Expert Systems With Applications 217 (2023) 119561

Contents lists available at ScienceDirect

Expert Systems With Applications

journal homepage: www.elsevier.com/locate/eswa

Review

An overview on state-of-the-art electrocardiogram signal processing

methods: Traditional to AI-based approaches
Venkata Anuhya Ardeti a, *, Venkata Ratnam Kolluru a, George Tom Varghese b,
Rajesh Kumar Patjoshi c
a
Department of Electronics and Computer Engineering, Koneru Laskshmaiah Education Foundation, Guntur, India
b
Department of Electronics and Instrumentation Engineering, St. Joseph’s College of Engineering and Technology, Palai, Kottayam, Kerala, India
c
Department of Electronics and Communication Engineering, National Institute of Science & Technology, Berhampur, Odisha, India

A R T I C L E I N F O A B S T R A C T

Keywords: Over the last decade, cardiovascular diseases (CVD’s) are the leading cause of death globally. Early prediction of
Biomedical processors CVD’s can help in reducing the complications of high-risk patients. The electrocardiogram (ECG) is an efficient
Classification aiding tool, provides accurate information about various cardiac conditions of the human heart. Evaluation and
Deep learning
interpretation of ECG signal has become the major goal in current research to identify and mitigate risky car­
Feature engineering
Internet of Things
diovascular conditions. The ECG signal is efficiently analysed and classified using signal processing techniques,
Machine learning ranging from traditional to machine learning approaches and its subbranches, such as deep learning are used for
Wearable devices the early detection and diagnosis of cardiac conditions and arrhythmias. The development of novel types of body
sensors increases the need for automated, low-cost, real-time, and efficient ECG monitoring systems that can be
used at home or in ambulatory settings. IoT in healthcare industry enables the doctors and specialists to diagnose
the patient status remotely in a smart and efficient manner. Motivated with all the research done in this area for
early diagnosis of cardiac abnormalities and save the life of high-risk patients, we surveyed numerous research
works reported in the literature and provide a glance on several aspects of ECG signal monitoring systems. We
first introduce the step-by-step framework of ECG signal analysis starting data acquisition to classification, de­
scribes each stage analysis of both traditional and advanced machine learning models that have been reported in
the literature. We provide a deep discussion on ECG signal acquisition, pre-recorded clinical ECG data taken from
the databases, signal processing and denoising, detection of feature based on feature engineering, and signal
classification along with the comparative assessment amongst reviewed studies. This work also presents the
detailed analysis of smart health care system that employs portable and wearable devices, as well as the Internet
of Things (IoT) and wireless technologies, as an innovative medium for data transmission that allows for remote
access to a patient’s health status at lower costs and with greater efficiency. Moreover, this work provides an
extensive knowledge on various hardware platforms that have been adopted for the development of biomedical
processors including microcontrollers, FPGA and ASIC. Additionally, the challenges and limitations are discussed
in this research field and directions for future work are suggested. This is the first review paper of its kind to
present a comprehensive discussion on all aspects of cardiovascular disease monitoring systems in one place.

1. Introduction of top 10 causes of the death globally given by WHO on 9th December
2020. Among 55.4 million deaths worldwide in 2019, the greater part
Recently, World Health Organization (WHO) reported that, larger (55 %) was because of these top 10 causes. Coronary heart disease and
numbers of deaths occur due to coronary heart disease and stroke for the stroke alone represents a consolidated of 15.2 million deaths in 2019.
year 2019. Cardiovascular diseases (CVD’s) arise due to inability of However, these top 10 causes are classified into 3 categories i.e., In­
heart to receive blood from the coronary arteries due to blood shortage juries, Communicable and Non-communicable. The first category, In­
(Wasimuddin and Elleithy, 2020). Fig. 1 illustrates the statistical report juries claims about 2.5 % of total deaths occur due to road accidents. The

* Corresponding author.
E-mail address: [email protected] (V.A. Ardeti).

https://doi.org/10.1016/j.eswa.2023.119561
Received 2 May 2022; Received in revised form 13 December 2022; Accepted 13 January 2023
Available online 20 January 2023
0957-4174/© 2023 Elsevier Ltd. All rights reserved.
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

second one is Communicable diseases take up about 10 % deaths glob­ Portable and wearable sensor-based systems (Soh et al., 2015) use
ally due to nutritional deficiencies and maternal causes. The third one, miniaturized bio-sensing devices, microelectronics and wireless
Non-Communicable diseases occur due to poor blood and air flow inside communication modules which provides an unobtrusive solution for the
the body and claims about 43.5 % of the deaths globally. AI investigating of remote patients by means of low cost. Recent de­
The ireegular or abnormal heartbeats is often referred as arrhythmia. velopments in electronics and communication technologies have
Cardiac arrest and sudden fall happen depend on the severiety of spawned an era of Internet of Things technology (IoT) (Haroon et al.,
arrhythmia. Timely diagnosis and accurate analysis of arrhythmia can 2016). IoT in medical sector drive the researchers globally to layout
save many lives worldwide. This prompted us to conduct a compre­ promising frameworks that could offer smooth scientific help to every-
hensive assessment of ECG. An electrocardiograph is a machine that one. It enables the physical-world elements to communicate both
records the electrical movements of atria and ventricles during physically and wirelessly (Ray, 2018). This paper presents a detailed
contraction and relaxation of heart, which propagate as pulsating elec­ survey on different aspects of ECG analysis such as data acquisition,
trical waves towards the skin. The recorded signal can be determined feature engineering, and classification from conventional to advanced
with the help of electrodes and proper apparatus. The change in the machine learning techniques, smart health monitoring devices, and real
electrical potential among the electrodes is commonly improved with time implementations with embedded architectures. Even though, there
the help of an operational amplifier. The signal is then denoised to are many survey articles in the study, they concentrate only on few as­
extract important information from the signal. This graphical depiction pects. Consequently, this state-of-the art survey provides the desired
is represented in the form of ECG. The typical ECG consists of a peri­ information about most of the aspects of ECG analysis and serves as one
odical pattern of P-QRS-T waves which represents different aspects best source for the researchers working in this field.
involved in the conduction of electrical signals inside the heart. A ‘P-
QRS-T-U’ wave corresponds to one cardiac cycle (one heartbeat) (Kar­ 1.1. Our contributions
pagachelvi et al., 2010). There is also a U wave which can be seen oc­
casionally. The periodical pattern of ECG signal shown in Fig. 2 is drawn This study contributes to the emerging field of research for diagnosis
by taking (Anuhya et al., 2020) as reference. Applications of ECG in­ of cardiac conditions and various arrhythmias by monitoring ECG signal
cludes cardiac beat rhythm testing, heart rate measurement, diagnosis of in real-time and exploring tele-health solutions and methodologies to
heart flaws, biometric identification (Merone et al., 2017) and prevent these abnormalities. Many survey articles have been reported in
emotional recognition. the literature ranging from real-time ECG signal analysis to remote
The analysis of heartbeat patterns, temporal and morphological monitoring applications, however, they are limited to certain aspects.
characteristics provide scientific knowledge to automatically recognize For instance, (Chi et al., 2010) discussed about dry-contact and non-
an ECG pattern. However, there is a difficulty with the automated contact electrodes used to record ECG signal for analysis. (Berkaya
classification of ECG signal, because the temporal and morphological et al., 2018; Wasimuddin and Elleithy, 2020; Merdjanovska et al., 2022)
characteristics may include noticeable differences for different in­ presents a survey on ECG signal analysis and methods. (Karpagachelvi
dividuals under distinct environmental assets (Banerjee et al., 2014). et al., 2010; Dewangan, 2017) concentrated only on feature extraction
The major drawback in identifying the coronary heart infections is, the techniques. (Taha et al., 2016) reviewed various classification methods
recorded ECG may vary from one individual to the other, and in some of ECG signals. (Vincent and Sreekumar, 2017) discussed both feature
cases distinct patients have diverse ECG morphologies for a similar extraction and classification techniques of arrhythmia analysis. (Serhani
sickness (Dilmac et al., 2015). In addition, two distinct diseases might et al., 2020) provides a comprehensive review on architecture, pro­
have relatively the same characteristics as the ECG signal (Mateo et al., cesses, and key challenges of ECG monitoring systems. (Pardesh et al.,
2016). Therefore, the electrical signal of individual heartbeat must be 2017; Chhabra et al., 2017) described various real-time IoT based health
analysed and classified to detect anomalies. monitoring systems. (Rizwan et al., 2020) comprehensively discusses
The process for the interpretation of long-term recorded ECG signals, traditional and machine learning models for automatic diagnosis of
particularly for bedside and Holter monitoring, might be very prob­ atrial fibrillation (AF) reported in the literature. In addition to previous
lematic for an individual because it takes lot much time to compute studies, the highlights of this work are discussed as follows:
(Elizabeth et al., 2017). The people residing in pastoral areas with low
income are facing the problem with increased cost of healthcare. From • Present an extensive overview on different stages of electrocardio­
the last decade, development of wearable (Personalized) monitoring gram analysis.
structures had drawn more attention to reduce the cost of healthcare.

Top 10 Causes of death, 2019

Coronary heart disease 9
Injuries
Stroke 6.2
Chronic obstructive… 3.2
Lower Respiratory Infections 3 Communicable,
Diseases

Lung cancers, Trachea,… Matemal, neonatal and

1.8 nutritional conditions
Alzheimer disease & other… 1.7
Non-Communicable
Diabetes mellitus 1.6 Diseases
Diarrhoeal diseases 1.6
Tuberculosis 1.3
Road Injury 1.2

0 5 10
Death Rate (in millions)

Fig. 1. World Health Organization Statistical Report.

2
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Fig. 2. Periodical pattern of ECG.

• Summarize the most notable developments of smart health moni­ implementation of health monitoring systems in real time. Section 7
toring systems. confers about the challenges and limitations of smart health care.
• Discussed the latest hardware implementations of real-time ECG Finally, section 8 concludes this paper and discusses the future di­
monitoring systems. rections of this research.
• Identifies challenges and limitations for various automatic ECG
diagnosis systems. 2. Technologies, algorithms and tools
• Outlines the future vision of next generation ECG monitoring systems
for healthcare. This section bullet out various technologies, algorithms and tools
used for the diagnosis of cardiovascular diseases.
The rest of the paper is organized as follows: Section 2 bullet out the
technologies, algorithms and tools used in this field of research. Section 2.1. Technologies
3 introduces the fundamental concepts of ECG electrode system and
medical devices available in the market to monitor ECG signal. Section 4 ECG technology has been used by doctors for years to gather vital
comprehensively surveys the state-of-art; preprocessing, feature engi­ information on the health of their patients. Technologies in medical
neering, feature optimization & classification techniques from tradi­ science can be applicable in both medical and consumer world (Neu­
tional to deep learning approaches reported in the literature. Section 5 rosky, 2015) as shown in Fig. 3. In medical applications, ECG technology
elaborates various prototypes developed for wearable health monitoring is typically divided into two categories. ECG for diagnostic purposes and
systems published in the literature. Section 6 provides a glance on ECG for monitoring purposes.
various embedded boards and biomedical processors designed for the

Fig. 3. Technologies related to medical science.

3
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

• When using ECG technology for diagnostic purposes, multiple leads used by all the categories of monitoring systems to transport the ECG
(usually 8–12 leads) of data are acquired via electrodes that are signals from sensors to the underlying servers and/or devices under
attached to left/right arms and legs, chest, etc. of a patient. Once which the ECG is processed and visualized.
collected, the doctor analyses the data and give the valuable infor­
mation related to the functioning of the heart and whether any ab­ 2.2. Algorithms
normalities have been observed. If there are any abnormalities, the
doctor can prescribe more tests and suggest appropriate treatment. There are numerous algorithms to diagnose cardiovascular diseases.
• ECG technology for monitoring purposes typically uses a single These algorithms can be traditional rule-based, advanced machine
channel connected to a chest strap or left/right arm electrodes. This learning and deep learning algorithms.
kind of ECG technology is used for monitoring heart rate and basic
functioning of the heart during surgeries or during hospitalization, in 1. Traditional Algorithms: These algorithms are simple rule-based tech­
emergency rooms or while transporting the patient in an ambulance niques mainly used in the preprocessing, feature extraction and
to provide alerts if any abnormal changes are detected in vital signs. classification stages in ECG diagnosis. Researchers use various filters
• Most of the medical ECG equipment uses special gel-based sensor such as the low pass (Castroflorio et al., 2013, Zhang et al., 2014,
electrodes and requires a trained technician to be correctly place on Chui et al., 2016), high pass (Nejadgholi et al., 2011, Chui et al.,
the body. ECG equipment for medical use generally requires regu­ 2016), band pass filter (Afkhami et al., 2016), adaptive filters
latory compliance. (Nuthalapati et al., 2019), notch filter (Li et al., 2016), and other
• In the consumer world, ECG technology is being used for health and filter-based approaches (Salman et al., 2018, Sulthana et al., 2018)
wellness applications by being incorporated into wearable and mo­ have been adopted by researchers in the recent studies to denoise the
bile technology. In these cases, a single channel ECG signal is ac­ ECG signal during preprocessing stage. Several finite impulse
quired by dry contact metal electrodes touching the skin, usually on response (FIR) (Carlos et al., 2019), derivative (Sadhukhan et al.,
the fingers or the wrist. Consumer ECG equipment can be used 2012), windowing (Oktivasari et al., 2019) and transformation-
without any specific training, and algorithms analyse the data to based algorithms (Fujita et al., 2015) are utilized to extract the
derive useful information for the wearer. This information can temporal, morphological features during feature extraction stage.
include heart rate, heart rate variability, stress level, respiration rate, Similarly various thresholding-based algorithms (Kumar et al., 2016)
and more. In consumer applications, the data obtained is often are used at classification stage for the classification of ECG signals.
combined with mobile apps to assist with long-term monitoring and However, traditional diagnosis processes are time-consuming,
analysis. Additionally, these apps can display information through expensive, and often require human intervention.
valuable health and wellness metrics that is easy to read, understand, 2. Machine Learning Algorithms: ML algorithms are typically divided into
and interpret. three groups, including supervised, unsupervised, and semi-
supervised algorithms. However, there are various subgroups of
(Serhani et al., 2020) comprehensively explained distinct ECG ML algorithms based on various learning strategies. Some of the
monitoring technologies, shown in Fig. 4 is divided into 5 groups which popular ML algorithms include support vector machines (SVM)
is considered for future generation ECG monitoring systems. These are (Venkatesan et al., 2018, Jha et al., 2020), decision tree (Exarchos
classified based on the monitoring context for which they are developed, et al., 2007), random forest (RF) (Li et al., 2016, Anuhya et al.,
the involved ECG technologies and devices that characterize these sys­ 2022), and naive Bayes (NB) (Jen et al., 2008), K-nearest neighbor
tems, the monitoring scheme and frequency adopted, the monitoring (KNN) (Goovaerts et al., 2019), and other learning algorithms
target and purpose for which these systems were implemented, and (Hadjem et al., 2016; Xiao et al., 2018) are used for feature extrac­
finally, the futuristic monitoring system that leverages new technolo­ tion and classification are ECG signals.
gies, such as AI, robotics, and nanotechnology, to shape emerging ECG 3. Deep Learning Algorithms: Deep learning (DL) is a subfield of machine
monitoring systems. In the bottom, the horizontal underneath level learning (ML) that employs multiple layers to extract both higher and
represents the communication protocol and the wireless technology lower-level information from input (i.e., images, numerical value,

Fig. 4. Classification of ECG Monitoring Technologies.

4
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

categorical values). Deep learning may be classified into three types: their detection and classification algorithms. Many libraries are avail­
supervised, semi-supervised, and unsupervised. In the recent years, able in computer-based programs like Matlab, Python, and Labview that
several deep learning models have been proposed to improve the can be used to evaluate algorithms. These tools offer ways to import pre-
accuracy of different learning tasks during feature engineering and recorded ECG signals from publicly accessible databases. However, the
classification which includes Artificial Neural Network (ANN) “R” tool itself can be used to examine datasets. On the other hand, there
(Maedeh et al., 2012), Convolutional Neural Network (CNN) (Sell­ exist emulation boards also known as open-source hardware such as
ami et al., 2019, Kamaleswaran et al., 2018), Recurrent Neural Arduino Mega 2560, the Duino Olimexino-5510, the TI MSP430-T5510,
Network (RNN) (Mostayed et al., 2018), Long ShortTerm Memory and many other tools that can be used for testing and analysis. For
(LSTM) (Chang et al., 2020), and Deep Belief Network (DBN) example, Arduino IDE and MSPSim. AD8232 ECG sensor use Arduino
(Mathews et al., 2018). emulator for signal acquisition and analysis. Another System on Module
(SOM) boards such as RK3188 and AM335X with ARM Cortex Quad-
2.3. Diagnostic and research tools Core processor on-board to provide embedded processing of algo­
rithms along with System on Chip (SOC) boards such as NXP Nexperia-
2.3.1. Diagnostic tools 8550. On open-source hardware platform and SOM boards with ARM
There are various tools used to diagnose heart arrhythmias from bed Cortex processors, android offers a platform for application develop­
side to wearable and portable devices. These include, ment. AliveCor, Fluke ProSim 8 ECG Patient Simulator, TriSmed
TSM3000B, and many other portable simulators are also available, and
1. Electrocardiogram (ECG): The electrical activity of the heart can be they can be used to gather ECG and carry out various tests in real-time.
detected by using electrodes attached to the chest or sometimes to
the arms or legs. 3. ECG electrode system
2. Echocardiogram: A non-invasive involves a hand-held instrument
(transducer) applied to the chest to obtain images of the heart’s size, The first human ECG was determined, and the electrical activity of
composition, and motion. the heart was recorded by A.D. Waller in 1887 (Besterman et al., 1979).
3. Holter Monitor: A portable ECG device that monitor the electrical However, the ability to diagnose the normal rhythm was difficult for
activity of the heart for a day to make diagnosis. conventional medical check-ups up to 1960′ s. In advance there may
4. Event recorder: A wearable ECG device can be worn for longer period have several methods to compute ECG. H. P. Da et al, in his literature
(usually 30 days) to detect sporadic arrhythmias. classified the ECG measuring approaches into in-the person, off-the
5. Stress test: Exercise can cause or make some arrhythmias worse. The person and on-the person (Silva et al., 2015). Equipment’s utilized for
heart’s activity is tracked as you walk on a treadmill or ride a sta­ in-the-person class were implantable apparatus that are located within
tionary bike during a stress test. the body in the form of pills. Contrasting to the former one, devices used
6. Tilt table test: Monitors the heart rate and blood pressure as you lie for the off-the-person class were intended to calculate ECG with no or
flat on the table. The doctor observes the heart and nervous system less contact of the skin. One example of this type is the capacitive based
that respond to the change in the angle of the table while tilting. (Chi et al., 2010) devices measure the change in the electric potential
stimulated by the body permits the signal estimation at a separation of
Traditional ambulatory Holter monitors (HM) are still widely used, approximately 2 cm. In most cases, the devices under on-the-person
but they have some limitations in varying times that can often make class are used and normally require the usage of few electrodes
them inconvenient and make it difficult to analyse the results (Hamad attached to the human body for physiological monitoring helps in
et al., 2021). In in recent years, technological advancements in ambu­ reducing the use of laboratory or hospitals. There may be other sta­
latory ECG revolutionized consumer world applications to develop tionary medical devices available in hospitals for monitoring of ECG.
mobile health (mHealth) devices, specifically, wearable technologies Bed based and holter monitors are the examples of stationary medical
have been considered as an innovative solution for replacing invasive, devices.
less efficient methods for monitoring cardiac arrhythmias. Biosensors On-body sensors require some electrodes to acquire the signal, of
and sophisticated artificial intelligence (AI) algorithms are now incor­ them one will acts as a guide for the remaining. Generally, that guided
porated in new iterations of smartwatch and smartphone technologies, electrode is located adjacent to right leg. Depending on the pair of
making them capable of diagnosing cardiac arrhythmias. The spectrum electrodes selected and placed on the skin, different patterns of signals
of these devices ranges from smart accessories to sensors embedded into are obtained and are indicated with the name of leads. The usual
clothing and shoes (Xintarakou et al., 2022). In specifically, patches are placement of electrodes depends on Einthoven’s triangle (Neha et al.,
leadless, wearable electronics applied to the patient’s chest that allow 2021). The standard bedside monitoring system is a twelve lead ECG
for ambulatory ECG monitoring over a period of days to weeks. The test, which is a powerful analytical tool greatly, used in hospitals to
wrist-worn devices such as fitness bands and smartwatches may track identify the cardiac patients upon sleeping on the bed.
heart rate in real time (). For example, the recent introduction of Fitbit A 12-lead ECG system uses 3 limb leads (I, II & III), 3 augmented
(Lubitz et al., 2021), Apple (Seshadri et al., 2020), and Samsung (Avram limb leads (aVR, aVL and aVF) and 6 pre-cordial (chest) leads (V or V1,
et al., 2021) watches to the health-conscious consumer has resulted in an V2, V3, V4, V5 and V6) calculated using 10 electrodes (Spano et al.,
increase in the sales of wearable technology. Physiologists are increas­ 2016). Of these, the leads II and V are mostly used for heartbeat and
ingly using single-lead ECG monitors, such as, ZioPatch (iRhythm arrhythmia classification. The techniques which use both these leads
technology) (Zhang et al., 2018) and a compact ECG device, Kardia provide the good results of data. But the limitations of this system
Mobile (Alivecor) (Ding et al., 2020) to identify paroxysmal rhythms. include criticality for doctors to make the right decision when diag­
The FDA has approved these current inventions, which are not only nosing the patient, long time usage of gel-based electrodes may cause
practical for patients to utilize but also serve as medical equipment to skin irritation, bulky, cost effective and the system immobility. To
diagnose cardiac diseases. Also, several accessories integrate heart rate overcome the limitation of immobility, a holter monitoring system was
sensors for real-time diagnosis such as, chest straps (Navalta et al., invented by Jeff holter in 1962. It is a portable ECG monitor that can be
2020), rings (Kinnunen et al., 2018), glasses (Constant et al., 2015) and worn by a patient for at least 24 to 48 h for continuously monitoring
even into textiles (Pagola et al., 2018). heart rhythm (Medline Plus, 2020). The major limitation of this method
is, at the end of the monitoring period the device must be returned to the
2.3.2. Research tools doctor to retrieve and analyse the information. For convenience, a
Researchers employ PC-based software to train, test, and evaluate wearable planar-fashionable circuit board-based shirt was designed to

5
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

remove the usage of wires from the conventional holter monitoring database records, data collection and its applications. In some instances,
system. Technological advances develop a 5-lead ECG configuration single-lead recordings can be as effective than 12-lead recordings.
composed of 5 electrodes, placed at RA (Right Atria), LA (Left Atria), RL Studies have shown that most researchers use MITDB database exten­
(Right Leg), and LL (Left Leg) and at chest (Moses et al., 2015) are widely sively for their research evaluation.
used for the diagnosis. The similar performance with lesser computa­
tional cost can also be achieved by reducing the usage of number of 4.2. Preprocessing
electrodes (Tomasic et al., 2017) compared to the conventional 12-lead
system using multiple leads. Although several acquisition devices The electrical impulses recorded from the human body have very less
available in the market are used to estimate the ECG, the accuracy and amplitude and are affected with various noises and artifacts. The main
performance of the device depends upon the processing, feature engi­ aim of preprocessing is to reduce such noises; artifacts or any offset ef­
neering and classification methods that we incorporate to acquire the fects present in a signal and determines the feature. Various forms of
signal. noises are classified as follows:

4. Stage-based ECG signal analysis • Power Line Noise/Interference: An interference with a frequency of
50/60 Hz and a bandwidth of less than 1 Hz is caused due to the
A complete automatic approach for detecting arrhythmias from ECG radiation of high voltage power lines and improper grounding of
signals can be classified into five stages shown in Fig. 5 (Wasimuddin electronic machinery etc.
et al., 2010), includes 1) data acquisition, 2) preprocessing, 3) feature • Base line wandering: Base line wander is a noise signal having a very
engineering, 4) feature optimization and 5) feature classification. The less frequency of (0.15–0.3) Hz arises because of offset voltages in
first stage describes the acquisition of ECG data from the different the electrode–skin impedance, breathing and body movements of an
sources such as, clinical database, wearable devices and can be experi­ individual.
mental. In the second step, we discuss various filtering and smoothing • Instrumental noise: It is the noise associated with each component of
techniques published in literature to eliminate the noise and normalize an instrument used to measure the ECG.
the ECG signal. The proper classification of various heart diseases de­ • Electrosurgical noise: the noise that results from other electronic
pends on the detection of fiducial points of an ECG signal. Identification equipment’s utilized in the patient surroundings at a frequency be­
and extraction of fiducial features are done at stage three. In stage four, tween (100 K-1 M) Hz.
repeated and redundant features are to be removed to enhance the • Electrode contact noise: noise arises because of failure of contact
classification accuracy of the model. The type or class of each heartbeat between skin and electrode.
can be classified from the set of optimized features. • Muscular noise: also called electromyography noise, occurs when the
other muscles aside to the heart gets contracted.
4.1. Data acquisition • Electrode motion or movement artifacts: It is the noise arises because
of impedance variations between the electrode and skin causing
ECG data can be acquired using 12-lead or to single lead electrode potential variations in the ECG due to movements.
system by placing the electrodes that are specifically designed to place at
specific points of the human body and records the heart electrical ac­ The impact of these noises on analysis of ECG may cause the physi­
tivity. Different governmental, academic, and private organisations cian to take wrong decisions about patient condition and recommends
recorded the ECG data and provide it as an accessible database. These an incorrect treatment. Therefore, it is necessary to keep the signal away
databases contain the attributes such as number of leads used for from these artifacts and noises. Signal denoising and preprocessing plays
recording, lead number, type, annotations, age, gender, and health a vital role to eliminate such noises and improves the accuracy of the
condition of the patient. These attributes help in extracting different signal.
temporal and morphological characteristics of the ECG signal for anal­ (S. Celin et.al., 2018) demonstrated that signal can be taken as a 1D,
ysis and classification. This section examines various publicly available 2D, 3D and so on up to MD. But 2 D, 3D and higher degrees require many
ECG data sources, and its specifications are tabulated in Table 1. The variables and become more complex to compute mathematically. Hence,
scales of the database’s are reported in terms of record length, lead usage 1D (1 dimensional) signals are preferred to process and analyse the data.
of each record, voltage, sampling frequency, number of classes that Preprocessing stage involves filtering, re-sampling, digitization,

Fig. 5. Physiological signal recognition system.

6
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 1
Publicly available ECG databases and its specifications.
Database [Ref] No. of No. of No. of Voltage Sampling Record Data collection Application
Leads Classes Records Frequency (fs) in Length
Hz

MIT-BIH Arrhythmia (Moody 2 (MLII 5 48 5 µV 360 30 min Holter machine Arrhythmia detection &
et al., 2001) &V1/2/4/ Signal denoising
5)
MIT-BIH Normal Sinus Rhythm ( 2 NM 18 NM 360 24hr Ambulatory ECG Biometric Identification
Mousavi et al., 2019) records
MIT-BIH Atrial Fibrillation (Li 2 4 25 10mv 250 10hr Ambulatory ECG Biometric identification/
et al., 2020) records
MIT-BIH Supraventricular 2 NM 78 NM 360 30 min Ambulatory ECG Arrhythmia detection
Arrhythmia (Elgendi et al., records
2013)
LTSTDB (Jager et al., 1998) 3 NM 86 20 mV 250 24hr Holter machine ST segment detection
PTBDB (Bousseljot et al., 1995) 12 7 549 16 mV 1K 512 sec NM MI detection
AHADB NM 8 80 10 mV 500 5 sec TeleMedCare Ventricular Arrhythmia
monitor detection
MITNSTDB (Moody et al., 1984) 2 NM 2 5 µV 360 30 min Taken from MIT- Noisy signal analysis
BIH
ESCDB (Taddei et al., 1992) 2 NM 48 5 µV 250 30 min Holter machine Ischemia, T-wave change
& ST-segment detection
QTDB (Laguna et al., 1997) 2 NM 105 5 µV 250 15 min Holter machine ECG Delineation
CSEDB (Smisek et al., 2017) 15 NM 125 1 µV 500 10 sec NM ECG diagnostics
ChallengeDB (Clifford et al., 1 4 3658 NM 300 9–61 sec AliveCor Atrial Fibrillation
2017) detection
INCARTDB (Goldberger et al., 12 9 75 NM 257 30 min Holter machine NM
2001)
TWADB (Moody et al., 2008) 2–12 NM 100 5 µV 500 2 min NM T-wave
CPSCDB (Liu et al., 2018) 12 9 6877 NM 500 6–60 sec NM NM

NM = Not Mentioned.

normalization, and artifacts removal for efficient assessment of a feature artifacts, base line wander and muscle noise are (0.05–40) Hz (Chua
is illustrated in Fig. 6. et al., 2011), (0.1–100) Hz (Afkhami et al., 2016; Ubeyli et al., 2009),
The main aim of filtering is to eliminate the noises and artifacts from (0.5–40) Hz (Wahabi et al., 2014), (0.5–50) Hz (Donoso et al., 2013),
an ECG signal. In the numerous surveys, researchers use different types (1–30) Hz (Fang et al., 2013), (1–40) Hz (Gutta et al., 2016), (1–100) Hz
of filters to decrease the noise levels and to improve the accuracy of the (Haseena et al., 2011), (1–120) Hz (Yang et al., 2013).
signal. Initially, the recorded ECG signal is gone through a band pass Noise removal can also be achieved with the help of basic low pass
filter before analysis. Band pass filter mainly attenuates muscular noise, and high pass filters. A low-pass filter (LPF) is basically used to filter the
artifacts due to movement of electrodes, PCI, base line wandering, high higher frequency components of a recorded signal. (Oresko et al., 2010)
and low frequency noise signals and provides anti-aliasing. (Pan et al., used LPF with an 11 Hz cut-off frequency to eliminates the noises
1985) uses digital band pass filter (BPF) at a band pass frequency of occurred at higher frequencies and power line noise. The same fre­
nearly (5–12) Hz to eliminate DC bias noise & base line wander. quency is also considered in (Chui et al., 2016). LPF with different fre­
Different frequency bands used for the band pass filter to remove motion quencies used in the study to reduce the high frequency components are
30 Hz (Li et al., 2016), 35 Hz (Zhang et al., 2014), 50 Hz (Alcaraz et al.,
2011), 70 Hz (Alcaraz et al., 2013), 90 Hz (Vullings et al., 2011), 100
(Castroflorio et al., 2013) Hz. (Mar et al., 2012; Nejadgholi et al., 2011)
also used LPF for the removal of noise, but the frequency used in the
literature was not mentioned. In contrast to LPF, a high-pass filter (HPF)
is intended to eliminate lower frequency components of the signa for
signal enhancement. Unlike LPF, HPF attenuates less amount of signal,
but it suffers with phase shift. (Chui et al., 2016) used a high pass filter
with a cut-off frequency of 5 Hz, to eliminate base line wander and any
DC offset present in the signal. The HPF with different frequencies
specified in the literature is 1 Hz (Li et al., 2016), 2.2 Hz (Nejadgholi
et al., 2011), and 0.5 Hz (Alcaraz et al., 2011; Alcaraz et al., 2013;
Vullings et al., 2011). However, the extensive use of these filters may
disturb the signal morphology in frequent & making it impracticable for
diagnosis.
Notch filters are greatly utilized to eradicate the DC offset’s present
in the signal. These filters mingle both low pass & high pass to remove
the frequency range. Most of the notch filters centred at a frequency of
50 Hz or 60 Hz to remove the noise (Chua et al., 2011; Li et al., 2016;
Castroflorio et al., 2013; Nejadgholi et al., 2011; Kutlu et al., 2012).
Median filters are also used to filter the recorded signals. A single me­
dian filter is used to remove special effects and any arbitrary noise
present in the signal (Afkhami et al., 2016; Oresko et al., 2010). A set of
2 median filters was used (Zhang et al., 2014; Mar et al., 2012; Mak
Fig. 6. Framework of Preprocessing.

7
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

et al., 2010) to remove base line wander noise. One filter of 200 ms Table 2
duration is used to remove P waves and QRS complexes. Another filter Performance Comparison of Various Filters.
has a duration of 600 ms is used to eliminate T wave. Then the signal is Reference Filter Type Frequency/ Performance
passed through a 12 valve LPF at 35 KHz frequency (Soria et al., 2009). Duration Metrics
Likewise, a set of 3 median filters is required to ease the isoelectric (Accuracy)
interference (Dima et al., 2013). Authors (Rahman et al., 2013; Salman (Chua et al., Digital Band (0.05–40) Hz, 60 100 %
et al., 2017) introduces the use of adaptive noise cancellation techniques 2011; Mak Pass, Notch, Hz, 200 ms and
to reduce the noise of an ECG signal. (Gowri et al., 2014) proposed a et al., 2010) Median 600 ms
(Afkhami et al., Digital Band Pass (0.10–101) Hz 99.7 %
leaky based normalized higher order filtering techniques to eliminate 2016)
the adaptive power line interference noise. (Ubeyli et al., 4th order (~0.1–101) Hz 95.24 %
After filtering, re-sampling and digitization is required to condense 2009) Butterworth
the usage of memory necessities with lesser computing rate. From the Band Pass
(Fang et al., Band Pass (1–30) Hz 99 %
last decade, the filtered ECG signal was re-sampled and digitized with a
2013)
frequency of 125 Hz (Li et al., 2014), 200 Hz (Donoso et al., 2013), 250 (Gutta et al., 4th order (1–40) Hz 96.2 %
Hz (Martis et al., 2013), 257 Hz (Kumar et al., 2017), 360 Hz (Ubeyli 2016) Butterworth
et al., 2009; Wen et al., 2009), and 1KHz (Shaolin et al., 2022). Wavelet Band Pass
transform technique is also employed in order to preserve the signal (Yang et al., Band Pass (1–100) Hz 96.2 %
2012)
properties and avoiding loss of information (Singh et al., 2006). An (Haseena et al., Band Pass (1–120) Hz 95.8 %
adaptive bionic wavelet transform is proposed by (Sayadi et al., 2007), 2011)
which is the modification of conventional wavelet to minimize noise. (Chui et al., Low pass, High 11 Hz, 5 Hz 97.01 %
(Wiklund et al., 2002) proposed multi resolution wavelet transform 2016) Pass
(Li et al., 2016) Low pass, High 30 Hz, 1 Hz 96.2 %
decomposes the noisy signal into different wavelet coefficients and
Pass
provides the valuable information at different extents. Some other (Zhang et al., Low Pass 35 Hz 86.6 %
methods used to eliminate base line wander noise are adaptive signal 2014)
enhancers (Salman et al., 2016), circular leaky adaptive algorithm (Alcaraz et al., Low pass, High 50 Hz, 0.5 Hz 99.6 %
(Nuthalapati et al., 2019), noise cancellers using variable step size 2011) Pass
(Alcaraz et al., Low Pass, High 70 Hz, 0.5 Hz, 50 88.84 %
(Salman et al., 2018), Zoned adaptive algorithm (Sulthana et al., 2019)
2013) Pass, Notch Hz
and kalman filters (Sulthana et al., 2018) are also used for the reduction (Castroflorio Low Pass 100 Hz 80 %
of noise. et al., 2013)
ECG signals are often corrupted with motion artifact noise due to the (Nejadgholi et al., High Pass 2.2 Hz 92.5 %
2011)
unstable contact between the skin and the surface of the electrodes,
(Mar et al., 2012) Median 200 ms and 600 94 %
muscle contraction and breathing. Patients who move their limbs during ms
test result in abrupt changes to ECG baseline, which may lead to pre­ (Dima et al., Median 200 ms and 600 89.22 %
mature contractions, interference with ECG wave patterns or other 2013) ms
ventricular or supraventricular arrhythmias (Andres Ricardo Perez-
Riera et al., 2018). A method for removing ECG movement artifacts
4.3. Feature engineering
based on empirical wavelet transform and wavelet thresholding is pro­
posed by (Xiaowen Xu et al., 2019). This method effectively removes the
Feature engineering plays a promising role in ECG signal analysis.
motion artefact while improving the signal to noise ratio (SNR) and
ECG classification requires the proper detection of features. The peri­
restoring the original ECG’s QRS complexes with less distortion. Motion
odical ECG signal pattern consists of various complex features. A feature
artifacts affect heart rate variability (HRV) by generating some outliers
is a characteristic or a structural component or a fiducial point that can
in the ECG signal. (Anuhya et al., 2022) reports efficient outlier detec­
be extracted from the ECG pattern. Features are of different types such as
tion and for removal of outliers. (Deepak Berwal et al., 2019) introduced
temporal, morphological, and statistical. Temporal features are obtained
a two-phase technique called stationary wavelet transform with level
from the signal time intervals, and the morphological features describe
thresholding (SWT-LT) to eliminate movement artifact from ECG signal.
the structure of a segment like ST. Statistical features are mostly
In comparison to other methods mentioned, the proposed methodology
calculated using time domain constraints. These features are extracted
effectively removes a variety of in-band motion artefacts with an
with the help of distinct parameters such as lowest value, peak value,
average correlation coefficient of 0.9337. (Egle Butkeviciute et al.,
average value, standard deviation, skewness etc. The performance of the
2019) introduces a novel approach for eliminating movement artifact
analysis depends upon the extracted features, thereby, it is necessary to
from EEG signals during sports exercises. The artifacts are removed
extract the feature accurately. The functioning of cardiac conduction
successfully by applying baseline estimation with sparsity (BEADS) filter
system can be evaluated based on the amplitudes and its intervals of
algorithm and ECG as a reference signal by preserving the spectral
these fiducial points. The standard values of the feature amplitudes and
characteristics of entire EEG signal.
its intervals according to AAMI (Association for the Advancement of
Furthermore, Amplitude normalization (Silva et al., 2015; Sulthana
Medical Instrumentation) for a normal sinus rhythm are tabulated in
et al., 2019) is considered as one of the important steps in preprocessing
Table 3.
stage. It avoids false positives and false negatives occurred due to
Features can be extracted in three distinct domains i.e., spatial, fre­
maximum amplitudes of R and T waves. The output of sampled and
quency and time–frequency domain. The morphological features are
digitized signal should be normalized to reduce the amplitude offsets,
determined in the spatial domain. While the temporal and statistical
DC offset and minimize the variations in signal and to eliminate the
features are obtained in time and frequency domains, respectively. From
influence of amplitude biases (Ebrahimzadeh et al., 2014). It also helps
the last decade, various conventional signal processing techniques and
in comparing the ECG signals of different patients visually. Table 2
machine learning models have been introduced to find the R-R interval,
summarizes the performances of various filters used in preprocessing
QRS complex, J-joint, P-R interval, ST segment, T-wave, and other fea­
stage.
tures as well. The previous methods stated in the literature earlier are
discussed in the following two sections.

8
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 3 Table 4
Standard values of ECG features for a normal healthy sinus rhythm. Traditional Feature Engineering (FE) Methods.
Features/Fiducial points Normal Value Tolerance [Ref] Database FE Method Fiducial Efficiency
points
P Width 110 ms ±20 ms
PR Interval 160 ms ±40 ms (Tadejko et al., MITDB Wavelet ST- Sen = 99 %
QRS Interval 100 ms ±20 ms 2007) transform Segment
QT Interval 400 ms ±40 ms (Fujita et al., MITDB DWT R-peak, Sen = 99.64 %,
RR Interval 800 ms ±200 ms 2015) QRS PP = 99.82 %, err
P Amplitude 0.15 mV ±0.05 mV = 0.54
QRS Amplitude 1.5 mV ±0.5 mV (Zidelmal MITDB, DWT-based J-point, ST [Acc = 96.35 %]
ST Level 0 mV ±0.1 mV et al., 2012) ESCDB windowing segment MITDB
T Amplitude 0.3 mV ±0.2 mV [Acc = 93.33 %]
U Amplitude 0.1 mV ±0.06 mV ESCDB
(Bulusu et al., LTSTDB LPT and KLT ST-segment [Acc = 90 %, Sen
2011) = 91 %] KLT;
4.3.1. Traditional feature engineering methods [Acc = 82 %, Sen
= 85 %] LPT
After preprocessing, the denoised ECG signal is subjected to feature
(Amon et al., ESCDB Time domain ST-segment NR
engineering stage, at which the fiducial points are to be extracted. It can 2016)
be done either by direct method or transformation methods. In direct (Oktivasari MITDB, Gaussian + QRS RMSE = 0.17
method, the width and height of the QRS complex’s, R to R interval and et al., 2019) QTDB SMM complex
ST segment can be obtained with the use of direct filters. For instance, (Madeiro et al., MITDB MM QRS DR = 85.76 %
2020) complex
(Olvera et al. 2006) utilized an improved matched filter method to
(Sadhukhan PTBDB Derivative QRS Sen = 99.8 %
detect ST segment for analysis. Later (Tadejko et al., 2007) formulated et al., 2012) method complex,
the well-known kohonen self-organizing maps to extract R-R intervals R-peak
and morphologies of ECG signal. The limitations associated with uti­ (Curtin et al., MITDB Windowing QRS Acc = 94.3 %,
lizing the direct methods is the morphological characteristics and fre­ 2018) technique complex, Sen = 96 %, PP =
R-peak 97.3 %, err = 0.3
quency content varies with time. Also, it is necessary to analyse the ECG (Umer et al., MITDB Windowing P, Q, R, S Overall Acc = 99
signal with respect to time to describe the characteristics of a signal 2014) technique and T %, P[Sen = 96.72
effectively. This justifies the use of time–frequency representation in the waves %], Q[Sen =
quantitative analysis of ECG signal. 97.12 %], R[Sen
= 96.04 %], S
Wavelet transform (WT) is an efficient tool that many researchers
[Sen = 97.32 %],
have employed to identify various features of ECG signal. It has an T[Sen = 97.56 %]
ability to analyse non-stationary ECG signal more accurately than (Martinez QTDB Phasor QRS Sen = 99.81 %,
Fourier Transform (FT) and Short Time Fourier Transform (STFT) ap­ et al., 2010) transform complex, P- PP = 99.89 %, err
proaches due to its time–frequency localization phenomenon. Compared wave, T- = 0.017
wave
to WT, the Discrete wavelet transform (DWT) provides signal informa­ (Subramanian MITDB Multi- QRS- Acc = 93.35 %,
tion with less computational time. This transformation become a pop­ et al., 2017) wavelet complex Sen = 98.5 %, PP
ular method for the analysis of biomedical signals such as ECG in both transform = 97 %, err =
time and frequency domain. It firmly decomposes the signal with 0.044
(Saadi et al., MITDB, FIR-based QRS [Sen = 99.9 %,
different levels of resolution and can be analysed further using various
2015) ESCDB adaptive complex PP = 99.87 %]
filters to extract different fiducial points. (Fujita et al., 2015) proposed thresholds MITDB; [Sen =
DWT approach for the detection of QRS complex and R-peak and ach­ 99.84 %, PP =
ieves a sensitivity of 99.64 %, positive predictivity of 99.82 % and with a 99.71 %] ESCDB
less error rate of 0.54. The fiducial points such as J-point, ST segment are
detected using DWT-based windowing technique in (Zidelmal et al.,
4.3.2. ML based feature engineering methods
2012) which is evaluated on multiple datasets. Bulusu et al. (2011)
Traditional signal processing approaches takes a considerable time to
introduced Karhunen-Loeve Transform (KLT) and Legendre Polynomial-
detect and classify various arrhythmias while considering computa­
based transformation (LPT) to extract ST-segment. However, this
tional constraints. This large processing time is critical to one’s life and it
method results with a lower sensitivity compared to (Fujita et al., 2015).
could be reduced if rapid techniques were evolved. Recent advance­
The time domain (Amon et al., 2016) methods such as gaussian window-
ments in technologies develop faster methods to detect various
based synthesized mathematical models (SMM) (Oktivasari et al., 2019),
abnormal conditions using mathematical computations and Neural
mathematical morphology (Madeiro et al., 2020), derivative-based
Networks (NN). These technologies, on the other hand, rely heavily on
(Sadhukhan et al., 2012), FIR-based smoothing (Carlos et al., 2019)
the accurate identification of features, of which QRS complex is the most
have produced the highest sensitivity of 99.81 % which is somewhat
predominant feature of the ECG signal. Several Artificial Intelligence
lesser than DWT-based techniques. Also, several windowing-based al­
(AI) techniques are proposed in the literature to detect various fiducial
gorithms proposed in (Curtin et al., 2018; Umer et al., 2014), to detect
points of ECG signal.
various fiducial points such as P, Q, R, S, T, and QRS complex reports 99
Combination of wavelet transform, and artificial neural networks
% accuracy which is analogous to DWT based techniques. A fully inte­
(ANN) was utilized by (Mazhar et al., 2006) for feature extraction and
grated digital ESP was introduced by (Nourhan Bayasi et al., 2016) for
classification of ECG. Test results showed up features are extracted using
the prediction of fiducial points for detection in real time. This method
wavelet transform and further classified with the help of ANN with an
was implemented using 65-nm technology consumed a power of 2.78
accuracy of 92 %. Instead of continuous wavelet transform, (Maedeh
microwatts with an accuracy of 86 %. Table 4 provides a list of con­
et al. 2012) utilized DWT and neural networks to improve the accuracy
ventional signal processing methods for obtaining fiducial points of ECG
to 96.5 %. (Jen et al., 2008) detect QRS complex with the support vector
along with the performance measures in terms of positive predictivity
machine (SVM), Naïve Bayes (NB), K-Nearest Neighbour (KNN) and
(PP), sensitivity (Sen), accuracy (Acc), detection rate (DR), error (err)
variational mode decomposition (VMD) approaches. Some of the ma­
and root mean square error (RMSE).
chine learning approaches (Afsar et al., 2008; Mehta et al., 2007; Had­
jem et al., 2016; Xiao et al., 2018) to detect features are summarized in

9
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 5 98 %. Later, an improved BFO algorithm i.e., an Adaptive BFO optimi­

ML based Feature Engineering Methods. zation technique is proposed by (Padmavathi et al., 2017) with variable
[Ref] Database FE Method Fiducial points Efficiency step-size and improves the accuracy of 98.74 %. Similar studies used
singular value decomposition (SVD) (Padhy et al., 2017), GA with PSO
(Mazhar MITDB WT + ANN Transformed Acc = 92 %
et al., Coefficients (Patro et al., 2020) optimization techniques for efficient optimization of
2006) features in literature. Table 6 summarizes various feature optimization
(Maedeh MITDB DWT + ANN Transformed Acc = 96.5 % techniques for classification.
et al., Coefficients
2012)
(Jen et al., QTDB VMD + KNN, QRS complex Sen = 86 %,
4.5. Abnormal classification
2008) NB, SVM and R-peak PP = 89.1 %
Acc = 88 % The optimized feature set is given as input to the classifier to
(Afsar et al., ESCDB Ensemble NN ST-deviation Sen = 90.74 %, recognize the information about cardiac diseases from ECG. Abnormal
2008) PP = 89.21 %
classification has become a valuable and promising technique for early
(Mehta CSEDB Entropy-based QRS complex Sen = 99.93 %,
et al., SVM PP = 99.13 % assessment of arrhythmia. By observing the ECG, experienced cardiol­
2007) ogists can easily detect abnormal beats from normal rhythm. However, it
(Hadjem ESCDB Decision tree T-wave and ST- Sen = 86 %, is still the challenging task for a computer that automatically detects
et al., segment PP = 89.1 %
abnormalities even in the presence of disturbed environment. Several
2016) Acc = 88 %
F1 = 0.6
conventional and machine learning-based classification techniques
(Xiao et al., LTSTDB Google’s Change in ST- Sen = 82.63 %, employed in the literature are discussed in this section.
2018) Inception V3 2- segment PP = 80.35 %
D CNN F1 = 0.87 4.5.1. Traditional classification methods
The classification of ECG beats into normal and abnormal by means
Table 5 and its performance is expressed with respect to sensitivity of conventional classification techniques are discussed in this section.
(Sen), detection rate (DR), positive predictivity (PP), accuracy (Acc) and Threshold-based algorithms are used in (Banerjee et al., 2013; Kumar
F1 score (F1) are reported. et al., 2016) to classify beats evaluated on different datasets. However,
these techniques result in optimal performance for classification. To
improve the performance, modified Pan-Tompkins (Faezipour et al.,
4.4. Feature optimization 2010) proposed an adaptive thresholding technique and results with an
improved accuracy of 95 %. Later a Multimodal decision learning
Reduction of feature dimensionality is one of the important tasks for technique (MDL) (Naik et al., 2016) is proposed for the classification of
pattern recognition, data mining, and classification. Redundant and normal and abnormal beats evaluated on MITDB and achieved the
irrelevant features need to be removed to develop robust learning highest sensitivity of 100 %. The studies (Firoozabadi et al., 2019)
models. Feature optimization technique fasten the classification process, discuss conventional classification methods to classify each class of ECG
improves classification accuracy with less computational time. This heartbeats. Several conventional classification models are tabulated in
section presents a few feature optimization methods employed for ECG Table 7.
analysis reported in the literature.
Studies (Martis et al., 2012) present a well-known Principal 4.5.2. ML and DL based classification techniques
Component Analysis (PCA) method that most researchers used for Machine learning, which is the subset of Artificial intelligence (AI)
reducing dimensionality in classifying ECG signal. It realises the input deals with intelligent behaviour of computers. It performs classification
vector with high dimensionality into a low-dimensionality vector with using a variety of supervised or unsupervised learning algorithms. Su­
components being uncorrelated. It computes principal components and pervised learning classifies the data based on the labelled/structured
projects the data into directions that have highest variability. These data. While unsupervised learning involves clustering/ categorial forms
uncorrelated points on the other hand, may not indicate the optimal way based on unstructured data. In this section several AI-NN based tech­
for getting highest possible discrimination to differentiate among niques reported in the literature are discussed for the classification of
different classes. Another well-known popular dimension reduction various abnormalities. (Zahra et al., 2020) some of the deep learning
method that is Linear Discriminant Analysis (LDA) (Tantawi et al., 2013) approaches for ECG arrhythmia classification are discussed and
is used to transform larger original dataset into smaller one. (Martis elaborated.
et al., 2013) reported the use of Independent Component Analysis (ICA) Artificial Neural Network’s (ANN) resolve the problems associated
method which optimizes the multi-dimensional data into features which with linear & non-linear classification of different network configura­
are statically independent with one another. The combination of PCA, tions. (Hirose et al. 2003) used a complex-valued ANN; extension of real-
LDA and probabilistic neural network (PNN) technique is introduced by valued network includes complex-valued operations to classify the
(Wang et al., 2013) to classify 5 types of beats and achieved an accuracy beats. (Exarchos et al., 2007) Decision Tree (DT) based Fuzzy clustering
of 99.71 %. is utilized to enhance the ANN performance. ANN with ensemble
The Bacteria Foraging optimization (BFO) is yet another technique empirical mode decomposition with adaptive noise system is developed
proposed by (Das., 2015) requires minimum knowledge and converges to classify different types of beats in (Abdalla et al., 2019). Support
quickly to reach global minimum solution. Author used BFO to reduce Vector Machine (SVM), (Venkatesan et al., 2018) which is a linear
the dimension of combined morphological and temporal features. (Li classifier that separates the classes linearly by creating a hyperplane
et al., 2017) proposed Genetic Algorithm (GA) based optimization from high-dimensional space. SVM can also capture the non-linear re­
technique with Back Propagation Neural Network (BPNN) classifier lationships of the ECG signal, detects the heartbeats, and classify the
which reduces 50 % of features than original dataset and achieves a data as normal/abnormal with high accuracy. Researchers proposed
classification accuracy of 97.78 %. An improved classification perfor­ several SVM based classification techniques to detect arrhythmias in
mance is obtained with Cuttlefish optimization algorithm (Ahmad et al., literature which involves Multi-class SVM (Jha et al., 2020), SVM with
2018) compared to (Li et al., 2017). (Padmavathi et al., 2015) intro­ NN (Sahoo et al., 2017), and ensembled SVM, a combination of multiple
duced the hybrid BFO and particle swarm optimization (PSO) technique SVM’s (Huang et al., 2020). However, due to its high dimensionality
with Scalar Conjugate Gradient Neural Network (SCGNN) for the se­ space, it suffers with computational constraints. The main limitation of
lection of optimal feature set and recorded the classification accuracy of these machine learning solutions is the use of heuristic hand-crafted

10
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 6
Classification of feature optimization methods.
[Ref] Optimization Technique Reduced dimensionality Classification Accuracy Database

(Martis et al., 2012) PCA + SVM 12 features 98.11 % MITDB

(Martis et al., 2013) ICA + PNN 12 features 99.28 %
(Wang et al., 2013) PCA + LDA + PNN NM 99.71 %
(Manab Kumar Das, 2015) BFO + SVM NM 98.2 %
(Li et al., 2017) GA + BPNN 50 % 97.78 %
(Ahmad et al., 2018) CFA + ANN NM 97.96 %
(Padmavathi Kora et al., 2015) BFPSO + SCGNN 20 best features from a total of 200 features 98.71 %
(Padmavathi Kora et al., 2017) ABFPSO + SCGNN 20 best features from a total of 200 features 98.74 %
(Padhy et al., 2017) SVD NM 95.3 % PTBDB
(Patro et al., 2020) GA + PSO NM 94.90 % recognition rate

2020) to extract both temporal and spatial features of ECG. Experi­

Table 7
mental analysis is carried out on PTB database in both intra-patient and
Traditional ECG classification techniques.
inter-patient manner and attains an accuracy of 99.93 % and 96.50 %
[Ref] Database No. of Classification Efficiency respectively. Similarly, to overcome the correlation problem between
Classes technique
context and gradient dispersion in conventional CNN (Jiang et al., 2021)
(Banerjee PTBDB 2 Thresholding Acc = 97.62 %, developed a hybrid attention-based deep learning network (HADLN) for
et al., 2013) Sen = 97.29 %,
arrhythmia classification. HeartNet, a new deep learning method com­
Spe = 98.8 %
(Kumar et al., ESCDB 2 Thresholding Sen = 98.13 %,
pressed by multi-head attention mechanism on the top of CNN model is
2016) Spe = 98.17 % proposed by (Rafi et al., 2022) for automatic ECG classification. In this
(Faezipour MITDB 2 Adaptive Acc = 99.51 %, work, the insufficient data label problem is solved using generative
et al., 2010) Thresholding err = 0.0049 adversarial network (GAN) with generating additional training samples.
(Naik et al., MITDB 2 MDL Acc = 93.33 %,
The performance of the model is tested using MIT-BIH dataset and at­
2016) Sen = 100 %, Spe
= 81.81 %, tains an accuracy of 99.67 ± 0.11 and 89.24 ± 1.71 MCC. However, this
PP = 90.47 % model may fail to converge its prediction if there’s too much negative
(Firoozabadi MITDB 2 PCA + ICA Acc = 99.28 %, samples. The non-Euclidean nature of ECG signals makes it difficult to
et al., 2019) Sen = 97.97 %, analyse them using standard CNNs. This motivates the use of trans­
PP = 99.21 %
former networks, a special category of graph neural networks (GNN).
Transformer networks compensate the shortcomings of CNN poor per­
features with trivial attribute learning designs. Also, some machine formance and captures both temporal features and context vectors using
learning solutions requires the manual selection of features in ECG attention mechanism. (Che et al., 2021) introduced the integration of
signal which in turn results in information loss. transformer network in CNN framework to capture temporal informa­
To overcome these limitations, one promising solution is the use of tion and a new link constraint is introduced to the loss function to
deep learning structural designs. Conventional Neural Networks (CNN) enhance the classification ability of the embedded vector. To solve the
are evolved from ANN that analyses much deeper and performs classi­ problem of inter-heartbeat dependencies, (Hu et al., 2022) developed a
fication. (Jyothi et al., 2021) reviews various segmentation approaches novel transformer based deep learning model ECG DETR. ECG-ViT, yet
in deep learning to detect cardia arrhythmia. DL differs from the con­ transformer-based deep learning model for ECG classification is pro­
ventional techniques particularly in three facets: first, it together opti­ posed by (Neha et al., 2022). Their developed solution obtained 99.3 %
mizes the representation and classification of features to optimize their accuracy and 99.7 % F1 score with tenfold cross validation.
integration; secondly, due to its strong significant power with large Deep learning models are popular since they yield high accuracy for
capability, making it suitable for handling huge amounts of data; lastly, classification problems. However, these models need special hardware
its deeper representation can capture higher level data variations that for training and have high time complexity. To address this problem,
can be shared various tasks. With the availability of GPU’s, DL is gaining researchers began to develop textural feature engineering based hand-
popularity now-a days. The applications of DL ranges from biometric crafted models to reduce the time burden. (Tuncer et al., 2019) intro­
application (Neung et al., 2022), object detection, image classification duced a method for automated classification of 17 arrhythmia classes.
(Jen et al., 2008), sleep apnea detection (Bahrami et al., 2022; Kapil They combined DWT with 1-D hexadecimal local pattern approach to
et al., 2021) and computer vision (CV) (Naz et al., 2021) technologies. attain high success rate with minimal processing cost. This study focused
Also, it involves numerous cardiological applications in ECG arrhythmia on the DWT-based preprocessing, 1D-HLP-based feature extraction,
detection using “Deep Belief Network with Restricted Boltzmann Ma­ feature concatenation, neighborhood component analysis (NCA)-based
chine” (DBN-RBM) (Mathews et al., 2018), Deep-CNN (Sellami et al., feature reduction, and 1 nearest neighborhood (1NN)-based classifica­
2019; Kamaleswaran et al., 2018; Wu et al., 2020), Deep Neural tion. Authors used 1000 10-second ECG signal fragments from 45 in­
Network (DNN) (Nurmaini et al., 2020), Residual CNN (Li et al., 2020), dividuals from the MIT-BIH arrhythmia data set and achieved a 95 %
Long-Short term Memory network (LSTM) (Chang et al., 2020), End-to- accuracy rate. A lightweight homeomorphically irreducible tree (HIT)
End CNN (Jafarian et al., 2020), CNN-LSTM (Chen et al., 2020) and graph pattern technique is developed by (Baygin et al., 2021) for the
RNN-LSTM (Mostayed et al., 2018). generation of handcrafted features. The most informative features were
In recent times, CNN uses attention-based architecture to increase its chosen by chi2 selector and then classified using support vector machine
interpretability for ECG classification. Attention networks only choose (SVM) classifier. This model was tested on 12-lead ECG signals taken
the most informative features required for final selection. (Liu et al., from 10,646 individuals of public ECG dataset and attained an accuracy
2021) introduced a new learning attention mechanism based on neural of more than 92 %. In an effort to identify arrhythmia, (Tuncer et al. in
network search (NAS) in CNN to detect the rhythm and morphological 2022) developed yet another enhancement to the ECG signal detection
abnormalities of a 12-lead China physiological signal challenge (CPSC) technique. The authors proposed a novel 1-D discrete wavelet based
dataset. A multi-lead attention (MLA) mechanism integrated with CNN concatenated mesh tree (DW-CMT) for preprocessing the signals. Later,
and bidirectional gated recurrent unit (BiGRU) is proposed by (Fu et al., a novel ternary chess pattern (TCP) technique is used for the extraction

11
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

of features. To select the more informative features from the extracted Table 8
ones, neighborhood component analysis (NCA) based feature selection ML and DL based ECG classification techniques.
method is used. In their study, two datasets, MITBIG arrhythmia dataset [Ref] Database No. of Classification Efficiency
and St. Petersburg dataset were used for experimental analysis. Their Classes technique
developed solution attains a maximum classification of 96.6 % for MIT- (Exarchos et al., ESCDB, 2 DT + Fuzzy Acc = 91.7
BIH dataset using KNN and 97.8 % accuracy for St. Petersburg ECG 2007) MITDB clustering %, Sen =
dataset using SVM. To achieve high recognition performance along with 91.2 %, Spe
high classification accuracy, (Subasi et al., 2021), proposed a novel = 92.2 %
(Abdalla et al., MITDB 5 ANN + Acc = 99.9
tower graph transformation model. In this work, a novel hexadecimal 2019) CEEMDAN %, Sen =
local adaptive binary pattern (HLABP) feature extraction technique is 99.7 %, Spe
presented, which is an improved version of (Tuncer et al., 2019). A = 99.9 %
hybrid iterative feature selection technique, RFINCA is implemented (Venkatesan MITDB 2 SVM Acc = 96 %
et al., 2018)
which is a combination of RelieF and NCA is implemented. Their
(Jha et al., MITDB Multiple SVM Acc =
developed solution was tested over ANN and DNN networks and obtains 2020) 99.27 %,
an accuracy of 95.70 % and 97.10 %. Deep CNN models have been Sen =
successfully used for solving numerous tasks in computer vision tech­ 96.23 %,
nology. In computer visualization, transfer learning is typically repre­ Spe =
99.58 %,
sented with pre-trained models. (Naz et al., 2021), proposed a (Sahoo et al., MITDB 4 SVM + NN Acc =
transformation-based approach for the detection of ventricular 2017) 98.39 %,
arrhythmia using 2D CNN. ECG signals are initially transformed into 2D Sen =
binary images. Transfer learning is used to train the model and extract 96.86 %,
Spe =
deep features from various output layers. The features are then fused
98.92 %,
and selected using entropy-based feature selection mechanism. Finally, err =
the selected features are usually trained with various supervised 0.0042 %
learning classifiers. The performance of the model was evaluated on (Huifang et al., MITDB Heart beats Ensemble SVM Acc = 98.6
MIT-BIH dataset and obtains an accuracy of 97.6 % for cubic SVM 2020) %, Sen =
98.5 %, Spe
classifier. Some popular classification models along with its perfor­
= 99.78 %,
mance are tabulated in Table 8. (Mathews et al., MITDB 2 DBN-RBM Acc =
We made the overall comparison of percentage share of each method 2018) 93.63 %
for arrhythmia classification among the total papers studied. Fig. 7 il­ Sen =
88.62 %,
lustrates the correlation between the percentage each model and the
err = 0.047
proportion of cardiac arrhythmia’s considered in the papers studied. %
Atrial fibrillation (AF), abnormal heartbeats, and supraventricular beats (Sellami et al., MITDB 5 Deep-CNN Acc =
ectopic beats/ventricular ectopic beats (SVEB/VEB) are the most often 2019) 99.79 %,
recognised arrhythmias, contributing 38 %, 20 %, and 18 %, respec­ Sen =
94.65 %,
tively. In addition, Fig. 8 summarises the accuracy performance of all
Spe =
the studied arrhythmias based on the classification methods reported. 99.36 %,
CNN-based classification has been shown to be successful for classifying PP = 97.71
arrhythmias, according to the best classification methods shown in %
(Kamaleswaran ChallengeDB 3 Deep-CNN Acc =
Fig. 8. Therefore, CNN is considered as one best learning tool for health
et al., 2018) 82.24 %, F1
applications. However, the robustness of classification system for long = 0.83 %
term reliability must be verified in this field of research. (Wu et al., ChallengeDB 4 Deep-CNN [err = 0.93]
2020) N, [err =
5. Smart health 0.82] AF,
[err = 0.79]
other
Now-a-days, due to the rapid growth in population, people having (Nurmaini PTBDB Heartbeat’s DNN Acc =
less income are suffered to consult the doctor regularly for medical et al., 2020) 99.34 %,
check-up. Consequently, there is a need to examine the health state of Sen =
93.83 %,
the patient wherever he is. Telehealth paving the way for the fulfilment
Spe =
of smart health care. It is being extensively used to expand the 99.57 %,
geographical coverage of health care services and improve access to PP = 89.81
care. It incorporates variety of practices and involves the interaction %
between patients and clinicians over the phone, email, video calls or (Li et al., 2020) MITDB 5 Residual CNN Acc =
99.06 %,
conferences, the Internet, streaming media, wireless connections, and Sen =
with remote devices. Remote patient monitoring is a type of telehealth 93.21 %,
service, which makes use of certain technology to enable communica­ PP = 96.76
tion between doctors and patients at home. Patient satisfaction with %
(Chang et al., CSEDB Multi-class LSTM Acc = 90 %
remote health care services is briefly explained by (Liliana et al., 2021).
2020)
Smart wearable’s enables the people to supervise a few critical condi­ (Jafarian et al., PTBDB 1 End-to-End Acc =
tions and assists the patients remotely to lessen the treatment cost. Fig. 9 2020) CNN 98.21 %,
explains the overall taxonomy of smart health care structure. The Sen = 97.5
medical devices utilized for the system is categorized into two types. The %, Spe =
98.01 %
stationary devices are typically used in hospitals and clinics to examine (Chen et al., MITDB Multi-class CNN + LSTM Acc =
the physiological signals of a user (Prabha et al., 2018). On-body ser­ 2020) 97.15 %,
vices are again divided into two categories. The in-vitro sensor devices (continued on next page)
are placed outside of the human skin without use of lab environment

12
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 8 (continued ) management services deal with storing, protecting, and analysing the
[Ref] Database No. of Classification Efficiency data, aids remote monitoring with value-based networks and ensures
Classes technique safety with access control, vision management, video surveillance at
Sen =
hospitals (Solanas et al., 2014). Also, it provides some set of services
95.40 %, such as semantic processing, context aware services to identify the ac­
Spe = curate location of the user etc. There is wide range of applications of
96.80 %, smart health care system starting from temperature monitoring of in­
PP = 95.56
fants to vital sign monitoring of elders. These are grouped into three
%
(Mostayed CPSCDB MITDB RNN + LSTM F1 = 74.15 intra-body, inter-body, and continuous remote monitoring applications
et al., 2018) % (Aditi et al., 2021). Finally, the end users are nothing but an individual
(Fu et al., 2020) MITDB 2 MLA + Acc = or user, hospitals, laboratories government organizations etc.
CNN_BiGRU 99.93 %intra- Smart health care system utilizes a wearable device used to monitor
patient,
Acc =
the physiological parameters of the user remotely. Fig. 10 shows the
96.50 %inter- generalized architecture of smart wearable health monitoring system
patient drawn by considering (Liang et al., 2016) as reference. Wearable sys­
(Taki et al., MITDB 5 GAN Acc = 99.67 tems generally comprise of four units i.e., capturing unit, processing unit
2022) ± 0.11
and transmission modules and end users. Data capturing unit may
MCC =
89.24 ± consist of several miniature sensors that capture the physiological sig­
1.71 nals from the human & sends the data to the processing unit through
(Sukla et al., MITDB 4 FPGA Acc = 99.3 wired or wireless connection. The data is then processed, analysed, and
2022) % stored by a PDA or microcontroller or cell phone and transmits to remote
F1 = 99.7
%,
services through a wireless communication module. A server or a
(Tuncer et al., MITDB 17 DWT-HLP Acc = 95 % database act as a gateway to send the information required to remote
2019) clinicians, hospitals etc providing remote monitoring of patients. An
(Baygin 12-lead 7, 4 HIT + Acc = 92 %7 alert signal is also sent to the ambulance or to hospitals in case of any
et al.,2021) public ECG Multilevel class
abnormalities. It should be noted that smart healthcare architectures
records pooling Acc =
97.15 %4 must be interoperable across various technologies. Patients who have
class
access to smartphones, laptops, tablets, and desktop computers can
(Tuncer et al., MITDB, St. 17 DWT-CMT Acc = 96.6 easily use telehealth programmes to connect with medical professionals
2022) Petersberg %MITBIH who may be able to diagnose, track, and treat a wide range of acute and
Acc = 97.8
chronic diseases. A mobile-based smart wearable system was developed
%St.
Petersberg by (Tewary et al., 2016) capable of detecting abrupt fall scenarios,
(Subasi et al., MITDB 17 1D-HLP Acc = 95.7 cardiac irregularities, and hypertension/hypotension, making it suitable
2021) %ANN for real-time monitoring, self-diagnostic, and remote diagnosis. (He
Acc =
et al., 2013) proposed a six layered cloud architecture for which can be
97.10 %
DNN able to deal with heterogeneous, multimodal, and non-stationary
physiological signals to provide the personalize healthcare services.
This model utilizes the message queue as a cloud engine and high per­
formance publish/subscribe mechanism for reliable communication.
Data compression is one of the important aspects need to be considered
upon transmitting the data in smart health care at higher rates. There­
fore, an energy efficient multiple modality data compression framework
has been deployed by (Ahmed et al., 2018) which is adaptive to more
network constraints. In addition to developing IoT in smart healthcare,
secured data transmission is significant and should withstand to handle
multiple attacks. In order to improve the security and privacy issues, a
novel multi-resolution 1D CNN has been proposed by (Zhang et al.,
2017).
In recent years, the world had developed many wearable continuous
medical monitoring prototypes in real time. These prototypes may be
hardware configuration based, textile-based, microcontroller based or
(Body Area Network) BAN-based. Table 9 summarizes the commercial
products developed in the literature during last 20 years. The techno­
logical advancements in ambulatory ECG revolutionized consumer
world applications to develop mobile health (mHealth) devices, specif­
Fig. 7. Percentage of cardiac diseases considered in this research.
ically, wearable technologies have been considered as an innovative
solution for replacing invasive, less efficient methods for monitoring
while the in-vivo sensor devices are implantable ones which are
cardiac arrhythmias. Biosensors and sophisticated artificial intelligence
attached interior to the skin for diagnosis.
(AI) algorithms are nowted in new iterations of smartwatch and
Connectivity technologies play a significant role in designing a sys­
smartphone technologies, making them capable of diagnosing cardiac
tem to increase the number of applications. These are the backbone for
arrhythmias. For example, the wrist-worn devices, AMON (Funded
good health system. There are various connectivity technologies in­
Project) (Anliker et al., 2004), Fitbit (Lubitz et al., 2021), Apple
cludes RFID (Belal et al., 2017), NFC (Mrunalinee et al., 2016), Zigbee
(Seshadri et al., 2020), and Samsung (Avram et al., 2021) watches to the
(Zhang et al., 2014) etc helps in transferring the data between physical
health-conscious consumer has resulted in an increase in the sales of
entities and the network. Tele health care provides management services
wearable technology. Physiologists are increasingly using single-lead
such as database, remote, network and security management. These
ECG monitors, such as, ZioPatch (iRhythm technology) (Zhang et al.,

13
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

105
100

Accuracy (%)
95
90
85
80
75
70

Cardiac Arrhythmias

Fig. 8. Best classification accuracy report of arrhythmia studied.

Smart Health Care System

Medical Connectivity System

Services Applications End Users
Devices Technologies Management

On body Stationary RFID Database Context aware Inter Body Hospitals

Sensors Medical BLE Management Sensing
Devices NFC Remote Semantic Individuals
Zigbee Management processing Intra Body
Wi-Fi Network Sensing Governament
UWB Management Community Organizations
In Vitro In Vivo Cellular Security Health Care Remote
Sensors Sensors Sigfox Management Monitoring Diagnostic
Z-wave Child Health Laboratories
Care etc.....
Clinical Research
Institutes

Fig. 9. Classification of Smart Health care System.

Fig. 10. General architecture of smart wearable health monitoring system.

2018) and a compact ECG device, Kardia Mobile (Alivecor) (Ding et al., (Zhang et al., 2018) and a compact ECG device, KardiaMobile (Alivecor)
2020) to identify paroxysmal rhythms. Physiologists are increasingly (Ding et al., 2020) to identify paroxysmal rhythms. Alivecor Kardia is
using single-lead ECG monitors, such as, ZioPatch (iRhythm technology) small, single channel, hand-held EKG device used to detect four most

14
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 9
Smart wearable devices and commercial products.
Literature Physiological Integrated Sensors Communication Hardware Description Applications Pros & Cons
Research Parameters Modules

Life Shirt (Heilman BT, R, ECG ECG sensor, Blood Bluetooth, wires Sensor embedded in vest, Whole day monitoring Low Power, less
et al., 2007) sensor PDA weight, comfortable
AMON (EU-Funded ECG, BP, Spo2, ECG sensor, two-axis Bluetooth, GSM Wrist worn device Emergency & Remote Multi parameters, low
Project) (Anliker HR accelerometer Monitoring power, bulky
et al., 2004)
Live Net (MIT) ( ECG, Spo2, R, T, ECG sensor, 3-D wires, 2.4 GHz PDA, microcontroller Parkinson symptom, Multi parameter
Sung et al., 2005) BP, EMG, GSR accelerometer, GSC radio, GPRS board behavior modeling etc. measurement, little
sensor, temperature complex
sensor
MagIC (Rienzo ECG, R, T ECG sensor, textile Bluetooth PDA, Vest with textile Records cardio- Washable,
et al., 2005) sensor, temperature sensors, custom respiratory & motion comfortable
sensor electronic board signals in clinical
environment and in0
daily life
Code Blue ( ECG, Spo2, A ECG sensor, 3-D Zigbee Sensor motes with custom Real time vital signal Fairness across
Shnayder et al., accelerometer processing boards monitoring with multiple paths,
2005) wearable sensors security is needed
Wearable ECG ( ECG ECG sensor Zigbee, GPRS, Wires Microcontroller board, Remote detection of Intelligent device
Fensli et al., PDA cardiac issues
2005)
Lifeguard (Mundt ECG, Spo2, R, T, Off-the-shell sensors Bluetooth, serial Custom microcontroller Medical monitoring in Multi parameter
et al., 2006) BP, A cables device & commercial extreme environments monitoring system,
biosensors low power
consumption
MyHeart (EU- ECG, R, A ECG sensor, Textile Bluetooth, GSM, PDA, textile & electronic Prevention & early Comfortable, highly
Funded Project) ( sensors, activity sensor conductive yarns sensors on clothes, heart diagnosis of CVD accurate
Habetha et al., belt
2006)
MERMOTH (EU- ECG, R, T, A ECG sensor, Textile conductive yarns, Garment with Knitted dry General health Low cost, knitted,
Funded Project) ( sensors, activity sensor RF link electrodes, PDA monitoring stretchable garment
Luprano et al.,
2006)
Human++ (IMEC) ECG, EMG, EEG ECG sensor, Zigbee Miniaturized low-power Autonomous wearable Can run for 3 months
(Torfs et al., electromyogram sensor BAN nodes, energy sensor networks for with the use of 2 AA
2006) scavenging general health batteries
monitoring
Health Gear (Oliver HR, Spo2 ECG sensor, pulse Bluetooth Custom-sensing board, Sleep monitoring to Less weight, little
et al., 2006) oximetry communication sensors, detect sleep apnea events complex
cell-phone
AUDABE (Katsis ECG, R, GSR, ECG sensor, GSC sensor Bluetooth, wires, Mask, glove, chest sensors Emotional state Wearable intelligent
et al., 2006) EMG Wi-Fi monitoring at extreme system
stress conditions of an
individual
WEALTHY (EU- ECG, R, T, EMG, ECG sensor, 3-D Bluetooth, Textile & electronic monitoring of Efficient artifacts
Funded Project) ( A accelerometer, conductive yarns, sensors on jacket rehabilitation & chronic removal, accurate
Parasido et al., temperature sensor GPRS diseases of elderly measurement, little
2005) patients complex
Smart vest ( ECG, BP, T, PPG, ECG sensor, Blood 2.4 GSM ISM RF, Vest with woven sensors, General remote health Multi-parameter
Pandian et al., GSR sensor, temperature woven wires microcontroller monitoring monitoring system,
2008) sensor comfortable
Body Area Network ECG, BP, R ECG sensor, BP monitor, Zigbee, Wi-Fi, GPRS Zigbee based motes & Forecast human physical Efficient integration,
(Monton et al., pulse oximeter Custom based devices state during daily complexity in custom
2008) activities design
WSN u-Healthcare ECG, BP, SPo2, A ECG sensor, 2 axis Zigbee, CDMA Custom tiny motes, cell Remote identification and Convenient to use,
system (Chung accelerometer phone, commercial health monitoring moderate cost
et al., 2008) sensors
Personal Health ECG, BP, A ECG sensor, Pulse Bluetooth, GPRS Cell phone & custom Heart-attack self-test for Simple user interface
Monitor ( monitor available biosensors CVD patients
Leijdekkers et al.,
2008)
HeartToGo (Jin ECG, A ECG sensor, 3-D Bluetooth, GPRS Cell phone & custom Individual remote CVD More accurate, easy to
et al., 2009) accelerometer available biosensors detection use
Bioharness ( ECG, R, T, A, P ECG sensor, textile Bluetooth or ISM RF Chest belt Remote monitoring of Multiple parameter
Zephyr, 2009) sensors human performance in monitoring, bulky
real time
Planar fashionable ECG Electrodes Wires Electrodes integrated in Continuous monitoring of Easy to measure,
circuit-based shirt ECG moderate performance
shirt (Jerald
et al., 2009)
BIOTEX (EU- ECG, SPo2, R, ECG electrode, PH Bluetooth Multi-parametric patch, Real time personal health Low power, huge
Funded Project) ( Sweat rate, sensor, solid sensor, vest band monitoring applications
Coyle et al.,
2010)
(continued on next page)

15
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Table 9 (continued )
Literature Physiological Integrated Sensors Communication Hardware Description Applications Pros & Cons
Research Parameters Modules

Proe-tex (EU- HR, BP, Spo2, T, Peizo electric sensors, Zigbee, Wi-Fi Microcontroller, flexible Real time monitoring in Intelligent, low power,
Funded Project) ( dehydration textile electrodes band harsh environment complex
Curone et al.,
2010)
Magic System (Kim SPo2, Posture, ECG electrode, 3D- Bluetooth PDA, vest with textile Real time monitoring in Easy to carry, lower
et al., 2016) ECG, R accelerator sensor altitude mount weight
Health wear (Liao Activity monitor, optical biomedical 3G Wearable garment for a Bio-signals are Less weight, High
et al., 2012) Pulse, SPo2 sensor, fabric electrode portable patient continuously monitored power consumption
wirelessly
Glucose Sugar levels in Glucose sensor RF Integration of glucose wireless contact-lens tear Low-current-noise,
Monitoring ( blood sensor, and its interface monitoring accurate detection
Meriggi et al., chip on polymer subtrate
2013)
Smart closes (Shyr BP, EMG, ECG, R Wearable sensing device Bluetooth Smart phone Long-time monitoring & Friendly user
et al., 2013) emergency interface, efficient
energy
Smart phone Spo2, HR ECG electrode, BT USB Smart phone, smart Continuous monitoring Low power, easy to use
healthcare ( sensor clothes
Milsis et al.,
2014)
Posture Monitoring Multi-Posture Accelerometer Bluetooth Microcontroller, Vest Long-time monitoring & Low power, accuracy,
vest (Lin et al., emergency moderate cost
2014)
Wearable sensing Movements Textile strain sensor Conductive yarns Gesture sensing device Flexion angle monitor Convenient to use,
device (Wilhelm and cables high precision
et al., 2014)
Smart Shirt (Tada ECG electrode Cables Smart Shirt, Holter monitor in ECG High stability,
et al., 2015) microcontroller washing comfort,
durability
Sweat monitoring ( Sweat Camera Wires smart phone, Monitoring sweat rates Low power, wired
Chen et al., microcontroller via image analysis connectivity
2015)
In-ear audio ECG Microphone Wire Microphone under HPD Mostly used in industrial User friendly,
wearable (Martin records breathing applications comfortable
et al., 2018) intensities

ECG = Electrocardiogram, BP = Blood Pressure, HR = Heart Rate, EMG = Electromyogram, BT = Body Temperature, R = Respiration, SPo2 = Pulse Oximetry Oxygen
Saturation, GSR = Galvanic Skin Response, A = Accelerometer, EEG = Electroencephalogram, PPG = Photoplethysmogram.

common arrhythmias such as atrial fibrillation, tachycardia, brady­ strip with peak-to-peak input range of 10 mV, paired with smart mobile
cardia and unclassified normal ECG beats by medical professional. (Ding known as KardiaMobile. The recorded signal is analysed for AF detection
et al., 2020) demonstrate the use of Alivecor Kardia Mobile technology and displayed in mobile interface. Doctors highly endorse it as an
shown in Fig. 11, for the detection of atrial fibrillation. The automatic excellent tool to monitor one’s heart health from the convenience of
analysis with Alivecor is performed by using single lead 30-sec rhythm one’s home, with over 20 Lakh valued consumers.
On the other hand, the textile based wearable devices such as Life
shirt, Magic IC, My Heart, Mermoth, Audabe, Wealthy, Biotex, Proetex,
Magic system etc., were specially developed to monitor the multiple
physiological signals. These devices are likely shirt’s, small garments
and on vest worn types. For instance, the “smart vest” is a stretchable,
flexible and washable shirt used to measure the BP, ECG and HR. Be­
sides, screen printable devices with embedded sensors which use
conductive links were also developed to monitor the abdominal pa­
rameters. These prototypes have great advantage that works at low cost
and helps individuals to a greater extent with high precision. The other
commonly prototypes developed based on microcontroller boards,
wireless technologies used in numerous applications to monitor the
physiological signals with higher efficiency and at lower rates. As there
are numerous devices available in the market, the higher quality life
increases the demand.
However, the emergence of wearable technology and smart health
monitoring systems has created a deluge of highly complex data, making
it challenging for researchers to analyse and interpret the data in many
circumstances. This gap limits the use of data collected from various
consumer-grade fitness applications and gadgets because it may contain
crucial information on the current state and development of one’s own
health. This problem can be encountered with the use of visual data
analytics which represents the information visually, allowing the expert
to understand it better, get deeper insight and make better data-based
decisions.
Fig. 11. Alivecor Kardia Mobile.

16
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Visualization is mainly used for the analysis of physiological signals domain techniques has been widely applied for visualization and anal­
acquired from the human body. For instance, periodogram, a frequency ysis of time-series ECG signals. However, it is challenging to interpret
domain technique (Klimov et al., 2010); spectrogram (Aguera P. E. et al., the results and requires extensive knowledge in signal processing. The­
2011) and wavelet, (Aboalayon et al., 2016) are the time–frequency meRiver (Havre et al., 2002), a method for visualizing two-dimensional

Table 10
Embedded Architectures Implementations in real time.
[Ref] Signals Acquired Communication Processing Unit Portability Purpose
technology

(Bashyam et al., ECG, Heart rate GSM AT89s52 Yes Real time supervising ECG system was developed to alert the
2017) doctors in case of any abnormal conditions through GSM.
(Varma et al., Heart Rate, GSM/GPRS LPC2148 Yes Acquired signals are lively transmitted to smart phone using
2013) Temperature GSM which in turn transmitted the data from smart phone to
server using GPRS. This proposed method was developed
mostly for the Mobile patients.
(Verma and Heart Rate, GPS, RF PIC16877A Yes A Wireless home care telemedicine was developed which
Gupta, 2012) Temperature allows the professionals to monitor the vital signs patients
remotely.
(Mahmud et al., ECG, Heart rate nRF Arduino Yes An ultra-low power & low cost architecture was developed
2015) in both contact and contact-less manners. This system can
also be used to observe the premature infants.
(Gupta et al., BP, Heart Rate, Internet/Wi-Fi Intel Galileo Development Yes To provide efficient health services to patients, an IoT based
2016) Temperature board (32-bit Intel Pentium HMS was implemented which enables the doctors to assist
Processor) the patient based on information stored in the cloud.
(Yogaraj et al., Temperature, Heat ESP8266 module Arduino Yes An E-health care monitoring system was proposed to reduce
2017) Beat, ECG the stress of the rural people to meet the doctors.
(Ibrahim et al., Temperature, Heat ESP8266 module, Arduino Yes A wearable vital monitoring system was developed that
2017) Beat, ECG GSM enables the doctors to constantly monitor the patient status
in his/her smart phone
(Hemalatha et al., BP, Temperature, Zigbee, WLAN Smart RFID tag No The proposed system uses RFID tag to monitor and store the
2017) Heartbeat, Blood sensed physiological data and transmits to local server using
glucose levels Zigbee. Later the doctor assists the patients using WLAN.
The developed system is mostly used in hospitals
(Sandeep et al., Heart rate, GSM LPC2148 Yes Proposed model enables to continuously monitor the
2016) Temperature temperature and heart rate of the patient. An alert signal will
be sent to doctor if any abnormal conditions present through
GSM.
(Rao et al., 2016) ECG Zigbee Arduino Yes Real time remote monitoring system prototype was
proposed to monitor the heart rate continuously.
(Bhoomika et al., Temperature, Oxygen GSM, ESP8266 PIC18F46k22 Yes An AES encrypted smart security framework is proposed. At
2016) levels module extreme times an alert signal will send to doctor through
GSM.
(Kumar et al., Heart rate, Inbuilt Wi-Fi Raspberry Pi Yes Physiological signals of the patient are monitored using
2016) Temperature, raspberry pi. Collected information is stored in the website
Respiration rate through inbuilt Wi-Fi module of raspberry pi.
(Joshi et al., ECG eZ430-RF MSP430 Yes The daughter board utilized in the projected system
2011) amplifies & filters the noisy ECG and sent to microcontroller
for processing. The data is then transmitted to server
through simplicity network protocol.

(Liu et al., 2013) ECG No Internet STM32 Yes A novel portable ECG monitoring was proposed using ARM
cortex M3 processor. This system uses automatic analysis
algorithms to interpret the heart rate.
(Park et al., ECG USB FPGA Yes An FPGA based cardiac monitoring is proposed. Heart rate
2016) calculation algorithm is used in smart phone app that
interprets the rate. The signals can be visually seen in
MATAB simulation environment.
(Pardesh and Heartbeat, GSM/Internet Raspberry Pi Yes The proposed method analyzed the raspberry pi-based
Sagar, 2017) Temperature, ECG health monitoring system to monitor and detect the
abnormalities if any.
(Omar et al., Body temperature Zigbee Arduino, Raspberry Pi Yes A dedicated embedded wireless transmission system was
2017) developed to transmit the data between two transceivers
over long ranges.
(Preejith et al., ECG Bluetooth ARM cortex M4 Yes Wearable ultra-low power real time supervising system has
2016) been developed for the recognition of cardiac events.
(Gay et al., 2007) ECG, oxygen levels, BP GSM, Wi-Fi Smart phone as processing Yes A personalized HMS using smart phone and different
unit wearable sensors wad developed to monitor the
physiological signals.
(Li et al., 2017) BP Bluetooth STM32 Yes Continuous Blood pressure monitoring system was
developed. This system uses simulation tool to verify the
results.
(Shaad et al., Heart rate Bluetooth RFduino Yes A real time wireless HMS was developed to measure the ECG
2017) using Mobile phone case.
(Kavya et al., ECG GSM/GPS DE2 Altera Cyclone II Yes FPGA based tele-cardiac system is developed to monitor
2014) board arrythmia.
(Jagadiswara Rao ECG, Heartbeat, Zigbee, Wi-Fi 8051, Raspberry Pi No An RTOS based supervisory system has been developed to
et al., 2016) Temperature telecast the video streaming of the patient status remotely.

17
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

time series data, is only capable of handling non-negative data values. noise brought on by magnetic fields in trays and vibrating accessories
Horizon graphs (Perin et al., 2013) can visualize both positive and (Elgendi, 2013). Processing time, high dimensional feature space and
negative values by performing color-coding but is less capable of classification accuracy are the three main considering challenging tasks
comparing time-varying values of different profiles. Seismic waves are in ECG signal analysis. Researchers use either feature selection or
plotted using wiggle traces (Emery et al., 1996), but it might be chal­ transformations methods to overcome the problems. Feature selection
lenging to compare different graphs when the coordinate axes are methods eliminate the redundant features and select the appropriate
different. Kite diagram (Garcia et al., 1997), Data Vases (Thakur et al., features set, whereas feature transformation achieves this goal by
2009), Eventogram (Elgendi M. et al., 2016), Stroscope (Cho et al., transforming the original space into a lower-dimensional subspace.
2014) are some of the visualization techniques that bridge the gap by Some approaches produce optimal results, while same may not give.
plotting the data values against temporal variables. Although, a variety One should select the approach that best meets the expectations for the
of data visualisation and interactive data inspection techniques have processing time and accuracy of the analysis because there is a trade-off
been developed for exploring physiological datasets, some techniques between the optimality of the result and the overall processing time of
might not effectively display long sequences of multidimensional data the employed method. Therefore, researchers have been interested in
values. (Damasevicius et al., 2018) provide a novel technique for two- optimization and development of traditional ECG signal processing
dimensional physiological time series visualisation based on Gramian systems.
Angular Field (GAF) representations of time series Hjorth parameters. Recently, AI techniques are typically very useful in ECG signal
The main advantage of this method is visual observation of specific interpretation. However, the integration of AI algorithms into clinical
features in 2D plot and compact representation of large series data in a practice is associated with numerous technical difficulties. The main
single image. challenge comes up with the standardization of data. The wide inter­
operability of ECG data is necessarily constrained by differences in the
6. Real time implementation with embedded architectures types of existing ECG input data, storage formats, and interpretation
statements. These formats vary depending on whether the data collected
Numerous research efforts have been undertaken over the last from ambulatory devices, bedside devices or mobile devices which could
decade to develop various real time health monitoring systems with lead to misunderstandings and differences in data analysis. The major
abnormality detection. This section provides a summary of the various limitation with AI algorithms is it requires large amount of high-quality
hardware implementations that are published in the literature as shown data to provide accurate results for developing algorithms in the man­
in Table 10. (Ying Wei et al., 2020) reviewed the pros and cons of AI- agement of rare diseases with limited sample size. Sometimes, the
based biomedical processors for hardware design and classification al­ quality of the data may be lowered by incomplete, heterogeneous, or
gorithms to design an algorithm for biomedical applications. To noisy data, which could lead to missing values, redundancy, or data
decrease the design expenditure, the processing unit utilized for the sparsity. On the other hand, AI models generally requires high end
development of good health caring systems has low-pace, lesser in-built processors to operate and have higher computational complexities.
memory and may not include any external security features. In contrast, Hence there is a trade off in designing a system.
the reliable good health care architecture should consume less electric Telehealth engages patients and clinicians with enhanced distant
potential, includes security mechanisms, and aids remote monitoring access, immediacy and accessibility over network, higher patient
and portable. engagement rates, lower hospitalization rate and better outcomes.
Despite the many benefits of leveraging telehealth, still there are certain
7. Challenges and limitations challenges and limitations in view of providers and patient. The biggest
challenges that telehealth provider is facing are the lack of significant
ECG signals obtained from individuals or taken from the databases reimbursement from Medicare and Medicaid programs, certain re­
may contain significant amounts of noise. Sometimes the signal data strictions, and regulations to offer telehealth benefits, and scalability of
may be large and massive in case of long-term monitoring. Conse­ infrastructure. If the organizations have lack of infrastructure, imple­
quently, signal processing and interpretation become relatively difficult menting and addressing telehealth benefits are extremely difficult. The
tasks. This increases the demand for automatic signal data analysis and telehealth challenges that patients often grapple with lack of awareness
interpretation. Accurate computer analysis of ECG signals is quite and technical skills to access telehealth services, security and privacy
challenging due to its low amplitude and extreme susceptibility to high issues while transmitting the patient data electronically and difficulty to
frequency noise and other distortions. The accuracy of ECG signal de­ visit every time remotely. The significant limitations with telehealth are
pends on the quality of the input data. Based on the past and current the inability to perform an in-person physical examination. The inability
research on ECG signal processing methods such as denoising, R-peak to weigh patients may result in incorrect treatments, for example
detection, signal compression, feature extraction and classification. we chemotherapy and paediatric drugs. Telehealth visits might not be
highlight the key challenges and practical issues of various signal quality suitable for all patients or all clinical circumstances. Because of the
assessment methods in this work. (Satija et al., 2017) state that a single digital barrier, some people may not have equal access to telemedicine,
signal processing method may not be adequate to eliminate all kinds of including those who live in remote areas with limited Internet connec­
noise and artifacts. Therefore, it is important to recognize the types of tivity, the elderly, and persons from different social and cultural
noise present in ECG recordings and select an appropriate signal pro­ backgrounds.
cessing methods for each type of noise. In some cases, the noise may Remote monitoring systems enables the electronic transmission of
appear during the first 10 s of the ECG signal. Therefore, it is important patient data to healthcare professionals in another location for medical
to determine the boundaries of noisy segment and eliminate them during examination. While telehealth also makes use of information and
preprocessing. communication technologies (ICT), it doesn’t employ any tools for data
Proper placement of electrodes in right position is important while collection. Remote health monitoring is specific application of tele­
taking electrocardiogram. Incorrect positioned electrodes affect the health, which makes use of certain technology to enable communication
measurement of P, QRS and T complexes in different leads and thus between doctors and patients at home. Even though remote monitoring
leads to inaccurate results. Traditional handcrafted algorithms have has numerous benefits, still there are some challenges and limitations
difficulty in detecting P-waves due to their low signal-to-noise ratio, need to be overcome to implement a successful remote health system.
which limits their ability to transform an ECG trace into a rich data Network availability and connectivity is one big challenge for remote
representation. Moreover, it is also challenging to accurately extract health systems especially when transmitting large data packets over
morphological information from ECG readings because of electrical network. For instance, network interruptions at critical times could

18
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

hamper the diagnosis when continuous monitoring of brain or heart will substantially increase in the coming years. The ongoing trend to­
conditions. Remote health systems have security and privacy issues ward long-term monitoring to improve detection performance, puts
since they are vulnerable to hacking while operating through digital pressure on healthcare systems to develop cost-effective solutions. The
networks. Security measures have been applied to remote health, to limitations associated with remote monitoring devices can be effectively
maintain a healthy clinic-patient relationship. Remote monitoring sys­ solved by keeping the device in various modes such as standy, sleep,
tems and equipment must adhere to food and drug administration (FDA) hilbernate and power save mode to reduce the power consumption and
regulation standards for usage in healthcare. FDA approval itself is a extends the battery life. For wireless connectivity solutions, wake-up
challenge to system providers as it involves the inspection of develop­ times and standby power consumption are also crucial factors. The
ment/manufacturing process, quality assurance and life cycle assess­ choice of buck-boost converters, and the package selection for wireless
ments. These process overheads and extra approval steps may lead to devices also help to reduce the size of the final product.
increased costs thus making remote monitoring systems expensive to As a future research direction, exploring the rapidly growing IoT and
adopt. Data quality is important for remote monitoring systems. These smart devices has a potential to transform the next generation ECG
systems need to periodically calibrate to maintain reasonable accuracy. monitoring systems for preventive healthcare. Most researchers use 12-
Apart from these, there are some design challenges need to be faced by lead technology for efficient diagnosis of CVDs such as atrial fibrillation,
patient monitors such as, portability, power consumption (or battery myocardial infraction, coronary artery disease etc. Till date, the ‘gold
life), patient safety and integration. Sensors and wearable devices of standard’ for assessing abnormalities of cardiac rhythm is 12-lead Hol­
different types available in various sizes have some constraints of suit­ ter, there has been an increasing demand for portable devices which
ability, usability, comfort, and convenience. Some devices may not work allow monitoring of cardiac rhythm in real-world settings such as home
near hot places, while some battery-operated devices may run out of or workplace. As like Alivecor, the integration of AI technology with 12-
power faster in cold weathers. So, the design of devices especially lead smart monitoring devices has a great potential in developing real-
wearable ones, may require revision from time to time. Patient monitors time smart health monitoring systems. Integration of robotics with
that measure vital signs use power and data isolation for patient safety. healthcare automation simplify robotic-assisted surgery for elderly and
In-home patient monitoring requires minimal integration effort to remote patients, paving the way for revolutionary healthcare of the
transfer data to cloud by enabling connectivity to a variety of cloud future.
vendors. The limitations of remote health systems include the limited To that end, we believe that this work presents a thorough, state-of-
accessibility of smart phones for elderly, lack of patient-provider the-art evaluation of ECG monitoring systems, with a detailed discussion
engagement may face difficulty of using requisite technologies. on several important research works. This paper serves as reference for
researchers in this field to understand, compare and evaluate the fea­
8. Conclusion and future work tures of ECG monitoring systems. In addition, it addresses the key
challenges faced by ECG signal analysis, AI, IoT, and telehealth moni­
With increasing world population, CVDs are the prominent occurring toring systems. Finally, it outlines a future vision of the next-generation
conditions to date. ECG is an essential tool for identifying the irregu­ ECG monitoring systems for healthcare.
larities in cardiac health conditions. Early diagnosis of heart abnor­
malities with CAD systems and AI based solutions can save many lives of CRediT authorship contribution statement
high-risk patients. This article presents a comprehensive review on
various aspects of denoising, feature engineering and features extraction Venkata Anuhya Ardeti: Data curation, Conceptualization, Writing
from traditional to AI based approaches. In addition, survey of existing – original draft, Writing – review & editing. Venkata Ratnam Kolluru:
literature on smart health ECG monitoring systems and various real-time Formal analysis, Supervision, Validation. George Tom Varghese:
hardware implementations are presented in this work. Moreover, this Visualization, Validation. Rajesh Kumar Patjoshi: Visualization,
survey addresses the challenges and limitations associated with several Validation.
ECG signal processing methods, AI based solutions, telehealth, and
remote monitoring applications. Declaration of Competing Interest
In the literature, various types of noises associated with ECG data can
be removed by applying different types of filters. Therefore, one should The authors declare that they have no known competing financial
be very careful in the selection of appropriate filter types and fre­ interests or personal relationships that could have appeared to influence
quencies since most of the databases do not include the detailed infor­ the work reported in this paper.
mation about the measuring environment. The Filter bands, cut-off and
centre frequencies of the adaptive filters can be carefully changed for the Data availability
elimination of different noises in ECG signals. Advancements in AI
technology offers a great solution for automatic diagnoses of healthcare. The authors are unable or have chosen not to specify which data has
The subfields of AI, such as the ML and DL algorithms effectively works been used.
for signal denoising, feature extraction and classification purposes.
Although ML-based approaches outperform traditional statistical Acknowledgement
methods for predicting cardiovascular events, deep learning algorithms
outperforms ML techniques in some instances. Ensemble learning might None. No funding to declare.
be an excellent option for handling multi-dimensional data. The use of
attention modules and transformer networks in deep learning make the
References
model more interpretable in diagnosing cardiac irregularities. However,
deep learning models are cost effective since it requires large amount of Abdalla, F. Y. O., Wu, L., Ullah, H., Ren, G., Noor, A., & Zhao, Y. (2019). ECG Arrhythmia
training data to produce accurate results. Classification using Artificial Intelligence and Nonlinear and Nonstationary
Decomposition. Signal, Image, and Video Processing, 13(7), 1283–1291. https://doi.
Enabling technologies in healthcare research offer huge opportu­
org/10.1007/s11760-019-01479-4
nities for the development of ECG monitoring systems. IoT enables Aboalayon, K. A. I., Faezipour, M., Almuhammadi, W. S., & Moslehpour, S. (2016). Sleep
remote, unconstrained connectivity and services that leverage data to Stage Classification Using EEG Signal Analysis: A Comprehensive Survey and New
assist timely, significant, and crucial decisions for a better lifestyle. With Investigation. Entropy, 2016(18), 272.
Aditi, S., & Nurmi, J. (2021). Systematic Reviews on Machine-Learning Algorithms used
the development and widespread use of sensors and wearables like in Wearable-Based eHealth Data Analytics. IEEE Access, 9, 112221–112235. https://
smartwatches, the amount of ECG data available for diagnostic analysis doi.org/10.1109/ACCESS.2021.3103268

19
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Afkhami, R. G., Azarnia, G., & Tinati, M. A. (2016). Cardiac Arrhythmia Classification Chang, K. C., Hsieh, P.-H., Wu, M.-Y., Wang, Y.-C., Chen, J.-Y., Tsai, F.-J., et al. (2020).
using Statistical and Mixture Modeling Features of ECG Signals. Pattern Recognition Usefulness of Machine-Learning-Based Detection and Classification of Cardiac
Letter, 70, 45–51. https://doi.org/10.1016/j.patrec.2015.11.018 Arrhythmias with 12-Lead Electrocardiograms. Canadian Journal of Cardiology., 37
Afsar, F. A., Arif, M., & Yang, J. (2008). Detection of ST segment deviation episodes in (1), 94–104. https://doi.org/10.1016/j.cjca.2020.02.096
ECG using KLT with an ensemble neural classifier. Physiological Measurement., 29(7), Che, C., Zhang, P., Zhu, M., Qu, Y., & Jin, B. (2021). Constrained Transformer Network
747–760. https://doi.org/10.1088/0967-3334/29/7/004 for ECG Signal Processing and Arrhythmia Classifcation. BMC Medical Informatics
Aguera, P. E., Jerbi, K., Caclin, A., & Bertrand, O. (2011). ELAN: A software package for and Decision Making, 21, 184.
analysis and visualization of MEG, EEG, and LFP signals. Comput. Intell. Neurosci., Chen, G. (2015). Algorithm for Heart Rate Extraction in a Novel Wearable Acoustic
158970. Sensor. Health Care Technology Letters, 2, 28–33.
Ahmad, A. S. S., & Salah, M. (2018). Features Optimization for ECG Signals Chen, C., Hua, Z., Zhang, R., Liu, G., & Wen, W. (2020). Automated Arrhythmia
Classification. International Journal of Advanced Computer Science and Applications, 9 Classification Based on a Combination Network of CNN and LSTM. Biomedical Signal
(11). https://dx.doi.org/10.14569/IJACSA.2018.091154. Processing and Control., 57(101819). https://doi.org/10.1016/j.bspc.2019.101819
Ahmed, S., et al. (2018). A Deep Learning Approach for Vital Signs Compression and Chhabra, M., & Kalsi, M. (2017). Real Time ECG monitoring system based on Internet of
Energy Efficient Delivery in mhealth Systems. IEEE Access, 6. https://doi.org/ Things (IoT). International Journal of Scientific and Research Publications, 7(8),
10.1109/ACCESS.2018.2844308 547–550.
Alcaraz, R., Hornero, F., & Rieta, J. J. (2013). Dynamic Time Warping Applied to Chi, Y. M., & Jung, T. P. (2010). Dry-Contact and Noncontact Bio-potential Electrodes:
Estimate Atrial Fibrillation Temporal Organization from the Surface Methodological Review. IEEE Reviews in Biomedical Engineering, 3, 106–119. https://
Electrocardiogram. Medical Engineering & Physics, 35(9), 1341–1348. https://doi. doi.org/10.1109/RBME.2010.2084078
org/10.1016/j.medengphy.2013.03.004 Cho, B., Kim, H., Bae, J., & Seo, J. (2014). Stroscope: Multi-Scale Visualization of
Alcaraz, R., & Sandberg, F. (2011). Classification of Paroxysmal and Persistent Atrial Irregularly Measured Time-Series Data. IEEE Transactions on Visualization and
Fibrillation in Ambulatory ECG Recordings. IEEE Transactions on Biomedical Computer Graphics., 20(5), 808–821.
Engineering, 58, 1441–1449. https://doi.org/10.1109/tbme.2011.2112658 Chua, T. W., & Tan, W. W. (2011). Non-singleton genetic fuzzy logic system for
Amon, M., & Jager, F. (2016). Electrocardiogram ST-segment morphology delineation Arrhythmias Classification. Engineering Applications of Artificial Intelligence, 24(2),
method using orthogonal transformations. PloS One, 11(2), e0148814. 251–259. https://doi.org/10.1016/j.engappai.2010.10.003
Anliker, U., & Ward, J. A. (2004). AMON: A Wearable Multi-Parameter Medical Chui, K. T., & Tsang, K. F. (2016). An Accurate ECG-Based Transportation Safety
Monitoring and Alert System. IEEE Transactions on Information & Technology in Drowsiness Detection Scheme. IEEE Transactions on Industrial Informatics, 12,
Biomedicine., 8(4), 415–427. https://doi.org/10.1109/TITB.2004.837888 1438–1452. https://doi.org/10.1109/TII.2016.2573259
Anuhya, A. V., Venkata Ratnam, K., George Tom, V., & Rajesh Kumar, P. (2022). An Chung, W. Y., Lee, S. C., & Toh, S. H. (2008). WSN Based Mobile U-Healthcare System
Outlier Detection and Feature Ranking based Ensemble Learning for ECG Analysis. with ECG, Blood Pressure Measurement Function. Proceedings 28th Annual
International Journal of Advanced Computer Science and Applications, 13(6). https International IEEE Engineering in Medicine and Biology Society Conference. 1533-
://doi.org/10.14569/IJACSA.2022.0130686. 1536, https://doi.org/10.1109/IEMBS.2008.4649461.
Anuhya, A. V., Venkata Ratnam, K., & Rajesh, K. P. (2020). A Novel Adaptive S- Clifford, G. D., Liu, C., Moody, B., Lehman, L.-W. H., Silva, I., Li, Q. Mark, R. G. (2017).
Transform based Electrocardiogram Signal Analysis of QRS-Peak Detection. Journal AF classification from a short single lead ECG recording: The PhysioNet/Computing
of Green Engineering, 10(9), 5530–5547. in Cardiology Challenge 2017. Computing in Cardiology, 44. https://doi.org/
Avram, R., Ramsis, M., Cristal, A. D., et al. (2021). Validation of an algorithm for 10.22489/CinC.2017.065-469.
continuous monitoring of atrial fibrillation using a consumer smartwatch. Heart Constant, N. Douglas-Prawl, O. Johnson, S., and Mankodiya, K. (2015). Pulse Glasses:An
Rhythm, 18(9), 1482–1490. https://doi.org/10.1016/j.hrthm.2021.03.044 unobtrusive, wearable HR monitor with Internet-of-things functionality. IEEE 12th
Bahrami, M., & Forouzanfar, M. (2022). Sleep Apnea Detection from Single-lead ECG: A International Conference on Wearable Implant Body Sensor Networks, BSN, 1-5,
Comprehensive Analysis of Machine Learning and Deep Learning Algorithms. IEEE https://doi.org/10.1109/BSN.2015.7299350.
Transactions on Instrumentation and Measurement, 71. https://doi.org/10.1109/ Coyle, S. (2010). BIOTEX-Biosensing Textiles for Personalized Healthcare Management.
TIM.2022.3151947 IEEE Transactions on Information Technology & Biomedicine, 14(2), 364–370. https://
Banerjee, S., & Mitra, M. (2013). Application of cross wavelet transform for ECG pattern doi.org/10.1109/TITB.2009.2038484
analysis and classification. IEEE transactions on instrumentation and measurement, 63 Curone, D., Secco, E. L., Tognetti, A., Loriga, G., Dudnik, G., Risatti, M., et al. (2010).
(2), 326–333. https://doi.org/10.1109/tim.2013.2279001 Smart Garments for Emergency Operators: The Proetex Project. IEEE Transactions on
Banerjee, S., & Mitra, M. (2014). Application of Cross-Wavelet transform for ECG Pattern Information Technology in Biomedicine: A Publication of the IEEE Engineering in Medicine
Analysis and Classification. IEEE Transactions on Instrumentation Measurements., 63, and Biology Society, 14(3), 694–701. https://doi.org/10.1109/TITB.2010.2045003
326–333. https://doi.org/10.1109/TIM.2013.2279001 Curtin, A. E., Burns, K. V., Bank, A. J., & Netoff, T. I. (2018). QRS complex detection and
Bashyam, S. D. B. (2017). Development of Real-Time ECG Signal Monitoring System for measurement algorithms for multichannel ECGs in cardiac resynchronization
Telemedicine Application. 3rd International Conference on Bio-signals, Images, and therapy patients. IEEE Journal of Translational Engineering in Health and Medicine, 6,
Instrumentation, 16-18, https://doi.org/10.1109/ICBSII.2017.8082285. 1900211. https://doi.org/10.1109/JTEHM.2018.2844195
N. Bayasi T. Tekeste H. Saleh Low-Power ECG-Based Processor for Predicting Ventricular Damasevicius, R., Rytis, M., Wozniak, M., & Poáap, D. (2018). In Visualization of
Arrhythmia IEEE Transactions on Very Large-Scale Integration Systems, https:// Physiologic Signals based on Hjorth Parameters and Gramian Angular Fields. Slovakia:
2016 doi.org/a0.1109/TVLSI.2015.2475119. Kosice and Herlany. https://doi.org/10.1109/SAMI.2018.8323992.
Baygin, M., Tuncer, T., Dogan, S., Tan, R., & Acharya, R. U. (2021). Automated M.K. Das Electrocardiogram Signal Analysis for Heartbeat Pattern Classification 2015
Arrhythmia Detection with Homeomorphically Irreducible Tree Technique using http://ethesis.nitrkl.ac.in/6662/1/Manab_Kumar_Das(Ph.D)_510EC305.
more than 10,000 Individual Subject ECG Records. Information Sciences., 575, Dewangan, N. K. (2017). A Survey on ECG Signal Feature Extraction and Analysis
323–337. Techniques. Biomedical Signal Processing. https://. https://doi.org/10.17148/
C. Belal K. Rajiv RFID Based Hospital Real Time Patient Monitoring System IEEE IJIREEICE.2015.3603
International Conference on Computer and Information Science 2017. Dilmac, S., & Korurek, M. (2015). ECG Heartbeat Classification Method based on
Berkaya, S. K., Uysal, A. K., Gunal, E. S., & Ergin, S. (2018). A Survey on ECG Analysis. Modified ABC Algorithm. Applied Soft Computing, 36, 641–655. https://doi.org/
Biomedical Signal Processing and Control, 43, 216–235. https://doi.org/10.1016/j. 10.1016/j.asoc.2015.07.010
bspc.2018.03.003 Dima, S. M., & Panagiotou, C. (2013). On the Detection of Myocardial Based on ECG/
Berwal, D., Vandana, C. R., Dewan, S., Jiji, C. V., & Baghini, M. S. (2019). Motion Artifact VCG Analysis. IEEE Transactions in Biomedical Engineering, 60, 3399–3409. https://
Removal in Ambulatory ECG Signal for Heart Rate Variability Analysis. IEEE Sensors doi.org/10.1109/tbme.2013.2279998
Journal (Volume: 19, Issue: 24). https://doi.org/10.1109/JSEN.2019.2939391 Ding, E. Y., Marcus, G. M., & McManus, D. D. (2020). Emerging Technologies for
Besterman, E., & Creese, R. (1979). Waller-Pioneer of Electrocardiography. Br. Heart Identifying Atrial Fibrillation. Circulation Research, 127, 128–142. https://doi.org/
Journal, 42(1), 61–64. 10.1161/CIRCRESAHA.119.316342
Bhoomika, D. B. (2015). Secured Smart Healthcare Monitoring System based on IoT. Donoso, F. I., & Figueroa, R. L. (2013). Atrial Activity Selection for Atrial Fibrillation
International Journal on Recent and Innovation Trends in Computing and ECG Recordings. Computers in Biology and Medicine, 43, 1628–1636. https://doi.org/
Communication, 3(7), 4958–4961. 10.1016/j.compbiomed.2013.08.002
Bousseljot, R., Kreiseler, D., & Schnabel, A. (2009). Nutzung der EKG-Signaldatenbank Ebrahimzadeh, A., Shakiba, B., & Khazaee, A. (2014). Detection of Electrocardiogram
CARDIODAT der PTB uber das Internet. Biomedizinische Technik. Biomedical Signals using an Efficient Method. Applied Soft Computing, 22, 108–117. https://doi.
engineering, 317–318. https://doi.org/10.1515/bmte.1995.40.s1.317 org/10.1016/j.asoc.2014.05.003
S.C. Bulusu M. Faezipour V. Ng M. Nourani L.S. Tamil S. Banerjee Transient ST-segment Egle Butkeviciute; Liepa Bikulciene; Tatjana Sidekerskiene; Tomas Blazauskas; Rytis
episode detection for ECG beat classification 2011 IEEE/NIH Life Science Systems Maskeliunas. Removal of Movement Artefact for Mobile EEG Analysis in Sports
and Applications Workshop (LiSSA) (2011, April). Bethesda, MD, USA 10.1109/ Exercises IEEE Access 7 2019 7206 7217 10.1109/ACCESS.2018.2890335.
lissa.2011.5754171. Elgendi, M. (2016). Eventogram: A Visual Representation of Main Events in Biomedical
Carlos, L., & Shmaliy, Y. S. (2019). Denoising and Features Extraction of ECG Signals in Signals. Bioengineering, 3(22).
State Space Using Unbiased FIR Smoothing. IEEE Access, 7. https://doi.org/10.1109/ Elgendi, M., & Talkachova, A. (2013). Fast QRS Detection with an Optimized Knowledge-
ACCESS.2019.2948067 Based Method: Evaluation on 11 Standard ECG Databases. Plos One, 8(9). https://
Castroflorio, T., & Mesin, L. (2013). Use of Electromyographic and Electrocardiographic doi.org/10.1371/journal.pone.0073557
Signals to Detect Sleep Bruxism Episodes in a Natural Environment. IEEE Journal in Elizabeth, A. V., & Sreekumar, K. (2017). A Survey on Approaches for ECG Signal
Biomedical Health, 17, 994–1001. https://doi.org/10.1109/JBHI.2013.2274532 Analysis with Focus to Feature Extraction and Classification. International Conference
Celin, S., & Vasanth, K. (2018). ECG Signal Classification Using Various Machine on Inventive Communication and Computational Technologies, IEEE,. https://doi.org/
Learning Techniques. Journal of Medical Systems, 42(12). https://doi.org/10.1007/ 10.1109/ICICCT.2017.7975175
s10916-018-1083-6

20
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Emery, D., and Myers, K. (1996). (eds.): Sequence Stratigraphy. Blackwell Publishing, Hu, R., Chen, J., & Zhou, L. (2022). A Transformer-Based Deep Neural Network for
Malden, 1996. Arrhythmia Detection using Continuous ECG Signal. Computers in Biology and
Exarchos, T. P., Tsipouras, M. G., Exarchos, C. P., Papaloukas, C., Fotiadis, D. I., & Medicine, 144, Article 105325.
Michalis, L. K. (2007). A Methodology for the Automated Creation of Fuzzy Expert Huang, H., & Hu, (2020). G. Ensemble of Support Vector Machines for Heartbeat
Systems for Ischaemic and Arrhythmic Beat Classification based on a Set of Rules Classification. Proceedings of 10th International Conference on Signal Processing,
Obtained by a Decision Tree. Artificial Intelligence in Medicine, 40(3), 187–200. Beijing, China, https://doi.org/10.1109/ICOSP.2010.5657034.
https://doi.org/10.1016/j.artmed.2007.04.001 Ibrahim, A. A., Yasin, M., & Wang, Z. (2017). IOT Patient Health Monitoring System.
Faezipour, M., Saeed, A., Bulusu, S. C., Nourani, M., Minn, H., & Tamil, L. (2010). International Journal of Engineering Research and Application, 7(10), 1–3. https://doi.
A patient-adaptive profiling scheme for ECG beat classification. IEEE Transactions on org/10.9790/9622-0710070103
Information Technology in Biomedicine: A Publication of the IEEE Engineering in Medicine Zephyr Incorporation. (2009). Available http://www.zephyrtech.co.nz/.
and Biology Society, 14(5), 1153–1165. https://doi.org/10.1109/ Jafarian, K., Vahdat, V., Salehi, S., & Mobin, M. (2020). Automating Detection and
TITB.2010.2055575 Localization of Myocardial Infarction using Shallow and End-To-End Deep Neural
Fang, S. C., & Chan, H. L. (2013). QRS Detection-Free Electrocardiogram Biometrics in Networks. Applied Soft Computing, 93. https://doi.org/10.1016/j.asoc.2020.106383
their Constructed Phase Space. Pattern Recognition Letters, 34(5), 595–602. https:// Jager, F., Moody, G. B., (1998). Taddei, A., Antolic, G., Emdin, M., Smrdel, A. Mark, R. G.
doi.org/10.1016/j.patrec.2012.11.005 (2002). A long-term ST database for development and evaluation of ischemia
Fensli, R., Gunnarson, E., & Gundersen, T. (2005). A Wearable ECG-Recording System for detectors. Computers in Cardiology 1998. 25, Cleveland, OH, USA. https://doi.org/
Continuous Arrhythmia Monitoring in a Wireless Tele-Homecare Situation. 10.1109/cic.1998.731793.
Proceedings 18th IEEE Symposium Computer-Based Medical Systems. 407-412, Jen, K., & Hwang, Y. (2008). ECG Feature Extraction and Classification using Cepstrum
https://doi.org/10.1109/CBMS.2005.22. and Neural Networks. Journal of Medical and Biological Engineering, 28(1).
Firoozabadi, R., Gregg, R., & Babaeizadeh, S. (2019). Modelling and classification of the Jha, C. K., & Kolekar, M. H. (2020). Cardiac Arrhythmia Classification using Tunable Q-
ST segment morphology for enhanced detection of acute myocardial infarction. 2019 Wavelet Transform based Features and Support Vector Machine Classifier.
Computing in Cardiology Conference (CinC). https://doi.org/10.22489/ Biomedical Signal Processing and Control, 59(101875), Article 101875. https://doi.
cinc.2019.005. org/10.1016/j.bspc.2020.101875
Fu, L., Lu, B., Nie, B., Peng, Z., Liu, H., & Pi, X. (2020). Hybrid Network with Attention Jiang, M., Gu, J., Li, Y., Wei, B., Zhang, J., Wang, Z., et al. (2021). HADLN: Hybrid
Mechanism for Detection and Location of Myocardial Infraction Based on 12-Lead Attention-Based Deep Learning Network for Automated Arrhythmia Classification.
Electrocardiogram Signals. Sensors (Basel), 20(4). https://doi.org/10.3390/ Front Physiology, 12, Article 683025.
s20041020 Z. Jin J. Oresko S. Huang A.C. Cheng HeartToGo: A Personalized Medicine Technology
Fujita, N., Sato, A., & Kawarasaki, M. (2015). Performance study of wavelet-based ECG for Cardiovascular Disease Prevention and Detection 2009 IEEE/NIH Life Science
analysis for ST-segment detection. 2015 38th International Conference on Systems and Applications Workshop. LSSA & (2009, April). Bethesda, USA 80 83
Telecommunications and Signal Processing (TSP). Presented at the 2015 38th 10.1109/lissa.2009.4906714.
International Conference on Telecommunications and Signal Processing (TSP), Joshi, A., Ravindran, S., & Miller, A. (2011). S. EKG-Based Heart-Rate Monitor
Prague, Czech Republic. https://doi.org/10.1109/tsp.2015.7296298. Implementation on the Launch Pad Using MSP430G2xx. Texas Instruments
Garcia, S. (1997). Indicators for Sustainable Development of Fisheries in ‘Land quality Application report, SLAA486B, Available at https://www.ti.com/lit/an/slaa486b/
indicators and their use in sustainable agriculture and rural development. FAO Land slaa486b.pdf?ts=1649496136373.
and Water Bulletin, 5(213). Jyothi, P., & Pradeepini, G. (2021). Review on Cardiac Arrhythmia Through
Gay, V. (2007). A Health Monitoring System using Smart Phones and Wearable Sensors. Segmentation Approaches in Deep Learning. Advances in Intelligent Systems and
International Journal of ARM., 8(2). Computing., 139–147. https://doi.org/10.1007/978-981-33-6176-8_15
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., & Ivanov, P. C. (2001). Predicting Kamaleswaran, R., Mahajan, R., & Akbilgic, O. (2018). A Robust Deep Convolutional
paroxysmal atrial fibrillation/flutter: The PhysioNet/Computing in cardiology Neural Network for the Classification of Abnormal Cardiac Rhythm using Single
challenge 2001. Computing. in Cardiology. Lead Electrocardiograms of Variable Length. Physiological Measurement, 39(3),
Goovaerts, G., Padhy, S., et al. (2019). A Machine-Learning Approach for Detection and Article 035006. https://doi.org/10.1088/1361-6579/aaaa9d
Quantification of QRS Fragmentation. IEEE Journal of Biomedical and Health Karpagachelvi, S., Dr, M., & Arthanari, M. (2010). ECG Feature Extraction Techniques- A
Informatics, 23(5), 1980–1989. https://doi.org/10.1109/JBHI.2018.2878492 Survey Approach. International Journal of Computer Science and Information Security, 8
Gowri, T., Swomya, I., Rahman, Z. U., & Dodda, R. K. R. (2013). Adaptive powerline (1).
interference removal from cardiac signals using Leaky based normalized higher Katsis, C. D., Ganiatsas, G., & Fotiadis, D. I. (2006). An Integrated Telemedicine Platform
order filtering techniques. Proceedings - 1st International Conference on Artificial for the Assessment of Affective Physiological States. Diagnostic Pathology, 1(1), 16.
Intelligence, Modelling and Simulation, 294-298, AIMS. https://doi.org/10.1109/ https://doi.org/10.1186/1746-1596-1-16
AIMS.2013.54. Kavya, G., & Thulasibai, V. (2014). VLSI Implementation of Tele-monitoring System for
Gupta, P., & Agarwal, D. (2016). IoT based Smart HealthCare Kit. International Conference High-Risk Cardiac Patients. Indian Journal of Science and Technology, 7(5), 571–576.
on Computational Techniques in Information and Communication Technologies. https:// https://dx.doi.org/10.17485/ijst/2014/v7i5.13.
doi.org/10.1109/ICCTICT.2016.7514585 Kim, Y., Lee, S., & Lee, S. (2016). Coexistence of ZigBee based WBAN and Wi-Fi for
Gutta, S., & Cheng, Q. (2016). Joint Feature Extraction and Classifier Design For ECG- Health Tele-monitoring Systems. IEEE Journal of Biomedical & Health. Information,
Based Biometric Recognition. IEEE Journal Biomedical Health, 20, 460–468. https:// 20(1). https://doi.org/10.1109/JBHI.2014.2387867
doi.org/10.1109/JBHI.2015.2402199 Kinnunen, H., Rantanen, A., Kenttä, T., & Koskimäki, H. (2020). Feasible assessment of
Habetha, J. (2006). The MyHeart Project-Fighting Cardiovascular Diseases by Prevention recovery and cardiovascular health: Accuracy of nocturnal HR and HRV assessed via
and Early Diagnosis. Proceedings 28th Annual International IEEE Engineering in ring PPG in comparison to medical grade ECG. Physiological Measurement, 41(4),
Medicine and Biology Society Conference, 6746-6749. 04NT01. https://doi.org/10.1088/1361-6579/ab840a
Hadjem, M., Nait-Abdesselam, F., & Khokhar, A. (2016, September). ST-segment and T- Klimov, Y. S., & Taieb-Maimon, M. (2010). Intelligent visualization and exploration of
wave anomalies prediction in an ECG data using RUSBoost. 2016 IEEE 18th time-oriented data of multiple patients. Artificial Intelligence in Medicine, 49, 11–31.
International Conference on e-Health Networking, Applications and Services Kumar, M., Pachori, R. B., & Acharya, U. R. (2017). Characterization of coronary artery
(Healthcom), Munich, Germany, https://doi.org/10.1109/ disease using flexible analytic wavelet transform applied on ECG signals. Biomedical
healthcom.2016.7749493. signal processing and control, 31, 301–308. https://doi.org/10.1016/j.
Hamad, A. K. S. (2021). New Technologies for Detection and Management of Atrial bspc.2016.08.018
Fibrillation. Journal of Saudi Heart Association, 33(2), 169–176. https://doi.org/ Kumar, R., & Pallikonda Rajasekaran, D. M. (2016). Raspberry Pi Based Patient Health
10.37616/2212-5043.1256 Status Observing Method Using Internet of Things. International Conference on
Haroon, A. W. (2016). Constraints in the IoT: The World in 2020 and beyond. Current Research in Engineering Science and Technology., 101–105.
International Journal of Advanced Computer Science and Applications, 7(11. https:// Kumar, A., & Singh, M. (2016). Ischemia Detection using Isoelectric Energy Function.
doi.org/10.14569/IJACSA.2016.071133 Computers in Biology and Medicine, 68, 76–83. https://doi.org/10.1016/j.
Haseena, H. H., Mathew, A. T., & Paul, J. K. (2011). Fuzzy Clustered Probabilistic and compbiomed.2015.11.002
Multilayered Feed Forward Neural Networks for Electrocardiogram Arrhythmia Kutlu, Y., & Kuntalp, D. (2012). Feature Extraction for ECG Heartbeats using Higher
Classification. Journal of Medical Systems, 35, 179–188. https://doi.org/10.1007/ Order Statistics of WPD Coefficients. Computer Methods and Programs in Biomedicine,
s10916-009-9355-9 105, 257–267. https://doi.org/10.1016/j.cmpb.2011.10.002
Havre, B., Hetzler, and Nowell, L. (2002). Themeriver (tm): in search of trends, patterns, Laguna, P., Mark, R. G., Goldberg, A., & Moody, G. B. (2002). A database for evaluation
and relationships. IEEE Trans Visualization Computer Graphics, 8:9-20. of algorithms for measurement of QT and other waveform intervals in the ECG.
He, C.h., & Fan, X. (2013). Toward Ubiquitous Healthcare Services with a Novel Efficient Computers in Cardiology 1997, Lund, Sweden. https://doi.org/10.1109/
Cloud Platform. IEEE Transactions on Biomedical Engineering, 60(1). https://doi.org/ cic.1997.648140.
10.1109/TBME.2012.2222404 P. Leijdekkers V. Gay A Self-Test to Detect a Heart Attack using a Mobile Phone and
K.J. Heilman S.W. Porges Accuracy of the Lifeshirt (R) Vivometrics in the Detection of Wearable Sensors 2008 21st IEEE International Symposium on Computer-Based
Cardiac Rhythms Biology in Psychology 75 3 2007 300 305 https://doi.org/0.1016/ Medical Systems 2008 Finland 93 98 10.1109/cbms.2008.59.
j.biopsycho.2007.04.001. Li, J., Gao, M., Zhang, M., et al. (2020). Treatment of atrial fibrillation: a comprehensive
Hemalatha, R., & Pavani, B. (2017). IoT Based Patient Monitoring System. International review and practice guide, 31(3), 153–158. https://doi.org/10.5830/CVJA-2019-064
Journal of Computing, Communications & Instrumentation Engineering, 4(1). https://doi Li, Q. J., & Li, R. (2018). Design of a Continuous Blood Pressure Measurement System
.org/10.15242/IJCCIE.H1216012. based on Pulse Wave and ECG Signals. IEEE Journal of Translational Engineering.
Hirose, A. (2003). Complex-valued Neural Networks: Theories and Applications. World Health and Medicine, 6.
Scientific Publishing Company Incorporated, 5. https://doi.org/10.1142/5345

21
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Li, Q., Rajagopalan, C., & Clifford, G. D. (2014). A Machine Learning Approach to Multi- Mehta, S., & Lingayat, N. (2007). Comparative study of QRS detection in single lead and
Level ECG Signal Quality Classification. Computer Methods and Programs in 12-lead ECG based on entropy and combined entropy criteria using support vector
Biomedicine, 117(3), 435–447. https://doi.org/10.1016/j.cmpb.2014.09.002 machine. Journal of Theoretical & Applied Information Technology, 3(2).
Li, Q., Rajagopalan, C., & Clifford, G. D. (2016). Ventricular Fibrillation and Tachycardia Merdjanovska, E., & Rashkovka, A. (2022). Comprehensive Survey on Computational
Classification using a Machine Learning Approach. IEEE Transactions on Biomedical ECG Analysis: Databases, Methods and Applications. Expert Systems with Applications,
Engineering, 61, 1607–1613. https://doi.org/10.1109/TBME.2013.2275000 203. https://doi.org/10.1016/j.eswa.2022.117206
Li, H., & Yuan, D. (2017). Genetic Algorithm for the Optimization of Features and Neural Meriggi, P. (2010). Polysomnography in Extreme Environments: The MagIC Wearable
Networks in ECG Signal Classification. Scientific Reports, Article, 41011. https://doi. System for Monitoring Climbers at Very-High Altitude on Mt. Everest slopes.
org/10.1038/srep41011 Proceedings in Computer Cardiology, 37, 1087–1090.
Li, Z., Zhou, D., Wan, L., Li, J., & Mou, W. (2020). Heartbeat Classification using Deep Merone, M., & Soda, P. (2017). ECG Databases for Biometric Systems: A Systematic
Residual Convolutional Neural Network from 2-Lead Electrocardiogram. Journal of Review. Expert System with Applications, 189–202. https://doi.org/10.1016/j.
Electrocardiology, 58, 105–112. https://doi.org/10.1016/j.jelectrocard.2019.11.046 eswa.2016.09.030
Liang, T., & Yong, J. (2016). Wearable medical monitoring systems based on wireless Milsis, A., & Katsaras, T. (2011). Clinical Effectiveness of the ‘Healthwear’ Wearable
networks: A review. IEEE Sensors Journal, 16(23), 8186–8199. System in the Reduction of COPD Patients Hospitalization, Wireless Mobile
Liao, Y. T., Yao, H., Lingley, A., Parviz, B., & Otis, B. P. (2012). A 3-μW CMOS Glucose Communication and Healthcare. Kos Island. Greece. Springer., 54–60. https://doi.org/
Sensor for Wireless Contact-Lens Tear Glucose Monitoring. IEEE Journal on Solid- 10.1007/978-3-642-29734-2_8
State Circuits, 47(1), 335–344. https://doi.org/10.1109/JSSC.2011.2170633 Monton, E., & Hernandez, J. F. (2008). Body Area Network for Wireless Patient
Liliana, H., Grazyna, G., & Bitkowska, A. (2021). The Research on Patient Satisfaction Monitoring. Telemedicine E-Health Communication Systems, 2, 215–222. https://doi.
with Remote Healthcare Prior to and during the COVID-19 Pandemic. International org/10.1049/iet-com:20070046
Journal of Environmental Research and Public health., 18(10), 5338. https://doi.org/ Moody, G. B., & Mark, R. G. (2001). The impact of the MIT-BIH arrhythmia database.
10.3390/ijerph18105338 IEEE Engineering in Medicine and Biology Magazine: The Quarterly Magazine of the
W.-Y. Lin M.-Y. Lee W.C. Chou The design and development of a wearable posture Engineering in Medicine & Biology Society, 20(3), 45–50. https://doi.org/10.1109/
monitoring vest 2014 IEEE International Conference on Consumer Electronics 51.932724
(ICCE), Las Vegas 2014 NV, USA 10.1109/icce.2014.6776027. Moody, G. B., Muldrow, W., & Mark, R. G. (1984). A Noise Stress Test for Arrhythmia
J. Liu Y. Zhou Design of A Novel Portable ECG Monitor for Heart Health 2013 IEEE Sixth Detectors. Computers in cardiology, 11(3), 381–384. https://doi.org/10.13026/
International Symposium on Computational Intelligence and Design 10.1109/ C2HS3T
ISCID.2013.178. G.B. Moody The Physionet/Computers in Cardiology challenge 2008: T-wave alternans.
Liu, F., Liu, C., Zhao, L., Zhang, X., Wu, X., Xu, X., et al. (2018). An open access database Computers in Cardiology 2008 Bologna 10.1109/cic.2008.4749089.
for evaluating the algorithms of electrocardiogram rhythm and morphology Moses, D., & Deisy, C. (2015). A Survey of Data Mining Algorithms used in
abnormality detection. Journal of Medical Imaging and Health Informatics, 8(7), Cardiovascular Disease Diagnosis from Multi-Lead ECG Data. Kuwait Journal of
1368–1373. Science, 42(2), 206–235.
Liu, Z., Wang, H., Gao, Y., & Shi, S. (2021). Automatic Attention Learning Using Neural Mostayed, A., Luo, J., Shu, X., & Wee, W. (2018). Classification of 12-lead ECG signals
Architecture Search for Detection of Cardiac Abnormality in 12-Lead ECG. IEEE with bi-Directional LSTM network. Computer Vision and Pattern Recognition. https://.
Transactions on Instrumentation and Measurement, 70. https://doi.org/10.1109/ https://doi.org/10.48550/arXiv.1811.02090
TIM.2021.3109396 Mousavi, S. H., Hijmans, J. M., Rajabi, R., et al. (2019). Kinematic risk factors for lower
Lubitz, A. A., Faranesh, A. Z., Atlas, A. J., et al. (2021). Rationale and design of a large limb tendinopathy in distance runners: A systematic review and meta-analysis. Giant
population study to validate software for the assessment of atrial fibrillation from & Posture, 69, 13–24. https://doi.org/10.1016/j.gaitpost.2019.01.011
data acquired by a consumer tracker or smartwatch: The Fitbit heart study. American Mrunalinee, P., & Smruti, C.h. (2016). Smart Healthcare using NFC. International Journal
Heart Journal, 238, 16–26. https://doi.org/10.1016/j.ahj.2021.04.003 of Innovative Research in Science and Engineering, IEEE,. https://doi.org/10.1109/
Luprano, J. (2006). European Projects on Smart Fabrics, Interactive Textiles: Sharing ICIS.2007.159
Opportunities and Challenges. Proceedings of Workshop on Wearable Technology Mundt, C. W., & Montgomery, K. N. (2005). A Multi-Parameter Wearable Physiological
Intel Textiles, Helsinki, Finland. Monitoring System for Space and Terrestrial Applications. IEEE Transactions on
Madeiro, D. V. J. P., Lobo Marques, J. A., Han, T., & Coury Pedrosa, R. (2020). Information Technology & Biomedicine, 9(3), 382–383. https://doi.org/10.1109/
Evaluation of mathematical models for QRS feature extraction and QRS morphology titb.2005.854509
classification in ECG signals. Measurement: Journal of the International Measurement G.R. Naik K.A. Reddy Classification of ECGs using Intensity Weighted Fire-Fly
Confederation, 156(107580). https://doi.org/10.1016/j.measurement.2020.107580 Optimization and Multimodal Decision Learning algorithm 2016 2nd IEEE
Maedeh, S., & Shahbahrami, K. (2012). A. Classification of ECG Arrhythmias using International Conference on Computer and Communications (ICCC) 2016 Chengdu.
Wavelet Transform and Neural Network. International Journal of Computer Science, China 10.1109/compcomm.2016.7924812.
Engineering and Applications, 2(1). Naz, M., Shah, J. H., Khan, M. A., et al. (2021). From ECG Signals to Images: A
Mahmud, S., & Wang, H. (2015). An Inexpensive and Ultra-Low Power Sensor Node for Transformation based Approach for Deep Learning. Peer J Computer Science, 7.
Wireless Health Monitoring System. 17th International Conference on E-health https://doi.org/10.7717/peerj-cs.386
Networking, Application & Services, https://doi.org/10.1109/ Neha, S., Anand, P., Anand, P., S., & Sanjeev, Ch, N. (2022). ECG-ViT: A Transformer-
HealthCom.2015.7454553. Based ECG Classifier for Energy-Constraint Wearable Devices. Journal of Sensors,
Mak, J. N. F., Hu, Y., & Luk, K. D. K. (2010). An automated ECG-artifact removal method Article: 2449956, https://doi.org/10.1155/2022/2449956.
for Trunk Muscle Surface EMG Recordings. Medicine in Engineering and Physics, 32, Neha, A., & Mishra, B. (2021). Origins of ECG and Evolution of Automated DSP
840–848. https://doi.org/10.1016/j.medengphy.2010.05.007 Techniques: A Review. IEEE. Access., 9. https://doi.org/10.1109/
Mar, T., Zaunseder, S., & Martinez, J. P. (2012). Optimization of ECG classification by access.2021.3119630
means of Feature Selection. IEEE Transactions in Biomedical Engineering, 58. https:// Nejadgholi, I., Moradi, M. H., & Abdolali, F. (2011). Using Phase Space reconstruction for
doi.org/10.1109/TBME.2011.2113395 Patient Independent Heartbeat Classification in Comparison with Some Benchmark
Martin, A., & Voix, J. (2018). In-Ear Audio Wearable: Measurement of Heart and Methods. Computers in Biology and Medicine, 41, 411–419. https://doi.org/10.1016/
Breathing Rates for Health and Safety Monitoring. IEEE Transactions on Biomedical j.compbiomed.2011.04.003
Engineering, 65(6), 1256–1263. https://doi.org/10.1109/tbme.2017.2720463 Neung, J., & Kwak, K. (2022). ECG-Based Biometrics using a Deep Network based on
Martinez, A., Alcaraz, R., & Rieta, J. J. (2010). Applicatiof the phasor transform for Independent Component Analysis. IEEE Access, 10. https://doi.org/10.1109/
automatic delineation of single-lead ECG fiducial points. Physiological Measurement, ACCESS.2022.3147807
31(11), 1467–1485. https://doi.org/10.1088/0967-3334/31/11/005 Nurmaini, S., Darmawahyuni, A., Sakti Mukti, A. N., Rachmatullah, M. N., Firdaus, F., &
Martis, U. J. R., & Acharya, L. C. (2013). ECG Beat Classification Using PCA, LDA ICA and Tutuko, B. (2020). Deep Learning-Based Stacked Denoising and Autoencoder for
Discrete Wavelet Transform. Biomed. Biomed. Signal Process, 8, 437–448. https://doi. ECG Heartbeat Classification. Electronics, 9(1). https://doi.org/10.3390/
org/10.1016/j.bspc.2013.01.005 electronics9010135
Martis, R. J., & Acharya, U. R. (2013). Automated Detection of Atrial- Fibrillation using Nuthalapati, S., & Goduguluri, V. Y. (2019). Artifact Elimination in Cardiac Signals using
Bayesian Paradigm. Knowledge Based Systems, 54, 269–275. https://doi.org/ Through Circular Leaky Adaptive Algorithms for Remote Patient Care Monitoring.
10.1016/j.knosys.2013.09.016 Indian Journal of Public Health Research and Development, 10(11). https://doi.org/
Martis, R. J., Acharya, U. R., Mandana, K. M., Ray, A. K., & Chakraborty, C. (2012). 10.5958/0976-5506.2019.03956.1
Application of principal component analysis to ECG signals for automated diagnosis Oktivasari, P., Hasyim, M., Hs, A., Freddy, & Suprijadi. (2019, Julie). A simple real-time
of cardiac health. Expert Systems with Applications, 39(14), 11792–11800. https:// system for detection of normal and myocardial ischemia in the ST segment and T
doi.org/10.1016/j.eswa.2012.04.072 wave ECG signal. 2019 International Conference on Information and
Mateo, J., Torres, A., Aparicio, A., & Santos, J. L. (2016). An Efficient Method for ECG Communications Technology (ICOIACT), Indonesia. https://doi.org/10.1109/
Beat Classification and Correction of Ectopic Beats. Computer Electronics Engineering, icoiact46704.2019.8938461.
53, 219–229. https://doi.org/10.1016/j.compeleceng.2015.12.015 Oliver, N., & Flores Mangas, F. (2006). HealthGear: A Real-Time Wearable System for
Mathews, S. M., Kambhamettu, C., & Barner, K. E. (2018). A Novel Application of Deep Monitoring and Analysing Physiological Signals. International workshop on Wearable
Learning for Single-Lead ECG Classification. Computers in biology and medicine, 99, and Implantable Body Sensor Networks. https://doi.org/10.1109/BSN.2006.27
53–62. https://doi.org/10.1016/j.compbiomed.2018.05.013 Olvera, F. E. (2006). Electrocardiogram Waveform Feature Extraction using the Matched
Mazhar, B., & Tayel, M. E. (2006). ECG Images Classification using Feature Extraction Filter. Statistical Signal Processing II.
Based on Wavelet Transformation and Neural Network. International Conference on Omar, A., & Rao, K. P. (2017). Dedicated Real-Time Monitoring System for Health Care
AIML, ICGST. using ZigBee. Healthcare Technology Letters, 4(4), 142–144.
Oresko, J. J., Jin, Z. P., & Duschl, H. (2010). A Wearable Smart Phone-Based Platform for
Real-time Cardiovascular Disease Detection Via Electrocardiogram Processing. IEEE

22
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Transactions on Information Technology in Biomedical, 14, 734–740. https://doi.org/ the 2016 IEEE International Conference on Wireless Communications, WiSPNET,
10.1109/TITB.2010.2047865 https://doi.org/10.1109/WiSPNET.2016.7566392.
Padhy, S., & Dandapat, S. (2017). Third-order Tensor-Based Analysis of Multilead ECG Salman, M. N., Trinatha Rao, P., & Ur Rahman, M. Z. (2017). Adaptive Noise Cancellers
for Classification of Myocardial Infarction. Biomed. Biomed. Signal Process, 31, 71–78. for Cardiac Signal Enhancement for IoT based Health Care Systems. Journal of
https://doi.org/10.1016/j.bspc.2016.07.007 Theoretical and Applied Information Technology, 95(10), 2206–2213.
Padmavathi, K., & Ramakrishna, S. K. (2015). Hybrid bacterial foraging and particle Salman, N. M., Trinatha Rao, P., & Ur Rahman, Z. (2018). Novel Logarithmic Reference
swam optimization for detecting Bundle Branch Block. Springer Plus, Article, 481), 4 Free Adaptive Signal Enhancers for ECG Analysis of Wireless Cardiac Care
(1, 1–19. Monitoring Systems. IEEE Access, 6, 46382–46395. https://doi.org/10.1109/
Padmavathi, K., & Ramakrishna, S. K. (2017). Detection of Bundle Branch Block using ACCESS.2018.2866303
Adaptive Bacterial Foraging Optimization and Neural Network. Egyptian Informatics Sandeep, K., & Prasad, V. V. G. S. (2016). Implementation of Heartbeat and Temperature
Journal, 18(1), 67–74. https://doi.org/10.1016/j.eij.2016.04.004 Monitoring System Per Minute Rate. International Research Journal of Engineering and
Pagola, J., Franscisco-Pascual, J., & Moya, A. (2018). Yield of Atrial Fibrillation Technology, 03(7), 1124–1127.
Detection with Textile Wearable Holter from the acute Phase of Stroke: Pilot Study of Satija, U., Ramkumar, B., & Manikandan, M. S. (2017). Noise-aware dictionary-learning-
Crypto-AF registry. International Journal of Cardiology, 251, 45–50. https://doi.org/ based sparse representation framework for detection and removal of single and
10.1016/j.ijcard.2017.10.063 combined noises from ECG signal. Healthcare Technology Letters, 4(1), 2–12. https://
Pan, J., & Tompkins, W. J. (1985). A real-time QRS detection algorithm. IEEE doi.org/10.1049/htl.2016.0077
Transactions on Bio-Medical Engineering, 32(3), 230–236. https://doi.org/10.1109/ Sayadi, O., & Shamsollahi, M. B. (2007). Multi-Adaptive Bionic Wavelet Transform:
TBME.1985.325532 Application to ECG Denoising and Baseline Wandering Reduction. EURASIP Journal
Pandian, P. S., & Mohanavelu, K. (2008). Smart Vest: Wearable Multi-parameter Remote on Advances in Signal Processing, 14, 1–11. https://doi.org/10.1155/2007/41274
Physiological Monitoring System. Medicine Engineering and Physics., 30, 466–477. Sellami, A., & Hwang, H. (2019). A Robust Deep Convolutional Neural Network with
https://doi.org/10.1016/j.medengphy.2007.05.014 Batch-Weighted Loss for Heartbeat Classification. Expert Systems with Applications,
Paradiso, R., Loriga, G., & Taccini, N. (2005). A Wearable Health Care System Based on 122, 75–84. https://doi.org/10.1016/j.eswa.2018.12.037
Knitted Integral Sensors. IEEE Transactions on Information Technology in Biomedicine, Serhani, M. A., Kassabi, H. T. E., Ismail, H., & Navaz, A. N. (2020). ECG Monitoring
9(3), 337–344. https://doi.org/10.1109/TITB.2005.854512 Systems. Review, Architecture, Processes, and Key Challenges. Sensors, 20(6). https://
Pardesh, V., & Sagar, S. (2017). Health Monitoring Systems using IoT and Raspberry Pi- A doi.org/10.3390/s20061796
Review. International Conference on Innovative Mechanisms for Industry Applications, Seshadri, D. R., Bittel, B., et al. (2020). Accuracy of Apple Watch for Detection of Atrial
IEEE.. https://doi.org/10.1109/ICIMIA.2017.7975587 Fibrillation. Circulation, 141, 702–703. https://doi.org/10.1161/
Park, J., Seong, K., & Noh, H. (2016). An ECG Monitoring System using Android Smart CIRCULATIONAHA.119.044126
Phone. IEEE International Conference on Consumer Electronics, Asia (ICCE-Asia), Shaad, M.d., & Wang, H. (2017). A Wireless Health Monitoring System using Mobile
https://doi.org/10.1109/ICCE-Asia.2016.7804828. Phone Accessories. IEEE Internet of Things Journal, 4(6). https://doi.org/10.1109/
Patro, K. K., & Prakash, J. A. (2020). An efficient Optimized Feature Selection with JIOT.2016.2645125
Machine Learning Approach for ECG Biometric Recognition. IETE Journal of Shaolin, R., & Yang, X. (2022). Homecare-Oriented ECG Diagnosis with Large-Scale Deep
Research. https://doi.org/10.1080/03772063.2020.1725663 Neural Network for Continuous Monitoring on Embedded Devices. IEEE Transactions
Pérez-Riera, A. R., Barbosa-Barros, R., Daminello-Raimundo, R., & Carlos de Abreu, L. on Instrumentation and Measurement, 71. https://doi.org/10.1109/
(2018). Main artifacts in electrocardiography. Annals of Noninvasive TIM.2022.3147328
Electrocardiology, 23(2), e12494. Shnayder, V., Chen, B.-R., Lorincz, K., Jones, T. R. F. F., & Welsh, M. (2005). Sensor
Perin, F. V., and Fekete, J. D. (2013). Interactive horizon graphs: improving the compact Networks for Medical Care. Proceedings of the 3rd international conference on
visualization of multiple time series. Proceedings of the SIGCHI Conf. on Human Embedded networked sensor systems - SenSys ’05. Presented at the 3rd international
Factors in Computing Systems (CHI ’13), 3217-3226. conference, San Diego, California, USA, https://doi.org/10.1145/
Medline Plus. Holter Monitor. National library of medicine, Retrieved from https:// 1098918.1098979.
medlineplus.gov/ency/article/003877.htm. Accessed January 6, 2016. Shyr, T.-W., Shie, J.-W., Jiang, C.-H., & Li, J.-J. (2014). A Textile-Based Wearable Sensing
Prabha, S., & Elias, K. (2018). Everything you Wanted to know about Smart Healthcare. Device Designed for Monitoring the Flexion Angle of Elbow and Knee Movements.
IEEE Consumer Electronics Magazine, 7(1), 18–28. https://doi.org/10.1109/ Sensors (Basel, Switzerland), 14(3), 4050–4059. https://doi.org/10.3390/
MCE.2017.2755378 s140304050
Preejith, S. P., & Dhinesh, R. (2016). Wearable ECG Platform for Continuous Cardiac Silva, H. P., & Carreiras, C. (2015). Off-the-person Electrocardiography: Performance
Monitoring. Annual International IEEE Engineering in Medicine and Biology Society Assessment and Clinical Correlation. Health Technology, 4, 309–318. https://doi.org/
Conference, IEEE.. https://doi.org/10.1109/embc.2016.7590779 10.1007/S12553-015-0098-Y
Rafi, H. T., & Woong-Ko, Y. (2022). HeartNet: Self Multi-Head Attention Mechanism via Singh, B. N., & Tiwari, A. K. (2006). Optimal Selection of Wavelet Basis Function Applied
Convolution Network with Adversarial Data Synthesis for ECG-based Arrhythmia to ECG Signal Denoising. Digital Signal Processing, 16(3), 275–287. https://doi.org/
Classification. medRxiv, https://doi.org/10.1101/2021.12.20.21268090. 10.1016/j.dsp.2005.12.003
Rahman, M. Z. U., & Karthik, G. V. S. (2013). An Efficient Cardiac Signal Enhancement Smisek, R., Marsanova, L., Nemcova, A., Vítek, M., Kozumplík, J., & Novakova, M.
using Time-Frequency Realization of Leaky Adaptive Noise Cancelers for Remote (2017). CSE database: Extended annotations and new recommendations for ECG
Health Monitoring Systems. Measurement: Journal of the International Measurement software testing. Medical & Biological Engineering & Computing, 55(8), 1473–1482.
Confederation, 46(10), 3815–3835. https://doi.org/10.1016/j. https://doi.org/10.1007/s11517-016-1607-5
measurement.2013.07.009 P.J. Soh A.E. Vandenbosch Wearable Wireless Health Monitoring: Current Developments
T. Rao A., Gopi, M., Prasad, M.V.S.S. Real Time ECG Signal Transmission for Remote 2015 IEEE Microwave Magazine Challenges and Future Trends 10.1109/
Monitoring International Journal of Engineering Development and Research MMM.2015.2394021.
348–354 2014 IJEDR1303068. Solanas, A., Patsakis, C., Conti, M., Vlachos, I., Ramos, V., Falcone, F., et al. (2014).
Rao, J., & Rao, M. K. (2016). An RTOS based Architecture for Patient Monitoring System Smart health: A Context-Aware Health Paradigm Within Smart Cities. IEEE
with Sensor Networks. Indian. Journal of Science and Technology, 9(17). https://dx. Communications Magazine., 52(8), 74–81. https://doi.org/10.1109/
doi.org/10.17485/ijst/2016/v9i17/92960. mcom.2014.6871673
Ray, P. P. (2018). A Survey on Internet of Things Architectures. Journal of King Saud Soria, M. L., & Martinez, J. P. (2009). Analysis of Multi-Domain Features for ECG
University-computer and Information Sciences, 30, 291–319. https://doi.org/10.1016/ Classification. Computer Cardiology, 561–564.
j.jksuci.2016.10.003 Spano, E., Di Pascoli, S., & Iannaccone, G. (2016). Low-Power Wearable ECG Monitoring
Rienzo, M., & Rizzo, F. (2005). MagIC system: A New Textile-Based Wearable Device for System for Multiple-Patient Remote Monitoring. IEEE sensors journal, 16(13),
Biological Signal Monitoring Applicability in Daily Life and Clinical Setting. 5452–5462. https://doi.org/10.1109/jsen.2016.2564995
Proceedings 28th Annual International IEEE Engineering in Medicine and Biology Subasi, A., Dogan, S., & Tuncer, T. (2021). A Novel Automated Tower Graph based ECG
Society Conference. 7167-7169, https://doi.org/10.1109/IEMBS.2005.1616161. Signal Classification Method with Hexadecimal Local Adaptive Binary Pattern and
Rizwan, A., Zoha, A., Mabrouk, I. B., et al. (2021). A Review on the State of the Art in Deep Learning. Journal of Ambient Intelligence and Humanized Computing. https://doi.
Atrial Fibrillation Detection Enabled by Machine Learning. IEEE Revisions in org/10.1007/s12652-021-03324-4
Biomedical Engineering, 14, 219–239. https://doi.org/10.1109/RBME.2020.2976507. Subramanian, B. (2017). ECG Signal Classification and Parameter Estimation using
https://neurosky.com/2015/05/ecg-technology-comparison-medical-vs-consumer- Multiwavelet Transform. International Journal of Medical Sciences, 28(7).
applications/ Sulthana, A., Rahman, M. Z. U., & Mirza, S. S. (2018). An Efficient Kalman Noise
Saadi, D. B., Tanev, G., Flintrup, M., Osmanagic, A., Egstrup, K., Hoppe, K., et al. (2015). Canceller for Cardiac Signal Analysis in Modern Telecardiology Systems. IEEE Access,
Automatic real-time embedded QRS complex detection for a novel patch-type 6, 34616–34630. https://doi.org/10.1109/ACCESS.2018.2848201
electrocardiogram recorder. IEEE Journal of Translational Engineering in Health and Sulthana, A., & Ur Rahman, M. Z. (2019). Artifact cancellation from cardiac signals in
Medicine, 3, 1900112. https://doi.org/10.1109/JTEHM.2015.2421901 health care systems using a zoned adaptive algorithm. International Journal of
Sadhukhan, D., & Mitra, M. (2012). R-peak Detection Algorithm for ECG Using Double Engineering and Advanced Technology, 8(5), 988–993.
Difference and RR Interval Processing. Procedia Technology, 4, 873–877. https://doi. Sulthana, A., & Zia-Ur-Rahman, M. (2019). Adaptive Signal Enhancement in Clinical
org/10.1016/j.protcy.2012.05.143 Cardiac Care Systems using Normalized Median LMS Variants. Indian Journal of
Sahoo, S., Kanungo, B., Behera, S., & Sabut, S. (2017). Multiresolution Wavelet Public Health Research and Development, 10(9), 441–446.
Transform Based Feature Extraction and ECG Classification to Detect Cardiac Sung, M., Marci, C., & Pentland, A. (2005). Wearable Feedback Systems for
Abnormalities. Measurement: journal of the International Measurement Confederation, Rehabilitation. Journal of Neuroengineering and Rehabilitation., 2(1), 17. https://doi.
108, 55–66. https://doi.org/10.1016/j.measurement.2017.05.022 org/10.1186/1743-0003-2-17
Salman, M. N., Rao, P. T., & Rahman, M. Z. U. (2016). Baseline Wander Removal in Tada, Y., Amano, Y., Sato, T., Saito, S., & Inoue, M. (2015). A Smart Shirt Made with
Cardiac Signals using Variable Step Size Adaptive Noise Cancellers. In Proceedings of Conductive Ink and Conductive Foam for the Measurement of Electrocardiogram

23
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

Signals with Unipolar Precordial Leads. Fibers, 3, 463–477. https://doi.org/10.3390/ Wu, Q., Sun, Y., Yan, H., & Wu, X. (2020). ECG Signal Classification with Binarized
fib3040463 Convolutional Neural Network. Computers in Biology and Medicine, 121(103800),
Taddei, A., Distante, G., Emdin, M., Pisani, P., Moody, G. B., Zeelenberg, C., et al. (1992). Article 103800. https://doi.org/10.1016/j.compbiomed.2020.103800
The European ST-T database: Standard for evaluating systems for the analysis of ST- Xiao, R., Xu, Y., Pelter, M. M., Fidler, R., Badilini, F., Mortara, D. W., et al. (2018).
T changes in ambulatory electrocardiography. European Heart Journal, 13(9), Monitoring significant ST changes through deep learning. Journal of
1164–1172. https://doi.org/10.1093/oxfordjournals.eurheartj.a060332 Electrocardiology., 51(6S), S78–S82. https://doi.org/10.1016/j.
P. Tadejko W. Rakowski Mathematical morphology-based ECG feature extraction for the jelectrocard.2018.07.026
purpose of heartbeat classification 6th International Conference on Computer Xiaowen Xu, Ying Liang, Pei He and Junliang Yang. (2019). Adaptive Motion Artifact
Information Systems and Industrial Management Applications (CISIM’07) 2007 Elk, Reduction Based on Empirical Wavelet Transform and Wavelet Thresholding for the
Poland 10.1109/cisim.2007.47. Non-Contact ECG Monitoring Systems. Sensors 2019, 19(13), 2916; https://doi.org/
Taha, E. T., El-Sayed, A., & Rafat, S. (2016). A Survey on Classification of ECG Signal 10.3390/s19132916.
Study. Communications on Applied Electronics, 6(5), 11. https://doi.org/10.5120/ Xintarakou, A., Sousoni, V., Asvestas, D., et al. (2022). Remote Cardiac Rhythm
cae2016652467 Monitoring in the Era of Smart Wearables: Present Assets and Future Perspectives.
Tantawi, M. M., Revett, K., Salem, A., & Tolba, M. F. (2013). Fiducial Feature Reduction Frontiers in Cardiovascular Medicine, 9. https://doi.org/10.3389/fcvm.2022.853614
Analysis for Electrocardiogram (ECG) based Biometric Recognition. Journal of Yang, H., Kan, C., Liu, G., & Chen, Y. (2013). Spatiotemporal Differentiation of
Intelligent Information Systems, 40(1), 17–39. https://doi.org/10.1007/s10844-012- Myocardial Infarctions. IEEE transactions on automation science and engineering: A
0214-7 publication of the IEEE Robotics and Automation Society, 10(4), 938–947. https://doi.
Tewary S., Chakraborty S., Majumdar J., Majumder R., Kundu D., Ghosh S., Das Gupta S. org/10.1109/tase.2013.2263497
A novel approach towards designing a wearable Smart Health Monitoring System Ying, W., Zhou, J., Wang, Y., et al. (2020). A Review of Algorithm & Hardware Design for
measuring the vital parameters and emergency situations in real-time and providing AI-Based Biomedical Applications. IEEE Transactions on Biomedical Circuits and
the necessary medical care through telemedicine; Proceedings of the 2016 IEEE Systems, 14(2). https://doi.org/10.1109/TBCAS.2020.2974154
Students’ Conference on Electrical, Electronics and Computer Science, SCEECS; Yogaraj, A., & Ezilarasan, M. (2017). IoT Based Smart Healthcare Monitoring System for
Bhopal, India. 5–6 March 2016; pp. 1–8. Rural/Isolated Areas. International Journal of Pure and Applied Mathematics, 114(12),
Thakur, and Rhyne, T. M. (2009). Data Vases: 2D and 3D Plots for Visualizing Multiple 679–688.
Time Series. Proceedings of the 5th Int. Symposium on Advances in Visual Zahra, E., & Loni, M. (2020). A Review on Deep Learning Methods for ECG Arrhythmia
Computing: Part II (ISVC ’09), Springer-Verlag, Berlin, Heidelberg, 929-938. Classification. Expert Systems with Applications, 7. https://doi.org/10.1016/j.
Tomasic, I., & Trobec, R. (2017). Electrocardiographic Systems with Reduced Numbers of eswax.2020.100033
Leads - Synthesis of the 12-lead ECG. IEEE Reviews in Biomedical Engineering, 7, Zerald, Y., & Kim, H. (2009). A Wearable ECG Acquisition System with Compact Planar-
126–142. https://doi.org/10.1109/rbme.2013.2264282 Fashionable Circuit Board-Based Shirt. IEEE Transactions on Information Technology in
Torfs, T., Leonov, V., Van Hoof, C., & Gyselinckx, B. (2006). Body-heat Powered Biomedicine, 13(6), 897–902. https://doi.org/10.1109/TITB.2009.2033053
Autonomous Pulse Oximeter. IEEE Proceedings International Conference on Sensors. Zhang, Z., Dong, J., Luo, X., Choi, K. S., & Wu, X. (2014). Heartbeat classification using
22-25, Daegu, Korea, https://doi.org/10.1109/ICSENS.2007.355497. disease-specific feature selection. Computers in Biology and Medicine, 46, 79–89.
Tuncer, T., Dogan, S., Pławiak, P., & Acharya, U. R. (2019). Automated arrhythmia https://doi.org/10.1016/j.compbiomed.2013.11.019
detection using novel hexadecimal local pattern and multilevel wavelet transform Zhang, Z., & Hu, X. (2014). Zigbee Based Wireless Sensor Networks and their use in
with ECG signals. Knowledge Based Systems, 186, Article 104923. Medical and Health Care Domain. International Conference on Sensing Technology,
Tuncer, T., Dogan, S., Plawiak, P., & Subasi, A. (2022). A Novel Discrete Wavelet- IEEE, https://doi.org/10.1109/ICSensT.2013.6727754.
Concatenated Mesh Tree and Ternary Chess Patrren based ECG Signal Recognition Zhang, M. J., Roetker, N. S., & Folsom, A. R. (2018). Feasibility of Using a Leadless Patch
Method. Biomedical Signal Processing and Control, 103331. https://doi.org/10.1016/j. Monitor in Community Cohort Studies: The Multiethnic Study of Atherosclerosis.
bspc.2021.103331 Pacing Clin Electrophysiol, 41(11), 1389–1390. https://doi.org/10.1111/pace.13507
Ubeyli, E. D. (2009). Adaptive Neuro-Fuzzy Inference System for Classification of ECG Zhang, Q., & Zhou, D. (2017). HeartID: A Multiresolution Convolutional Neural Network
Signals using Lyapunov Exponents. Computer Methods and Programs in Biomedicine, for ECG-Based Biometric Human Identification in Smart Health Applications. IEEE
93, 313–321. https://doi.org/10.1016/j.cmpb.2008.10.012 Access, 5. https://doi.org/10.1109/ACCESS.2017.2707460
Umer, M., Bhatti, B. A., Tariq, M. H., Zia-ul-Hassan, M., Khan, M. Y., & Zaidi, T. (2014). Zidelmal, Z., Amirou, A., Adnane, M., & Belouchrani, A. (2012). QRS detection based on
Electrocardiogram Feature Extraction and Pattern Recognition using a Novel wavelet coefficients. Computer Methods and Programs in Biomedicine, 107(3),
Windowing Algorithm. Advances in Bioscience and Biotechnology, 05(11), 886–894. 490–496. https://doi.org/10.1016/j.cmpb.2011.12.004
https://doi.org/10.4236/abb.2014.511103
Varma, M., Rao, Y., & Ramaraju, J. S. S. (2013). GSM-GPRS Based Intelligent Wireless
Mobile Health Monitoring System for Cardiac Patients. International Journal of
Engineering Research & Technology, 2(10). https://doi.org/10.15623/ijret.201 Ms. A. Venkata Anuhya received her B.Tech degree in the
Department of Electronics & Communication Engineering
5.0404024.
Venkatesan, C., Karthigaikumar, P., Paul, A., Satheeskumaran, S., & Kumar, R. (2018). under Jawaharlal Nehru Technological University, India, 2012.
ECG Signal Preprocessing and SVM Classifier-Based Abnormality Detection in She received M.Tech degree in Embedded Systems from K L
Remote Healthcare Applications. IEEE Access: Practical Innovations, Open Solutions, 6, University, Andhra Pradesh, India in 2015. She worked as As­
9767–9773. https://doi.org/10.1109/access.2018.2794346 sistant Professor in Electronics and Communication Depart­
ment for St. Mary’s Women’s Engineering College from 2015-
Verma, S., & Gupta, N. (2012). Microcontroller based Wireless Heart Rate Telemonitor
for Home Care. IOSR Journal of Engineering, 2(7), 25–31. https://doi.org/10.9790/ 2016 and Vignan’s Nirula Institute of Science and Technology
3021-02722531 from 2016-2017. She is currently doing research work as a Full-
Vincent, A. E., & Sreekumar, K. (2017). A Survey on Approaches for ECG Signal Analysis time research scholar in the department of Electronics &
With focus to Feature Extraction and Classification. International Conference on Computer Engineering in Koneru Lakshmaiah Education
Foundation, Andhra Pradesh, India. Her research interests are
Inventive Communication and Computational Technologies (ICICCT). https://doi.org/
10.1109/ICICCT.2017.7975175 on Embedded Systems, Internet of things (IoT), Machine
Vullings, R., de Vries, B., & Bergmans, J. W. M. (2011). An Adaptive Kalman Filter for Learning, Deep Learning and Smart Health monitoring.
ECG Signal Enhancement. IEEE Transactions on Bio-Medical Engineering, 58(4),
1094–1103. https://doi.org/10.1109/TBME.2010.2099229
Wahabi, S., Pouryayevali, S., Hari, S., & Hatzinakos, D. (2014). On evaluating ECG Dr. K. Venkata Ratnam received the M.Sc degree from
biometric systems: Session-dependence and body posture. IEEE Transactions on Acharya Nagarjuna University, Andhra Pradesh in 2003, M.
Information Forensics and Security, 9(11), 2002–2013. https://doi.org/10.1109/ Tech degree from Bharath University, Chennai, Tamilnadu in
tifs.2014.2360430 2006 and Ph. D degree in Electronics and Communication En­
Wang, J. S., Chiang, W.-C., Hsu, Y. L., & Yang, Y. T. C. (2013). ECG Arrhythmia gineering from NIT-Rourkela, India in 2016. He has been with
Classification using a Probabilistic Neural Network with a Feature Reduction the Department of Electronics and Communication Engineering
Method. Neurocomputing, 116, 38–45. https://doi.org/10.1016/j. and worked as a faculty member for Bharath University,
neucom.2011.10.045 Chennai, India, from 2006-2008; as an Associate Professor for
Wasimuddin, M., & Elleithy, K. (2020). Stages-Based ECG Signal Analysis from QISCET, Ongole, Andhra Pradesh, from 2008-2010 and for
Traditional Signal to Machine Learning Approaches: A Survey. IEEE Access, 8. RISE Gandhi Ongole, Andhra Pradesh, from 2015-2016. He is
https://doi.org/10.1109/ACCESS.2020.3026968 currently working as an Associate Professor in the Department
Wen, C., Lin, T. C., Chang, K. C., & Huang, C. H. (2009). Classification of ECG Complexes of Electronics and Computer Engineering in Koneru Laksh­
Using Self-Organizing. CMAC Measurement, 42, 399–407. https://doi.org/10.1016/j. maiah Education Foundation, Andhra Pradesh, India. His cur­
measurement.2008.08.004 rent research interests include IoT, VLSI &Embedded Systems, Solar Cells and MPPT. He
Wiklund, U., & Akay, M. (2002). Wavelet Decomposition of Cardiovascular Signals for published 8 research papers in different international conferences, 26 research papers
Baroreceptor Function Test in Pigs. IEEE Transactions on Biomedical Engineering, 49, published in various SCI and SCOPUS indexed international journals. Currently his h-index
651–661. https://doi.org/10.1109/tbme.2002.1010848 is 7 and i10 index is 3.
Wilhelm, F. H., Roth, W. T., & Sackner, M. A. (2003). The LifeShirt: An Advanced System
for Ambulatory Measurement of Respiratory and Cardiac Function. Behavior
Modification, 27(5), 671–691. https://doi.org/10.1177/0145445503256321

24
V.A. Ardeti et al. Expert Systems With Applications 217 (2023) 119561

George Tom Varghese received B.Tech. in Electronics & Rajesh Kumar Patjoshi received the B.Tech degree in Elec­
Communication Engineering from college of engineering tronics and Telecommunication engineering from the Amar­
chengannur (CEC), Kerala in 2006, India, M.Tech and Ph.D avati University, India and the M.Tech and Ph.D. degrees in
from National institute of Technology, Rourkela, Odisha, India electronics and communication engineering from the National
in 2009 and 2015 respectively. Presently he is working as an Institute of Technology, Rourkela, India, in 2010 and 2015,
Associate professor and HoD in the department of Electronics respectively. He is currently an Associate Professor in the
and Instrumentation engineering in St. Joseph’s College of Department of Electronics and communication Engineering,
Engineering and Technology, (SJCET) Palai, Kottayam, Kerala, National Institute of Science and Technology, Berhampur. Her
India. His research interest includes Analog and mixed signal research interests include, VLSI Design for Artificial Intelli­
design and low power VLSI Design. gence (Neuromorphic IC Design), Low Power VLSI and Radio
Frequency IC Design, Network Control system, Embedded
Systems, Power Quality Enhancement in Grid Integrated
Renewable Energy System.

25