Advancements in Artificial Intelligence for Cardiovascular Disease Diagnosis and Risk Stratification

Full text

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
480
Vol. 2 No. 5 (2024)
Advancements in Artificial Intelligence for Cardiovascular
Disease Diagnosis and Risk Stratification
Muhammad Hammad u Salam1*
Department Computer Science &Information Technology, University
of Kotli, Azad Jammu and Kashmir. Corresponding Author Email:
hammad.salam@uokajk.edu.pk
Shujaat Ali Rathore2
Department Computer Science &Information Technology, University
of Kotli, Azad Jammu and Kashmir
Dr. Nasrullah3
Department of Computer Science & IT, University of Jhang,35200,
Jhang
Muhammad Azhar Mushtaq4
Department of Information Technology, University of Sargodha,
Sargodha Pakistan.
Tahir Abbas5
Department of Computer Science, TIMES Institute, Multan, 60000,
Pakistan
Abstract
The article analyses the use of artificial intelligence (AI) in the different
branches of cardiology including its use in predictive judgement,
diagnostics, and risk evaluation. It examines narrow (applied) AI and
general (strong) AI with special attention to the machine learning
methods applied to data from ECG, echocardiography, sonography,
CT, MRI, and PET scans. It also attempts to show how AI can improve
diagnostic accuracy, minimize medical blunders, and improve AI-
assisted treatment planning. It demonstrates how AI can be used for
the patient telephone calls in emergency medical services and, as part

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
481
Vol. 2 No. 5 (2024)
of preventive cardiology, for the analysis of enormous databases
received from tonometry, pulse wave velocity, and biochemical
investigations. Attention is also given to the issues concerning AI
integration, such as using LLMs for clinical documentation. Further,
ethical issues relating to AI including data privacy and culpability for
the actions of AI are discussed. Further, it has been highlighted that
the considered aspect needs more attention, therefore, patient data
security and AI tools integration in medical practice requires precise
regulation to ensure that facilitated patient outcomes do not come at
the cost of patient data breach. Necessary measures for successful
integration of AI tools into practice of cardiology directed to
improvement of patient care quality and healthcare systems
effectiveness are given in conclusion.
Keywords: Artificial Intelligence, Machine Learning, Cardiovascular
Diseases, Predictive Analytics, Diagnostic Tools, Risk Stratification,
Electrocardiography, Echocardiography, MRI, CT, Large Language
Models, Healthcare Integration, Ethical Issues.
Introduction
The technologies of artificial intelligence (A I) and machine learning
(ML) have been regarded as innovative tools in the diagnosis and risk
assessment of cardiovascular disease (CVD). Currently, the healthcare
sector is undergoing significant transformation with the incorporation
of AI, especially deep learning, owing to advanced systems in
cardiovascular diagnosis and patient outcome prediction. AI facilitates
the assimilations of multiple sources of clinical and imaging data,

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
482
Vol. 2 No. 5 (2024)
thereby increasing the accuracy and efficiency in the diagnoses and
risk assessments of heart diseases [1].
Many studies AI has recently begun incorporating machine learning
algorithms into risk prediction models at the cardiovascular level
utilizing traditional risk factors combined with modern clinical data.
These models were found to outperform many other models such as
the Framingham risk score [2]. For example, electrocardiographic data
is being fed into machine learning algorithms to automatically
identify specific heart disorders and predict likelihood of major
cardiovascular events such as myocardial infarction and stroke [3].
Apart from diagnostics, AI has also demonstrated its ability to
forecast patient outcomes such as long-term mortality and disease
progression involving the integration of biomarker, imaging, and
genetic information. Machine learning models using historical data
can automatically find structures in data that mere statistical
evaluations cannot [4]. In addition, AI as a continuously evolving
technology makes it possible to personalize treatment approaches for
CVD [5].
Despite the significant potential for AI in cardiovascular care,
there are competing issues that must be dealt with. There are ethical
issues surrounding patient privacy, data security, and transparency of
the algorithms used with respect to the AI's application in clinical
routines. The deployment of AI systems needs to be done carefully to
mitigate the risks of automated biases and misclassification of
patients with protected characteristics [26]. In addition, these models
need to be tested for validity and reliability when implemented in

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
483
Vol. 2 No. 5 (2024)
different populations and clinical settings to make sure they actually
work in the real world [27].
Recent developments in AI include automated systems for analyzing
medical images, including CT scans, MRIs, and even echocardiograms.
These systems which have been trained on massive datasets can pick
up the minute indications of diseases which human clinicians would
miss when looking at the images. AI is now being utilized more and
more in imaging cardiology for detecting coronary artery diseases
(CAD), heart failures, and valvular disorders with great precision [28].
Moreover, AI has also been demonstrated to have a role in stroke risk
estimation in patients suffering from atrial fibrillation or other forms
of arrhythmias, with promising results in clinical practice [29].
While headway is being made, some skepticism still remains
about placing so much trust in AI for clinical decision-making,
especially with something as sensitive as CVD. AI does have the ability
to help clinicians by making inferences closely based on data, but it is
crucial to remember that just like every other clinical decision, it
needs constant evaluation. Every patient’s case is different, and
granular details cannot be captured through algorithms and models
alone [30].
Literature Review
In recent years, substantial efforts have been put into the
development of artificial intelligence (AI) technology for its use with
the diagnosis and management of cardiovascular diseases (CVD). The
application of different machine learning (ML) models such as deep
learning and neural networks is now common in ECG,

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
484
Vol. 2 No. 5 (2024)
echocardiography, and other cardiac imaging, which has resulted in
improved accuracy and early diagnosis of the disease [6].These AI
models achieve enhanced clinical diagnosis by assisting human
experts with aid of trained medical images on large datasets for
pattern recognition that would otherwise not be possible manually [7].
AI-assisted interpretation of echocardiograms has enabled increased
detection of heart conditions that include but are not limited to,
myocardial infarctions and heart valve diseases, thus minimizing the
need for some diagnostic procedures to be performed invasively [8].
In addition, almost all of these processes have relied on AI to perform
risk stratification by using medical history, lifestyle, and available
biomarkers to predict adverse cardiovascular events for improving
treatment [9].
Support vector machines (SVM), random forests, and other
machine learning algorithms have been utilized to predict heart
failure and the onset of arrhythmias, enabling clinicians to make
important treatment and patient management decisions [10]. AI is
increasingly embedded into clinical workflows to assist healthcare
practitioners by providing in-depth analysis of patient information,
thereby speeding up diagnosis and primary management [11]. Such
systems are exceptionally helpful in emergency care situations where
a quick decision is needed. Moreover, AI is demonstrating its utility in
population health management by predicting severe cardiovascular
events, allowing healthcare practitioners to identify and intervene
within high-risk populations before the events occur [12].

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
485
Vol. 2 No. 5 (2024)
Nevertheless, the implementation of AI in cardiovascular care comes
with many hurdles. One major issue is the absence of quality data
that is essential for high-performance models. Several AI solutions are
built with non-generalizable data, which creates biases in diagnosis
and even treatment suggestions [13]. Alongside these concerns, the
incorporation of AI into healthcare also poses ethical and legal issues
such as privacy of patient data, understanding of the algorithms used,
and responsibility for the outcomes of AI interventions [14]. It is
always absolutely necessary to add the adoption of protocols with
stringent verification checks on AI systems to ensure reliability and
safety [15]. In any case, while anticipating the improvements with AI
adoption, it should not be overlooked that its use should always
augment and never substitute human clinical judgement. The
complexities surrounding the issue guarantee that AI tools will be
utilized alongside the insight of seasoned medical professionals for
the absolute welfare of the patients [16].
The possibilities of Artificial Intelligence seem to be endless. As
it progresses, its potential for transforming the detection and
management of cardiovascular ailments becomes more apparent.
Cardiovascular care is also promising with further developments in
the machine learning methods, big data analysis, and the
incorporation of AI in healthcare systems. Realizing the full benefits of
AI for patient outcomes will require addressing technical, ethical, and
regulation barriers [17].
AI, ML, and DL technologies are providing cardiology with
advanced tools for cardiovascular diagnosis, risk assessment, and

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
486
Vol. 2 No. 5 (2024)
treatment using sophisticated technologies that analyze vast datasets
like medical images and patient files to reveal patterns that are
usually unnoticed by physicians [18]. Especially, cardiac imaging has
witnessed the successful application of deep neural networks (DNNs)
that can facilitate the early diagnosis of arrhythmias and coronary
artery disease [19]. In addition, ML algorithms are being increasingly
used to integrate clinical data, biomarkers, and other variables for
predicting unfavorable patient outcomes such as heart failure and
stroke [20]. One of the prominent innovations is using AI-powered
tools for echocardiogram and electrocardiogram analysis, which
enhances the accuracy of diagnosis through the elimination of
tedious manual interpretation [21]. AI can also improve the quality of
clinical decisions made during patient care by making timely,
evidence-based recommendations and, in turn, minimize human error
and neglect [22]. Even with such advancements, integrating AI into
cardiology brings along some difficulties.
Certain challenges persist especially regarding the auditability
and transparency of AI systems for important healthcare operations
[23]. Moreover, the aspects concerning data protection, consent, and
the ethical boundaries of the use of AI in the medicine field are still
extensively searched and debated [24]. Owing to the myriad
opportunities that AI provides, achieving its success in cardiology is
only a matter of time, but the challenges have to be resolved to
optimize its clinical practice [25].
There are two categories of artificial intelligence (AI): weak AI and
strong AI. Weak AI, also known as narrow AI, is focused on

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
487
Vol. 2 No. 5 (2024)
performing a specific set of tasks. Specialized AI is currently the most
widespread form of AI, this approach to research and development
assumes that AI is a simulation and will always remain a simulation of
human cognitive functions. In this view, computers may only appear
to think but do not actually possess consciousness. Specialized AI
simply follows predefined rules imposed by its operator and cannot
go beyond these rules. A good example of applied AI is the
characters in video games, who behave realistically within the context
of their game characters but cannot perform any actions outside the
ones defined by the developers.
On the other hand, general AI or universal AI represents a form
of AI that closely mimics the general intelligence of the human mind,
capable of reasoning and formulating original thoughts. While
researchers are still studying universal AI, it remains a theoretical
benchmark, with no publicly known examples. However, for the sake
of clarity in the review, the types of AI will be differentiated based on
the strength of their intelligence.
AI and ML hold transformative potential in cardiovascular care,
offering improvements in diagnosis, treatment, and risk management.
However, the integration of these technologies into routine clinical
practice requires careful consideration of ethical, technical, and
regulatory challenges.
Methodology Of The Research
AI and Cardiology
The connection between AI and cardiology is currently evolving into a
comprehensive integration. Discussions among medical professionals

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
488
Vol. 2 No. 5 (2024)
about AI mostly focus on its potential applications in clinical decision-
making. These discussions primarily relate to algorithms used in
diagnosing and treating diseases, which predict certain outcomes
using diverse (multimodal) data sources.
Objective of the Review
The goal of the review is to identify specific applications of both
specialized and universal AI in the field of cardiology.
Stage I: Justification for the Study
Identifying and Analyzing Existing Systematic Reviews
The study opens with a systematic review of systematic reviews on
the application of artificial intelligence in cardiology. The scope of this
study is to find out what has already been covered in the research
and what gaps exist, as well as how the sophistication of the AI
methods used in cardiovascular diagnosis and risk stratification has
evolved over time. This includes reviewing published articles,
conducting clinical trials, and creating meta-analyses to thoroughly
comprehend the particular technology’s effectiveness in cardiology.
Defining the Basis and Characteristics of the Research Problem
The central research problem is to explore how AI can transform
cardiovascular disease (CVD) diagnosis and management. We aim to
identify both narrow AI applications (specific machine learning
models like support vector machines and artificial neural networks)
and general AI (large language models and automated decision-
making systems) used within clinical cardiology. This research
examines the full spectrum of AI’s role from diagnostic imaging to
predictive analytics and treatment personalization.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
489
Vol. 2 No. 5 (2024)
Stage II: Protocol Development
Developing a Clear Research Query
The research question in relation to the application of artificial
intelligence reads: "What is the application of artificial intelligence in
the diagnosis, risk assessment and treatment planning of patients
suffering from cardiovascular diseases?" The question is meant to
direct the systematic review and literature synthesis to examine and
evaluate the Application of AI in cardiology in regards to effectiveness,
challenges, and prospects.
Strategy for Identifying Primary Research Studies
The investigations center around primary peer-reviewed articles,
clinical trials, and preprint posters on the subject of AI applications in
cardiology and thus a comprehensive search strategy is employed.
Subsequently, the most important resources such as PubMed,
MEDLINE, SpringerLink, MedRXiv, and ResearchGate were used to
search for required materials. Some of the phrases that were searched
for include: "AI in cardiology," "machine learning in cardiovascular
diagnosis," “deep learning in CVD,” and ”AI risk stratification
cardiovascular.”
Inclusion/Exclusion Criteria
The inclusion criteria for the research are:
Publication Years: 1990-2024.
Source Type: Peer-reviewed articles, clinical studies, conference
proceedings, reviews, and preprints from recognized medical and
academic publishers.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
490
Vol. 2 No. 5 (2024)
Study Focus: Research that focuses on AI in diagnostic imaging,
risk prediction, and management of CVD.
Exclusion criteria:
Studies not related to AI applications in CVD.
Non-peer-reviewed studies or unpublished work that is not
available in public databases.
Data Extraction Scheme
Data will be extracted from the identified literature through an
extensive review of the articles. The focus will be on gathering
information on:
AI methodologies employed (e.g., deep learning, support vector
machines, neural networks).
Applications in various cardiology fields such as
electrocardiography (ECG), echocardiography, and magnetic
resonance imaging (MRI).
AI’s impact on improving diagnostic accuracy, early detection,
and risk prediction in cardiovascular diseases.
Development of Summarization Strategy
The categorization of AI applications in cardiology for purposes of
summarization will be done methodically. It will take into
consideration types of diagnostic tools, treatment predictions, and
healthcare management. Main highlights from the studies will be
captured under different headings such as AI for ECG interpretation,
AI in MRI and CT scans, and AI in personalized medicine and risk
mitigation.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
491
Vol. 2 No. 5 (2024)
Review of Collected Information by the Research Team
In order to thoroughly analyze and review the collected data, a team
approach will be taken to ensure that all perspectives on the
information are captured. The research team will evaluate and
integrate all the synthesized AI's application data focused on its use
in cardiology, while diagnosing major concerns like scarcity of clinical
AI validation and data bias in AI-supported systems.
Stage III: Literature Review and Data Study
Reviewing Literature on Clinical Studies
As part of this section, an AI application review specifically in
cardiovascular medicine will be conducted. This involves a literature
review of the selected articles and research papers including the
relevant clinical studies of the use artificial intelligence in
cardiovascular science. Primarily these studies will be evaluated based
on the research design, AI applications processes, and results. Specific
focus will be given to studies that assess the effectiveness of AI-based
diagnosis and risk assessment of cardiovascular diseases compared to
more traditional techniques.
Selection of Relevant Sources
The research team shall check all sources and, in particular, their
studies’ quality and relevance. Selected papers will be reviewed in full
text to determine whether they fall within the scope of the research
questions.
Data Extraction and Synthesis
The final step of the methodology will involve synthesizing the
extracted data and organizing it into themes based on specific

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
492
Vol. 2 No. 5 (2024)
cardiology domains. The collected data will be analyzed for trends in
AI usage, effectiveness, and areas where AI application still faces
limitations or challenges.
Machine Learning Technologies, Specialized AI In Cardiology
Machine Learning (ML) is the basic practice of using algorithms for
prediction by analyzing data and learning from them. Such models
can self-learn based on previous experiences or accumulated data.
Algorithms can extract important tasks from operator commands that
need to be executed by generalizing examples provided as training
datasets.
There are currently various types of ML algorithms. These are
grouped by learning style (i.e., supervised learning, unsupervised
learning, and semi-supervised learning), similarity, or function (i.e.,
classification, regression, decision trees, clustering, deep learning,
etc.). All ML algorithms consist of three different components:
Representation: A set of classifiers in a form understandable to the
computer.
Evaluation: The goal defined for the classifier model (algorithm),
scores.
Optimization: The method of finding the classifier with the highest
number of points.
The primary goal of ML algorithms is to generalize beyond the
training sets they are given, which implies successful interpretation of
data they have never processed before. The main difference between
ML and AI is that ML works to increase accuracy, while AI aims to
increase the chances of success.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
493
Vol. 2 No. 5 (2024)
The three main types of ML methods are: supervised learning,
unsupervised learning, and semi-supervised learning. The choice of
algorithm and learning type can be based on various approaches,
such as the problem at hand, the volume of data involved, or the
types of data available. ML demonstrates a dynamic role in medical
diagnostic applications, as it involves creating self-learning algorithms.
Figure.1 Model of AI Integration into Healthcare Institutions
Providing Cardiology Care
The Figure 1 shows one model of how artificial intelligence (AI) can be
implemented in the operations of healthcare facilities, particularly in
cardiology. The diagram has several interrelated clusters representing
specific areas where AI augments healthcare services. The AI
techniques cluster (yellow shaded) contains a number of machine
learning models including supervised / unsupervised, reinforcement /
self-learning that are necessary for the processing of healthcare data
and improving clinical predictive accuracy. The big language models

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
494
Vol. 2 No. 5 (2024)
division presents prospects of AI utilization in natural language
understanding to achieve better communication and decision-making
processes in medicine.
The healthcare integration cluster (blue section) shows how AI
can be embedded in health systems for data sharing and processing
using blockchain technology, and graphic data mining to improve
diagnosis. This integration is particularly important in making
informed clinical decisions in cardiology, one of the areas that require
accurate large scale data processing and personalized predictions and
care.
The cluster of healthcare tools and technologies (colored in
pink) indicates AI's role in the development of diagnostic tools such
as ECG, MRI, CT imaging, SPECT imaging, and echocardiography,
which are important for diagnosing heart issues. The role of AI in
telemedicine and wearable devices allows for remote supervision and
ongoing patient care, further improving accessibility and outcomes.
The primary results section (marked in red) depicts the value AI could
provide on vertical augmentation of patient care AI such as in
precision medicine, cost efficiency, and decision intelligence. AI, when
incorporated into treatment strategies, allows healthcare
professionals to deliver more personalized care in a cost effective
manner, which correlates to better patient outcomes over time. Lastly,
the other key areas in purple feature one of the major advancements
AI has revolutionized, automatic diagnosis and remotely supervised
continual monitoring, which is crucial for timely health evaluations as
well as health emergencies.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
495
Vol. 2 No. 5 (2024)
This model highlights the importance of cardiology AI applications
spanning from diagnosis, treatment and follow-up telemonitoring
providing integrated healthcare system for better medical services
and patient outcomes.
Cardiovascular Disease Prediction (CVD)
This involves supervised learning, as labeled data such as cardiograms
are required to train a weak AI model.
AI in ECG
The goal of implementing AI in ECG is to fully automate the analysis
of the cardiogram taken by medical personnel to save time for the
functional diagnostics doctor and reduce labor costs for specialists.
The use of both supervised and unsupervised learning for ECG
analysis and interpretation has shown significant potential.
Controlled ML, such as artificial neural networks and support vector
machines, can be trained for classification functions based on labeled
data. In contrast, unsupervised ML can detect potential correlations in
unlabeled data. In addition to ML, the recent emergence of deep
learning-based ECG analysis can assist doctors in various clinical
scenarios related to CVD diagnosis, prognosis, and risk stratification.
Instead of manually created vectors, deep learning systems employ
an end-to-end learning strategy that programs the system to learn
necessary functions from raw data. The advantage of deep neural
networks lies in their ability to identify new interdependencies
independent of manually selected feature extraction, opening vast
arrays of previously unexplored data.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
496
Vol. 2 No. 5 (2024)
AI in Echocardiography (EchoCG)
As with traditional methods slice recognition in echocardiography
remains an essential process in estimating the structure of the heart.
Identifying the sections may, however, pose a problem to doctors
that lack experience. The process of heart image classification has
been attempted to be automated with the help of Artificial
Intelligence employing convolutional neural networks. Investigations
indicate that certain deep learning methods are capable of classifying
EchoCG images with acceptable accuracy. The use of slices
recognition through AI will increase the speed of evaluation, improve
detection, and increase accuracy. The technology is known to help in
areas where there is a shortage of trained personnel. A functional
assessment of the left ventricle (LV) is an important aspect in the
diagnostics of EchoCG. Commercial software packages ensure that
measurements have maximum accuracy and good correlation with
MRI measurements.
New algorithms AI-based can now measure LV in record time,
by analyzing standard apical views with global longitudinal strain
imaging. Ai also aids in diagnostic imaging for young specialists and
significantly more accurate on mortality estimation compared to
manual measurement. For clinical diagnosis, precise segmentation of
the heart cavities is very essential. Automatic segmentation software
makes it possible to decrease the amount of manual work and
increase the objectivity of the obtained data. Artificial Intelligence is
now used to identify the endocardial wall and optimally estimate the
dimensions of the heart chambers. The LV is often segmented for the

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
497
Vol. 2 No. 5 (2024)
purposes of measuring the ejection fraction (EF) and evaluating
myocardial motion.
This process is time-consuming concerning data segmentation
needed for analysis. On the other hand, the automatic method suffers
from rapid movements, respiratory noise, and ultrasound artifacts
which distort the images.
Segmentation of the right ventricle (RV) is particularly difficult
because of the geometry of the heart, the thinner wall structure, and
lower quality of the images captured RV is problematic, however, it is
crucial both for the diagnosis and understanding what is happening
at the clinical level. Sparse matrix transformation and wall thickness
constraints serve as one of the potential solutions to combat poor
image quality. Attempts have also been made toward multichamber
models to enhance segmentation and incorporate the
communications between heart chambers. In lower quality images, RV
segmentation precision relies strongly on previously recorded data
and its correlation with other areas of the heart. At this time, utilizing
AI technology for better slice recognition, functional assessment
evaluation, and segmentation in EchoCG seems to yield faster and
better outcome accuracy for post diagnosis evaluation.
Cardiosonography and Auscultation, Heart Murmur Analysis
Using AI
AI is used to optimize the diagnostic potential of EchoCG. Below are
several ways AI is used in EchoCG. AI can accelerate the assessment of
patient conditions by automating deformation signal calculations, EF,
and other measurements. This type of automation can increase the

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
498
Vol. 2 No. 5 (2024)
throughput of ultrasound examination rooms. A blind randomized
study on heart function assessment using EchoCG and AI showed that
AI-guided workflow for initial heart function evaluation outperformed
the initial evaluation performed by a specialist. There has been a
significant increase in interest in using AI technology for heart
auscultation to identify coronary artery disease (CAD). AI algorithms
have made significant progress in detecting heart murmurs. However,
numerous clinical studies are required to confirm their accuracy
before recommending clinical use. Previous studies confirmed the
accuracy of diagnoses made by AI, although their sample sizes were
small. The largest study recorded heart tones of 1,362 patients, which,
to our knowledge, is the largest sample size. The authors compared
AI-assisted auscultation with auscultation performed by experienced
cardiologists and found that AI correctly identified heart tone
changes with 97% sensitivity and 89% specificity. AI auscultation has
high sensitivity and specificity for detecting abnormal heart sounds,
similar to those determined by cardiologists, making AI auscultation a
potentially useful screening tool for CAD. Compared to traditional
algorithms, the method uses convolutional neural networks to study
feature representation.
AI Analysis of CT and MRI Results
Accurate diagnosis is crucial for predicting the onset, progression,
and outcomes of CVD, which in turn reduces associated risks. The
most common and vital diagnostic tests include MRI and CT. High-
quality heart images are obtained using MRI and CT, but these have
long acquisition times, limited availability, and require radiation use.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
499
Vol. 2 No. 5 (2024)
AI in MRI Decoding
MRI is a vital non-invasive tool for assessing heart diseases. However,
it is time-consuming due to the need for high-quality images and
consideration of heart and respiratory movements. Methods such as
parallel imaging and compressed sensing have accelerated MRI by
collecting less data and using prior knowledge. Currently, AI,
particularly deep learning, is being explored to further speed up MRI.
Before AI can be widely used, technical challenges need to be
overcome. Bustin et al. (2024) provided a detailed review of recent AI
advancements assisting in MRI, focusing on deep learning for 2D and
3D images. The paper discusses limitations, issues, and potential
future directions of these methods. Of particular interest to
cardiologists is MRI phenotyping of heart failure and CAD.
AI in CT Imaging
The use of AI in heart CT is likely to play a key role in the future,
reducing reporting time, providing information on coronary
atherosclerotic plaques, and detecting myocardial ischemia by
evaluating perfusion.
AI for Heart PET Scan Analysis
Positron Emission Tomography (PET) provides an absolute
quantitative evaluation of myocardial perfusion. Results are usually
summarized in topological representations known as polar maps.
Currently, deep learning is improving classification performance
through iterative learning of complex multidimensional patterns.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
500
Vol. 2 No. 5 (2024)
Universal AI and its Applications in Cardiology
Although there are numerous examples of AI models developed for
medical use, in recent months, there has been a rise in public
awareness about large language models (LLMs), such as generative
pre-trained transformers like "ChatGPT," which was released in
November 2022 [32]. ChatGPT has been proclaimed a revolution [32]
in the field of AI. Trained on large datasets of text available on the
internet, AI models like LLMs seem to provide high-level answers. This
type of AI can demonstrate medical knowledge and generate
recommendations [33].
With advanced deep learning and natural language processing
capabilities, such as ChatGPT and subsequent versions, like the
generative pre-trained transformer-4, "GPT-4," they generate
conversational questions and answers resembling human-like
interactions and can be used in surveys for clinical use [34]. However,
it is important for users, including doctors and patients, to know how
to critically assess such LLMs, understand the regulatory environment,
and be generally aware of the potential weaknesses and hidden
dangers of using them [35].
Table 1 Processes that AI can Automate in Cardiology Institutions
[31]
Automatable
Processes
Activity Details
Medical Literature
Analysis
Evaluating and collating new work in the area
of treatment and therapy of cardiology.
Providing cardiologists with the most recent

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
501
Vol. 2 No. 5 (2024)
applicable research works.
Assistance in
Diagnosis and
Treatment Planning
Formulating possible diagnoses (including
differential diagnosis) and treatment
modalities considering his/her symptoms and
medical history.
Patient-Specific
Education
Developing individualized educational
materials for patients with prevalent
conditions such as myocardial infarction or
heart failure.
Risk Prediction and
Disease Progression
Estimating the risk in an individual for
development of cardiovascular diseases as
well as progression of these diseases based
on textual interpretation of x-ray findings.
Report and
Document
Generation
Creating reports and documents including
discharge summaries, referral summaries, and
transfer notes.
Identification of
Patterns in
Electronic Medical
Record Data
Detecting trends in patient data including
treatment queries raised regarding blood
pressure results and cholesterol levels.
Real-Time Language
Translation
Offering direct language interpretation during
consultations with patients from abroad.
Immediate translation for telemedicine
consultations with colleagues from other
countries.
Automated
Preparing interpretive summaries of patient

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
502
Vol. 2 No. 5 (2024)
Consultation
Summaries
consultations for doctors’ notes in regards to
the major points and anticipated actions.
Personalized
Lifestyle and
Medication
Recommendations
Advising patients on pertinent changes in
his/her lifestyle and treatment plans based on
his/her medical history, current condition and
local and international practices using lifestyle
medicine.
This Table.1 provides a detailed overview of the various processes in
cardiology institutions that can be automated with AI, covering
aspects from diagnosis assistance to real-time language translation.
LLMs in Cardiological Care Organizations
In healthcare institutions focused on cardiology, there are numerous
administrative AI applications. While AI use in this field is somewhat
less revolutionary compared to its use in diagnostics and patient care,
it can still significantly enhance the efficiency of healthcare and
preventive institutions (Table 1). On average, a nurse spends 25% of
their shift on administrative and regulatory tasks. Technologies most
likely to be related to this goal include universal AI and chatbots. They
can be used for various applications in healthcare, such as processing
claims, clinical documentation, economic activity management in
healthcare institutions, and managing automated medical record
systems, as well as appointment scheduling [31, 37].
Some organizations have used chatbots to interact with
patients within telemedicine frameworks. These natural language
models can be useful for simple tasks such as prescription updates or
appointment bookings. However, in a survey of 500 American users of

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
503
Vol. 2 No. 5 (2024)
the five most popular healthcare chatbots, patients expressed
concerns about privacy issues, discussing complex health conditions,
and the low convenience of use [31, 36, 37].
AI Chatbots in Cardiological Care Practices
As a strategic investment, AI indicates that changes in the current
healthcare delivery models will have to be made, if not new ones
created. The initial implementing of AI boasts low expenses, however,
the re-skilling of employees and renovation of hospitals and clinics
alongside the initial investment may raise the overall expenditure of
incorporating AI into clinical practices. Strategies are needed to shift
healthcare spending away from what currently accounts for a large
share of national GDP towards new developing technologies that are
impactful and cost-efficient. In simpler terms, when there is a scarcity
of resources available, the additional expenses that new technologies
incur have to be offset with their potential benefits and clinical
improvements such as an enhanced quality of life or longer lifespan.
These considerations are weighed against the expenditures that
socially acceptable cost-effectiveness ratios put forth for healthcare.
Few comprehend the economic aspects of cardiovascular AI
technologies [38, 39]. The speed at which AI technologies develop
and progress may need a new approach in assessment that is much
more effective than the traditional technology evaluation methods
used in the development of healthcare policy.
Discussion
The realm of cardiology Artificial Intelligence (AI) is progressing
rapidly and becoming an integral part of clinical practice. The

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
504
Vol. 2 No. 5 (2024)
diagnostic panorama of cardiovascular medicine is already changing
because of the widespread use of machine learning (ML) based
algorithms. Significant strides are already being seen in AI driven
electrocardiology (ECG), echocardiography, sonogrphy, and even
radiology. Machine learning algorithms assist in the analysis of
massive data sets, aid in uncovering previously undetectable patters,
and are superseded in the diagnosis of cardiovascular diseases (CVD)
using traditional techniques. For instance, AI is proving particularly
valuable in the more accurate interpretation of ECG and
echocardiogram images with assistance in detection of arrhythmias,
heart valve diseases, and even myocardial infarctions. Additionally,
the automation of such processes not only improves efficiency, but
also reduces human error enabling quicker and more reliable clinical
decision making.
The development of general artificial intelligence built on large
language models (LLMs) represents one of the most critical
advancements in AI technology. These models have great potential
regarding continuous medical education (CME), as they can provide
clinicians with summaries of relevant updates in research, treatment
techniques, and other useful information. These features can
tremendously increase the availability and relevance of information
for specialists, making it possible for them to improve their decision-
making while significantly decreasing the time incurred on literature
evaluations. LLMs may help in improving the quality of patient care
by facilitating the creation of tailored evidence-based treatment
strategies for patients. Such systems enable specialists to keep pace

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
505
Vol. 2 No. 5 (2024)
with contemporary developments, a concern that is often
encountered in rapidly advancing areas like cardiology, where
physicians are compelled to practice medicine on outmoded
information.
Another possible area of development is AI’s potential to
enhance processes such as completing and assessing medical
documentation. AI-powered tools are enhancing the delivery of
healthcare services by lessening the administrative work needed of
clinicians, thus, increasing the time available for actual patient
healthcare. AI has the potential to relieve human workload and
reduce errors that stem from manual data entry by automating
monotonous documentation activities. Furthermore, AI is being
incorporated into the automation of processes within different
departments, patient data collection and processing, as well as
enhancing diagnostic and prognostic activities through personalized
predictive medicine. This is critical since health systems are trying to
improve productivity due to the increasing patient load and limited
available resources.
However, this development in technology does come with a
catch, as there are several barriers that need to be sorted before LLMs
or any other AI technology can be used in clinical practice. One of the
major problems has to do with the ethical AI implications concerning
patient care. The use of the AI system raises serious concerns of data
privacy, data security, and possible algorithmic biases that may
detrimentally influence the quality of care given to various patient
groups. It is essential that the AI algorithms are adequately trained on

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
506
Vol. 2 No. 5 (2024)
a representative corpus to prevent the perpetuation of inequalities,
be it in diagnosis or treatment based on age, gender, or ethnicity.
Additionally, the need for clarifying reasoning behind decisions made
by AI systems is crucial to win the confidence of clinicians and
patients, with the former being particularly sensitive to such
“explanations.” The ability of many AI models to remain opaque raises
barriers to clinical acceptance, as healthcare providers may not be
willing to accept the validity of decisions, diagnoses, and treatment
options rendered by AI that is not understood by them.
Concerns also arise regarding the integration of AI in clinical
practice. Healthcare professionals seem to be skeptical when it comes
to fully adapting to AI-enabled tools as they do not trust their real-
world accuracy or even usefulness. Clinicians tend to fear that AI
might relieve them from judgment duties, especially in complicated
cases that require vast clinical knowledge. The acceptance of AI
technologies will progress only when the systems are deemed
intuitive enough not to obfuscate clinical problems. It is clear that AI
should not make decisions independently in which case the decision-
making process should be collaborative in nature. This way the skill of
clinicians is preserved along with their critical importance in the
healthcare system – decisions remain within the realm of
professionals.
Even with these challenges AI in cardiology looks promising. As
AI systems become more robust and diagnostic procedures,
treatment, and ethical considerations are refined, the prevention and
management of cardiovascular diseases can drastically improve. The

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
507
Vol. 2 No. 5 (2024)
application of AI in Clinical Decision Support Systems, Electronic
Health Records, and Predictive Modeling will enhance granularity of
patient care and health outcomes. However, for widespread AI
adoption in clinical practice, issues regarding data quality, model
explainability, and clinician-centeredness must be improved.
Insights Formed:
Increased Diagnostic Accuracies: AI applications in cardiology,
particularly ECG, echocardiogram videos, and medical imaging have
great improvements in CVDs detection rates.
Modern Continuing Medical Education: LLMs can change the
landscape of medical education by actively providing new and
updated information to clinicians regarding their work and field of
expertise.
Increased Productivity: Administrative tasks like medical
documentation, one of the more burdensome tasks in medicine, AI
powered technologies are optimistic in promoting efficiency.
Ethical and Professional Concerns: The challenges of data
confidentiality, algorithmic bias, and skepticism from professionals
need to be solved to ensure the integration of AI into clinical practice
takes place smoothly.
Future Potential: AI can currently assist in various activities, but
it has the most promise in decision support systems in personalized
cardiology, which then would lead to better patient outcomes.
Conclusion
AI technologies utilizing neural networks and similar constructs are
already in use by many cardiologists, radiologists, and functional

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
508
Vol. 2 No. 5 (2024)
specialists across the globe. In clinical practice, machine learning is
spreading to almost all areas of practice, including diagnostic imaging
and prognostic marking. However, AI applications and technologies
that have been clinically tested for prognosis and diagnosis are scarce
despite the increased uptake. Moreover, randomized controlled
studies investigating the effectiveness of these AI technologies are
still in their infancy.
Albeit the hope brought forth byAAI and “strong” AI systems
like large language models (LLMs) regarding the improvement of
service and care in the health sector, there are many technical and
ethical issues that still need to be dealt with before it can be put into
full practice. For LLMs, the following issues are still pertinent:
1. Data Collection and Recording: Algorithms intended for
standardization and recording of medical data often lack the
necessary precision. This combined with insufficient fine-tuning of the
algorithms leads to generation of clinical information that is either
inappropriate or irrelevant and thus renders the algorithms useless in
clinical settings.
2. Human Judgment and Experience: Although LLMs are capable of
providing assistance in clinical summaries or offering education to
patients, they are not able to account for the sophisticated human
judgment which is indispensable for making important medical
decisions.
3. Ethical Considerations: AI use in scientific research writing and
grant proposals raise the question of ethical authorship, responsibility,
and abuse of AI. The responsibility for the accuracy and integrity of

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
509
Vol. 2 No. 5 (2024)
AI-generate content lies with the experts and as such, requires
verification by control algorithms before use.
In summary, the use of AI in cardiology is very promising, but it
will require tackling issues related to technicalities, ethics, and
professional conduct before it can be easily integrated into clinical
practice. For success of AI in medicine, technological advancements
along with strong regulations will be vital.
References
[1]. M. Bhagawati, S. Paul, S. Agarwal, "Cardiovascular
disease/stroke risk stratification in deep learning framework: a
review," Cardiovascular Disease, 2023, doi:
10.1016/j.cvd.2023.03.005.
[2]. S. Awasthi, N. Sachdeva, Y. Gupta, AG Anto, "Identification and
risk stratification of coronary disease by artificial intelligence-
enabled ECG," eClinical Medicine, 2023, doi:
10.1016/j.eclinm.2023.03.010.
[3]. C. Ekhator, S. Ahmad, SN Mohsin, "The role of artificial
intelligence in prediction, risk stratification, and personalized
treatment planning for congenital heart diseases," Cureus, 2023,
doi: 10.7759/cureus.13479.
[4]. P. Mathur, S. Srivastava, X. Xu, "Artificial intelligence, machine
learning, and cardiovascular disease," Clinical Medicine, 2020, doi:
10.1177/1179546820927404.
[5]. JS Suri, S. Paul, MA Maindarkar, "Cardiovascular/stroke risk
stratification in Parkinson's disease patients using atherosclerosis

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
510
Vol. 2 No. 5 (2024)
pathway and artificial intelligence paradigm," Metabolites, 2022,
doi: 10.3390/metabo12040312.
[6]. Md Tanvir Rahman Tarafder, Md Masudur Rahman, Nisher
Ahmed, Tahmeed-Ur Rahman, Zakir Hossain, Asif Ahamed,
"Integrating Transformative AI for Next-Level Predictive Analytics
in Healthcare," 2024 IEEE Conference on Engineering Informatics
(ICEI-2024), Melbourne, Australia, 2024.
[7]. Asif Ahamed, Md Tanvir Rahman Tarafder, S M Tamim Hossain
Rimon, Ekramul Hasan, Md Al Amin, "Optimizing Load Forecasting
in Smart Grids with AI-Driven Solutions," 2024 IEEE International
Conference on Data & Software Engineering (ICoDSE-2024),
Gorontalo, Indonesia, 2024.
[8]. A. Nuthalapati, "Building Scalable Data Lakes For Internet Of
Things (IoT) Data Management," Educational Administration:
Theory and Practice, vol. 29, no. 1, pp. 412–424, Jan. 2023, doi:
10.53555/kuey.v29i1.7323.
[9]. SM T. H. Rimon, Mohammad A. Sufian, Zenith M. Guria, Niaz
Morshed, Ahmed I. Mosaddeque, Asif Ahamed, "Impact of AI-
Powered Business Intelligence on Smart City Policy-Making and
Data-Driven Governance," International Conference on Green
Energy, Computing and Intelligent Technology (GEn-CITy 2024),
Johor, Malaysia, 2024.
[10]. Abdullah Al Noman, Md Tanvir Rahman Tarafder, S M Tamim
Hossain Rimon, Asif Ahamed, Shahriar Ahmed, Abdullah Al Sakib,
"Discoverable Hidden Patterns in Water Quality through AI, LLMs,
and Transparent Remote Sensing," The 17th International

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
511
Vol. 2 No. 5 (2024)
Conference on Security of Information and Networks (SIN-2024),
Sydney, Australia, 2024, pp. 259-264.
[11]. Ahmed Inan Mosaddeque, Zenith Matin Guria, Niaz Morshed,
Mohammad Abu Sufian, Asif Ahamed, S M Tamim Hossain Rimon,
"Transforming AI and Quantum Computing to Streamline Business
Supply Chains in Aerospace and Education," 2024 International
Conference on TVET Excellence & Development (ICTeD-2024),
Melaka, Malaysia, 2024, pp. 231-236, doi:
10.1109/ICTeD62334.2024.10844659.
[12]. Asif Ahamed, Nisher Ahmed, Jamshaid Iqbal Janjua, Zakir
Hossain, Ekramul Hasan, Tahir Abbas, "Advances and Evaluation of
Intelligent Techniques in Short-Term Load Forecasting," 2024
International Conference on Computer and Applications (ICCA-
2024), Cairo, Egypt, 2024.
[13]. T. Abbas, J. I. and M. Irfan, "Proposed Agricultural Internet of
Things (AIoT) Based Intelligent System of Disease Forecaster for
Agri-Domain," 2023 International Conference on Computer and
Applications (ICCA), Cairo, Egypt, 2023, pp. 1-6, doi:
10.1109/ICCA59364.2023.10401794.
[14]. J. I., A. Sabir, T. Abbas, S. Q. Abbas and M. Saleem, "Predictive
Analytics and Machine Learning for Electricity Consumption
Resilience in Wholesale Power Markets," 2024 2nd International
Conference on Cyber Resilience (ICCR), Dubai, United Arab
Emirates, 2024, pp. 1-7, doi: 10.1109/ICCR61006.2024.10533004.
[15]. Nadeem, N., Hayat, M.F., Qureshi, M.A., et al., "Hybrid
Blockchain-based Academic Credential Verification System (B-

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
512
Vol. 2 No. 5 (2024)
ACVS)," Multimed Tools Appl 82, 43991–44019, 2023. doi:
10.1007/s11042-023-14944-7.
[16]. A. Rehman, F. Noor, J. I., A. Ihsan, A. Q. Saeed, and T. Abbas,
"Classification of Lung Diseases Using Machine Learning
Technique," 2024 International Conference on Decision Aid
Sciences and Applications (DASA), Manama, Bahrain, 2024, pp. 1-7,
doi: 10.1109/DASA63652.2024.10836302.
[17]. Asif Ahamed, Hasib Fardin, Ekramul Hasan, S M Tamim Hossain
Rimon, Md Musa Haque, & Abdullah Al Sakib. (2022). Public
Service Institutions Leading The Way With Innovative Clean Energy
Solutions. Journal of Population Therapeutics and Clinical
Pharmacology, 29(04), 4477-4495.
[18]. J. I., M. Nadeem and Z. A. Khan, "Machine Learning Based
Prognostics Techniques for Power Equipment: Comparative
Study," 2021 IEEE International Conference on Computing (ICOCO),
Kuala Lumpur, Malaysia, 2021, pp. 265-270, doi:
10.1109/ICOCO53166.2021.9673564.
[19]. B. Y. Almansour, A. Y. Almansour, J. I. Janjua, M. Zahid, and T.
Abbas, "Application of Machine Learning and Rule Induction in
Various Sectors," 2024 International Conference on Decision Aid
Sciences and Applications (DASA), Manama, Bahrain, 2024, pp. 1-8,
doi: 10.1109/DASA63652.2024.10836265.
[20]. A. M. A. Al-Tarawneh, R. A. AlOmoush, T. ul Islam, J. I. Janjua, T.
Abbas, and A. Ihsan, "Current Trends in Artificial Intelligence for
Educational Advancements," 2024 International Conference on

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
513
Vol. 2 No. 5 (2024)
Decision Aid Sciences and Applications (DASA), Manama, Bahrain,
2024, pp. 1-6, doi: 10.1109/DASA63652.2024.10836340.
[21]. M. A. Sufian, S. M. T. H. Rimon, A. I. Mosaddeque, Z. M. Guria, N.
Morshed, and A. Ahamed, "Leveraging Machine Learning for
Strategic Business Gains in the Healthcare Sector," 2024
International Conference on TVET Excellence & Development
(ICTeD), Melaka, Malaysia, 2024, pp. 225-230, doi:
10.1109/ICTeD62334.2024.10844658.
[22]. Khan, T.A., Abbas, S., Ditta, A., Khan, M.A., Alquhayz, H., Fatima,
A. and Khan, M.F., 2020. IoMT-Based Smart Monitoring
Hierarchical Fuzzy Inference System for Diagnosis of COVID-19.
Computers, Materials & Continua, 65(3).
[23]. T. M. Ghazal, J. I. J, W. Abushiba, and S. Abbas, "Optimizing
Patient Outcomes with AI and Predictive Analytics in Healthcare,"
2024 IEEE 65th International Scientific Conference on Power and
Electrical Engineering of Riga Technical University, Riga, Latvia,
2024, pp. 1-6, doi: 10.1109/RTUCON62997.2024.10830874.
[24]. S. B. Nuthalapati, M. Arun, C. Prajitha, S. Rinesh and K. M.
Abubeker, "Computer Vision Assisted Deep Learning Enabled Gas
Pipeline Leak Detection Framework," 2024 5th International
Conference on Smart Electronics and Communication (ICOSEC),
Trichy, India, 2024, pp. 950-957, doi:
10.1109/ICOSEC61587.2024.10722308.
[25]. J. I, S. Zulfiqar, T. A. Khan and S. A. Ramay, "Activation Function
Conundrums in the Modern Machine Learning Paradigm," 2023
International Conference on Computer and Applications (ICCA),

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
514
Vol. 2 No. 5 (2024)
Cairo, Egypt, 2023, pp. 1-8, doi:
10.1109/ICCA59364.2023.10401760.
[26]. A. Lin, M. Kolossváry, M. Motwani, "Artificial intelligence in
cardiovascular imaging for risk stratification in coronary artery
disease," Radiology, 2021, doi: 10.1148/ryct.2021200512.
[27]. Abbas, T., Khan, A. H., Kanwal, K., Daud, A., Irfan, M., Bukhari, A.,
& Alharbey, R. (2024). IoMT-Based Healthcare Systems: A Review.
Computer Systems Science & Engineering, 48(4).
[28]. S. Shu, J. Ren, J. Song, "Artificial intelligence for disease
diagnosis and risk prediction in nuclear cardiology," Journal of
Nuclear Cardiology, 2022, doi: 10.1007/s12149-022-01653-w.
[29]. P. Maurovich-Horvat, "Artificial intelligence: practical primer for
clinical research in cardiovascular disease," Journal of the
American Heart Association, 2019, doi: 10.1161/JAHA.119.012788.
[30]. S. Shrestha, D. Gupta, L. Saba, "Machine learning-aided risk
stratification system for the prediction of coronary artery disease,"
Computers in Biology and Medicine, 2020, doi:
10.1016/j.compbiomed.2020.104043.
[31]. I. Skalidis, A. Cagnina, and S. Fournier, "Use of large language
models for evidence-based cardiovascular medicine," Eur. Heart J.
Digit. Health, vol. 4, no. 5, pp. 368–369, 2023. doi:
10.1093/ehjdh/ztad041.
[32]. A. Sarraju, D. Ouyang, and D. Itchhaporia, "The opportunities
and challenges of large language models in cardiology," JACC:
Adv., vol. 2, no. 7, p. 100438, 2023. doi:
10.1016/j.jacadv.2023.100438.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
515
Vol. 2 No. 5 (2024)
[33]. A. Sarraju, D. Bruemmer, E. Van Iterson, et al., "Appropriateness
of cardiovascular disease prevention recommendations obtained
from a popular online chat-based artificial intelligence model,"
JAMA, vol. 329, pp. 842–844, 2023. doi: 10.1001/jama.2023.1044.
[34]. P. Lee, S. Bubeck, and J. Petro, "Benefits, limits, and risks of GPT-
4 as an AI chatbot for medicine," N. Engl. J. Med., vol. 388, pp.
1233–1239, 2023. doi: 10.1056/NEJMsr2214184.
[35]. E. Wu, K. Wu, R. Daneshjou, et al., "How medical AI devices are
evaluated: limitations and recommendations from an analysis of
FDA approvals," Nat. Med., vol. 27, pp. 582–584, 2021. doi:
10.1038/s41591-021-01312-x.
[36]. T. Davenport and R. Kalakota, "The potential for artificial
intelligence in healthcare," Future Healthc. J., vol. 6, no. 2, pp. 94–
98, 2019. doi: 10.7861/futurehosp.6-2-94.
[37]. Abbas, T., Fatima, A., Shahzad, T., Alharbi, M., Khan, M. A., &
Ahmed, A. (2024). Multidisciplinary cancer disease classification
using adaptive FL in healthcare industry 5.0. Scientific Reports,
14(1), 18643.
[38]. M. Smith, A. Sattler, G. Hong, and S. Lin, "From code to bedside:
implementing artificial intelligence using quality improvement
methods," J. Gen. Intern. Med., vol. 36, no. 4, pp. 1061–1066, 2021.
doi: 10.1007/s11606-020-06394-w.
[39]. L. Petersson, I. Larsson, J. M. Nygren, et al., "Challenges to
implementing artificial intelligence in healthcare: a qualitative
interview study with healthcare leaders in Sweden," BMC Health
Serv. Res., vol. 22, no. 1, p. 850, 2022.

Spectrum of Engineering Sciences
Online ISSN
3007-3138
Print ISSN
3007-312X
516
Vol. 2 No. 5 (2024)