607.00455.pdf.
[Accessed: 20- Jun- 2022].
[35] L. Yu, X. Yang, H. Chen, J. Qin, and P. A. Heng, “Volumetric ConvNets with Mixed Residual
Connections for Automated Prostate Segmentation from 3D MR Images”. Thirty-First AAAI
Conference on Artificial Intelligence 2017; 66–72.
[36] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image
Segmentation.” MICCAI 2015;234–41.
[37] M. Gao, Z. Xu, L. Lu, A. P. Harrison, R. M. Summers, and D. J. Mollura, “Holistic Interstitial Lung
Disease Detection using Deep Convolutional NeuralNetworks: Multi-label Learning and Unordered
Pooling.” arXiv [Internet] 2017;9352:1–9. Available from: http://arxiv.org/abs/1701.05616/.
[38] A. Jamaludin, T. Kadir, and A. Zisserma, “SpineNet: Automatically pinpointing classification
evidence in spinal MRIs.” Vol. 9901 of Lecture Notes Comput Sci 2016:166-75.
[39] B. Jacobsmeyer, Focus: tracking down an epidemic’s source Physics 5: 89; 2012.
[40] World Health Organization, and WHO Telemedicine, “Opportunities and Developments in Member
States, Reports on the Second Global Survey on eHealth 2009”, Global Observatory for eHealth
Series, Volume 2, 2010
[41] Q. A. Qureshi, I. Ahmad, and A. Nawaz, “Readiness for E-Health in the Developing Countries Like
Pakistan,” Gomal Journal of Medical Sciences, Volume 10, No.1, Peshawar, Pakistan, 2012.
 Computer Science & Information Technology (CS & IT) 243
AUTHORS

Ozioma Collins Oguine is a Graduate Research Assistant at the University of Abuja,

Nigeria. He graduated from the same University with First Class Honors (Summa cum
Laude), top 1% from the Department of Computer Science. His research interests are
Machine/Deep Learning, Computer Vision, Robotics, and Human-Computer Interaction
(HCI). He is a Member of the Intelligent Automation Network (IAN), Black in AI,
Black in Robotics, an illustrious member of the International Society of Engineers
(IAENG) in Artificial Intelligence and Computer science, and an Associate Member of
the British Computing Society (BCS).

Kanyifeechukwu Jane Oguine is a Graduate Research Assistant at the University of

Abuja, Nigeria. She graduated from the same University with First Class Honors
(Summa cum Laude), top 2% from the Department of Computer Science. Her research
interests are Machine/Deep Learning, Computer Vision, Computational Algorithm, and
Human-Computer Interaction (HCI). She is a Member of the Intelligent Automation
Network (IAN), Black in Robotics, also a notable member of the International Society of
Engineers (IAENG) in Artificial Intelligence and Computer science, and an Associate
Member of the British Computing Society (BCS).