1
Evolution, Contemporary Approaches and Application of Artificial Intelligence in
Health Care
Student’s Name
Professor
Institution
Course
Date
2
Evolution, Contemporary Approaches and Application of Artificial Intelligence in
Health Care
1.0 Introduction
The purpose of this report is threefold: to give an account of the development of artificial
intelligence (AI) from its beginnings to this day, describe how it enables shifts in the healthcare
industry, and identify its potential and future. Therefore, this study aims to discuss the historical
development of AI, review the current approaches to AI, and assess the specific application of AI
in healthcare. Thus, this report presents readers with the possible advantages and limitations of
AI-based technologies when it comes to solving multifaceted issues in practice, specifically
within health care.
In the following sections, it is important to first look at the historical overview of AI and
the causes for its periods of popularity over the past half a century. The report discusses the
present main categories of AI, including machine learning and deep neural networks. It focuses
on healthcare, where the currently possible use of AI in diagnostics, treatment planning, and drug
discovery will be discussed. Last, this report presents a concrete case study of an AI application
in healthcare, presenting the technique behind its functioning, its efficacy, and its significant
concerns. By employing this systematic approach, we intend to give the readers an appreciation
of what AI can and cannot do in healthcare today and soon.
2.0 Historical Overview of AI
AI can be defined as the imitation of human intellect by machines, and the historical
origins of AI systems can be traced back to the period between 1950 and 1960. It was in 1956 at
the Dartmouth Conference, formulated and initiated by John McCarthy, Marvin Minsky,
Nathaniel Rochester, and Claude Shannon, that the term ‘Artificial Intelligence’ was used for the
3
first time and is considered today as the genesis of AI (Kaul et al., 2020, pp. 807-812). The first
AI research area was based on the symbolic processing of knowledge for reasoning, solving
problems, and analyzing natural language. One of the significant contributions during this period
was Arthur Samuel's checker-playing program, which represented an essential concept of
machine learning based on experience, and also Joseph Weizenbaum’s ELIZA, which was a
simple yet rather sophisticated natural language processing program that behaved like a person
or a psychologist as it responded based on pattern matching and substitution methodology.
However, the optimism of the AI movement decreased in the early 1970s, which led to
what is referred to as the first AI winter. It was marked by reduced funding and researchers'
attention to artificial intelligence development because of unfulfilled expectations and weak
computational capacities at that time. Many promises made at the beginning of the 1960s did not
come true, creating skepticism about AI's outlook. Difficulties in expanding early AI systems to
include real-world complexity were realized (Surden, 2019). Early approaches to AI proved to
have certain drawbacks, not least the inability to solve problems quite quickly for the human
brain, such as natural language understanding or visual perception. This time of disillusionment
saw the focus move to more realistic and specific objectives in the research being conducted in
artificial intelligence.
The 1980s became characterized by the AI resurgence due to the success of expert
systems and government and industry funding. Expert systems that encode the expertise of a
human into a rule-based pattern could be used in practical fields such as diagnosis and search for
mineral products (Kaul et al., 2020, pp. 807-812). New initiatives for funding computer science
research and development were triggered by the Japanese government's announcement of the
Fifth Generation Computer Systems project in 1982. This period also saw the end of an NP-
4
complete problem for training neural networks with the emergence of the backpropagation
algorithm, allowing for the training of multi-layer networks. This period saw an exponential
increase in companies' formation to exploit AI business ventures (Surden, 2019).
However, the optimism was short-lived, and the advancement of AI encountered another
roadblock in the late 1980s and early 1990s, resulting in the second AI winter. These created
problems of maintainability and update, rendering their applications rigid in the sense that they
could not be used for solving general problems (Haleem et al., 2019, pp. 231-237). Worse still,
there was a dismal show in the AI hardware market, most notably the Lisp machine segment,
thus aggravating the situation. Moreover, funding for the military and AI decreased with the end
of the Cold War. This period directed researchers to reconsider their approaches and slowly shift
to a more realistic view of the possibilities of AI.
The modern history of AI dates back only to the nineties of the twentieth century when
the focus was on the usage of data and machine learning. The availability of large datasets and
massive improvements in the power of computers are some of the reasons for changes in
problems in AI. The concepts of supervised learning were then introduced in various algorithms
applications (Haleem et al., 2019, pp. 231-237). The progress of science, especially with the
advent of the Internet network, allowed for filling AI systems with extensive data. During this
period they also included probabilistic methods and the appearance of Bayesian networks, which
provided new strategies for dealing with uncertainty in AI systems.
Several factors explain the variations in people's attitudes toward AI. This problem has
heavily contributed to the previous AI winters, as excessive promise led to a crash after the
failure to meet expectations. Historical AI systems had certain restrictions because of the
imperfect technologies they contained at that time regarding computing and available data. The
5
reasons were quite clear: the complexity of real-life problem-solving regularly defeated the
existing AI techniques of that period. Budget swings in response to economic conditions and
shifts in focus on the part of governments also helped to inflate and deflate the practice of AI
application (Haleem et al., 2019, pp. 231-237). However, each cycle is beneficial because it
brings more people into the rational expectations of what AI could do and what could be
reasonably expected of AI technologies.
3.0 Current Main AI Approaches
3.1 Machine Learning
Supervised machine learning (SML) has become one of the most important subfields of
contemporary artificial intelligence (AI). In its simplest form, it is about making a system learn
from experience to do a particular task better without being programmed to do so (Serban et al.,
2020, pp. 1–12). This statistical approach has received much success in a broad community in
areas such as image recognition and natural language processing.
Supervised learning is one of the two main ML categories and involves learning on data
already tagged with labels. In this particular approach, the algorithm aims to identify the
relationship between inputs and output labels to predict the required outputs given new inputs
that have not been previously seen. Its application is mainly seen in image recognition, spam
detection, and even predictions (Jiang et al., 2020, pp. 675–687; Shinde and Shah, 2018, pp. 1–
6). Supervised learning models like support vector machines, random forests, and gradient
boosting models have received high accuracy in most learning problems.
On the other hand, unsupervised learning works on raw data that is not tagged or
categorized and seeks to identify unknown relationships or structures in the data given.
Similarity-based clustering techniques, for example, K-means and DBSCAN, bring together data
6
points that are very similar, while other methods like PCA aid in the visualization of input data
with many dimensions (James et al., 2023, pp. 503-556). Exploratory data analysis and feature
learning are good areas in which to apply unsupervised learning.
Reinforcement learning (RL) is another category in which an agent strives to make a
decision and perform it in an environment. It focuses on the agent acquiring reinforcements or
penalties regarding the activities to gain efficiency in the long run. Seismic success in gameplay,
robotics, and autonomous systems has evidenced the capability of RL. New methods such as Q-
learning and policy gradient methods have brought AI to a level where it can now solve
previously thought impossible tasks (Ernst and Louette, 2024, pp. 111–126).
3.2 Deep Neural Networks
DNNs have been central to the advances of modern AI because they can dramatically
improve performance in areas such as image and speech recognition, natural language
processing, and game playing. A type of ANN, DNNs are composed of numerous interlinked
layers of neurons that can acquire increasingly complex data features. CNNs are now the go-to
model for most tasks involving images. Their basic architecture, which has convolution acts for
detecting local areas and pooling layers for reducing the spatial dimension, makes it perfect for
space-like data such as images. CNNs have reached super-human-level performance regarding
image recognition and have been applied in fields such as facial recognition, MRI analysis, and
self-driving cars.
RNNs, on the other hand, are meant to work on sequential data, making them helpful in
exercising time series and natural language processing. In contrast to feedforward networks,
RNNs have feedback connections that can either store ' memory' or self-recurrent connections,
which is why they can handle sequences of different lengths and catch long-term dependencies.
7
One kind of RNN, LSTM networks, has been widely useful in some tasks, including language
modelling, machine translation, and speech recognition.
Transformers are a robust architecture that can improve the performance of neural
networks, especially in natural language processing. As proposed in 2017 in the paper "Attention
Is All You Need," transformers utilize the self-awareness process to ascertain the relevance of
some parts of the input data. This eliminates the difficulties in processing long-range
dependencies in sequences, making it a better model than RNNs for many natural language
processing tasks. Recently, NLP models such as BERT, GPT, and T5 based on transformer
architecture have been benchmarks in several NLP tasks and are helpful in translation,
abstracting, and question-answering.
3.3 Natural Language Processing
Natural Language Processing (NLP): Here, the study revolves around how computers
process natural languages, or, in other words, the ability to comprehend, analyze, and even create
content in natural languages. The most beneficial developments Nevertheless, there are still
issues related to common-sense understanding, referring to a conversation, and the misuse of
these efficient language models. Deep learning, especially with transformer models, has
contributed to vast improvements in NLP. These include machine translation, sentiment analysis,
named entity recognition, and question answering. Many large language models, such as GPT-3,
have shown exemplary skills in setting realistic language content and diverse language-based
tasks without much further training. Nevertheless, issues are still related to common-sense
understanding, referring to a conversation, and the misuse of these efficient language models.
3.4 Computer Vision
8
Computer vision is one of the branches of AI whose ultimate goal is to allow computers
to view and comprehend data from the outside world. CNN has played a significant role in
progressing this field by attaining the capabilities of human-level performance in the following
tasks: image classification, object detection, and face recognition (Khan et al., 2021, pp. e4–e4).
The latest advancements are instance segmentation, where the objects in an image are detected
and where they begin and end at the pixel level. Generative models such as GANs, which are
Generative Adversarial Networks, provided new possibilities for image synthesis and
manipulation. Computer vision technologies are used in various fields, such as self-driving cars,
health and medical care, virtual reality, security, and others; however, some of their implications
include the violation of users' privacy and the introduction of prejudicial patterns into image
recognition data sets (Khan et al., 2021, pp. e4–e4).
3.5 Robotics and Automation
Robotics and autonomous systems are manifestations of artificial intelligence in the
physical environment, consisting of perception, learning, and controlling resources. With the
recent development in machine learning, especially in reinforcement learning, robots can learn
tasks on their own on a trial-and-error basis (Ribeiro et al., 2021, pp. 51–58). Thus, the relatively
fresh field of soft robotics, borrowing concepts from real life, has introduced new concepts in
safe human-robot interaction. Self-driving cars are no longer fiction; computer vision, sensors,
and decision-making make such vehicles closer to reality in autonomous cars. Nevertheless,
certain issues arise in navigating through an unstructured environment, making decisions based
on ethical approaches, and guaranteeing non-hazardous and steady performance in emergencies
(Ribeiro et al., 2021, pp. 51–58).
4.0 AI in Healthcare: A Contemporary State of Art
9
4.1 Diagnostic Applications
Artificial intelligence has introduced improved and quick diagnostic solutions in all
medical disciplines. Deep learning models have achieved high-performance levels in medical
image analysis, such as X-rays, MRI, and CT scans, and sometimes even higher than
radiologists' effects (Jelonek et al., 2020, pp. 319–327). These AI algorithms can quickly and
accurately spot previously unapparent irregularities in a patient's scans, doing their job better and
faster than a human observer could. Besides image interpretation, AI is used to diagnose and
manage intricate medical information: the big data of patients comprising electronic health
records, genetics, and lifestyles. Machine learning techniques can assess risk factors for different
diseases, which can help preventive measures be planned early. For instance, the models may be
used to estimate the probability of certain diseases like diabetes, heart disease, or specific types
of cancer given a patient's medical history and genetic makeup (Wang and Preininger, 2019, pp.
016-026; Haleem et al., 2019 pp. 231-237). This capability enables medical professionals to offer
early treatment to patients, thus increasing the chances of their recovery and, at the same time,
decreasing the total health cost.
4.2 Treatment Planning and Personalized Medicine
Artificial intelligence is gradually disrupting how treatments are approached with an eye
toward the unique and nuanced character of every patient's needs. Blood tests coupled with the
input of data on a patient's family history, general health status, behavior, reactions to
medications, and other aspects create a basis for developing personalized approaches to
treatment, bringing the healthcare industry closer to the management of patients as unique
individuals (Haleem et al., 2019 pp. 231-237). In oncology, AI systems use bioinformatics to
consider the genomics of a patient's tumor to suggest the best treatment for the patient and then
10
compare the genetic information with the patient's trial and results. In mental health, artificial
intelligence machines are used to evaluate the patient’s speech, social media output, and disease
data to help with disorders such as depression and anxiety disorder (Park et al., 2020). For
chronic conditions like diabetes, the system learns from data collected from continuous glucose
monitors, insulin injections, food intake, and physical activity for tailored recommendations for
insulin administration and behavioral changes. These specificities have the capability of
enhancing the overall quality of patient care and productivity of the procedures, thus enhancing
the probability of improved results among the patients (Jelonek et al., 2020, pp. 319–327;
Nielsen et al., 2020, pp. 791–798).
4.3 Drug Discovery and Development
The pharma industry uses AI to enhance drug discovery and development prospects and
possibly cut the time and expenses of bringing new drugs to market. AI systems help in library
screening for drug lead compounds by considering large databases on various chemical
compounds, proteins, and biological activities superior to conventional methods (Park et al.,
2020). Machine learning algorithms used in this regard help researchers foresee a new
compound's efficacy, toxicity, and other side effects and decide which of the compounds
predicted are worth further analysis. This capability has also empowered the discovery of new
antibiotics and possible treatments for some diseases. AI improves the process of selecting
patients and defining trials since it analyzes data on patients and outcomes of previous trials,
which may enhance the success rates of clinical trials and decrease the time and costs necessary
for trials (Reddy et al., 2019, pp. 22–28). AI also helps track trials and detect possible safety
issues. This progression could enhance the delivery of new treatments and help get promising
drugs to sufferers within a relatively short time.
11
4.4 Medical Imaging and Radiology
AI has created strong effects in medical imaging and radiology; deep learning algorithms,
especially convolutional neural networks (CNNs), have presented a superlatively excellent
performance in image classification, segmentation, and anomaly detection. AI systems of this
sort have demonstrated performances that are as effective as those of radiologists in tasks such as
the identification of breast cancer in mammograms, lung nodules in chest X-rays, and brain
tumors in MRI scans (Shinde and Shah, 2018, pp. 1–6). Besides detection, AI in medical images
helps measure and describe the images acquired. In neurology, AI algorithms segment structures
in the brain and quantify their volumes, which can help in reaching diagnosis and tracking
ailments such as Alzheimer's disease. In cardiology, AI learns how to interpret sonograms of the
heart muscles and their performance to ascertain if they are performing optimally or have any
complications (Reddy et al., 2019, pp. 22–28). AI integration pertains to improving the radiology
process and presenting fewer human errors and superior-quality diagnoses that can positively
impact the patient's condition (Jelonek et al., 2020, pp. 319–327).
4.5 Patient Care Tracking and Prognosis
Machine learning in patient surveillance and predictive analytics are already being
realized in the healthcare system to manage adverse events. Some of the applications of AI
include those in intensive care units, where the AI system constantly monitors and interprets data
obtained from a patient from different monitors and informs healthcare providers of any changes
that may be indicative of a patient's deterioration (Reddy et al., 2019 pp. 22–28; Secinaro et al.,
2021 pp. 1–23). Authorized in chronic disease management, AI-based remote monitoring enables
patients to be treated at home while in touch with their clinicians. These systems process
information collected from wearable devices, smartphones, and home-monitoring devices to
12
monitor patients’s conditions and signal an early-stage complication to clinicians. For instance,
in the case of heart failure, AI learns from data collected from implantable cardiac devices to
avoid hospitalization (Secinaro et al., 2021, pp. 1–23). At a population health level, AI models
analyze big health data to diagnose and see patterns of diseases in a given population. During the
COVID-19 crisis, AI was applied to predict infection trends, hospital bed usage, and the efficacy
of different interventions, proving the role of AI in imperative control of the highest public health
(Reddy et al., 2019, pp. 22–28).
4.6 Challenges and Ethical Considerations of AI in Healthcare
AI has great potential in healthcare; however, it has its issues and concerns when applied
in practice. One consideration is that bias can be introduced into the ML algorithms; this means
that if the training data does not represent various populations' needs, it may worsen the
inequality seen in the healthcare sector (Gerke et al., 2020, pp. 295–336). Another challenge is
data privacy and security because, in developing AI tools, it can be necessary to have access to
patients' data. Some algorithms for machine learning are black boxes that will create difficulties
in understanding the results and setting responsibility, which is fatal in healthcare. Some social
issues include the risk of deskilling human healthcare professionals due to reliance on AI systems
or overdependence on the machine's recommendations. Though governments and legislators
worldwide are developing regulations, there are no perfect rules regarding the development,
appraisal, and deployment of AI applications in clinical practice. Further related issues that must
be considered include the organization of the healthcare staff and healthcare as a whole, as well
as the patient-doctor interaction (Gerke et al., 2020, pp. 295–336). The confrontation of these
ethical and practical issues is necessary to unleash the potential of using AI to enhance health by
preserving confidence and fairness in the healthcare setting.
13
5.0 Critique of a Particular AI Solution in Healthcare
5.1 Description of the chosen AI solution
IBM Watson for Oncology is a web-based platform and clinical decision support system
developed with artificial intelligence to help oncologists develop the nuts and bolts of treatment
for cancer patients. As a concept, it was created with the support of IBM and Memorial Sloan
Kettering Cancer Center; its goal is to introduce reliable treatments based on patients'
characteristics (Shaheen, 2021). Watson for Oncology considers patients’s electronic health
records, which consist of clinical notes, lab results, and imaging studies, and then compares this
data with an extensive library of published literature, oncology guidelines, and specialists’
knowledge.
The system uses NLP to extract and analyze free-text medical data, while ML algorithms
analyze the data and find patterns that may not exist in the human eye of a clinician. It then
derives a ranked list of possible treatments and their corresponding evidence from literature and
clinical trial databases (Shaheen, 2021). The purpose is to enhance the oncologist's knowledge by
providing direct links to the latest treatments and protocols, which may improve the quality and
homogeneity of cancer treatment in various clinical settings.
5.2 Technical Analysis of the Used Approach
IBM Watson for Oncology fundamentally employs natural language processing (NLP),
machine learning, and cognitive computing methodologies. The relevance of the NLP part is to
identify and comprehend unstructured medical contents like clinic notes or research papers. It
contains complex programming that allows the tool to comprehend medical meanings and strip
the context to specific information to refer it to over a medical database (Darko et al., 2020, p.
14
103081). This helps the system handle medical literature and patient records and apply analytical
reasoning.
The machine learning component of Watson depends on data involving the various cancer
cases, treatments, and knowledge from their archives. The first applies supervised and
unsupervised learning algorithms to determine the regularities, relations, and possible treatment
strategies (Antonopoulos et al., 2020, p. 109899). The decision support issues allowed the system
to consider several elements and come up with more than one treatment solution, bearing in mind
the justification for doing so. This approach tries to mimic the decision-making process of
experienced oncologists, and at the same time, the agent has an extensive database that would be
hard for a human being to learn and update.
5.3 Benefits and Limitations
However, one of the primary uses of IBM Watson for oncology is its proficiency in
processing and analyzing large amounts of medical data, which could decrease the time
clinicians spend on the literature search and, respectively, increase the chances of making a
decision based on the latest evidence. It also can provide uniformity in the treatment being
offered in various performances, thereby minimizing quality differences in healthcare facilities
(Antonopoulos et al., 2020, p. 109899). Oncologists practicing in resource-constrained
environments or treating patients with rare diseases can benefit from Watson since they cannot
find similar clinical decision-support solutions. Nonetheless, it also has several drawbacks:
Similar to many AI and machine-learning algorithms, the relevancy and accuracy of Watson's
recommendations largely depend on the quality and representativeness of the official train data.
Some of the challenges arise from possible bias in the recommendations given by the system,
primarily to patient groups, which are poorly represented in the data used in the training process.
15
Moreover, some of Watson's algorithms, for example, the ones that involve deep analysis of text,
work in a 'black box' model, and therefore, it is hard for clinicians to verify the logic behind
particular suggestions (Darko et al., 2020, p. 103081). There are also questions about the
system's capacity to keep up with the fast pace of changes in medical knowledge and learn from
real-life clinical practice efficiently.
5.4 Real-World Performance and Case Studies
The adoption of IBM Watson for oncology on actual patients has shown varying
outcomes. As for the effectiveness of physician-patient communication, some previous research
findings suggest positive results. For example, an analytical study using the real-life test subjects
of the Manipal Comprehensive Cancer Centre in India revealed 93 percent compliance of Watson
treatment advice with the actual plan that the hospital's multidisciplinary tumor board came up
with regarding breast cancer cases (Serban et al., 2020, pp. 1–12). This implies that Watson could
help provide a second opinion, especially in practices where multidisciplinary tumor boards are
inaccessible.
However, other implementations have had issues. For example, the use of Watson at MD
Anderson Cancer Center in Texas was criticized due to system integration problems with clinical
practice and the system's capability to track the constantly changing developments in cancer
treatment (Shinde and Shah, 2018, pp. 1–6). Several oncologists have claimed that Watson
recommends radical treatment choices that ignore certain significant aspects of the patient's
status. Evaluations of AI solutions used in real-life medical environments also stress the constant
dynamic adjustment and fine-tuning of AI-based tools and the necessity of proper public
awareness about the strengths and weaknesses of AI as a tool that may facilitate clinical
decisions (Shinde and Shah, 2018, pp. 1–6).
16
5.5 Ethical and Regulatory Factors
Using AI systems such as IBM Watson for oncology presents various ethical and
regulatory issues. One primary concern is that AI could increase healthcare inequalities if the
training data and the algorithms are not diverse enough to capture a range of patients' characters
(Shinde and Shah, 2018, pp. 1–6). There are also issues with protecting data and information,
mainly where the health-related data handled by these systems is sensitive. The integration of AI
in healthcare brings problems of responsibility. If a patient has a poor outcome, it is unclear
whether this is an error of the AI system, healthcare provider, or hospital.
From a legal standpoint, the application of AI systems in the healthcare sector is
somewhat ambiguous. In the United States, through the FDA, efforts have been made to
determine the suitable regulatory models that can be applied to AI and machine learning-based
software as medical devices (Shinde and Shah, 2018, pp. 1–6). In today's climate, there is a risk
of slow and ineffective regulation of AI technology that can potentially harm patients. There is
also a lack of specification of procedures and protocols about the validation and subsequent
monitoring of AI within the context of clinical practice, along with a lack of consensus regarding
the requirements of transparency regarding the source and reasoning behind the AI-presented
recommendations.
5.6 Future Potential and Areas for Improvement
The future application possibilities of such an AI system as IBM Watson for oncology are
great. As the systems develop, the idea is to include real-time data intake from e-health records,
wearable devices, and genetic tests to individualize the treatment further. Another has to do with
its possibilities to expand its application in such aspects as treatment prognosis, risk detection,
and clinical trial choice (Serban et al., 2020, pp. 1–12).
17
Nevertheless, the actualization of this opportunity calls for efficiency improvement in key
areas for enhancement. Improving how the AI algorithm functions is important for developing
trust from care providers and users (Serban et al., 2020, pp. 1–12). The need for better
preparation data to make the systems more inclusive implies that proper representation of
training data can contribute significantly to AI's ability to deliver fair care to diverse patient
groups. A better integration of the artificial intelligence systems and the existing clinical
practices and electronic health record systems is also required. Nevertheless, future studies are
still essential for assessing the multiple effects of AI-based evidence in decision-making on the
total quality of the treatment, expenditures, and patients' conditions.
6.0 Conclusion
Artificial intelligence has come a long way, changing the healthcare industry through
improved diagnostic methods, tailored therapy, and new drugs. Artificial intelligence,
specifically machine learning and deep neural networks, form the basis of imaging, patient
monitoring, and predictive analysis. AI can be a powerful tool for medicine; however, its
application in oncology calls for ethics, better data, and compliance with current healthcare
structures. The future of AI is expected to help in the efficient treatment and recovery of patients,
enhance the organizational working of clinics and hospitals, invent new solutions in the medical
sciences, and much more. In the coming decades, continual improvement and development in
artificial intelligence will improve and probably revolutionize healthcare delivery by being
patient-centered, efficient, and accessible.
Reference list
Antonopoulos, I., Robu, V., Couraud, B., Kirli, D., Norbu, S., Kiprakis, A., Flynn, D., Elizondo-
Gonzalez, S. and Wattam, S., 2020. Artificial intelligence and machine learning
18
approaches to energy demand-side response: A systematic review.LRenewable and
Sustainable Energy Reviews,L130, p.109899.
Darko, A., Chan, A.P., Adabre, M.A., Edwards, D.J., Hosseini, M.R. and Ameyaw, E.E., 2020.
Artificial intelligence in the AEC industry: Scientometric analysis and visualization of
research activities.LAutomation in construction,L112, p.103081.
Ernst, D. and Louette, A., 2024. Introduction to reinforcement learning.LFeuerriegel, S.,
Hartmann, J., Janiesch, C., and Zschech, P.(2024). Generative ai. Business &
Information Systems Engineering,L66(1), pp.111-126.
Gerke, S., Minssen, T. and Cohen, G., 2020. Ethical and legal challenges of artificial
intelligence-driven healthcare. InLArtificial intelligence in healthcareL(pp. 295-336).
Academic Press.
Haleem, A., Javaid, M. and Khan, I.H., 2019. Current status and applications of Artificial
Intelligence (AI) in medical field: An overview.LCurrent Medicine Research and
Practice,L9(6), pp.231-237.
Haleem, A., Javaid, M. and Khan, I.H., 2019. Current status and applications of Artificial
Intelligence (AI) in medical field: An overview.LCurrent Medicine Research and
Practice,L9(6), pp.231-237.
James, G., Witten, D., Hastie, T., Tibshirani, R. and Taylor, J., 2023. Unsupervised learning.
InLAn introduction to statistical learning: with applications in PythonL(pp. 503-556).
Cham: Springer International Publishing.
Jelonek, D., Mesjasz-Lech, A., Stępniak, C., Turek, T. and Ziora, L., 2020. The artificial
intelligence application in the management of contemporary organization: Theoretical
assumptions, current practices, and research review. In Advances in Information and
19
Communication: Proceedings of the 2019 Future of Information and Communication
Conference (FICC), Volume 1L(pp. 319-327). Springer International Publishing.
Jiang, T., Gradus, J.L. and Rosellini, A.J., 2020. Supervised machine learning: a brief
primer.LBehavior therapy,L51(5), pp.675-687.
Kaul, V., Enslin, S. and Gross, S.A., 2020. History of artificial intelligence in
medicine.LGastrointestinal endoscopy,L92(4), pp.807-812.
Khan, A.A., Laghari, A.A. and Awan, S.A., 2021. Machine learning in computer vision: a
review.LEAI Endorsed Transactions on Scalable Information Systems,L8(32), pp. e4-e4.
Nielsen, A.N., Barch, D.M., Petersen, S.E., Schlaggar, B.L. and Greene, D.J., 2020. Machine
learning with neuroimaging: evaluating its applications in psychiatry.LBiological
Psychiatry: Cognitive Neuroscience and Neuroimaging,L5(8), pp.791-798.
Park, C.W., Seo, S.W., Kang, N., Ko, B., Choi, B.W., Park, C.M., Chang, D.K., Kim, H., Kim,
H., Lee, H. and Jang, J., 2020. Artificial intelligence in health care: current applications
and issues.LJournal of Korean medical science,L35(42).
Reddy, S., Fox, J. and Purohit, M.P., 2019. Artificial intelligence-enabled healthcare
delivery.LJournal of the Royal Society of Medicine,L112(1), pp.22-28.
Ribeiro, J., Lima, R., Eckhardt, T. and Paiva, S., 2021. Robotic process automation and artificial
intelligence in industry 4.0–a literature review.LProcedia Computer Science,L181, pp.51-
58.
Secinaro, S., Calandra, D., Secinaro, A., Muthurangu, V. and Biancone, P., 2021. The role of
artificial intelligence in healthcare: a structured literature review.LBMC medical
informatics and decision making,L21, pp.1-23.
20
Serban, A., Van der Blom, K., Hoos, H. and Visser, J., 2020, October. Adoption and effects of
software engineering best practices in machine learning. InLProceedings of the 14th
ACM/IEEE International Symposium on Empirical Software Engineering and
Measurement (ESEM)L(pp. 1-12).
Shaheen, M.Y., 2021. Applications of Artificial Intelligence (AI) in healthcare: A
review.LScienceOpen Preprints.
Shinde, P.P. and Shah, S., 2018, August. A review of machine learning and deep learning
applications. InL2018 Fourth international conference on computing communication
control and automation (ICCUBEA)L(pp. 1-6). IEEE.
Surden, H., 2019. Artificial intelligence and law: An overview.LGeorgia State University Law
Review,L35(4).
Wang, F. and Preininger, A., 2019. AI in health: state of the art, challenges, and future
directions.LYearbook of medical informatics,L28(01), pp.016-026.
