Open Access, Volume - 2
Feature Selecon Based Fuzzy Expert System for Ecient Diagnosis of
Coronary Artery Disease
*
Case Report
*Corresponding Author: LJ Muhammad, Computer Science Depart-
ment, Federal University of Kashere, Gombe State, Nigeria.
Email: lawan.jibril@fukashere.edu.ng
Received Date : Feb 11, 2022
Accepted Date : Mar 04, 2022
Published Date : Mar 18, 2022
Archived : www.jcmimagescasereports.org
Copyright : © LJ Muhammad 2022
Abstract
Coronary artery disease (CAD) is one of the most dangerous diseases which lead to sudden cardiac death. According to Word
Health Organizaon, CAD is the number one killer in the developed world, with over 7.4 million deaths aributed it. Before,
CAD is not common disease in Nigeria, however, is at this moment gaining much popularity in the country following the rising
number of health issues related to CAD diseases, including higher death rate, which is mostly due to lack of proper awareness
among the common people. The diagnosis of CAD is very expensive and me consuming which made computer sciensts to
use arcial techniques such as expert system to diagnose CAD’s paents. Therefore, in this work, fuzzy based expert system
for ecient diagnosis of coronary artery disease has been developed, implemented and evaluated. Hence, the system has
archived 90.08% overall accuracy which is very excellent, thus the accuracy determines the proporon of the total number
of predicons that were correct. At the same me, the system has 91.30% accuracy to classify of normal paents correctly by
the system (specicity) and 90.24% accuracy to classify abnormal paents correctly by the system (sensivity). This showed
that, the system performed eciently and excellently to diagnose CAD.
Keywords: CAD; Arcial Intelligence; Expert System; Fuzzy Logic; dataset; diagnosis.
Volume 2 | Issue 2 | 2022 1
Introducon
Coronary Artery Disease (CAD) is one of the deadliest diseases
in the world. It has been esmated that nearly one half of all
middle-aged men and one third of middle-aged women in the
United States have been aected with the CAD disease [46,
50]. In the developed countries CAD is one of the number one
killers with over 7.4 million deaths aributed to [29, 48]. It has
been esmated that, CAD is one in every seven deaths in the
United States is due to heart disease. CAD is the primary cause
of death in women, taking more lives than all cancers com-
bined The proporon of deaths in the United States that are
due to CAD has been decreasing slowly but connuously over
the past half century. Nonetheless, CAD remains the single
most common cause of death in the United States, according
[16, 19, 46]. In Nigeria, CAD is at the moment gaining much
popularity following the rising number of health issues related
to the disease, including higher death rate, which is mostly
due to lack of proper awareness among the common people.
According to the latest WHO data published in May 2014
Coronary Artery Disease Deaths in Nigeria reached 53,836
or 2.82% of total deaths. The rising gures of health issues
(cases) and mortality rate related to CAD in Nigeria recently
are alarming [2, 7, 26-27]. Research outputs have suggested
a rather increasing occurrence of instances of the CAD in the
Nigerian community, which however, the populace seems not
to be well informed and/or alarmed about it, and also, given
the health emergencies response level of Nigeria, the Nigerian
health systems do not seem very readily capable to deal with
the menace of CAD in both an immediate and a connuous
strategy to lower and/or overcome its adverse eects on the
populaon [26-27]. However, the major problem in tradional
method of medical diagnosis is inadequate guarantee of pre-
cision and accuracy. There are huge data management tools
available within health care systems, but analysis tools are not
sucient to discover hidden relaonships amongst the data.
Most of medical informaon is vague, imprecise and uncer-
tain [8]. Extracng correct informaon from this data is con-
sidered an art [16, 12-13]. It can be said to be an art because
it is complicated by many factors and its soluon involves liter-
ally all of a human’s abilies including technical experse and
jcmimagescasereports.org
Citaon: Y Atomsa, LJ Muhammad, FS Ishaq, Yusuf Abdullahi. Feature Selecon Based Fuzzy Expert System for Ecient Diag-
nosis of Coronary Artery Disease. J Clin Med Img Case Rep. 2022; 2(2): 1103.
Volume 2 | Issue 2 | 2022 2
intuion. Hence, it becomes necessarily essenal to leverage
compung technologies that support analycal processing
of this data to extract hidden correlaons and intelligences
within the data in order to improve accuracy and precision of
medical diagnosis [15].
Expert systems have been specically applied in a variety of
life sciences support systems development, ranging from stor-
age and retrieval of medical records, diagnoscs, up to expert
knowledge/decision support systems [3-5, 9]. Expert system
dened as an intelligence system that extracted its knowledge
using appropriate technique with the percepon of human ex-
pert, to solve the problems or make decision as human being
does [10, 44]. Expert System is dened as an intelligence sys-
tem which uses extracted knowledge from past domain expert
decision making reasoning in form of rules to solve problems
that ordinarily require human experse for their soluon, and
has the capability to update its rule-base as new knowledge
is discovered. There are many applicaon areas of expert
system such as medicine, educaon, agriculture, oil and gas,
environment, law, manufacturing, telecommunicaon and
power systems etc. [11]. This research work aims to develop
a fuzzy based expert system for supporng and compleng
human experse in the diagnosis of Coronary Artery Disease.
The focus is leveraging compung technology systems for the
ecient, precise and re-accessible diagnosis procedure for the
Coronary Artery Disease.
Related Work
In this study research arcles and conference papers pub-
lished by reputable pub¬lishers that employed expert system
for diagnosis of coronary artery disease were searched and
reviewed in this secon. In work [47] an expert system for
the diagnosis of the level of coronary heart disease by tak-
ing into account the problem of data imbalance developed.
In the study of [14], a fuzzy so sets expert system to predict
paents suer coronary artery disease was developed. The re-
search was a pioneering approach in applying fuzzy so sets to
a medical diagnosis problem in the form of predicng paents
who may be suering from coronary artery disease. In study
of [39], a web based Fuzzy Logic-based Expert System for the
diagnosis of heart failure disease was developed. An evolu-
onary fuzzy expert system is proposed for the diagnosis of
the Coronary Artery Disease (CAD) based on Cleveland clinic
foundaon datasets for heart diseases in the study of [49]. In
the study of [1], paents with coronary artery disease were
idened and classied through the neuro-fuzzy network with
the capacity of automacally extracng fuzzy rules. Fuzzy ex-
pert system was implemented using facilies and funcons of
MATLAB soware (7.12.0 version). A fuzzy expert system for
diagnosis of coronary artery disease by a non-invasive pro-
cedure was implemented in the study of [33]. The adapve
neuro fuzzy inference system and Advanced fuzzy resoluon
mechanism has been proposed in the study of [17] to diag-
nose the heart disease. The work of [38] has developed a com-
puter intelligent based approach for the diagnosis of heart dis-
eases. A fuzzy rule-based system which concentrated only on
accuracy and interpretability of the system was proposed by
Ref [36] and system that provided a heart disease paent with
background for suitable diagnosis and treatment. Ref. [42]
developed a weighted fuzzy rule-based Clinical Decision Sup-
port System (CDSS) for computer-aided diagnosis of the heart
disease. The dierent data mining techniques such as neural
networks, decision trees and naive bayes has been proposed
in the work of [45] for the study of heart disease predicon
system. A coronary artery disease fuzzy expert system for mi-
croarray data classicaon using a novel Genec Swarm Al-
gorithm, has been proposed in the study of [18] for obtaining
near rule set and membership funcon tuning. In the study of
[43] screening system has been developed for the early detec-
on of Coronary Artery Disease. In the study of [33], a fuzzy
rule-based system was designed to serve as a decision support
system for diagnosis Coronary heart disease. In the work [6]
a Fuzzy Expert System for heart disease diagnosis using V.A.
Medical Center, Long Beach and Cleveland Clinic Foundaon
database was designed and system is being designed with in
Matlab soware and it is viewed as an alternave for exisng
methods to disnguish of heart disease presence.
SN Feature Units Range
1 Age Years 1 – 150
2 Sex Male (1), Female (0) 0,1
3 Family History Yes (1), No (0) 0,1
4 Smoking Yes (1), No (0) 0,1
5 Diabetes Yes (1), No (0) 0,1
6 Hypertension Yes (1), No (0) 0,1
7 Hyperlipimedia Yes (1), No (0) 0,1
8 Blood Pressure mmHg 90 – 190
9 Glucose mg/dL 37 – 295
10 Cholesterol mg/dL 128 – 575
11 Triglyceride mg/dL 40 – 690
12 HDL mg/dL 10.6 – 73
13 LDL mg/dL 10 – 220
14 Creanine mg/dL 0.6 – 3.3
15 Body mass index kg/m
2
20.28 – 40.25
16 Heart rate Bpm 42 – 124
17 Chest pain
Typical Angina (4),
Atypical Angina(3),
Non- Anginal pain(2),
Asymptomac (1)
1 – 4
18 Diagnosis of CAD
Posive (1), Nega-
ve (2)
0,1
NB: mmHg stands for millimeters of mercury, mg/dL stands for milligrams
per deciliter, kg/m
2
stands for Kilogram-Meter Squared and Bpm stands
for beats per minute
Methods and Materials
Dataset
The medical expert diagnosc dataset for coronary artery
disease obtained at the Federal Teaching Hospital, Gombe
State, Nigeria were prepared in the appropriate format with
the helped of medical experts in the hospitals and only data
instances of the dataset without missing values were consid-
ered and collected. Therefore, there are only one thousand
two hundred and one data instances of the dataset without
missing value. The dataset is labeled one with eighteen fea-
tures including demographic, history and clinical features of
the paent’s CAD. The feature of the dataset are age, sex, CAD
family history, smoking, type of the chest pain, diabetes, glu-
cose, hypertension, blood pressure, cholesterol, Hyperlipid-
emia, high density lipoprotein (HDL), Triglyceride, low density
lipoprotein (LDL), Creanine, BodyMass, HeartRate and Diag-
nosis (Table 1) shows the descripon of the dataset features.
Feature Selecon with Correlaon Analysis
Correlaon coecient analysis was carried out on the de-
pendent and independent features of the CAD Dataset [20].
Correlaon coecient is used to determine the strength rela-
onship that exists between the dependent and independent
features which can either be posive or negave. The r value
is a set of innite number between -1 to +1 which show the
exisng relaonship either posive or negave between the
dependent and independent features [22-24]. The feature can
be evaluated by the equaon (1) below:-
Volume 2 | Issue 2 | 2022 3
jcmimagescasereports.org
(1)
Where the Importance is the correlaon coecient between
dependent feature set and independent feature and is the
ranking criteria for evaluang the set of feature, (avg(corrfc))
is the average of the correlaon between the dependent fea-
ture and the independent feature, avg(corr) is the average
of the correlaon between feature set, and k is the number
of features. Correlaon coecient analysis was carried out on
the dependent features of the CAD dataset which include Age,
Sex, Family History, Smoking, Chest Pain, Diabetes, Glucose,
Hypertension, Blood Pressure, Cholesterol, Hyperlipidemia
HDL, Triglyceride, LDL, Creanine, Body Mass and Heart Rate
and Diagnosis feature which is an independent features of the
CAD Dataset. (Table 2) and shows the r value of dependent
feature against the independent feature of the dataset while
(Figure 1) shows enre the correlaon coecient analysis ma-
trix of the dataset features.
SN Dependent Feature Independent feature r value correlaon coecient relaonship
1 Age Medical Diagnosc Result 0.42 Moderate uphill posive correlaon coecient relaonship
2 Sex Medical Diagnosc Result 0.50 Moderate uphill posive correlaon coecient relaonship
3 FamilyHistory Medical Diagnosc Result 0.48 Moderate uphill posive correlaon coecient relaonship
4 Smoking Medical Diagnosc Result 0.24 Weak uphill posive correlaon coecient relaonship
5 ChestPain Medical Diagnosc Result 0.58 Moderate uphill posive correlaon coecient relaonship
6 Diabetes Medical Diagnosc Result 0.61 Strong uphill posive correlaon coecient relaonship
7 Glucose Medical Diagnosc Result 0.55 Moderate uphill posive correlaon coecient relaonship
8 Hypertension Medical Diagnosc Result 0.65 Strong uphill posive correlaon coecient relaonship
9 BLoodPressure Medical Diagnosc Result 0.53 Moderate uphill posive correlaon coecient relaonship
10
Cholesterol
Medical Diagnosc Result 0.44 Moderate uphill posive correlaon coecient relaonship
11 Hyperlipidemia Medical Diagnosc Result -0.50 Moderate uphill negave correlaon coecient relaonship
12 HDL Medical Diagnosc Result -0.20 weak uphill negave correlaon coecient relaonship
13 Triglyceride Medical Diagnosc Result 0.28 Weak uphill posive correlaon coecient relaonship
14 LDL Medical Diagnosc Result 0.35 Moderate uphill posive correlaon coecient relaonship
15 Creanine Medical Diagnosc Result 0.40 Moderate uphill posive correlaon coecient relaonship
16 BodyMass Medical Diagnosc Result 0.50 Moderate uphill posive correlaon coecient relaonship
17 HeartRate Medical Diagnosc Result 0.53 Moderate uphill posive correlaon coecient relaonship
We remove the all independent aributes that have less than
0.50 posive correlaon coecient relaonships with depen-
dent aributes of the dataset as shown in (Table 3).
Mining Dataset with selected features
Interpreng mined paerns: in this phase, the visualizaon
the extracted hidden paerns or knowledge using C4.5 algo-
rithm was generated and it is called decision tree. However,
only independent aributes with more with 0.50 posive cor-
relaons and above were used to build the decision tree with
C4.5 algorithm. Figure 2 has shown the visualizaon decision
tree of the algorithm. C4.5 is one of the data mining classica-
on algorithms and it is an extension of Quinlan’s earlier ID3
algorithm. C4.5 algorithm was proposed by Ross Quinlan in
1993 to overcome some of the limitaons of ID3 algorithm
[25, 28, 51]. One the limitaon of ID3 algorithm overcomes
by C4.5 is ID3 sensivity to features with number of values
[30, 41].
The decision tree generated with C4.5 algorithm was con-
verted or transformed into or crisp rules. Below are the cor-
responding crisp rules of generated from the decision tree in
(Figure 2).
IF (HeartRate <99.5 mg/dl and BP <152.5 mg/dl and Hyperten-
sion =No and Diabates < 0.15 and Sex < 47 ) THEN Negave
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension= Yes and Diabates < 0.15 and Sex > 47 ) THEN Posive
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension = Yesand Diabates <2.85 ) THEN Negave
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension = No and Diabates =>1.52 and Chest pain= non_angi-
nal ) THEN Negave
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension =No and Diabates =>2.2 and Chest pain= asymt and
BMI >=19) THEN Negave
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension =Yes and Diabates = Yes and Chest pain= asymt and
BMI >=19 and Sex < 65) THEN Posive
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension >=152.2 and Diabates = Yes and Chest pain= atyp_an-
gina and Glucose < 69.5) THEN Posive
• IF (HeartRate <99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension >=152.2 and Diabates = Yes1 and Chest pain= atyp_an-
gina and Glucose >= 69.5) THEN Posive
• IF (HeartRate <=99.5 mg/dl and BP < 152.5 mg/dl and Hyper-
tension >=152.2 and Diabates = No and Chest pain= atyp_an-
gina and Glucose >= 69.5) THEN Negave
SN Dependent Feature Independent feature r value Correlaon coecient relaonship
1 Sex Medical Diagnosc Result 0.50 Moderate uphill posive correlaon coecient relaonship
2 ChestPain Medical Diagnosc Result 0.58 Moderate uphill posive correlaon coecient relaonship
3 Diabetes Medical Diagnosc Result 0.61 Strong uphill posive correlaon coecient relaonship
4 Glucose Medical Diagnosc Result 0.55 Moderate uphill posive correlaon coecient relaonship
5 Hypertension Medical Diagnosc Result 0.65 Strong uphill posive correlaon coecient relaonship
6 BLoodPressure Medical Diagnosc Result 0.53 Moderate uphill posive correlaon coecient relaonship
7 BodyMass Medical Diagnosc Result 0.50 Moderate uphill posive correlaon coecient relaonship
8 HeartRate Medical Diagnosc Result 0.53 Moderate uphill posive correlaon coecient relaonship
a) Sex: This input has two instances; either the paent is male
or female. Male = 1 and Female = 0 Hence, there is no fuzzi-
ness or overlap for this input. The membership funcons of
the linguisc variables of the sex input is shown is (Figure 3).
Discussion
Autoimmune abnormalies are a known complicaon of im-
munotherapy drugs, which can be used to treat a wide range
of malignancies including lung and breast cancer and mela-
noma, to name a few [2]. Specically, autoimmune endocrine
adverse eects resulng from cancer immunotherapy drugs
include hypophysis, hyperthyroidism, hypothyroidism and
adrenal insuciency. There are several immune checkpoint
inhibitor drugs already on the market for cancer immunother-
apy, such as the an-CTLA-4 (ipilimumab and tremelimumab)
and an-PD1 anbodies (pembrolizumab, nivolumab and pi-
dilizumab), all of which have been shown to have autoimmune
side eects. Ipilimumab, pembrolizumab and nivolumab,
however, are the most commonly used and studied for their
Input (Variable) Range Linguisc Term
Sex
1
2
Male
Female
Chest Pain
1
2
3
4
Typical angina
Atypical angina
Non-angina
Asymptomac
Diabetes
1
0
Yes
No
Glucose
<108 mg/dL
100–126 mg/dL
>120 mg/dL
Low (Normal)
Normal (Prediabetes)
High (Diabetes)
Hypertension
1
0
Yes
No
Blood Pressure (BP)
< 134 mmHg
128 - 154
mmHg
> 147 mmHg
Low (Hypotension)
Normal (Nomotension)
High (Hypertension)
Body Mass Index
(BMI)
< 10 kg/m
2
8- 25 kg/m
2
> 22 kg/m
2
Underweight
Normal
Obese
Heart Rate
(HR)
< 50 bpm
45 - 75 bpm
> 70 bpm
Low
Normal
Fast
Diagnosis Result
< 4
2 - 6
4- 8
> 6
Healthy
Mild
Moderate
Severe
a) Sex: This input has two instances; either the paent is male
or female. Male = 1 and Female = 0 Hence, there is no fuzzi-
Volume 2 | Issue 2 | 2022 4
jcmimagescasereports.org
Fuzzicaon
Moreover, aer determining the linguisc variable of each at-
tribute, converng crisp value into fuzzy values, the crisp set
rules generated earlier were converted into fuzzy set of rules.
Fuzzicaon is the process of transforming crisp values into
grades of membership for linguisc terms of fuzzy sets. The
membership funcon is used to associate a grade to each lin-
guisc term. Below are some of the fuzzy rules.
• IF (HR is Normal and BP is Normal and Glucose is Low
and Diabetes is Yes and Sex is Female ) THEN Healthy
• IF (HR is Normal and BP is Low and Glucose = Normal
is High and Diabetes = Yes and Sex is Male ) THEN Mild
• IF (HR is Low and BP is Normal and Glucose is Norma
and Diabetes is Yes and Sex is Male ) THEN Moderate
• IF (HR is Low and BP is Low and Glucose is Low and
Diabetes is No and Sex is Female) THEN Severe
• IF (HR is Normal and BP is Normal and Glucose is High
and Diabete ) THEN Healthy
• IF (HR is Low and BP is Low and Glucose is High and
Diabetes is Yes) THEN Mild
• IF (HR is Normal and BP is Normal and Glucose is
High and Diabetes is No and Chest pain is non_anginal) THEN
Healthy
• IF (HR is Low and BP is Low and Glucose is High and
Diabetes is Yes and Chest pain is non_anginal ) THEN Mild
• IF (HR is Low and BP is Normal and Glucose is High
and Diabetes is No and Chest pain is asymt and and BMI is
Normal) THEN Healthy
• IF (HR is Normal and BP is Low and Glucose is High
and Diabetes is No and Chest pain is asymt and and BMI is
High)THEN Mild
• IF (HR is Low and BP is Normal and Glucose is High
and Diabetes is No and Chest pain is asymt and and BMI is
Normal and Sex is Female ) THEN Moderate
Knowledge inference or Knowledge reasoning
Knowledge inference or Knowledge reasoning: this involves
applicaon of logical rules to the knowledge to deduce new
informaon. The inference engine draws conclusions from the
replicated human experse in the knowledge base of the ex-
pert system, which is the hallmark of the expert system [37,
40]. Mamdani inference technique is used to smulate rea-
soning of expert physicians in eld of diagnosis of coronary
artery disease in this work.. The usage of Mamdani technique
ness or overlap for this input. The membership funcons of
the linguisc variables of the sex input is shown is (Figure 3).
a. Chest pain: This input has four Chest Pain types: Typical
Angina, Atypical Angina, NonAngina, and Asymptomac. One
Paent can have only one type of Chest Pain at a me. To rep-
resent Chest Pain, 1 = Typical Angina, 2 = Atypical Angina, 3 =
Non-Angina and 4 = Asymptomac. Hence, there is no fuzzi-
ness or overlap for this input, thus there are crisp set of values
because the paent paent has just one chest pain at a me
and the membership funcons of the linguisc variables of
the chest pain input is shown is (Figure 4).
a. Diabetes: This input has two instances; either the paent
is posive or negave. Posive = 1 and Negave = 0 Hence,
there is fuzziness or overlap for this input. The membership
funcons of the linguisc variables of the diabetes input is
shown is (Figure 5).
b. Result of Diagnosis (Output): The output consists of four
fuzzy sets and their linguisc variables are Healthy, Mild, Mod-
erate and Severe. Each Linguisc variable has membership
funcon associated with it. The membership funcons of the
linguisc variables of Diagnosis output is shown is (Figure 6).
Volume 2 | Issue 2 | 2022 5
jcmimagescasereports.org
is shown in (Figure 7).
Defuzzicaon
Defuzzicaon is the process of converng the output of infer-
ence engine (fuzzy values) into crisp values. As the name im-
plies, defuzzicaon is the opposite operaon of fuzzicaon.
Since in the rst procedure the crisp values of input variables
are fuzzied into degree of membership with respect to fuzzy
sets, the last procedure extracts a precise quanty out of the
range of fuzzy set to the output variable [34-35]. The Defuzzi-
er technique employed in this work is Centroid. Centroid de-
fuzzicaon returns the center of area under the curve [31].
Centroid Method (also called center of area or center of grav-
ity) which is the most prevalent and physically appealing of all
the defuzzicaon methods. It is adopted in this study. It is
given by the algebraic expression below:-
(2)
Where z is the output variable, and (z) is the membership
funcon of the aggregated fuzzy set A with respect to z. the .
Fuzzy Based Expert System for Ecient Diagnosis of Coronary
Artery Disease
Like any other Fuzzy Inference System, 4.6. Fuzzy Based Expert
System for Ecient Diagnosis of Coronary Artery Diseasehas
been implemented in MATLAB and it has ve primary graphi-
cal user interfaces (GUIs) can all interact and exchange in-
formaon to each other as shown in (Figure 14). Any of the
interfaces can read and write both to the workspace and to
the disk, (the read-only viewers can sll exchange plots with
the workspace and/or the disk). Like any fuzzy inference sys-
tem, any or all of these ve GUIs can be opened. (Figure 8)
shown GUI of the expert system and other GUIs of the system
include Membership Editor Viewer, Rule Editor Viewer, Rule
Viewer and Surface Viewer. GUI of Membership Editor Viewer
is shown in (Figure 9), GUI of Rule Editor Viewer is shown in
(Figure 10), GUI of Rule Viewer is shown in (Figure 11 and Fig-
ure 12) has shown the GUI of Surface Viewer.
Volume 2 | Issue 2 | 2022 6
jcmimagescasereports.org
Performance Evaluaon of the Fuzzy Based Expert System
There are four classes of linguiscs variable of the system out-
put which include Healthy, Mild, Moderate, and Severe. Both
Healthy, Mild paents are considered as normal paents and
both Moderate and Severe are considered as abnormal (CAD)
paents while evaluang the performance of the fuzzy expert
system model. For the performance evaluaon of the system
were considered based on the following techniques:-
i. True posive (TP): It denotes the number of abnor-
mal paents correctly classied by the framework.
ii. True negave (TN): It denotes the number of normal
paents correctly classied by the framework.
iii. False posive (FP): It denotes the number of normal
paents wrongly classied as abnormal paents by the frame-
work.
iv. False negave (FN): It denotes the nu+mber of ab-
normal paents wrongly classied as normal paents by the
framework.
v. Specicity: It is denes as percentage of normal pa-
ents classied correctly by the framework. It is determined
as
i. Sensivity: It is dened as the percentage of abnormal pa-
ents classied correctly by the model. It is determined as
ii. Accuracy: It denotes the percentage of correctly classied
paents. In the present work (4-class problem) it is deter-
mined as:-
Where h is the number of correctly classied rules as healthy,
i the number of correctly classied rules as mild, j the number
of correctly classied rules as moderate, j the number of cor-
rectly classied rules severe and N is the total number of rules
in the system knowledge base.
Class
Total num-
ber of rules
No. of rules Correctly
classied
No. of rules wrong-
ly classied
Healthy 44 38 4
Mild 48 46 2
Moderate 42 36 6
Severe 40 38 2
Grand total 174 158 14
Class (Normal
Paent)
Total number
of rules
No. of rules Correctly
classied
No. of rules
wrongly classied
Healthy 44 38 4
Mild 48 46 2
Grand total 92 84 6
Class (Abnormal
Paent)
Total number
of rules
No. of rules Cor-
rectly classied
No. of rules wrongly
classied
Moderate 42 36 6
Severe 40 38 2
Grand total 82 74 8
The knowledge base of the Fuzzy System has 174 rules with
four classes as shown in (Table 8). These rules were validated
against the human expert driven data collected Federal Teach-
ing Hospital, Gombe State. (Table 5) has shown the general
classes of paents. There were forty four rules classied as
healthy and forty eight rules classied as mild which were
considered for normal or healthy paents. So total number
of rules classied for healthy paent is ninety two as shown
in (Table 6). while for abnormal paents, forty two rules were
classied as moderate and forty rules as severe. So total num-
ber of rules classied for abnormal paents are forty-one as
shown in (Table 7).
Out of forty four rules classied as healthy, thirty eight rules
were correctly in conformity with the expert system, so only
four rules were not in conformity with the data. While Out
of forty eight rules classied as mild, forty six rules were cor-
rectly in conformity with the expert system, so only two rules
were not in conformity with the data. Therefore, the system
correctly classied ninety two normal rules and however only
six rules were wrongly classied as normal. While Out of the
forty two rules classied as moderate, thirty six rules were
correctly in conformity with data of the expert system, so only
six rules were not in conformity with the data. While Out of
forty rules classied as severe, thirty eight rules were correctly
in conformity with data of the expert system, so only two rules
were not in conformity with the data. Therefore, the system
correctly classied seventy four abnormal rules and however
eight rules were wrongly classied as abnormal. The valida-
on result of the expert system is as follows:-
True posive (TP) = 37
True negave (TN) = 84
False posive (FP) = 4
False negave (FN) = 6
Specicity =84/92 = 91.30%
Sensivity =37/41 = 90.24%
Accuracy =158/174 = 90.08%
Therefore, the system has achieved 90.08% overall accuracy
which is very excellent, thus the accuracy determines the pro-
poron of the total number of predicons that were correct.
At the same me, the system has 91.30%accuracy to classify
of normal paents correctly by adopng the proposed frame-
work (specicity) and 90.24% accuracy to classify abnormal
paents correctly by adopng the proposed framework (sen-
sivity). This showed that, the system performed ecient and
excellently to diagnose CAD.
Conclusion
Coronary Artery Disease (CAD) is one of the deadliest diseases
in the world and in Nigeria, the CAD is at the moment gaining
Volume 2 | Issue 2 | 2022 7
jcmimagescasereports.org
much popularity following the rising number of health issues
related to the disease, including higher death rate, which is
mostly due to lack of proper awareness among the common
people. Expert systems have been specically applied in a va-
riety of life sciences support systems development, ranging
from storage and retrieval of medical records, diagnoscs, up
to expert knowledge/decision support systems. In this work,
a fuzzy based expert system for supporng and compleng
human experse in the diagnosis of CAD has been developed
and evaluated. The system archived 90.08% accuracy, 91.30%
specicity and 90.24% sensivity respecvely, which showed
that, the system performed eciently and excellently to di-
agnose CAD and it can deployed and used in the hospitals in
Nigeria.
Acknowledgement
This work was supported by the Terary Educaon Trust Fund,
Nigeria (TETFUND), as an Instuon Based Research Fund
(IBR) for Federal University of Kashere, Gombe State, Nigeria.
References
1. A Saeed and K Asieh. “Idencaon and Classicaon of
Coronary Artery Disease Paents using Neuro-Fuzzy Inference
Systems”, Journal of mathemacs and computer Science.
2014; 13:136-141.
2. AA Haruna, LJ Muhammad, BZ Yahaya et al. “An Improved
C4.5 Data Mining Driven Algorithm for the Diagnosis of Coro-
nary Artery Disease”. Internaonal Conference on Digizaon
(ICD), Sharjah, United Arab Emirates. 2019; 48-52.
3. AV Senthil Kumar. “Diagnosis of heart disease using ad-
vanced fuzzy resoluon mechanism”, Journal of Arcial Intel-
ligence. 2013; 1-9.
4. AS Noor, PA Venkatachalam and FH Ahmad. “Diagnosis of
Coronary Artery Disease Using Arcial Intelligence Based
Decision Support System”, Proceedings of the Internaonal
Conference on Man-Machine Systems (ICoMMS). 2019; 11-
13.
5. A Abraham, “Rule-based Expert Systems” Handbook of
Measuring System Design, Wiley & Sons, Ltd, 2005.
6. A Ali and N Mehdi “A Fuzzy Expert System for Heart Disease
Diagnosis” Proceedings of the Internaonal Mul Conference
of Engineers and Computer Sciensts. 2010; 1:17-19.
7. American Heart Associaon (AHA). Heart disease and stroke
stascs -at a glance Retrieved from hps://www.heart.
org/idc/groups/ahamah-public/@wcm/@sop/@smd/docu-
ments/downloadable/ucm_480086.pdf Date: 16th January,
(2019).
8. CU Nwaneli, “Changing Trend in Coronary Heart Disease in
Nigeria”, Africa medical Journal. 2010; 1(1):1-4.
9. C Enrico. “The Guide to Health Informacs”, 3rd Edion.
London, Arnold, 2003.
10. D Resul, T Ibrahim and S Abdulkadir. “Eecve diagnosis
of heart disease through neural networks ensembles”, Expert
Systems with Applicaons. 2009; 36:7675-7680.
11. E Turban and E Jay. “Decision Support Systems and Expert
Systems”. 6th Edion, Prence Hall, Upper Saddle River, NJ,
2011.
12. FS Ishaq, LJ Muhammad, YZ Yahaya, et al. Fuzzy-Based Ex-
pert System for Diagnosis of Diabetes Mellitus. Internaonal
Journal of Advanced Science and Technology. 2020; 136:39-50
13. FS Ishaq, LJ Muhammad, BZ Yahaya, Y Atomsa. “Data min-
ing driven models for diagnosis of diabetes mellitus: a survey”,
Indian Journal of Science and Technology. 2018; 11:41.
14. H Nasruddin, RS Osama, MK Ahmed and AG Mohammed.
“Fuzzy So Expert System in Predicon of Coronary Artery Dis-
ease”, Internaonal Journal of Fuzzy Systems, 2016.
15. J Giarratano and G Riley. “Expert Systems: Principles and
Programming”, PWS-Kent Publishing Co, Boston, M.A, 1989.
16. KP Debabrata, AV Aruldoss, P Renukadevi, D Devaraj. “De-
sign of fuzzy expert system for microarray data classicaon
using a novel Genec Swarm Algorithm”, Expert Systems with
Applicaons. 2012; 39:1811-1821.
17. K Humar and A Novruz. “Design of a Hybrid System for the
Diabetes and Heart Diseases”, Expert Systems with Applica-
ons. 2008; 35:82-89.
18. K P Ganesh, A V Aruldoss, P Renukadevi and D Devaraj.
“Design of fuzzy expert system for microarray data classica-
on using a novel Genec Swarm Algorithm”, Expert Systems
with Applicaons. 2012; 39:1811-1821.
19. LN Rani, S Det, LJ Muhammad. “Determinaon of Stu-
dent Subjects in Higher Educaon Using Hybrid Data Mining
Method with the K-Means Algorithm and FP Growth”, Inter-
naonal Journal Of Arcial Intelligence Research. 2021; 5(1).
20. LJ Muhammad, EA Algehyne. “Fuzzy based expert system
for diagnosis of coronary artery disease in Nigeria”, Health
Technol. 2001; 11:319-329.
21. LJ Muhammad, EJ Garba, ND Oye, GM Wajiga, AB Garko.
“Fuzzy rule-driven data mining framework for knowledge ac-
quision for expert system”. In: Translaonal Bioinformacs
in Healthcare and Medicine. Elsevier, Academic Press. 2021;
201-214.
22. LJ Muhammad, AH Ahmad, AM Ibrahim, A Mansir, B Ba-
ture, MA Jamila. “Performance Evaluaon of Classicaon
Data Mining Algorithms On Coronary Artery Disease Dataset”,
IEEE 9th Internaonal Conference on Computer and Knowl-
edge Engineering (ICCKE 2019), Ferdowsi University of Mash-
had, October, 2019.
23. LJ Muhammad et al. “Deep Learning Models for Predicng
COVID-19 Using Chest X-Ray Images”. In Trends and Advance-
ments of Image Processing and Its Applicaons. EAI/Spring-
er Innovaons in Communicaon and Compung. Springer,
Cham, 2022.
24. LJ Muhammad, EA Algehyne, SS Usman. “Predicve Super-
vised Machine Learning Models for Diabetes Mellitus”. Spring-
er Nature Computer Science, 2020.
25. LJ Muhammad, M Besiru Jibrin, BZ Yahaya, IA Moham-
med Besiru Jibrin, A Ahmad and JM Amshi, “An Improved C4.5
Algorithm using Principle of Equivalent of Innitesimal and
Arithmec Mean Best Selecon Aribute for Large Dataset,”
2020 10th Internaonal Conference on Computer and Knowl-
Volume 2 | Issue 2 | 2022 8
jcmimagescasereports.org
edge Engineering (ICCKE), Mashhad, Iran, 2020; 006-010.
26. LJ Muhammad, EJ Garba, ND Oye, and GM Wajiga. “On the
Problems of Knowledge Acquision and Representaon of Ex-
pert System for Diagnosis of Coronary Artery Disease (CAD)”,
Internaonal Journal of u- and e- Service, Science and Tech-
nology. 2018; 11(30):49-58.
27. LJ Muhammad, I Al-Shourbaji, AA Haruna et al. “Machine
Learning Predicve Models for Coronary Artery Disease”. SN
COMPUT. SCI. 2021; 2:350.
28. LJ Muhammad et al. Using Decision Tree Data Mining Al-
gorithm to Predict Causes of Road Trac Accidents, its Prone
Locaons and Time along Kano –Wudil Highway. Internaonal
Journal of Database Theory and Applicaon, 2017; 10:197-
208.
29. LJ Muhammad, SS Usman. Power of Arcial Intelligence
to Diagnose and Prevent Further COVID-19 Outbreak: A Short
Communicaon 2020.
30. LJ Muhammad, MM Islam, Usman SS. et al. Predicve Data
Mining Models for Novel Coronavirus (COVID-19) Infected Pa-
ents’ Recovery. Springer Nature Computer Science, 2020.
31. LJ Muhammad, BZ Yahaya, A Garba. et al. Mul Query Op-
mizaon Algorithm Using Semanc and Heurisc Approach-
es, Internaonal Journal of Database Theory and Applicaon,
2016; 6(9).
32. L Adel, N Raja, Z Rozia and B Awang. “Design of a Fuzzy-
based Decision Support System for Coronary Heart Disease
Diagnosis”, Journal of Medical System, 2012.
33. M Zahra and SA Mohammad. “CADICA: Diagnosis of Coro-
nary Artery Disease Using the Imperialist Compeve Algo-
rithm”, Journal of Compung Science and Engineering, 2014;
8(2):87-93.
34. MG, Tsipouras, TP Exarchos, DI Foadis, A Kotsia, A Naka
and LK Michalis, “A Decision Support System for the Diagnosis
of Coronary Artery Disease”, Proceedings of the IEEE Sympo-
sium on Computer-Based Medical Systems, 2008.
35. M Islam, S Mahmud, LJ Muhammad. et al. Wearable Tech-
nology to Assist the Paents Infected with Novel Coronavirus
(COVID-19). SN COMPUT. SCI. 2020; 1:320.
36. M Idris, A Pervez, JA Tariq and SZ Syed. “Fuzzy Rule Based
Classicaon for Heart Dataset using Fuzzy Decision Tree Al-
gorithm based on Fuzzy RDBMS” World Applied Sciences Jour-
nal. 2013; 28(9):1331-1335.
37. ND Wong. “Epidemiological studies of CHD and the evo-
luon of prevenve cardiology.”. Nature reviews. Cardiology.
2014; 11(5):276-89.
38. N Jesmin and I Tasadduq. “Associaon rule mining to de-
tect factors which contribute to heart disease in males and
females”, Journal of Expert Systems with Applicaons. 2013;
40:1086–1093.
39. OC Akinyokun, GB Iwasokun, SA Arekete and RW. Samuel
Fuzzy logic-drive expert system for the diagnosis of heart fail-
ure disease. Arcial Intelligence Research, 2015; 4(1).
40. OA Sarumi, O Aouedi, LJ Muhammad. “Potenal of Deep
Learning Algorithms in Migang the Spread of COVID-19”. In
Understanding COVID-19: The Role of Computaonal Intelli-
gence. Studies in Computaonal Intelligence, 2021; 963.
41. O Alsayed, MSM Rahim, IA Bidewi, et al. Selecon of the
Right Undergraduate Major by Students Using Supervised
Learning Techniques. Appl, 2021.
42. PK Anooj. “Clinical decision support system: Risk level pre-
dicon of heart disease using weighted fuzzy rules, Journal of
Computer and Informaon Sciences”, 2012; 24:27-40.
43. P Debabrata, KM Mandana, P Sarbajit, S Debranjan and
C Chandan. “Fuzzy expert system approach for coronary ar-
tery disease screening using clinical parameters”, Journal of
Knowledge-Based Systems, 2012; 36:162-174.
44. RA Soltan, MZ Rashad and B El-Desouky. “Diagnosis of
Some Diseases in Medicine via computerized Experts”, Inter-
naonal Journal of Computer Science & Informaon Technol-
ogy (IJCSIT), 2013; 5(5).
45. SD Chaitrali and SA Sulabha,“Improved Study of Heart Dis-
ease Predicon System using Data Mining Classicaon Tech-
niques, Internaonal Journal of Computer Applicaons, 2012;
47(7):0975 -888 .
46. TA Ibrahim, AS Adejoke, A Moses, CA Tony, BZ Ibrahim,
and HW Muhammad. “ICT Knowledge, Ulizaon and Percep-
on among Healthcare Providers at Naonal Hospital Abuja,
Nigeria”, American Journal of Health Research. Special Issue:
Health Informaon Technology in Developing Naons: Chal-
lenges and Prospects Health Informaon Technology. 2015;
3(1):47-53.
47. WM Wiharto, K Hari, and H Herianto. “Intelligence System
for Diagnosis Level of Coronary Heart Disease with K-Star Al-
gorithm”. Healthcare Informacs Research, 2016; 22(1):30-38.
48. World Health Organizaon (WHO). (2015). Cardiovascular
diseases, factsheet#317.Retrieved from hp://www.who.int/
mediacentre/factsheets/fs317/en/ Date: 16th January, 2019.
49. YD Niranjana and S Anto. “An Evoluonary-Fuzzy Expert
System for the Diagnosis of Coronary Artery Disease”, Interna-
onal Journal of Advanced Research in Computer Engineering
& Technology, 2014; 3(4).
50. Y Atomsa, LJ Muhammad, FS Ishaq, et al. “Expert system
for diagnosis of coronary artery disease: A survey” Journal of
Clinical Images and Medical Case Reports, 2021; 2.
51. Z Yahaya, LJ Muhammad, N Abdulganiyyu, FS Ishaq. “An
improved C4. 5 algorithm using L’Hospital rule for large data-
set”, Indian Journal of Science and Technology, 201.
Volume 2 | Issue 2 | 2022 9
jcmimagescasereports.org
