See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/347439149

Medical diagnostic systems using AI algorithms

Article in IEEE Access · December 2020

CITATIONS READS

0 640

4 authors, including:

Jimmy Singla Deepak Prashar

CT University Lovely Professional University
85 PUBLICATIONS 778 CITATIONS 97 PUBLICATIONS 915 CITATIONS

SEE PROFILE SEE PROFILE

Gyanendra Prasad Joshi

Sejong University
151 PUBLICATIONS 1,985 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jimmy Singla on 17 December 2020.

The user has requested enhancement of the downloaded file.

 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI

Medical Diagnostic Systems Using

Artificial Intelligence (AI) Algorithms:
Principles and Perspectives
SIMARJEET KAUR1 , JIMMY SINGLA1 , LEWIS NKENYEREYE2 , SUDAN JHA1 , DEEPAK
PRASHAR1 , GYANENDRA PRASAD JOSHI3 , SHAKER EL-SAPPAGH4,* , MD. SAIFUL ISLAM5 ,
S. M. RIAZUL ISLAM3,*
1
School of Computer Science and Engineering, Lovely Professional University, Punjab 144001, India
2
Department of Computer and Information Security, Sejong University, Seoul 05006, South Korea
3
Department of Computer Science and Engineering, Sejong University, Seoul 05006, South Korea
4
Information Systems Department, Faculty of Computers and Artificial Intelligence, Benha University, Banha 13518, Egypt
5
School of Information and Communication Technology, Griffith University, QLD 4222, Australia
*Corresponding author: Shaker El-Sappagh (E-mail: [email protected]) and S. M. Riazul Islam (E-mail: [email protected])
This work was supported in part by National Research Foundation of Korea-Grant funded by the Korean Government (Ministry of Science
and ICT-NRF-2020R1A2B5B02002478), and in part by Sejong university through its faculty research program.

ABSTRACT Disease diagnosis is the identification of an health issue, disease, disorder, or other condition
that a person may have. Disease diagnoses could be sometimes very easy tasks, while others may be a
bit trickier. There are large data sets available; however, there is a limitation of tools that can accurately
determine the patterns and make predictions. The traditional methods which are used to diagnose a
disease are manual and error-prone. Usage of Artificial Intelligence (AI) predictive techniques enables
auto diagnosis and reduces detection errors compared to exclusive human expertise. In this paper, we
have reviewed the current literature for the last 10 years, from January 2009 to December 2019. The
study considered eight most frequently used databases, in which a total of 105 articles were found.
A detailed analysis of those articles was conducted in order to classify most used AI techniques for
medical diagnostic systems. We further discuss various diseases along with corresponding techniques of
AI, including Fuzzy Logic, Machine Learning, and Deep Learning. This research paper aims to reveal
some important insights into current and previous different AI techniques in the medical field used in
today’s medical research, particularly in heart disease prediction, brain disease, prostate, liver disease, and
kidney disease. Finally, the paper also provides some avenues for future research on AI-based diagnostics
systems based on a set of open problems and challenges.

INDEX TERMS Big Data Analytics, Artificial Intelligence, Machine Learning, Deep Learning, Soft
Computing, Chronic Disease, Diagnosis, Health Care Prediction

I. INTRODUCTION the method of figuring out which disease is based on an

In the field of healthcare, the study of disease diagnosis individual’s symptoms and signs [5], as shown in Fig. 1. The
plays a vital role. Any cause or circumstances that lead to data gathered from medical history physical examination
pain, illness, dysfunction, or eventually, a human being’s of the individual having medical pathology constitutes the
death is called a disease. Diseases may affect a person knowledge required for diagnosis. Often, at least one diag-
physically and mentally, and it considerably manipulates nostic procedure, such as medical tests, is done during this
the living style of the affected person. The causal study of procedure. To form an honest diagnosis, a medical doctor
disease is called the pathological process [1]. A disease is will perform a process that involves several steps, allowing
made by signs or symptoms that are interpreted by clinical them to collect the maximum amount of information as
experts [2]–[4]. Diagnosis has been defined as the method possible [6]. Diagnosis of diseases is the most challenging
of identifying a disease from its signs and symptoms to process at the same time, a very pivotal phenomenon for a
conclude its pathology. Diagnosis can also be defined as medical care professional as before reaching the conclusion.

VOLUME 4, 2016 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

The diagnostic process could be very tiresome and complex. detect lung cancer, breast cancer, Stroke in less time than
To minimize the uncertainty in medical diagnosis health, the a trained radiologist. Various AI algorithms help doctors
care experts collect empirical data to ascertain a patient’s to analyze medical images such as MRIs, x-rays, and CT
disease. The patient’s correct treatment may be adjourned scans and diagnose specific diseases by just spotting signs.
or missed with serious health issues due to making fault in Detection of disease and providing correct treatment is
the diagnosis process. Unfortunately, all doctors don’t have always a tricky and complex process since some diseases
expert knowledge in each domain of the medical field. have very similar signs. Using medical expert systems,
Hence, there was a need of automatic diagnostic system doctors can diagnose patients more accurately and prescribe
that provides benefits from both human knowledge and the most suitable treatment. Using AI tools, doctors can not
accuracy of the machine [7]. A suitable decision support only detect the disease but can also classify the types of
system is needed to achieve accurate results from the diag- different fatal diseases. Modern AI algorithms already help
nosis process with reduced costs. Classification of diseases doctors in arranging a comprehensive approach to disease
depending upon various parameters is a complex task for management. Moreover, they are often used to improve
human experts but AI would help to detect and handle such surgical robots that execute highly complex operations. The
kinds of cases. Currently, various AI techniques have been contributions of the this paper is three folds
used in the field of medicine to accurately diagnosis sick- • We first describe the existing elements that affect the
nesses. AI is an integral part of computer science by which initial outbreak of disease detection.
computers become more intelligent. The vital need for any • We latter discuss how AI techniques have been altered
intelligent system is learning. There are various techniques for initial disease diagnosis
in AI that are based on Learning like deep learning, machine • We provide a thorough analysis through a systematic
learning, etc. Some specific AI methods that are significant review for medical diagnostic systems. We make use
in the medical field named as a Rule-based intelligent sys- of the well know PRISMA approach.
tem, provides a set of if-then rules in healthcare, which act as • We then provide a summary for all the selected articles;
a decision support system. Gradually, intelligent systems are the diseases which were targeted,the AI techniques
being replaced in the medical field by AI-based automatic which were used, the articles’ research goals along with
techniques where human intervention is very less [8], [9]. their findings. We also present a thorough discussion
The neural network or artificial neural network (ANN) is a of the reviewed articles followed by future research
large collection of neural units designed based on biological directions.
neurons connected in the brain. It is a simulation of the The rest of this paper is organized as follows. We present
human brain and works exactly like it. Each neural unit is the related works on AI applied methods for medical diag-
linked with many other neurons approximately similar to nostic systems in section II. Whereas Section III discusses
the bipartite graph [10]. These kinds of systems learn and fuzzy logic-based medical diagnosis, Section IV and Section
are trained automatically. V present diagnostic systems using machine learning and
Finding the possibilities and predictions regarding health deep learning algorithms, respectively, In Section VI, we
issues is a tedious task for doctors and surgical experts. In present the review guidelines using the Prisma technique.
some cases, ANN provides decisions regarding healthcare at Research findings, discussion, and future research directions
rapid speed wherein the systems can collect data, understand are included reported in Section VII, whereas the final
it, and detect pieces that will play a vital role in prediction Section concludes this review.
[11], [12]. Deep learning, a subset of machine learning and
also based on algorithms, is used in the medical field to II. RELATED WORKS
assist specialists for the examination of any illness. Thus, In this section we discuss current applied AI techniques
resulting in better medical decisions. Deep learning provide which are used for disease diagnostic process, relevant
benefits in different fields such as drug discovery, medical survey articles on diagnostic process and our contribution
imaging, Genome, detecting Alzheimer’s disease [13]. In in regards to the existing work.
this paper, we primarily focus on the three main branches Van Mourik et al. [14] carried out a survey on automated
of AI: Fuzzy logic, Machine learning, Deep Learning. The surveillance techniques for healthcare-associated infections.
major trend in healthcare using deep learning is to detect In this review, authors have described how automatic
breast cancer. In a recent study conducted by a cancer surveillance systems based on machine learning algorithms
institute, it is clear that the accuracy of Automatic breast bring enhanced performance and reliability compared to
cancer is equal/high than a human radiologist. Moreover, manual surveillance methods. Another finding of this review
AI trained itself continuously and have greater chances to is that the use of regression models can improve the effi-
produce more accurate results than before. Another signifi- ciency and sensitivity of surveillance programs. There are
cant application of AI is the Internet of Medical Things that some challenges that need to be addressed in the near future
helps to collect healthcare data using IOT Devices. AI-based such as post discharge surveillance, case-mix adjustment,
software detects the disease even before its occurrence by quantification of device utilization. BRONCHIOLITIS is a
sensing its symptoms. Neural networks can be trained to lung infection that is commonly seen in younger children
2 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 1. Block diagram of the diagnosis process

and infants. Luo et al. [15] reviewed this disease along with examination report. These sorts of frameworks provide an
respiratory syncytial virus (RSV), an infection that can be a instant and straightforward strategy for clinical assessment.
root cause of bronchiolitis. The systematic review provides They are also useful where an expert or clinical specialist
some insights into predictive modeling and also reported is absent. These frameworks give an outcome depending on
how machine learning can use to overcome limitations of the knowledgebase incorporated within or from specialists
predictive modeling. SEPSIS is a life-threatening condition or experts in the field. Various clinical diagnoses systems
that occurs due to your body’s response to infection, which created depend on the fuzzy set model and applied in the
causes inflammation that result in multiple organ failures medical field [22]. The word fuzzy refers to things that are
at the same time. Bhattacharjee et al. [16] performed a ambiguous. Sometimes we face a circumstance when we
systematic review to investigate the current trends in sepsis are uncertain about whether the state is valid or invalid,
detection in hospitals. Authors have investigated various wherein fuzzy logic provides reasoning for such conditions
sepsis detection scoring systems and screening tools along as depicted in Fig. 2. It is a rule-based method. Fuzzy Rule-
with their pros and cons in general hospital wards. Finally, Based System (FRBS) is a frequently used technique in
they observed biomarkers and electronic health records can healthcare that drives from Fuzzy Inference Systems (FIS).
have a huge effect in predicting sepsis. One more study on FRBS applies IF-THEN rules for information portrayal [23].
sepsis was performed by Sinha et al. [17]. They reported Besides this, clustering and classifying techniques are also
some drawbacks in routine blood culture testing for sepsis used in the medical domain. Also, FIS and FDSS are
detection. To analyzed suitable automatic sepsis detection determined as the most common techniques in the area of
methods that they investigated seven molecular technologies medicine [24]. The main feature of fuzzy logic is that it can
that utilize blood samples. In this study, they have discussed alleviate the inaccuracies and uncertainties of any situation.
the various present and future trends. In addition, they have There is no logic for the absolute valid and absolute invalid
also analyzed the impact of machine learning algorithms value, but partially true and partially false intermediate value
with electronic medical records in sepsis detection. They exists in a fuzzy logic system. Let’s take the following
conclude that by merging various technologies can improve example to show how fuzzy logic works.
the detection process and minimize the risk of using the In the past few years, Fuzzy logic is consistently gaining
wrong antibiotic. popularity in diagnosing disease based on different parame-
To the best of the knowledge, this is the first attempt ters. For instance, coronary illness is a sort of malady caused
that provide a comprehensive survey for disease prediction due to a damage or blockage of veins in the heart, thus
using the techniques of fuzzy logics, machine learning influencing less oxygen supply to heart organs. Common
and deep learning. In addition, contrary to existing survey heart diseases are heart failure, artery blockage, heart attack,
articles available in the literature, this work has focused on a stroke, etc [25]. Fuzzy logic is continually developing to
particular range of sicknesses including heart disease, brain distinguish heart patients all through the world with the
disease, prostate, liver disease and kidney disease assistance of growing new AI techniques.
Lots of articles have been published to detect coronary
III. FUZZY LOGIC AND DISEASE DIAGNOSIS
disease by utilizing Fuzzy logic. Sari and Gupta [26]
In this section we first summarize the current related work discussed coronary disease detection using a neuron-fuzzy
which are based on fuzzy logic. We later describe the fuzzy integrated system and their results reached a similar level
logic process for disease diagnosis. of doctor’s opinion in case of high/low cardiac risk. Junior
et al. [27] presented a cardiovascular arrhythmia grouping
A. EXISTING WORKS USING FUZZY METHODS framework utilizing fuzzy classifiers to recognize the par-
Fuzzy logic provides dynamic methods that deal with diffi- ticular point of the electroencephalogram utilizing network
cult problems. Fuzzy logic is assumed to be a solid tool fuzzy Rules. In their system, the total time of ECG signal
for decision-making systems, such as expert systems or processing is reduced by a sequence of samples, without
Pattern classification systems [18]–[21]. Fuzzy logic plays any essential loss. The ECG signals are imposed into the
a vital role in the medical evaluation as it provides an exact framework that implements cleaning, and afterward utilizes
VOLUME 4, 2016 3

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 2. Process of Fuzzy Logic

a clustering algorithm “Gustafson–Kassel fuzzy” for the solution. Another representation was given by Chen et
signal classification and correlation. Their study suggested al. [31] to introduce a productive brain problem detection
that common heart diseases like myocardial infarct, arte- system by the use of fuzzy k-closest neighbour or SVM
rial coronaria and angina diseases can easily be detected for Parkinson’s disease diagnosis. A comparative analysis
by their system. According to the obtained results, their was performed between SVM and FKNN. The experimental
method provided better disease diagnosis for Pulse Pressure outcome showed that the FKNN technique worked better
Variation compared to other reported systems. Ebola Virus over the SVM classifier. The accuracy obtained by the
Disease is a fatal infectious disease also known as the “Ebola FKNN was 96.07 which is more than the SVM method.
hemorrhagic fever”. Hence, a secure method of diagnosis Different diseases such as neuro diseases, cancer, diabetes,
has been investigated. Oluwagbemi et al. [28] described that heart diseases, thyroid disorder, asthma disease were also
Ebola fuzzy informatics system was designed to diagnose diagnosed by using various ANN mechanisms. The neuro-
EVD. They utilized fuzzy logic as its inference engine along fuzzy model has been proposed by Patra and Thakur [32]
with a collection of rules. A knowledgebase was created to for the proper diagnosis of adult Asthma disease. The
help provide a diagnosis of the Ebola Virus Disease (EVD). dataset was collected from various hospitals. Three learning
The method used as a fuzzy inference method was Root algorithms were used: ANN with Self Organizing Maps
Sum Square. According to the performance of their system, (SOM), ANN with Learning Vector Quantization (LVQ) and
we can say that their system is a valuable addition to fight ANN with Backpropagation Algorithm along with NF tool
against Ebola. BRAIN DISEASE or disorder is a condition to produce accurate results. Fuzzy inference was then used
where a person loses the capability of reasoning, loss of to classified data to diagnosis a disease.
memory; change personality, mild seizures, and twitching
are common symptoms. The brain is the central control of Fuzzy logic is also capable to detect dangerous diseases
the body. When brain problems occur, the results can be like cancer, especially BREAST CANCER. Breast cancer is
devastating. Brain diseases such as stroke, brain tumours, a sort of sickness caused by bumps found in the breast that
Alzheimer’s disease can cause problems like vision loss, frames the cells. Cancer appears when cells start to grow
weakness, and paralysis, etc [29]. Early detection of these out of control. Miranda and Felipe [33] inter-operated on
problems is very necessary for a doctor as well as a patient the Fuzzy Omega algorithm, an automated tool to detect
in order for the treatment to be started. Gopal and Karnan breast lesions. The user availed elements like contour,
[30] proposed a system for diagnosing Brain Tumor. A size, and density and the system suggested the BI-RADS
system designed to diagnose brain tumors using MRI images classification. Their method achieved an accuracy of 76.6 %
by the use of the Fuzzy C Means clustering algorithm. for nodules and 83.34% for calcifications. Another approach
The tools used along with Fuzzy C means algorithms was given by Nilashi [34] for early diagnosis to tackle
are Genetic Algorithm and Particle Swarm Optimization. the disease. The authors designed an information-based
The suspicious block is fragmented by the use of two architecture for the classification of breast cancer disease
algorithms GA and PSO. Computer-aided System is then using Clustering, and classification approaches. They used
utilized for verification and correlation of brain tumor in Expectation-Maximization for clustering the data. Fuzzy
the diagnosis algorithm. Fuzzy C Means helped to determine rules extracted from Classification and Regression Trees
the adaptive threshold for brain tumor fragmentation. The were used for the classification of breast cancer disease.
results of previous techniques were compared with existing Their method can be used as a decision support system for
outcomes. Their results indicated that it improves the overall disease diagnosis. The liver ailment is also a sort of hepatic
performances of the fragmentation and can find the optimal sickness that makes the liver stop its working partially or
completely. Most of the factors of liver ailment are due
4 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

to an alcoholic or hereditary nature. The most well-known to be valuable since most diagnosis processes have been
kind of liver illness is fatty liver. In order to diagnose, a performed based on the probability of medical findings [43].
liver disease, Satarkar S.L, and Ali M.S worked to form an The power of human thinking and decision-making ability
expert system that cooperated with fuzzy logic. According develop a clinical proof-based theory to make the process of
to the authors, the portrayal was provided by the Mamdani diagnosis better [44]. Due to the demonstrated viability of
approach to recognize the risk factors. Their system could be applying fuzzy methods in the field of healthcare to display
used to make predictions of cirrhosis and avoid the need for uncertainty, it has been used in the finding procedure with
liver biopsy [35]. DIABETES is a kind of sickness which various applications as per the kind of illness and targets
is caused by the increase of blood glucose levels in the of the researchers [45]. The main rule of this framework in
body. Apart from that, this disease decreases insulin level in medical science has two major elements in which symptoms
body cells and cause type 1, type 2, or gestational diabetes. are used as input and the disease as output. Generally, the
An excessive amount of sugar level in the body prompts Fuzzy logic process to disease diagnosis as described in Fig
different issues like harming the kidney and nerves. Kalpana 3 is made by the following steps:
and kumar [36] focused on developing a model to analyze
diabetes malady using a fuzzy determination mechanism. • Fuzzifier: The Fuzzification process is done by a
To decide whether a person has the possibility of diabetic Fuzzifier. It is a process of changing a crisp input value
or not, the author used the fuzzy determination system to to the fuzzy set. Hence Fuzzifier is used as a mapping
asses rules with the fuzzy operator in their study and portray from observing input to fuzzy value.
knowledge with descriptions. Lukmanto [37] proposed an • Inference engine: After completing the fuzzification
intelligence system by using a fuzzy hierarchical model that process, fuzzy value processed by the inference engine
can perform initial diagnosis against diabetes. The proposed using a set of rules act as a collection of rules to the
model was implemented on 311 relevant data and acquired knowledge base.
an accuracy of 87.46 % as equivalent to a medical doctor’s • Knowledgebase: This is the main component of the
statement. Another proposal was given by Rajeswari et al. fuzzy logic system. The overall fuzzy system de-
[38] on diabetic diagnosis using an associative classification pends on the knowledge base. Basically, it consists of
method based on fuzzy logic to tackle the problem of the rules, structured and unstructured data also named the
boundary value confusion while partitioning risks. Tooth database.
Decay, Periodontal Disease, Gingivitis, Dental Plaque, etc • Defuzzifier: The process of converting the output from
are diseases that occur in teeth, and are commonly termed the inference engine into crisp logic. Fuzzy value is an
as DENTAL DISEASES. Allahverdi and Akcan analyzed input to the defuzzification that maps fuzzy value to
based on periodontal dental disease around 164 fuzzy rules crisp value.
taken with some inputs. The prime goal of their study was
Fuzzy Logic is taken into account among the techniques
to decrease the time taken for early recognition of dental
for AI, where intelligent behavior is achieved by creating
disease [39]. Son et al. [40] designed a system called Dental
fuzzy classes of some parameters. The rules and criteria
Diagnosis System to find out dental problems which depend
are understandable by humans. These rules and the fuzzy
on the hybrid technique of fragmentation, classification and
classes are defined by a domain expert mostly. Therefore,
decision making. They investigated that the accuracy of
a great deal of human intervention is required in fuzzy
DDS in dental problem detection is 92 % approximately that
logic. The actual processing of data basically provides a
is higher than any other systems like fuzzy inference system
presentation of the information in fuzzy logic. One of such
(89%), fuzzy k-nearest neighbor (80%), prim spanning tree
representations can be done using machine learning in the
(58%) and Kruskal spanning tree (58%). Bacterial diseases
medical field even in a much better way than fuzzy logic.
like Cholera arises after swallowing polluted or infected
The statistical model used for estimation is not capable
water. This kind of disease can prompt drying out, diarrhea
to produce good performance results. Statistical models
and can also become the reason for death, if not handle at
fail to detect missing values, large data values and hold
the perfect time. Uduak and Mfon proposed a system based
categorical data [46]. All the above-mentioned reasons can
on Mamdani fuzzy approach. Centriod method was used as
be achieved through machine learning (ML). ML plays
a deffuzifier and performed better in MATLAB simulation
an essential role in numerous applications such as natural
[41]. Another representation was given by Okpor M.D,
language processing, information mining, image detection,
they classified their investigation on cholera using fuzzy
and disease detection. In all the above-mentioned domains,
classification. The results were satisfactory for tackling
ML provides appropriate solutions as per the problem.
cholera as compared to previous applications [42]
Thus, ML also facilitates advanced diagnosis systems and
B. MEDICAL DIAGNOSIS PROCESS USING FUZZY
treatment options in healthcare. In the following section, we
LOGIC
describe how ML was used for disease diagnostic systems
[47].
Fuzzy logic has the ability to portray information and
outcomes in the form of semantic articulation. It tends
VOLUME 4, 2016 5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 3. Flow chart of the Fuzzy logic process

IV. MACHINE LEARNING AND DISEASE DIAGNOSIS very little human intervention or domain expert’s service)
In this section we first present the current related work where we are not aware of what will be the results. A
which are based on machine learning. We then describe the machine is trained to learn a concept by giving examples
ML process for disease diagnosis. and creating pattern models that are supposed to differentiate
between two or more objects. In the medical field, machine
A. EXISTING WORKS USING FUZZY ML learning assist the experts to handle large and complicated
medical data and also helps to investigate the results. The
Machine learning is a field that comes within the broader
output of this process can be used for further research.
area of AI in which by training, a machine learns itself and
Therefore, when machine learning is applied in healthcare,
perform tasks. In machine learning, there are algorithms for
it increases the trust level of patients in medical science
supervised learning (under the control and “guidance” of a
in order to predict a disease by implementing machine
human expert) in which we are initially aware about both
learning algorithms. Sometimes, illness is not early detected
input and results, as well as unsupervised learning (requiring
6 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

by human experts, in such types of cases machine learning to identify the hidden designs of benign and malignant
can be used to detect early stages of the disease before tumors. Afterward, SVM was utilized to get the classifier
its occurrence or it becomes dangerous to someone. In this to differentiate the incoming tumors. Their system improves
way, it can help to prevent future problems as “Prevention accuracy up to 97% approximately. In another study, Asri
is better than cure”. The popularity of machine learning et al. [53] classified their analysis on breast cancer using
in different areas has tended it towards machine learn- different methods of machine learning. The authors have
ing algorithms that produce correct outcome as compared done comparatively performance based analysis between
to traditional models with little processing of raw data. ML methods such as SVM, k Nearest Neighbors, Decision
Machine learning algorithms like Decision trees, Support Tree using the Breast cancer dataset. The prime objective
vector machine, Multilayer perception, Bayes classifiers, was to evaluate the accuracy in classifying data relating to
K-Nearest Neighbour, Ensemble classifier techniques, etc each algorithm in terms of correctness, precision, sensitivity.
are used to determine various ailments. Using machine Results produced by those algorithms showed that SVM
learning algorithms can lead to rapid disease prediction with provided the highest accuracy. Soreness of one or more
high accuracy. The learning process begins with observa- joints, the reason for pain and stiffness that can become
tions or information, such as examples, direct experience worsen with age is referred as ARTHRITIS. Various sorts of
or instruction. In particular, the algorithms look for data arthritis exist such as osteoarthritis and rheumatoid arthritis.
patterns and makes better decisions. The key goal is to Each type has a different way of treatment. ARTHRITIS
allow the machines to learn automatically without human reduces the quality of life of a person. Hence, early detection
interference and adjust the response accordingly [48]. The of arthritis is necessary which can be achieved using ML
intended contribution of AI in the field of medical science [54]. Neeraj et al. [55] presented a system to classify patients
is to develop programs that can help a medical expert in with arthritis dataset which was taken from Koch. Their
practicing expert and more accurate diagnosis. The forecast system classified the data with features such as identity,
for diseases plays an important role in machine learning. gender, age and treatment with an algorithm CART to find
Various types of diseases can be predicted using ML tech- out true or false rates. DIABETES is a chronic disease that
niques. Here, we examine how machine learning techniques appears when the pancreas is not able to make insulin. To
are used to predict various disease types. We focused on predict diabetes disease, Nahla and Bradely [56] worked on
the prediction of some chronic diseases like kidney disease, diagnosis by classifying based on a blood test to diagnose
diabetes, heart disease, and breast cancer, lung disorders, diabetic disease using SVM classification. SVM prediction
etc. accuracy of 94%, the sensitivity of 93%, and specificity
KIDNEY DISEASE is a common word for diverse dis- of 94% were achieved. Kandhasamy and Balamurali [57]
orders affecting the kidney’s structure and working. The compared machine learning classifiers Random Forest, K-
definition of chronic kidney disease is centered on kidney Nearest Neighbors, J48 Decision Tree and SVM to classify
damage or reduced kidney function for three months or patients who have symptoms of diabetes. These techniques
more. Kidney failure is among the most serious outcomes have been tested with data taken from the UCI data
of chronic renal disease, with complications of decreased repository. The results of the algorithms have been tested
kidney function being the primary reason [49]. Sinha and with noisy data and dataset set without noisy data and
Sinha [50] proposed a decision support framework to diag- compared in terms of specificity, sensitivity and accuracy.
nose kidney disease. They compared the performance of two Their investigation concluded that the decision tree J48 clas-
classier, SVM, KNN. The comparison was based on accu- sifier got higher efficiency than the other three classifiers.
racy, precision and execution time of both algorithms. From PARKINSON’S DISEASE is a disorder responsible for the
the investigation they observed that KNN works better than dysfunction of nervous system progress and its movement.
SVM. In another study, Charleonnan et al. [51] classified Gradually symptoms arise may be some time starting from
his analysis on performing a comparative analysis based tremor in just one hand. Sriram et al. [58] proposed a
on four ML techniques KNN, SVM, logistic regression system in which the tools used for experimentation analysis
(LR), and decision tree classifiers to detect diagnosis kidney included classification and evaluation using Orange along
disease. In order to pick the best technique, they compared with weka tools. UCI Machine learning repository pro-
their performance with each other. It was observed that the vided Voice dataset for Parkinson’s disease. Classification
SVM method is best than the rest of others and gives a algorithm such as Random Forest showed good accuracy
maximum accuracy of 98.3 %. BREAST CANCER which (90.26) compared to all remaining algorithms like KNN,
is a chronic disease for females, is the most common SVM (88.9%) and Random Forest(90.26). Naïve Bayes has
cancer disease and a leading cause of death. In recent shown the least accuracy (69.23). In 2014 Salvatore [59]
years, machine learning was used a helpful tool in the supervised a machine learning algorithm which was used to
detection of breast cancer. Zheng et al. [52] focused on diagnose patients with Parkinson’s disease and Progressive
developing a model to diagnose breast cancer based on Supranuclear Palsy. They took 28 MRI image records of
the extracted tumor features. To extract useful information both PD and PSP patients based on feature extraction
and diagnose the tumor, the K-means algorithm was used technique and SVM was used as a classifier. The algorithm
VOLUME 4, 2016 7

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

was able to differentiate PD patients from PSP patients at an 3) Afterward, the selection of attributes, then select and
individual level. Respiratory system-nose, throat, and lungs process the dataset.
affected by a viral infection is known as Influenza. Pineda 4) Various classifications methods as mentioned in the
et al. [60] investigated seven different classifier of ML for diagram can be applied to preprocess dataset to eval-
detection of influenza and compared their results within built uate the accuracy of prediction of disease
influenza Bayesian classifier. Their study demonstrated that 5) The performance of different classifiers compared
ML had the power to provide a diagnosis of irresistible with each other in order to select the best classifier
sicknesses. Concerning the occurrence of cancer in liver with the highest accuracy.
cells, Sandeep et al. [61] proposed a model for Lung images In Machine learning, all the features extracted by a
which can be classified into normal or dangerous categories. domain specialist to minimize the complications of data and
According to the authors, by following this mechanism to develop patterns in such a way that would easily visible
results could be achieved with high accuracy. Through the to ML algorithms. However, deep learning based technique
use of electronic records, ML can predict various diseases. can extract features manually without human intervention,
the only condition is to make precise decisions in which the
B. MEDICAL DIAGNOSIS PROCESS USING ML testing data could be accurate. This technique eliminates the
requirement of a domain expert for feature extraction. In the
Machine learning has granted computer systems new abili-
following section, we describe how deep learning has been
ties that we could have never thought of. Machine learning
used for disease diagnostic system.
is a field of AI that gives machines to power to learn itself by
examples [62] in order to analyze how to different models
V. DEEP LEARNING AND DISEASE DIAGNOSIS
perform in ML without using human judgment. The working
In this section we first present the current related work
of ML are explained step by step as follow [63] as shown
which are based on deep learning. We further describe how
in Fig. 4.
deep learning is used for disease diagnosis processes.
1) Data Collection: The very first step is to collect data.
It is a very critical step as quality and quantity affect A. EXISTING WORKS USING DEEP LEARNING
the overall performance of the system. Basically it is An artificial intelligence technique that mimics the workings
a process of gathering data on targeted variables. of the human brain and creating patterns for decision
2) Data Preparation: After the collection of data, the making is known as Deep Learning. While machine learning
second step is data preprocessing. It is a process to methods required to break down a problem statement into
change raw data to useful data, on which a decision different parts first and then their outcome to be integrated
could be made. This process is also called data clean- at the final stage; the Deep Learning method’s objective is to
ing. solve the issue end to end. In medical science, deep learning
3) Choose a Model: To represent preprocessed data achieves better results than traditional machine learning
into a model, one chooses an appropriate algorithm models [65].
according to the task. Deep learning has got great interest in each field and
4) Train the Model: ML use supervised learning to train especially in medical image analysis. The term deep learning
a model to increase the accuracy of decision making refers to utilize of deep neural network models. The main
or doing predictions. component of the neural network is the simulation of the
5) Evaluate the Model: To evaluate the model, a number human brain in the form of neurons. It works on the
parameters is needed. The parameters are driven from scenario in which different signals use as input, join them
the defined objectives. Also, one needs to capture the suing weights and pass those joined signals to produce
performance of the model with the previous one. output [66]. The AANs (artificial neural networks) and
6) Parameter Tuning: This step may include: number- deep learning can be differentiated by the variations in a
ing of training steps, performance, outcome, learning number of hidden layers and their inter-connectivity and the
rate, initialization values, and distribution, etc. efficiency to yield a suitable result of the inputs. The ANNs
7) Make Predictions: To evaluate the developed model are generally constituted of three different layers and are
with the real world, it is indispensable to predict some instructed to retrieve well-structured information that could
outcome on the test dataset. If that outcome will match be suitably utilized only to perform the specialized task
with domain expert or opinions nearer to it, then that [67]. On the other hand, in Deep learning, physical and
model can be used for further predictions. clinical examination of the patient are determined through
the nature of the diseases. Though there are many tools
The basic steps of for disease detection using ML is
and techniques that are available for diagnosis of diseases,
described as follows [64][62]
a certain degree of inaccuracy and uncertainty still persists
1) Collect test data with patient details. in the diagnosis process. It is quite evident from various
2) The feature extraction process picks attributes which analytics survey that using machine learning techniques has
are useful for disease prediction its own limitations. In addition to that, the present system
8 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 4. Machine Learning System

of diagnosis only considers attributes to determine diseases. diagnosis of breast cancer that has been modeled using a
The conventional way of selecting attributes which is used deep belief network. In their technique, the unsupervised
for disease prediction some time yield erroneous result. path followed by back propagation supervised path with
In contrary to machine learning, deep learning is capable “Liebenberg Marquardt’s learning function” and weights
to select the most relevant attributes out of the database were initialized using the deep network path. Their function
which in turn leads to the prediction of diseases with a was tested on Breast Cancer Data and provided a correctness
great degree of precision [68]. A considerable number of in results up to 99% greater than previous approaches.
diagnosis systems using deep learning can be found in the Charan et al. [72] used CNN for breast cancer diagnosis.
literature. A total 322 mammograms records extracted for testing in
Skin diseases may affect human skin and mostly seem to which 189 were used and showed negative results and 133
be an external disease as it originates and affects the layers were of abnormal breast records. Their results showed the
of the skin. But sometimes it gives very important clues effectiveness of deep learning for breast cancer diagnosis
to diagnose underlying causes of internal diseases. There for mammogram images.
is a variety of skin disease that can be acne, skin cancer, Diabetes is a metabolic illness influencing people groups
rashes, etc. Early detection of skin disease is important around the world. Its frequency rates are expanding alarm-
as a preventive measure of future skin problems. Liao ingly and consistently. Goutham et al. [73] proposed a model
[69] proposed a system to classify different skin diseases for the classified diabetic and normal Heart rate signals
using deep convolutional neural networks. Using 2300 skin with help of deep learning system. They utilized CNN for
disease images taken from Dermnet and OLE dataset, the extracting features and HRV data was used use as input.
proposed system was able to train the CNN model and Classification of features was done by SVM. Their proposed
assess its results. Their system could achieve Top-1 accuracy system is predicted to help medical doctors to diagnose di-
of 73.1%. Another classification was given by Shoieb et abetes using ECG signals with very high accuracy. Another
al [70] to diagnose skin cancer. Their model detected the representation was given by Sisodia and Sisodia [74] on the
infected part of skin and CNN which is used for feature early detection of diabetics. The main aim of their research
extraction. Their model used SVM as a classifier and utilize was to develop a system that can predict the possibility
CNN to train the model suing skin image data. Their results of diabetics with maximum correctness. Hence three ML
represented significant improvement and accuracy compared algorithms SVM, Naive Bayes, and Decision Tree were used
to previous ones in skin diagnosis. Chronic disease such to diagnose diabetes at an early stage. The Pima Indians
as breast cancer when detected using deep learning get Diabetes Database was used to perform experiments. The
higher accuracy compared to other techniques. Zaher and performances of these algorithms were assessed on various
Eldeib [71] proposed a system CAD approach for the measures like Accuracy, Precision, F-Measure. Their results
VOLUME 4, 2016 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 5. Difference between Machine Learning and Deep Learning

indicated that Naive Bayes performed better with the highest high temperature, cough, difficulty in breathing. COVID-
accuracy of 0.76 compared to previous models. Heart dis- 19 has killed millions of people across the world. Due
ease classification was done by Rubin et al. [75] and their to the increasing number of cases and limited test kits,
study identified variabilities heart sounds using an automatic it becomes difficult to detect the presence of COVID-19.
cardiac auscultation system. Their algorithm collected the Here at this point, the need for other alternatives such as
time-frequency rate of heart sounds and classified with the X-ray has been arisen. When researchers use X-RAYS with
help of a deep convolutional neural network. The motive AI techniques it becomes easy to detect COVID-19 [78],
of their research was to determine normal and abnormal [79]. Using deep learning authors have developed a model
heart sounds. The authors achieved high specificity score out with four phases: data augmentation, preprocessing, stage-
of all entries. Miao and Miao [76] developed an enhanced I, and stage-II deep network model designing. The model
deep neural network (DNN) to diagnose heart disease. The has been implemented on 1215 X-RAY images. Initially,
designed deep neural network model was based on a deeper in stage1 model differentiates induced pneumonia, bacteria-
multilayer perceptron framework. Their model classified the induced pneumonia, and normal/healthy people with 93.01%
data based on the training set. To investigate the perfor- accuracy. After that images detected with viral-induced
mance of this model, 303 test data were taken from patient pneumonia are sent to stage2 for detection of COVID-
with coronary disease. Their model achieved accuracy of 19 that has gained 97.22% accuracy. Overall results of
83%, sensitivity of 93% approximately. For Liver cancer, this model were accurate, reliable, and fast [80]. Most
Sun et al. [77] developed three deep learning algorithms often COVID-19 disease makes doctors confused with lungs
implemented using Convolutional Neural Network, Deep infection in this condition and diagnosis become a difficult
Belief Networks and Stacked Denoising Autoencoder to task here. For this, quick diagnosis is required that can
diagnosis lung cancer diagnosis. They compared the per- be possible with different deep models. Authors introduced
formance of all three algorithms on 28 image features of a novel Convolutional CapsNet using chest X-ray images.
the lungs dataset. SVM was used for classification. CNN, This model provides accurate results with the binary classi-
DBNs, and SDAE provided accuracies of 0.7976, 0.8119, fication of 97.24% and multi-class classification of 84.22%
and 0.7929, respectively. [81]. In this study, a Pre-trained deep neural network was
COVID-19 (Coronavirus) disease is an infectious virus. used to diagnose COVID-19 on chest CT images.
It spreads when an infected person coughs, sneezes, and Brain Hemorrhage refers to bleeding within the brain,
his generated droplets are transmitted to other persons. it can happen due to a brain tumor, clot, or hypertension.
Most people who get infected by COVID-19 experience Whenever a Hemorrhage occurs, oxygen cannot be able to
10 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

reach the brain cells and eventually brain cells die rapidly method, the least collection of elements is based upon
[82]. A novel convolutional neural network based on ResNet evidence and meta-analyses that summarize and analyze
to diagnose and predict the type of brain hemorrhage is also scientific reliable literature by utilizing a structure method
developed [83]. 752,803 DICOM files have been collected based on predetermined queries that can be used by various
to conduct this study. This model obtained an accuracy of researchers. Different findings and ideas which are pub-
93.3%. lished in the conventional papers by different researchers can
be investigated with a correct and comprehensive analysis in
B. MEDICAL DIAGNOSIS USING DEEP LEARNING a systematic review method. With the help of the PRISMA
As mentioned earlier, the conventional automated diagnostic method, an investigator can perform systematic reviews
method used a machine-learning algorithm in that clinical and meta-analyses with a degree of accuracy that can lead
expert manually fetched features in diagnosis reports. But research in a well-structured manner.
sometimes it became difficult to extract features from large
dataset [47] [48]. Hence, those methods suffered with accu- A. LITERATURE SEARCH
racy and efficiency as depicted in Fig. 5. In this study, various 8 databases were extracted for accu-
Absence of important information is a considerable ob- rate review: BMC, Springer, ACM, IEEE, Elsevier, Google
stacle for deep learning models. Presently, medical research Scholar, Wiley digital library and ACM were selected based
use electronic health records, but there is no predictable on our research questions. Based on predefined questions
technique to evaluate the EHRs, which implies that accuracy and goals, the literature search was done by utilizing
of diagnostic process using automated system could be the keywords including “fuzzy logic”, “machine learning”,
limited. If the system fail to collect accurate data, the “deep learning”, “disease prediction”. Previous useful ar-
model will not able to diagnose a disease precisely, which ticles were extracted and recognized by a search strategy.
makes it complicated to show accurate prediction. To tackle The search strategy has shown for every journal described in
this kind of problem, the authors in [64], [65] developed Table 1 150 articles from 2009 to 2019 have been extracted
effective deep learning model for early correct detection and the detailed selection process is displayed using the
of various diseases. In conventional approach, a Deep CNN PRISMA diagram in Fig. 7.
model is used to detect diseases. Then the neural system
utilizes approaches to data expansion. Each layer inside B. DISTRIBUTION OF PAPERS BY JOURNALS
CNN filters the raw data in the image to get a specific In the study, 7 reputed journals have been selected to search
pattern. The few initial layers find the large feature set like for papers. Following table shows the various database
diagonal lines and the next few layers are used to get better providers. Table 1 shows the names of publishers which
details, organize them into complicated features. The most were selected, the number of articles selected and corre-
final layer works as an ordinary neural network and the sponding percentages. As shown in the Table 1, IEEE,
network becomes fully connected [66], [67]. Then it put Elsevier, Springer ranked first with 35.29%, 23.40% and
together highly specific features like various symptoms of 19.60% respectively.
the disease and as a result, perform the prediction of the
disease. The authors in [64], [84] rectified the approach in C. STUDY SELECTION AND ELIGIBLE PAPERS
order to solve the issue of lacking information or missing In this segment, the outcome of 80 research articles was
values. Afterward, a deep learning model trained by the taken into consideration. As shown in Fig. 7 those research
processed data have proved their efficiency as shown in Fig articles were chosen or taken into consideration based on
6. inclusion and exclusion criteria. As per exclusion criteria,
only qualified articles were chosen and chapters from book,
VI. RESEARCH METHODOLOGY thesis, and summary reports. Journal editorials, newsletter
PRISMA (Preferred Reporting Items for Systematic Re- and papers which were not in English were excluded.
views and Meta-Analysis) method was used for a systematic According to inclusion criteria, we considered the follow-
review. This method was invented by Moher et al. [85], ing criteria: reference of the author, year of publication,
[86]. In this method, a survey is carried out on basis of where it belongs to a journal or conference proceeding,
a predefined question by the virtue of which data from the definition of the diseases; its types and complications,
the studies that are included in the survey, are collected objectives, a loophole in the research, type of fuzzy methods
and subsequently analyzed systematically and evaluated used, type of machine learning methods used, type of deep
critically. Meta-analysis is a statistical, formal, quantitative learning methods used, results and concluding remarks. In
study design technique used for systematic evaluation and connection with this, 15 academic papers were excluded and
integrating the results of the included studies or previous 105 articles included. After reviewing all collected articles,
study to derive the conclusion. Both systematic review and only 80 papers qualified the eligibility criteria from where
meta-analyses are an integral part of research to summarize relevant articles were chosen for in-depth analysis and study.
evidence relevant to the efficacy and safety of medical care Furthermore, we proceeded by scrutinizing the abstract
interventions precision and certainty. In a systematic review and summary of the chosen articles to investigate whether
VOLUME 4, 2016 11

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 6. Deep Process to diagnosis a disease

TABLE 1. Different published articles selected for the literature review along data extraction form which was formulated helped to a great
with frequency
extend to accomplish the desire results and draw a suitable
Publisher Articles Percentage conclusion. The criteria which were incorporated included
Elsevier 11 23.40 the reference of the author, its year of publication, whether it
IEEE 18 35.29 belongs to a journal or conference proceeding, the definition
Springer 10 19.60 of the diseases; its types and complications, objectives,
ACM 5 9.80
BMC 1 1.96 loophole in the research, methods used fuzzy logic, machine
IOSPress 1 1.96 learning, and deep learning methods, results, finding and
BioMed central 3 5.88 positive impact on diagnosis process. Fig. 8 indicates a
Wiley online library 2 3.92
chart related to classification. After reviewing all collected
Total 51 100
papers, 80 academic research papers from 30 international
scientific journals and 10 conferences proceeding which
the selected articles fully satisfy the inclusion criteria. All were published from the year 2009 to 2019 were taken into
insignificant and unrelated articles were discarded in this account in this systematic research. We thoroughly reviewed
stage. Similarly, all academic research papers which did all selected article and finally retained those articles which
not match the inclusion criteria of disease diagnosis were applied fuzzy logic, machine learning, and deep learning for
discarded while selecting appropriate articles. In total, 89 diagnosis of a disease. Notwithstanding, though adopting
articles were qualified as per the inclusion criteria and were the PRISMA method and selecting articles accordingly is a
found to be compatible with our study and were taken time-consuming process, still this method is a most suitable
into account in this systematic review. Table 2 contains the method for carrying out research as it is a structured method
primary keywords used to search the relevant contents. for which we have to include only those articles in the study
which were explicit to the subject of the systematic review.
D. EXTRACTION AND SUMMARIZING OF DATA
In the last stage, we reviewed all the papers which consisted VII. DISCUSSIONS AND FUTURE RESEARCH
of 95 articles in order to complete the final study and achieve DIRECTIONS
the desired result. The articles which were extracted for the A. RESEARCH FINDINGS AND LESSONS LEARNED
research were vetted meticulously to find out the answer to This study was conducted to inspect the effect of AI meth-
the crucial questions as per the requirement of the research. ods in the diagnosis of various diseases. Very few review
A form was formulated for the extraction of data that make articles have been published to evaluate how AI methods
the necessary classification, inspection, and incorporation of are effective in Healthcare. In this review article, we were
the included articles in the light of the present criteria. The capable to locate and review 51 papers that used different
12 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 7. Article Selection Process

TABLE 2. Search strategy in different databases

Database Search Strategy

IEEE (Fuzzy Logic, Machine learning, Deep learning) AND (Disease Diagnosis)
Elsevier Pub-date > 2009 AND (AI Techniques) AND (Disease Diagnosis)
Springer (fuzzy logic, Machine learning, Deep learning] AND (Disease Diagnosis) AND Publication Date: (01/01/2009 TO 01/11/2020)
BMC AI techniques in Abstract AND Disease Diagnosis in Abstract
Taylor & Francis (AI methods) AND (disease diagnosis)
Google Scholar “AI methods” AND “Disease Diagnosis” anywhere in articles

TABLE 3. The distribution of AI methods by medical disciplines articles published this year. Firstly, in this research, we iden-
tify which mechanisms had the greatest effect on the disease
Disease Name Fuzzy Logic Machine Learning Deep Learning
diagnosis and conducted our research based on this purpose.
Cardiology 2 2 2
Neurology 2 2 2 Hence we examined appropriate classifications for the study
Dermatology 0 2 2 of AI methods such as fuzzy logic, machine learning, and
Breast Cancer 2 2 2 deep learning are reviewed basis on our analysis. The results
Diabetics 2 2 2
Kidney disease 2 2 2 show that deep learning is very popular among the present
Arthritis 0 1 0 researchers especially in the area of medical science as
Liver cancer 0 0 1 shown in Table 3. Another division is about classifying
Thyroid 0 0 1
Dental disease 2 2 0 the percentage of articles published every year. Over the
Ebola 1 0 0 past few years, the rapid growth in the study of AI in
Asthma 1 0 0 diagnosing diseases is reflected in the analysis of papers in
Cholera 2 0 0
Influenza 0 1 0 our research. The results indicate that the average published
Skin cancer 0 0 2 volume from 1% in 2009 reached 20% in 2019 as shown
in Fig. 9. AI helps doctors to early detection of disease and
produce beneficial results in improving the diagnosis process
AI methods for diagnosing various diseases from January that’s why AI is popular among investigators. Based on our
2009 to December 2019. The review considered eight most investigation, three widely used AI techniques (fuzzy logic,
used databases, in which a total of 105 articles were found. machine learning, and deep learning) were considered to
A detailed analysis of those articles was conducted in order review the articles in the field of healthcare and produced re-
to classify most used AI techniques for medical diagnostic sults based on this research. We considered the classification
systems. Since when were writing this review paper we were of articles that recorded the positive effect of AI methods
at the beginning of 2020, so we could not determine the in diagnosing complex diseases that can cause a serious
VOLUME 4, 2016 13

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 8. PRISMA methods for review

14 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 4. Fuzzy logic method review

Study Disease Research Goals Methodology Findings

[26] Heart Disease To measure the accu- Neuron-Fuzzy Integrated System and The results produced by this study by
racy of the framework tool used is MATLAB applying neuron-fuzzy integrated sys-
system for heart dis- tems matched with expert’s opinion.
ease
[27] Cardiology The main goal of this The algorithm is used for the sig- Only twenty samples to determine dis-
research is to reduce nal classification and the correlation is ease using fuzzy clustering algorithm.
time in processing of Gustafson Kassel fuzzy clustering This algorithm use. It takes a little
ECG signal by mini- measure of memory to store the infor-
mizing the number of mation and requires littler processing,
data samples, without in this manner producing the reaction
losing any essential in a small period.
bit and analysis the
heart signals
[28] Ebola The prime objective Fuzzy expert system was used along The accomplishment of Ebinformatics
of this study is to with The Root Sum Square shows the possibilities of applying an
develop a fuzzy sys- informatics device as a method for en-
tem to diagnose Ebola couraging basic tasks, for example, un-
with useful recom- derstanding finding, forecast, and suit-
mendations able suggestions.
[30] Brain Tumor To improve the seg- clustering algorithms Fuzzy C Means The results have provided substantial
mentation process. is used along with tools Genetic Algo- evidence that for brain tumor segmen-
rithm (GA), and Particle Swarm Opti- tation is The performance accuracy
mization (PSO) is 0.95 and rate of error reduced to
13%,tumor diagnosis reached to very
high rate 98%
[31] Parkinson disease the fuzzy k-nearest The proposed FKNN-based system is FKNN based framework get accuracy
neighbour system was compared with the support vector ma- up to 96.07% .this method can ensure a
used for detecting chines based methods. To further en- reliable diagnostic model for detection
Parkinson disease hance the diagnostic correctness for the of PD
detection of PD
[32] Asthma Disease This prime goal of Self-Organizing Map (SOM),Learning Tool The effectiveness of this model
this study is to design Vector Quantization, Back-Propagation assed based on dataset of asthmatic
a system that assist Algorithm, along with Neuron-Fuzzy patients. And use Back-Propagation to
to diagnosis of adult Fitting produce results and best validation per-
Asthma by use of formance at 535 samples
Neuron-Fuzzy Logic.
This model is helpful
in diagnose Asthma
and will get the accu-
rate result
[33] Breast cancer To achieve high They proposed fuzzy inference system. Initially domain expert investigates the
accuracy in results we Fuzzy Omega algorithm was used to images and produced results assess into
applied Fuzzy Logic generate membership function. Fuzzy Omega algorithm. The results
to automatic BI- are compared with previous study and
RADS categorization achieve very high accuracy of 0.76.
of breast lesions
[34] Breast cancer The developed Clustering method Expectation- Two data sets to evaluate breast cancer
information-based Maximization used for clustering. from UCI data repository. The outcome
framework used as Fuzzy rules are designed based produced from this dataset indicates
a medical decision on Classification and Regression fuzzy rule-based techniques achieved
support assistant to Trees. The rules generated by CART good prediction accuracy for breast
assist medical experts algorithm used to classy breast cancer cancer
in the healthcare. disease
[35] Liver disease Objective of this re- Mamdani Fuzzy Logic was parameters centroid technique improve outcome
search is to develop along with Fuzzy Logic tools Identify with defuzzification
a fuzzy expert sys- risk status with max-min method
tem for the detection
of Cirrhosis that is
one of the most com-
monly diseases arises
in liver
[36] Diabetes Disease To diagnosis diabetic The fuzzy mechanism consists of fuzzy The developed model obtain much ac-
a Fuzzy Expert inference, Collection of rules. also use curate results than previous models.
System used to fuzzy T-norm and T-Conorm operators.
investigate the To convert fuzzy values into crisp set
diabetes data.it is set Defuzzification has been used.
off fuzzy membership
functions and rules

VOLUME 4, 2016 15

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

[37] Diabetes Disease This model is pro- Fuzzy Hierarchical Model IS designed The outcome based on 311 records
posed to get exact re- to measure the results. equivalent to clinical expert.
sults as Clinical ex-
pert concluded related
Diabetes Mellitus
[38] Diabetes Disease Early detection of the Associative classification technique de- The proposed technique can make sense
prediabetic condition pend on fuzzy logic-based of the specific hazard factors like Age,
and compare the re- Glucose, DPF, BMI, and BP alongside
sults with the crisp its right unsafe estimations to anticipate
method. prediabetes in an enhanced manner than
the crisp method.
[39] Dental Disease The research aims to Mamdani Fuzzy Logic with 5 input Dental Disease identification time min-
propose a Fuzzy Ex- variables and 164 rules along with a imized
pert System for de- fuzzy logic toolbox
tection of periodontal
dental illness by ex-
amining and build up
a PC program to help
dental specialists for
examination, finding,
and treatment of the
disease.
[40] Dental Disease The designed system Perform segmentation process they use Accuracy of the used DDS algorithm
for dental diagnosis semi-supervised fuzzy clustering algo- is compared with other algorithms.DDS
problem was DDS. rithm achieved 0.92 highest accuracy greater
than other algorithms like prim span-
ning tree, fuzzy k-nearest neighbour,
affinity propagation clustering.
[41] Cholera fuzzy expert system Mamdani Fuzzy Logic parameter. uti- MATLAB achieved outcome 0.05 bet-
for the diagnosis and lize MATLAB with max-min approach ter as indicated in calculated results
monitoring of cholera and centroid
[42] Cholera Analysis cholera Fuzzy Classifier Algorithm With five Simulation results were satisfactory for
based on the FC input parameters and one Outcome pa- diagnosing cholera
algorithm. rameter Fuzzy Logic Toolbox

FIGURE 9. Article Selection per Year

16 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 5. Machine learning method review

Study Disease Research Goals Methodology Findings

[50] Kidney Diseases The main objective is to make a Classification, SVM, KNN. From the outcome of the test it
comparative study between the per- is evident that the KNN give
formance of SVM and KNN and better result in comparison to
make classification on the basis of SVM in term of classification.
its precision, correctness and com-
pletion time for the predication of
CKD.
[51] Kidney Disease Provide assistance to doctors for So far 4 types of machine SVM shows maximum in
deciding the accurate treatments learning methods are invented comparison to other three
methods for the chronic diseases viz: KNN (K-nearest neigh- methods at 0.99 whereas
related to kidney by using clinical bors), SVM (support vector the sensitivity of Logistic
data. machine), decision tree clas- recorded at 0.94, Decision
sifiers and LR(logistic regres- Tree recorded at 0.93 and
sion) KNN recoded at 0.96.
[52] Breast cancer The aim of this particular study is Algorithms generated related 30 investigation on the
to identify diseases related to breast to K-SVM (K-means and sup- Wisconsin Diagnostic Breast
cancer based 23 and determine the port vector machine) Cancer (WDBC) data set
charactertics of the tumour. carried out in this method and
recorded 97.38% accuracy
which my be considered as
better result.
[53] Breast cancer In this comparative study of WEKA data mining tool. It’s observed that SVM per-
Performance among various form better with highest de-
machines learning algorithms done gree of accuracy at 97.13%
viz: SVM, Decision Tree, NB and with the minimum uncertainty.
k-NN on the Wisconsin Breast
Cancer datasets. Subsequently
effectiveness and efficacy of all
algorithms with respect to its
precision, accuracy, and specificity
in classifying data is recorded
[55] ARTHRITIS Design a model to analyze the On dataset of arthritis Simple This method provides assis-
predicted improvement in arthritis CART Algorithm and WEKA tance to treat more critical
patient technique are used in tandem. cases for Arthritis or solve its
queries.
[56] Diabetics Early detection of Diabetes Melli- SVM SVM is a useful device to ex-
tus to decrease the death rates. amine diabetes and result yield
94% accuracy and its sensitiv-
ity and specificity recorded at
93%, and 94% respectively.
[57] Diabetics The objective of this research is KNN, SVM, RF, J48Decision Research result indicates that
to make a comparative study be- Tree; Decision tree J48 gives better
tween algorithms and it perfor- outcomes with 73.82% accu-
mances which are useful in the racy in comparison to other
detection of diabetes. methods.
[58] Parkinson’s disease Voice data examination is a very In this study for statisti- In this technique Parkinson’s
critical aspect in the current decade cal analysis Orange v2.0b disease are diagnosed with the
for studying and determining and weka v3.4.10 has been help of voice dataset using ma-
suitable diagnostic techniques for used whereas for classification chine learning algorithms.
Parkinson’s disease. SVM, KNN classifier have
been used
[59] Parkinson’s disease The main objective of this study is SVM The algorithm provides excel-
to analyze the viability of a over- lent discrimination of PD pa-
see machine learning algorithm for tients from PSP patients at
the assisted diagnosis of patients an individual level, Accuracy,
who already tested positive to Pro- Specificity, and Sensitivity >
gressive Supranuclear Palsy and 90%
Parkinson’s disease and
[60] Influenza To reduced uncertainty and vague Machine learning classifiers In the case of data with miss-
data in the detection process by ing values ML constantly pro-
applying fuzzy methods vide enhanced performance
[61] Liver cancer developed a model to detect the neural network classifier, fuzzy This model provide a mecha-
normal and abnormal region. c means, multinomial multi- nism to get results with more
variate Bayesian accuracy.

VOLUME 4, 2016 17

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 6. Deep Learning Methods

Study Disease Research Goals Methodology Findings

[69] Skin disease The study aims to analyze the 23,000 skin disease images ex- the system obtained high accu-
efficiency of universal building tracted from the Dermnet and racy in detection of skin dis-
skin disease detection system using OLE dataset to asses the per- ease
DCNN. formance of CNN framework.
[70] Skin cancer It is difficult to recognize A classifier system has been The outcome of this system
melanoma during its initial stages developed based on a mul- shown very high performance
by manual examination.hence ticlass linear Support Vector related to accuracy, specificity
development of automated system Machine.This system trained and sensitivity.
was required to assist early with CNN features extracted
diagnosis of skin cancer from skin images dataset
[71] Breast Cancer To diagnose breast cancer CAD Backpropagation neural net- The proposed model provide
strategy has been designed based work an accuracy of 0.99 in results
on deep belief network along with over previously-published
backpropagation supervised path models.
[72] Breast Cancer This method helps to detect the This study utilized different K The outcome described that K
region of tumor. Means, Closing, Dilation and Means algorithm is more accu-
Canny Edge Detection algo- rate in diagnosing tumor
rithms.
[73] Diabetes This study proposed methodology Heart rate variability, ECG, Developed to assist doctors in
to classify diabetic disease HRV CNN, LSTM diagnose diabetes using ECG
signals using deep learning frame- signals.it achieve accuracy up
works. to 0.95.
[74] Diabetes The goal of this study is to pro- Decision Tree, SVM, and Results obtained show Naive
posed a framework that able to Naive Bayes Bayes outperforms with accu-
prognosticate the possibility of di- racy of 0.76.
abetes in patients with maximum
accuracy
[75] Heart disease The study shown the how deep deep convolutional neural net- The overall results achieved
learning is used to diagnosis auto- work with this model is 0.83
mated cardiac recognizing variabil-
ities in heart sounds
[76] Heart disease The objective of this study to de- Deep neural network learning The model acquire accuracy of
signed deep neural network learn- 0.83,sensitivity of 0.93,preci-
ing to diagnose the patients and aid sion 0.79.
the clinical expert .it enhance the
correctness and reliability of heart
disease detection and prognosis in
patients
[1] Liver cancer The objective is to investigate lung CNN, DBNs, Stacked Denois- The correctness of CNN,
cancer using deep learning algo- ing Autoencoder DBNs, and SDAE are 0.79,
rithms on LIDC database 0.81, and 0.79, respectively.
[78] COVID-19 The objective is to minimize the Deep learning with four phases Overall results of this model to
need of test kits by providing al- and implemented in Python detect the presence of COVID-
ternatives with FastAi library 19 were accurate, reliable, and
fast. In stage1 with 93.01%
and stage2 97.22% with accu-
racy.
[80] COVID-19 To Provide a model for fast screen- A novel artificial neural net- This model provides accurate
ing in detection of COVID-19 work, Convolutional CapsNet results with the binary classi-
fication of 97.24% and multi-
class classification of 84.22%.
[81] COVID-19 To use Pre-trained neural networks DenseNet201 based deep This model obtained an accu-
for diagnosis of coronavirus using transfer learning racy of, 99.82% in training,
Chest CT Images 96.25% in testing, and 97.4%
in the validation.
[87] COVID-19 To conduct a comparative study on VGG16, VGG19, InceptionResnetV2 and Den-
deep learning models and find out DenseNet201, Inception- snet201 provide better results
a model with high accuracy ResNetV2, InceptionV3, in terms of accuracy 92.18%
Resnet50, and MobileNetV2 and 88.09%, respectively.
[82] Brain Haemorrhage To Analyse acute brain haemor- Aidoc (Tel Aviv, Israel) neural Aidoc software efficiently as-
rhage using non-contrast CT im- network software sist the doctors in making de-
ages cisions.
[83] Brain Haemorrhage Detection of brain haemorrhage in A novel convolutional neural Accuracy of this model is
less time network based on ResNet 93.3%.

18 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

health issue when detection is too late. We can say that tion by utilizing alternative methods, such as information
AI is a boom in such complex cases. From the relevance of augmentation and picture synthesis. However, it is not
these approaches concerning healthcare, we have reviewed fully clear whether they are suitable for AI-based medical
various AI techniques. The most common approach was diagnostics [89], [90].
associated with the use of AI technology in the area of
cardiovascular. We discussed cardiology through three main 3) Clinical Translation
AI techniques namely fuzzy logic, machine learning and The development in AI research used in medical diagnos-
deep learning. Since cardiology covers a wide variety of tics is indeed rapid, and their possible adoption has been
cardiovascular diseases, therefore we reviewed only a few of shown by systems including the detection of various can-
them. Different AI methods have some limitations in some cer metastasis [91], brain recognition [92], and diagnosing
disease detection or produce effective results. However, diseases in retinal pictures. Nevertheless, the adoption of
we tried to review the best articles with various diseases AI-based system in clinical settings will undergo various
like cardiology, Alzheimer’s disease, dental problems, Ebola transformations and phases and many methods still to come
virus, cholera, liver cancer, breast cancer with various AI [89]. As mentioned before, present studies focus mainly on
techniques. This study enhanced our understanding of the optimizing the performance of complex machine learning
effectiveness of AI techniques in diagnostic tests. Thus, this models, while disregarding their explainability. As a result,
research provides information for people working in clinical physicians struggle to interpret these models, and feel it
disciplines. Moreover, based on the results produced by this is hard to trust them. Therefore, reliable and trustworthy
research, we were able to find out the areas and diseases communications between medical experts and AI model
which used AI techniques and which have been ignored. experts is also highly important to transform the AI-based
A brief description of reviewed articles is also provided in diagnostic potentials into clinical practice.
Table 4, Table 5, and Table 6 for available articles which
used fuzzy logics, machine learning, and deep learning, 4) Medial Data Characteristics
respectively. Since the medical data is the ultimate basis of mining
knowledge required for disease diagnosis, the information
B. OPEN PROBLEMS AND FUTURE TRENDS should be of high quality. Moreover, the volume of medical
As evident from the progress and discussion presented in data is usually very high, the data sources are diverse, and
this paper, AI algorithms are potential to provide a signif- the data is often coming from real-time sensors. Therefore,
icant contribution to medical diagnostic systems. Nonethe- preserving the data quality is a challenging task. With more
less, in order to obtain the maximum potential of AI for and more mobile sources used for medical data, with com-
mining novel insights from the associated medical data, AI- plex applications that need remote access to healthcare data,
based diagnostic systems must address some major issues having it stored on the cloud seems a more viable option
as follows. [93]. Although various solutions have been introduced to
solve issues with cloud storage, none of them can handle
1) Explainable Diagnosis all aspects of medical data characteristics precisely, because
AI models are often criticized because of its internal unclear of the additional need to maintain the compliances with
decision-making process. In this regard, explainable AI medical data security policies.
deals with the implementation of clarity and reasoning of the
behaviors of statistical black-box AI learning methods, par- 5) Standardization and Interoperability
ticularly deep learning. As such, in addition to uncovering In the diagnosis context, there are many ways that vendors
the pattern recognition problems, AI systems should come can manufacture a diverse range of diagnostic products
with causal models of the world supporting explanation while integrating a set of AI algorithms selected from
and understanding. This is even more important when we many possible methods. However, they may not follow
seek for the applications of AI in medical diagnostics. standard rules and regulations for compatible interfaces and
Researchers argue that it is essential to look at even beyond associated protocols across diverse computing frameworks.
explainable AI [88]. Causability will eventually results in This prompts interoperability issues. To address system
explainable diagnosis covering measurements for the quality diversity, immediate efforts are required to set the technical
of explanations. standards for AI-based medicine and diagnosis [94]. In
this regard, various technical and medical organizations
2) Quality of Training including the AI group run by the international organization
The performances of machine learning and deep learning for standardization, world health professions alliance, and
algorithms largely depend on the availability of high-quality world health organization can work together [95], [96].
training models to achieve the required diagnostic capability.
Moreover, the problem of data scarcity is very central since 6) Secure Diagnosis
data are at the key of AI-based medical applications. There AI methods in general and deep learning techniques in par-
exist some efforts to create additional annotated informa- ticular are vastly application-specific where a model trained
VOLUME 4, 2016 19

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

for diagnosing one disease might not be able to work well (Fuzzy logic, machine learning, and deep learning) that are
for another diagnosis. The algorithms usually need to be widely used in healthcare and we produce our results using
retrained with respective medical data to be utilized for other these three methods. Also, the effect of every AI technique
diseases; otherwise, false diagnosis will be unavoidable. based on the frequency of influence recorded by papers was
Also, improper selections of hyper parameters, by even analyzed. Major medical areas that we have reviewed were
slight change, can invoke large change in model’s perfor- related to cardiology, neurology, cancer, kidney disease,
mance resulting in bad diagnosis. For example, whereas diabetics, cholera, and dental disease respectively using AI
supervised learning is considered stable due to fixed data diagnostic criteria. Besides this, we also discovered that
sets, reinforcement learning is not stable at all [97]. On that, the papers differed significantly depending on the type of
more insights are required for AI algorithms to be optimized disease. In this study we observe that AI is not limited
for particular disease diagnosis. Another important aspect to identify any specific disease, we can utilize various AI
of secure diagnosis is that the diagnostic systems must techniques to detect any kind of disease or to improve the
be protected from wrongdoers. The attackers exploit the diagnosis process for all diseases. Therefore we can say that
features of the AI algorithms to break the system. For this survey will be helpful in future research. Moreover,
example, an adversary can play with the training parameters in this research paper, we observe that over 91% of AI
and mislead the diagnostic system to learn the opposite of methods reported a positive impact on disease diagnosis.
what it is supposed to do. It is, therefore, very vital to deeply The efficiency to detect disease by AI cannot be ignored.
investigate the characteristics of AI algorithms, reexamine Another significant finding in this review is that most of
the respective roles in diagnostic systems, and address the the researchers use tools like MATLAB, Python, Java, C#
respective challenges. for designing AI architecture. This research also has some
limitations. PRISMA method analyzed the articles published
VIII. CONCLUDING REMARKS only in a specific decade in terms of healthcare using
Recent advancements in AI techniques lead to successful AI techniques. Although some selected articles published
applications of AI in healthcare. Even it has become a in 2020 were considered in this survey, the main review
hot topic of discussion whether AI expert systems will focus was on the articles published from 2009 to 2019. For
eventually replace human doctors. Still, we consider the future studies, we project to consider the diagnosis in a
fact the AI expert system can assist the human doctor to broader sense to indicate the applicability of AI methods
make a better decision or even replace human judgment in in Alzheimer’s disease [98] and Parkinson’s Disease [99]
some cases. Different AI techniques can help to find out Diagnosis. Moreover, the roles of AI techniques for the
relevant information from a large amount of clinical data. diagnostics systems using sensors-based computing frame-
Also, AI methods are trained in such a way that can have the works [100], [101] will also be investigated. An in-depth
ability of self-learning, error-correcting, and they produce assessment of the economic impact of AI in health care
results with high accuracy. This survey is about the use of [102] is also a part of our future works.
three AI approaches in disease diagnosis. In this review, we
assess the impact of the AI methods and their constancy on REFERENCES
disease diagnosis to minimize the errors in misdiagnosis, [1] J. L. Scully, “What is a disease?” EMBO reports, vol. 5, no. 7, pp. 650–
with the PRISMA method. To accomplish the primary goal, 653, 2004.
we developed a search scheme. In this prospect, different [2] R. Leaman, R. Islamaj Doğan, and Z. Lu, “Dnorm: disease name normal-
ization with pairwise learning to rank,” Bioinformatics, vol. 29, no. 22,
scientific journals including Google Scholar, IEEE, Science pp. 2909–2917, 2013.
Direct, Web of Science, Wiley Online Library, and Elsevier [3] N. Armstrong and P. Hilton, “Doing diagnosis: whether and how clini-
were chosen to fetch the published scientific papers from cians use a diagnostic tool of uncertain clinical utility,” Social Science &
Medicine, vol. 120, pp. 208–214, 2014.
the years 2009 to 2019. [4] A.-L. Barabási, N. Gulbahce, and J. Loscalzo, “Network medicine: a
All the retrieved papers are distributed based on authors, network-based approach to human disease,” Nature reviews genetics,
published years, various AI tools, the fuzzy methods, ma- vol. 12, no. 1, pp. 56–68, 2011.
[5] R. H. Scheuermann, W. Ceusters, and B. Smith, “Toward an ontological
chine learning methods and deep learning methods various treatment of disease and diagnosis,” Summit on translational bioinfor-
kinds of diseases, results and lastly the influence of AI meth- matics, vol. 2009, p. 116, 2009.
ods that are applied in disease detection. The results have [6] P. Croft, D. G. Altman, J. J. Deeks, K. M. Dunn, A. D. Hay, H. Heming-
way, L. LeResche, G. Peat, P. Perel, S. E. Petersen et al., “The science
shown that the frequency of paper publishing in the medical of clinical practice: disease diagnosis or patient prognosis? evidence
field has rapidly enhanced. Another aim of this study was to about “what is likely to happen” should shape clinical practice,” BMC
investigate which AI method was most effective for disease medicine, vol. 13, no. 1, p. 20, 2015.
[7] E. Choi, M. T. Bahadori, A. Schuetz, W. F. Stewart, and J. Sun, “Doctor
diagnosis according to most of the researchers. Based on our
ai: Predicting clinical events via recurrent neural networks,” in Machine
study we concluded that applied methods of AI in healthcare Learning for Healthcare Conference, 2016, pp. 301–318.
provide beneficial results by improved diagnosis process [8] C.-C. Lee, “Fuzzy logic in control systems: fuzzy logic controller. i,”
and to detect the disease in early stages which follows to IEEE Transactions on systems, man, and cybernetics, vol. 20, no. 2, pp.
404–418, 1990.
pick the suitable treatment plan. The other key concept to [9] J. Yen and R. Langari, Fuzzy logic: intelligence, control, and information.
keep in mind is that we investigated three AI techniques Prentice Hall Upper Saddle River, NJ, 1999, vol. 1.

20 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

[10] H. D. Beale, H. B. Demuth, and M. Hagan, “Neural network design,” [32] S. Patra and G. Thakur, “A proposed neuro-fuzzy model for adult asthma
Pws, Boston, 1996. disease diagnosis,” Comput Sci Informa Technol, vol. 3, pp. 191–205,
[11] C.-H. Weng, T. C.-K. Huang, and R.-P. Han, “Disease prediction with 2013.
different types of neural network classifiers,” Telematics and Informatics, [33] G. H. B. Miranda and J. C. Felipe, “Computer-aided diagnosis system
vol. 33, no. 2, pp. 277–292, 2016. based on fuzzy logic for breast cancer categorization,” Computers in
[12] M. Chen, Y. Hao, K. Hwang, L. Wang, and L. Wang, “Disease prediction biology and medicine, vol. 64, pp. 334–346, 2015.
by machine learning over big data from healthcare communities,” Ieee [34] M. Nilashi, O. Ibrahim, H. Ahmadi, and L. Shahmoradi, “A knowledge-
Access, vol. 5, pp. 8869–8879, 2017. based system for breast cancer classification using fuzzy logic method,”
[13] J. Betancur, F. Commandeur, M. Motlagh, T. Sharir, A. J. Einstein, Telematics and Informatics, vol. 34, no. 4, pp. 133–144, 2017.
S. Bokhari, M. B. Fish, T. D. Ruddy, P. Kaufmann, A. J. Sinusas et al., [35] S. Satarkar and M. Ali, “Fuzzy expert system for the diagnosis of
“Deep learning for prediction of obstructive disease from fast myocardial common liver disease,” International Engineering Journal For Research
perfusion spect: a multicenter study,” JACC: Cardiovascular Imaging, & Development, vol. 1, no. 1, pp. 2–7, 2015.
vol. 11, no. 11, pp. 1654–1663, 2018. [36] M. Kalpana and D. A. S. Kumar, “Fuzzy expert system for diagnosis of
[14] M. S. van Mourik, A. Troelstra, W. W. van Solinge, K. G. Moons, diabetes using fuzzy determination mechanism,” International Journal of
and M. J. Bonten, “Automated surveillance for healthcare-associated Science and Applied Information Technology, vol. 2, no. 1, pp. 354–361,
infections: opportunities for improvement,” Clinical infectious diseases, 2011.
vol. 57, no. 1, pp. 85–93, 2013. [37] R. B. Lukmanto and E. Irwansyah, “The early detection of diabetes mel-
litus (dm) using fuzzy hierarchical model,” Procedia Computer Science,
[15] G. Luo, F. L. Nkoy, P. H. Gesteland, T. S. Glasgow, and B. L. Stone, “A
vol. 59, pp. 312–319, 2015.
systematic review of predictive modeling for bronchiolitis,” International
[38] A. Rajeswari, M. S. Sidhika, M. Kalaivani, and C. Deisy, “Prediction
journal of medical informatics, vol. 83, no. 10, pp. 691–714, 2014.
of prediabetes using fuzzy logic based association classification,” in
[16] P. Bhattacharjee, D. P. Edelson, and M. M. Churpek, “Identifying patients
2018 Second International Conference on Inventive Communication and
with sepsis on the hospital wards,” Chest, vol. 151, no. 4, pp. 898–907,
Computational Technologies (ICICCT). IEEE, 2018, pp. 782–787.
2017.
[39] N. Allahverdi and T. Akcan, “A fuzzy expert system design for diagnosis
[17] M. Sinha, J. Jupe, H. Mack, T. P. Coleman, S. M. Lawrence, and of periodontal dental disease,” in 2011 5th International Conference on
S. I. Fraley, “Emerging technologies for molecular diagnosis of sepsis,” Application of Information and Communication Technologies (AICT).
Clinical microbiology reviews, vol. 31, no. 2, 2018. IEEE, 2011, pp. 1–5.
[18] M. S. M. Aras, F. A. Ali, F. A. Azis, S. M. S. S. A. Hamid, and [40] T. M. Tuan, H. Fujita, N. Dey, A. S. Ashour, V. T. N. Ngoc, D.-T. Chu
M. F. H. M. Basar, “Performances evaluation and comparison of two et al., “Dental diagnosis from x-ray images: an expert system based on
algorithms for fuzzy logic rice cooking system (matlab fuzzy logic fuzzy computing,” Biomedical Signal Processing and Control, vol. 39,
toolbox and fuzzytech),” in 2011 IEEE Conference on Open Systems. pp. 64–73, 2018.
IEEE, 2011, pp. 400–405. [41] U. A. Umoh and M. M. Ntekop, “A proposed fuzzy framework for
[19] F.-M. E. Uzoka, J. Osuji, F. O. Aladi, and O. U. Obot, “A framework for cholera diagnosis and monitoring,” International Journal of Computer
cell phone based diagnosis and management of priority tropical diseases,” Applications, vol. 82, no. 17, pp. 1–10, 2013.
in 2011 IST-Africa Conference Proceedings. IEEE, 2011, pp. 1–13. [42] M. Okpor, “Using fuzzy classifier for cholera analysis,” Intern J Sci Res,
[20] P. Nesteruk, L. Nesteruk, and I. Kotenko, “Creation of a fuzzy knowledge vol. 3, pp. 314–317, 2012.
base for adaptive security systems,” in 2014 22nd Euromicro Interna- [43] S. S. Godil, M. S. Shamim, S. A. Enam, and U. Qidwai, “Fuzzy logic: A
tional Conference on Parallel, Distributed, and Network-Based Process- “simple” solution for complexities in neurosciences?” Surgical neurology
ing. IEEE, 2014, pp. 574–577. international, vol. 2, 2011.
[21] G. Licata et al., “Employing fuzzy logic in the diagnosis of a clinical [44] P. G. Brust-Renck, V. F. Reyna, E. A. Wilhelms, and A. N. Lazar, “A
case,” Health, vol. 2, no. 03, p. 211, 2010. fuzzy-trace theory of judgment and decision-making in health care: ex-
[22] M. Rana and R. Sedamkar, “Design of expert system for medical diagno- planation, prediction, and application,” in Handbook of Health Decision
sis using fuzzy logic,” International Journal of Scientific & Engineering Science. Springer, 2016, pp. 71–86.
Research, vol. 4, no. 6, pp. 2914–2921, 2013. [45] M. I. Roveri, E. de Jesus Manoel, A. N. Onodera, N. R. Ortega, V. D.
[23] L. A. Zadeh, “Fuzzy sets,” Information and control, vol. 8, no. 3, pp. Tessutti, E. Vilela, N. Evêncio, and I. C. Sacco, “Assessing experience in
338–353, 1965. the deliberate practice of running using a fuzzy decision-support system,”
[24] N. H. Phuong and V. Kreinovich, “Fuzzy logic and its applications in PloS one, vol. 12, no. 8, 2017.
medicine,” International journal of medical informatics, vol. 62, no. 2-3, [46] E. Hüllermeier, “Does machine learning need fuzzy logic?” Fuzzy Sets
pp. 165–173, 2001. and Systems, vol. 281, pp. 292–299, 2015.
[47] A. C. Jamgade and S. Zade, “Disease prediction using machine learning,”
[25] E. Braunwald, “Heart disease,” 1988.
International Research Journal of Engineering and Technology, vol. 6,
[26] A. Ansari and N. K. Gupta, “Automated diagnosis of coronary heart
no. 5, pp. 6937–6938, 2019.
disease using neuro-fuzzy integrated system,” in 2011 World Congress on
[48] G. Zhang, “A modified svm classifier based on rs in medical disease
Information and Communication Technologies. IEEE, 2011, pp. 1379–
prediction,” in 2009 Second International Symposium on Computational
1384.
Intelligence and Design, vol. 1. IEEE, 2009, pp. 144–147.
[27] H. H. de Carvalho Junior, R. L. Moreno, T. C. Pimenta, P. C. Crepaldi, [49] A. S. Levey and J. Coresh, “Chronic kidney disease,” The lancet, vol.
and E. Cintra, “A heart disease recognition embedded system with fuzzy 379, no. 9811, pp. 165–180, 2012.
cluster algorithm,” Computer methods and programs in biomedicine, vol. [50] P. Sinha and P. Sinha, “Comparative study of chronic kidney disease
110, no. 3, pp. 447–454, 2013. prediction using knn and svm,” International Journal of Engineering
[28] O. Oluwagbemi, F. Oluwagbemi, and O. Abimbola, “Ebinformatics: Research and Technology, vol. 4, no. 12, pp. 608–12, 2015.
Ebola fuzzy informatics systems on the diagnosis, prediction and rec- [51] A. Charleonnan, T. Fufaung, T. Niyomwong, W. Chokchueypattanakit,
ommendation of appropriate treatments for ebola virus disease (evd),” S. Suwannawach, and N. Ninchawee, “Predictive analytics for chronic
Informatics in Medicine Unlocked, vol. 2, pp. 12–37, 2016. kidney disease using machine learning techniques,” in 2016 Management
[29] T. R. Insel and B. N. Cuthbert, “Brain disorders? precisely,” Science, vol. and Innovation Technology International Conference (MITicon). IEEE,
348, no. 6234, pp. 499–500, 2015. 2016, pp. MIT–80.
[30] N. N. Gopal and M. Karnan, “Diagnose brain tumor through mri using [52] B. Zheng, S. W. Yoon, and S. S. Lam, “Breast cancer diagnosis based on
image processing clustering algorithms such as fuzzy c means along feature extraction using a hybrid of k-means and support vector machine
with intelligent optimization techniques,” in 2010 IEEE International algorithms,” Expert Systems with Applications, vol. 41, no. 4, pp. 1476–
Conference on Computational Intelligence and Computing Research. 1482, 2014.
IEEE, 2010, pp. 1–4. [53] H. Asri, H. Mousannif, H. Al Moatassime, and T. Noel, “Using machine
[31] H.-L. Chen, C.-C. Huang, X.-G. Yu, X. Xu, X. Sun, G. Wang, and S.- learning algorithms for breast cancer risk prediction and diagnosis,”
J. Wang, “An efficient diagnosis system for detection of parkinson’s Procedia Computer Science, vol. 83, pp. 1064–1069, 2016.
disease using fuzzy k-nearest neighbor approach,” Expert systems with [54] K. Sinusas, “Osteoarthritis: diagnosis and treatment,” American family
applications, vol. 40, no. 1, pp. 263–271, 2013. physician, vol. 85, no. 1, pp. 49–56, 2012.

VOLUME 4, 2016 21

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

[55] N. Bhargava, R. Purohit, S. Sharma, and A. Kumar, “Prediction of [78] G. Jain, D. Mittal, D. Thakur, and M. K. Mittal, “A deep learning ap-
arthritis using classification and regression tree algorithm,” in 2017 2nd proach to detect covid-19 coronavirus with x-ray images,” Biocybernetics
International Conference on Communication and Electronics Systems and Biomedical Engineering, vol. 40, no. 4, pp. 1391–1405, 2020.
(ICCES). IEEE, 2017, pp. 606–610. [79] N. El-Rashidy, S. El-Sappagh, S. Islam, H. M. El-Bakry, and S. Abdel-
[56] N. Barakat, A. P. Bradley, and M. N. H. Barakat, “Intelligible support razek, “End-to-end deep learning framework for coronavirus (covid-19)
vector machines for diagnosis of diabetes mellitus,” IEEE transactions detection and monitoring,” Electronics, vol. 9, no. 9, p. 1439, 2020.
on information technology in biomedicine, vol. 14, no. 4, pp. 1114–1120, [80] S. Toraman, T. B. Alakus, and I. Turkoglu, “Convolutional capsnet: A
2010. novel artificial neural network approach to detect covid-19 disease from
[57] J. P. Kandhasamy and S. Balamurali, “Performance analysis of classifier x-ray images using capsule networks,” Chaos, Solitons & Fractals, vol.
models to predict diabetes mellitus,” Procedia Computer Science, vol. 47, 140, p. 110122, 2020.
pp. 45–51, 2015. [81] A. Jaiswal, N. Gianchandani, D. Singh, V. Kumar, and M. Kaur, “Clas-
[58] T. V. Sriram, M. V. Rao, G. S. Narayana, D. Kaladhar, and T. P. R. Vital, sification of the covid-19 infected patients using densenet201 based deep
“Intelligent parkinson disease prediction using machine learning algo- transfer learning,” Journal of Biomolecular Structure and Dynamics, pp.
rithms,” International Journal of Engineering and Innovative Technology 1–8, 2020.
(IJEIT), vol. 3, no. 3, pp. 1568–1572, 2013. [82] D. T. Ginat, “Analysis of head ct scans flagged by deep learning software
[59] C. Salvatore, A. Cerasa, I. Castiglioni, F. Gallivanone, A. Augimeri, for acute intracranial hemorrhage,” Neuroradiology, vol. 62, no. 3, pp.
M. Lopez, G. Arabia, M. Morelli, M. Gilardi, and A. Quattrone, “Ma- 335–340, 2020.
chine learning on brain mri data for differential diagnosis of parkinson’s [83] T. Lewick, M. Kumar, R. Hong, and W. Wu, “Intracranial hemorrhage
disease and progressive supranuclear palsy,” Journal of Neuroscience detection in ct scans using deep learning,” in 2020 IEEE Sixth Inter-
Methods, vol. 222, pp. 230–237, 2014. national Conference on Big Data Computing Service and Applications
[60] A. L. Pineda, Y. Ye, S. Visweswaran, G. F. Cooper, M. M. Wagner, (BigDataService). IEEE, 2020, pp. 169–172.
and F. R. Tsui, “Comparison of machine learning classifiers for in- [84] F.-C. Chen and M. R. Jahanshahi, “Nb-cnn: Deep learning-based crack
fluenza detection from emergency department free-text reports,” Journal detection using convolutional neural network and naïve bayes data fu-
of biomedical informatics, vol. 58, pp. 60–69, 2015. sion,” IEEE Transactions on Industrial Electronics, vol. 65, no. 5, pp.
[61] S. A. Dwivedi, R. Borse, and A. M. Yametkar, “Lung cancer detection 4392–4400, 2017.
and classification by using machine learning & multinomial bayesian,” [85] D. Moher, L. Shamseer, M. Clarke, D. Ghersi, A. Liberati, M. Petticrew,
IOSR Journal of Electronics and Communication Engineering (IOSR- P. Shekelle, and L. A. Stewart, “Preferred reporting items for systematic
JECE), vol. 9, no. 1, pp. 69–75, 2014. review and meta-analysis protocols (prisma-p) 2015 statement,” System-
[62] S. Marsland, Machine learning: an algorithmic perspective. CRC press, atic reviews, vol. 4, no. 1, p. 1, 2015.
2015. [86] A. Liberati, D. G. Altman, J. Tetzlaff, C. Mulrow, P. C. Gøtzsche, J. P.
[63] I. Kononenko, I. Bratko, and M. Kukar, “Application of machine learning Ioannidis, M. Clarke, P. J. Devereaux, J. Kleijnen, and D. Moher, “The
to medical diagnosis,” Machine Learning and Data Mining: Methods and prisma statement for reporting systematic reviews and meta-analyses of
Applications, vol. 389, p. 408, 1997. studies that evaluate health care interventions: explanation and elabora-
[64] K. Suzuki, “Overview of deep learning in medical imaging,” Radiological tion,” Annals of internal medicine, vol. 151, no. 4, pp. W–65, 2009.
physics and technology, vol. 10, no. 3, pp. 257–273, 2017. [87] K. Elasnaoui and Y. Chawki, “Using x-ray images and deep learning
for automated detection of coronavirus disease,” Journal of Biomolecular
[65] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” nature, vol. 521,
Structure and Dynamics, no. just-accepted, pp. 1–22, 2020.
no. 7553, pp. 436–444, 2015.
[88] A. Holzinger, G. Langs, H. Denk, K. Zatloukal, and H. Müller, “Caus-
[66] H. Suzuki, H. Ohsaki, and H. Sawai, “A network-based computational
ability and explainability of artificial intelligence in medicine,” Wiley In-
model with learning,” in International Conference on Unconventional
terdisciplinary Reviews: Data Mining and Knowledge Discovery, vol. 9,
Computation. Springer, 2010, pp. 193–193.
no. 4, p. e1312, 2019.
[67] R. Miotto, F. Wang, S. Wang, X. Jiang, and J. T. Dudley, “Deep learning
[89] N. Elazab, H. Soliman, S. El-Sappagh, S. Islam, and M. Elmogy, “Ob-
for healthcare: review, opportunities and challenges,” Briefings in bioin-
jective diagnosis for histopathological images based on machine learning
formatics, vol. 19, no. 6, pp. 1236–1246, 2018.
techniques: Classical approaches and new trends,” Mathematics, vol. 8,
[68] R. Anderson, A. Biong, and D. A. Gómez-Gualdrón, “Adsorption no. 11, p. 1863, 2020.
isotherm predictions for multiple molecules in mofs using the same deep [90] D. Tellez, G. Litjens, P. Bándi, W. Bulten, J.-M. Bokhorst, F. Ciompi, and
learning model,” Journal of Chemical Theory and Computation, 2020. J. van der Laak, “Quantifying the effects of data augmentation and stain
[69] H. Liao, “A deep learning approach to universal skin disease classifica- color normalization in convolutional neural networks for computational
tion,” University of Rochester Department of Computer Science, CSC, pathology,” Medical image analysis, vol. 58, p. 101544, 2019.
2016. [91] S. K. Saha, S. R. Islam, K.-S. Kwak, M. S. Rahman, and S.-G. Cho,
[70] D. A. Shoieb, S. M. Youssef, and W. M. Aly, “Computer-aided model “Prom1 and prom2 expression differentially modulates clinical prognosis
for skin diagnosis using deep learning,” Journal of Image and Graphics, of cancer: a multiomics analysis,” Cancer Gene Therapy, vol. 27, no. 3,
vol. 4, no. 2, pp. 122–129, 2016. pp. 147–167, 2020.
[71] A. M. Abdel-Zaher and A. M. Eldeib, “Breast cancer classification using [92] S. El-Sappagh, T. Abuhmed, S. R. Islam, and K. S. Kwak, “Multimodal
deep belief networks,” Expert Systems with Applications, vol. 46, pp. multitask deep learning model for alzheimer’s disease progression detec-
139–144, 2016. tion based on time series data,” Neurocomputing, vol. 412, pp. 197–215,
[72] S. Charan, M. J. Khan, and K. Khurshid, “Breast cancer detection in 2020.
mammograms using convolutional neural network,” in 2018 International [93] X. Yaya and Z. Bi-Geng, “Research on medical image storage and
Conference on Computing, Mathematics and Engineering Technologies retrieval system based on hadoop,” in Journal of Physics: Conference
(iCoMET). IEEE, 2018, pp. 1–5. Series, vol. 1544, no. 1. IOP Publishing, 2020, p. 012119.
[73] G. Swapna, R. Vinayakumar, and K. Soman, “Diabetes detection using [94] T. Zielke, “Is artificial intelligence ready for standardization?” in Euro-
deep learning algorithms,” ICT Express, vol. 4, no. 4, pp. 243–246, 2018. pean Conference on Software Process Improvement. Springer, 2020,
[74] D. Sisodia and D. S. Sisodia, “Prediction of diabetes using classification pp. 259–274.
algorithms,” Procedia computer science, vol. 132, pp. 1578–1585, 2018. [95] S. R. Islam, D. Kwak, M. H. Kabir, M. Hossain, and K.-S. Kwak, “The
[75] J. Rubin, R. Abreu, A. Ganguli, S. Nelaturi, I. Matei, and K. Sricharan, internet of things for health care: a comprehensive survey,” IEEE access,
“Recognizing abnormal heart sounds using deep learning,” arXiv preprint vol. 3, pp. 678–708, 2015.
arXiv:1707.04642, 2017. [96] M. Afzal, S. R. Islam, M. Hussain, and S. Lee, “Precision medicine
[76] K. H. Miao and J. H. Miao, “Coronary heart disease diagnosis using deep informatics: principles, prospects, and challenges,” IEEE Access, vol. 8,
neural networks,” Int. J. Adv. Comput. Sci. Appl., vol. 9, no. 10, pp. 1–8, pp. 13 593–13 612, 2020.
2018. [97] F. Hussain, R. Hussain, S. A. Hassan, and E. Hossain, “Machine learning
[77] W. Sun, B. Zheng, and W. Qian, “Computer aided lung cancer diagnosis in iot security: current solutions and future challenges,” IEEE Communi-
with deep learning algorithms,” in Medical imaging 2016: computer- cations Surveys & Tutorials, 2020.
aided diagnosis, vol. 9785. International Society for Optics and Pho- [98] S. El-Sappagh, H. Saleh, R. Sahal, T. Abuhmed, S. R. Islam, F. Ali,
tonics, 2016, p. 97850Z. and E. Amer, “Alzheimer’s disease progression detection model based

22 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3042273, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

on an early fusion of cost-effective multimodal data,” Future Generation

Computer Systems, vol. 115, pp. 680–699.
[99] W. Wang, J. Lee, F. Harrou, and Y. Sun, “Early detection of parkinson’s
disease using deep learning and machine learning,” IEEE Access, vol. 8,
pp. 147 635–147 646, 2020.
[100] F. Ali, S. El-Sappagh, S. R. Islam, A. Ali, M. Attique, M. Imran, and K.-
S. Kwak, “An intelligent healthcare monitoring framework using wear-
able sensors and social networking data,” Future Generation Computer
Systems, vol. 114, pp. 23–43, 2020.
[101] F. Ali, S. El-Sappagh, S. R. Islam, D. Kwak, A. Ali, M. Imran, and K.-S.
Kwak, “A smart healthcare monitoring system for heart disease predic-
tion based on ensemble deep learning and feature fusion,” Information
Fusion, vol. 63, pp. 208–222, 2020.
[102] J. Wolff, J. Pauling, A. Keck, and J. Baumbach, “The economic impact
of artificial intelligence in health care: Systematic review,” Journal of
Medical Internet Research, vol. 22, no. 2, p. e16866, 2020.

VOLUME 4, 2016 23

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
View publication stats