Major Project-III Report

On
Artificial Intelligence of Things System for Cardiac Disease
Detection

Submitted in partial fulfillment of the requirement of

University of Mumbai for the Degree of

Bachelor of Technology
In
Information Technology

Submitted By
Prasanna Laxman Pawar
Pranav Shashikant Bandal
Vedant Mohan Phatangare
Anit Roy Payyappilly

Supervisor
Prof. Krishnendu S Nair

Department of Information Technology

Pillai College of Engineering, New Panvel – 410 206

UNIVERSITY OF MUMBAI

Academic Year 2023–24

 DEPARTMENT OF INFORMATION TECHNOLOGY
Pillai College of Engineering
New Panvel – 410 206

CERTIFICATE
This is to certify that the requirements for the B.Tech Project Report entitled ‘Artificial
Intelligence of Things System for Cardiac Disease Detection’ have been successfully completed
by the following students:
Name Roll No.
Prasanna Laxman Pawar A638
Pranav Shashikant Bandal A602
Vedant Mohan Phatangare A665
Anit Roy Payyappilly A664
In partial fulfillment of Bachelor of Technology of Mumbai University in the Department of
Information Technology, Pillai College of Engineering, New Panvel – 410 206 during the
Academic Year 2023–24.

_____________________
Supervisor
Prof. Krishnendu S Nair

_____________________ _____________________
Head of Department Principal
Dr. Satishkumar L. Varma Dr. Sandeep M. Joshi

i
 DEPARTMENT OF INFORMATION TECHNOLOGY
Pillai College of Engineering
New Panvel – 410 206

PROJECT APPROVAL FOR B.TECH

This Project entitled “Artificial Intelligence of Things System for Cardiac Disease Detection” by
Prasanna Laxman Pawar, Pranav Shashikant Bandal, Vedant Mohan Phatangare, and Anit Roy
Payyappilly are approved for the degree of B.Tech. in Information Technology.

Examiners:

1. ________________

2. ________________

Supervisors:

1. ________________

2. ________________

Chairman:

1. ________________

Date:

Place:
ii
 DEPARTMENT OF INFORMATION TECHNOLOGY
Pillai College of Engineering
New Panvel – 410 206

Declaration

We declare that this written submission for the B.Tech Project entitled “Artificial Intelligence
of Things System for Cardiac Disease Detection” represents our ideas in our own words and where
others' ideas or words have been included. We have adequately cited and referenced the original
sources. We also declared that we have adhered to all principles of academic honesty and integrity and
have not misrepresented or fabricated or falsified any ideas / data / fact / source in our submission. We
understand that any violation of the above will cause disciplinary action by the institute and also evoke
penal action from the sources which have thus not been properly cited or from whom paper permission
has not been taken when needed.

Project Group Members:

Prasanna Laxman Pawar: __________________________

Pranav Shashikant Bandal: __________________________

Vedant Mohan Phatangare: __________________________

Anit Roy Payyappilly: __________________________

Date:

Place:
iii
 Table of Contents

Abstract................................................................................................................................ i

List of Figures...................................................................................................................... ii

List of Tables....................................................................................................................... iii

1. Introduction................................................................................................................. 1

1.1 Fundamentals................................................................................................... 1

1.2 Objectives........................................................................................................ 1

1.3 Scope............................................................................................................... 2

1.4 Outline……………………………….............................................................. 3

2. Literature Survey......................................................................................................... 4

2.1 Introduction……. ……………………............................................................ 4

2.2 Literature Review ………………………………………................................ 4

2.3 Summary of Literature Survey.…………………………................................ 9

3. Implemented System………………………………........................................................... 12

3.1 Overview…………………….......................................................................... 12

3.1.1 Existing System Architecture………………………………………. 12

3.1.2 Proposed System Architecture……………………………………… 14

3.2 Implementation Details………...................................................... 17

3.2.1 Algorithms / Techniques….................................................................. 17

3.2.2 Use Case Diagram / Activity Diagram…........................................... 18

3.2.3 Sample Dataset…............................................................................... 20

3.2.4 Hardware and Software Specifications…........................................... 22

4. Result and Discussion………….................................................................................. 24

4.1 Sample Inputs, Outputs and GUI Screenshots…………………………........ 24

4.2 Evaluation Parameter …................................................................................ 27

4.3 Performance Evaluation…………………………………………………….. 29

5. Conclusion and Future Scope........................................................................................ 31

5.1 Conclusion…………………………………………………………………….. 31

5.2 Future Scope…………………………………………………………………. 31

References............................................................................................................................ 33

Journal Paper…………………………………………………………………………….. 34

Copyright Certificates…………………………………………………………………… 42

Acknowledgement……………………………………………………………………......... 43
 Abstract
As the leading cause of illness and death in the current world, cardiovascular disease (CVD) can be
significantly decreased thanks to significant advancements in diagnosis and treatment. Unfortunately, due
to the limitations of the skills of extracted characteristics, standard ECG problem identification
algorithms exhibit high rates of misdiagnosis. Therefore, initiatives and approaches for awareness and
care must be developed to prevent the untimely death of people with coronary heart disorders.This study
proposes an artificial intelligence of things (AIoT) system for electrocardiogram (ECG) analysis and
cardiac disease detection. The system includes a front-end IoT-based hardware, a user interface on smart
device’s application (APP), a cloud database, and an AI platform for cardiac disease detection. The
front-end IoT-based hardware, a wearable ECG patch that includes an analog front-end circuit and a
Bluetooth module, can detect ECG signals. The APP on smart devices can not only display users’
real-time ECG signals but also label unusual signals instantly and reach real- time disease detection.
These ECG signals will be uploaded to the cloud database. The cloud database is used to store each
user’s ECG signals, which forms a big-data database for an AI algorithm to detect cardiac disease. The
algorithm proposed by this study is based on convolutional neural networks and the average accuracy is
94.96%. This device will help individuals to monitor their heart condition and will notify when signal
variation happens.

vi
 List of Figures

Figure 3.1 Existing system architecture 12

Figure 3.2 Proposed system architecture 15

Figure 3.3 Use case diagram 18

Figure 3.4 Activity Diagram 19

Figure 3.2.3.1 Sample Ecg input Image 22

Figure 4.1 Result: 24

4.1.1.ECG Sensing Device 24

4.1.2.Wire Connections 24
4.1.3.Home page 25
4.1.4.Login Page 25
4.1.5.Menu Option page 25
4.1.6.ECG Graph (Result) 25
4.1.7. Health Tips 26
4.1.8.Graph image upload page 26
4.1.9.Analyzed Result with treatment 26
4.1.10. Manually detection of disease 26

Figure 4.3.1 Confusion matrix of the predictive model 29

Figure. 4.3.2 Accuracy of the model 30

vii
 List of Tables

Table 1.1 Title of the table 1

Table 2.1 Summary of literature survey 9

Table 3.1 Sample Dataset Used for Experiment 20

Table 3.2 Hardware details 22

Table 3.3 Software details 23

viii
 Chapter 1
INTRODUCTION

1.1 Fundamentals

The artificial intelligence of things (AloT) platform proposed in this study aims at analyzing
real-time ECG signals to reduce the risks of severe arrhythmias. For real-time detection,
low-power consumption, and long duration of use, a complete system structure, including a
wearable front-end ECG sensing device, a user interface on smart device APP, a cloud
database, and an A1-based algorithm for cardiac disease analysis, is presented and described
in the following sections.

1.2 Objectives
1. Early detection of cardiac disease: The primary objective of the wearable device
should be to detect early signs of cardiac disease. This could include monitoring heart
rate, blood pressure, and other vital signs to identify abnormalities that may indicate
the presence of a heart condition.
2. Continuous monitoring: The wearable device should be designed to continuously
monitor the user's vital signs, providing real-time feedback on any changes or
abnormalities.
3. User-friendly design: The wearable device should be designed to be comfortable and
easy to use, so that users will be willing to wear it for extended periods of time.
4. Accurate and reliable: The device should be accurate and reliable in detecting and
tracking changes in the user's vital signs. This is important to ensure that users can
trust the information provided by the device.
5. Data collection and analysis: The device should be capable of collecting and
analyzing data over time, providing insights into the user's health and any trends or
patterns that may indicate the presence of a heart condition.
6. Notification and alert system: The device should be equipped with a notification and
7. alert system to notify users of any abnormalities in their vital signs that may require
medical attention.

1
 8. Integration with healthcare providers: The device should be designed to integrate with
healthcare providers, allowing for remote monitoring and telemedicine consultations
as needed. This can help to improve patient outcomes and reduce healthcare costs.

1.3 Scope
The scope of a wearable device for cardiac disease detection project will depend on various
factors such as the specific goals of the project, available resources, and technological
capabilities. However, some possible areas of scope for such a project could include:

1. Design and development of the wearable device: This could involve designing the
physical form and user interface of the device, as well as developing the necessary
sensors and software for monitoring and analyzing vital signs.
2. Integration with mobile applications and cloud-based platforms: The device could be
integrated with mobile applications and cloud-based platforms to allow for remote
monitoring and data analysis, as well as for the storage and sharing of patient
information.
3. Clinical testing and validation: The device would need to undergo clinical testing to
validate its accuracy, reliability, and safety for use in detecting cardiac disease.
4. Regulatory compliance: The project would need to ensure compliance with relevant
regulatory requirements and obtain necessary approvals before the device can be
marketed and sold.
5. User training and education: Users would need to be educated on how to use the
device and interpret the data it provides, as well as on the importance of monitoring
their vital signs for early detection of cardiac disease.

2
 1.4 Outline
The report is organized as follows: The introduction is given in Chapter 1. It describes the
fundamental terms used in this project. It motivates to study and understand the different
techniques used in this work. This chapter also presents the outline of the objective of the
report. Chapter 2 describes the review of the relevant various techniques in the literature
systems. It describes the pros and cons of each technique. Chapter 3 presents the Theory and
proposed work. It describes the major approaches used in this work. The societal and technical
applications are mentioned in Chapter 4. The summary of the report is presented in Chapter 5.

3
 Chapter 2
Literature Survey

2.1. Introduction

A literature survey is a critical aspect of any project report. It involves reviewing and
analyzing existing literature, research, and other sources related to the project topic.It
Provides a comprehensive understanding of the project topic: Literature survey helps to
gain a deeper understanding of the research topic by exploring previous work and
identifying research gaps.Helps to define research objectives and questions: A literature
survey provides the necessary background information to refine research objectives and
questions. It Offers a theoretical framework: The survey can provide a theoretical
framework and a foundation for the research, helping to guide the study and make it more
effective. It Identifies research methodology: Literature survey can assist in determining the
most appropriate research methodology and data analysis techniques for the
research.Supports validity and reliability: A literature survey is essential in establishing the
validity and reliability of the research by comparing the project findings with previous
work.

2.2. Literature Survey

Paper 1:Artificial intelligence based early detection of cardiac diseases using smart
wearables.Arti Rana, Vineet Kishore Shrivastava.2022

As the leading cause of illness and death in the current world, cardiovascular disease (CVD)
can be significantly decreased thanks to significant advancements in diagnosis and treatment.
Unfortunately, due to the limitations of the skills of extracted characteristics, standard ECG
problem identification algorithms exhibit high rates of misdiagnosis.This chapter discusses
how to manage cardiovascular disorders by prompt diagnosis, treatment, and surveillance.The
study suggests using artificial intelligence (AI) in the Internet of Things (IoT) to recognize
heart disorders as well as ECGs. In this article, we described a recently employed cloud-based
artificial intelligence framework for atrial fibrillation and briefly characterized the results of
the experiments.[1]

4
Paper 2:Cardiac monitoring with novel low power sensors measuring upper thoracic
electrostatic charge variation for long lasting wearable devices.Kanika Dheman,David
Werder.Michele Mango.2022

This paper presents the evaluation of a novel low-power sensor exploited to measure the
electrostatic charge variation in the upper thorax to provide an energy-efficient and accurate
detection of the electrical activity of the heart. The sensor is investigated for measuring the
heart activity in terms of the QRS complex. The paper presents the design of a wearable
sensor device, optimization of electrode positions and incorporation into a wearable chest
strap that can be integrated seamlessly under clothes. Due to the low power consumption of
the sensor, the sensor node consumes only 87.3 μW of power and can provide multiple weeks
of operation using a coin cell battery while providing the same functionality as that of
commercially available sensors such as photoplethysmography and electrocardiogram (ECG)
ICs. In addition, the chest strap was also characterized for different use scenarios during
sedentary and activity periods. They evaluated both the signal quality and the power
consumption compared with other sensor technology showing a power save of an order of
magnitude when compared with photoplethysmography sensors.[2]

Paper 3:Design of Low Power VLSI Architecture for Classification of Arrhythmic Beats
Using DNN for Wearable Device Applications. Authors: Meenali Janveja, Mayank.2021

This paper proposes a low power VLSI architecture that facilitates the classification of
Electrocardiogram (ECG) into normal and other seven types of arrhythmia beats using a Deep
Neural Network (DNN). Unlike the existing methods for heartbeat classification, in which
handcrafted ECG features are utilised, the proposed design leverages DNN for the
classification of arrhythmia using a complete ECG beat. This obviates the need to extract
ECG features separately and produces an accurate and power op-timised design for
arrhythmia classification. Evaluation of the proposed methodology on the MIT-BIH dataset
exhibits the accuracy and specificity of 97.01% and 99.09%, respectively, which is
comparable or better with respect to other algorithms implemented on software or ASIC based
platforms. The proposed architecture is realized at 180nm CMOS technology having
0.624mm2 area and exhibits 6.82× less power consumption at 1kHz as compared to other
methods.[3]

5
Paper 4:Severe Analysis of Cardiac Disease Detection using the Wearable Device by
Artificial Intelligence Authors: Maryam Khan, Qareena Yaseen, Asia Mumtaz, Ayesha
Saleem, Seemab Ishaq, Haseeb Udeen.2020

In the current era of technology, ArtificialIntelligence (AI) is playing a vital role in the health
care sector especially cardiac disease detection which is a major cause of sudden death. Both
the elderly and young are at the risk of sudden cardiac death at the ratio of 1-2% all around the
world. Although AI technology with wearable technology is being used to detect heart
diseases for quite some time now, sometimes it fails due to multiple reasons which include
algorithm failure, high cost of treatment, limited battery time wearable device, data training
issues, security and privacy issue in IoT, slow working of devices, poor internet or patients
don’t reach the hospital on time. Which gives rise to false results. Security and privacy issues
in the old devices are the biggest flaws due to which old devices work slowly and the internet
issues are common, it helps us to check their heart parameters anytime and anywhere in the
world which reduces the hospital's workload, cost issues and to line onward. Meanwhile, these
problems can be overcome by using modern models such as ECG assessment, AI-based
guidelines, Visy’s model which can recognize five critical diseases. A Wearable ECG patch is
a very lightweight model that provides high accuracy and efficiency. These devices are trained
by using a machine learning algorithm, and AI plays a prime role to detect the diseases. It
helps us to check their heart parameters anytime and anywhere in the world which reduces the
hospital's workload and cost issues, and the devices provide updated information as real-time
data is stored online and secured with firebase authentication. It is concluded that all modern
devices are more efficacious, cost-effective, user friendly, and more secure.[4]

6
Paper 5:Artificial Intelligence of Things Wearable System for Cardiac Disease Detection
Authors:Yu-Jin Lin, Chen-Wei Chhuang, Chun-Yueh Yen, Sheng-Hsin Huang,Peng-Wei
Huang.2019

This study proposes an artificial intelligence of things (AloT) system for electrocardiogram
(ECG) analysis and cardiac disease detection. The system includes a front-end IoT-based
hardware, a user interface on smart device's application (APP), a cloud database, and an A1
platform for cardiac disease detection. The front-end IoT-bascd hardware, a wearable ECG
patch that includes an analog front-end circuit and a Bluetooth module, can detect ECG
signals. The APP on smart devices can not only display users' real-time ECG signals but also
label unusual signals instantly and reach real- time disease detection. These ECG signals will
be uploaded to the cloud database. The cloud database is used to store each user's ECG
signals. which forms a big-data database for A1 algorithm to detect cardiac disease. The
algorithm proposed by this study is based on convolutional neural networks and the average
accuracy is 94.96%. The ECG dataset applied in this study is collected from patients in Tainan
hospital, Ministry of Health and Welfare. Moreover, signal verification was also performed by
a cardiologist.[5]

Paper 6:Early Detection of Cardiovascular Diseases Using Wearable Ultrasound Device

Authors: Sumaiya Shomaji, Domenic Forte, Alex Roman,Swarup Bhunia.2019

Early detection of symptoms is considered an effective approach to prevent deaths due to

cardiovascular disease (CVD). Routine cardiac check-ups are therefore suggested by
clinicians to detect CVD at an early stage. Unfortunately, existing methods for cardiac
diagnosis are costly and time-consuming as in most of the cases, these facilities are available
only in hospitals or clinics. In this article, a novel wearable ultrasonic imaging assembly is
proposed for routine monitoring of the carotid arteries in an easy-to-use and economical way.
Using standard B-mode ultrasound, which is suitable for wearable form factors, the device
monitors intima–media thickness (IMT), which is a proven indicator of cardiovascular
disease. The design parameters for all the essential hardware components of the proposed
wearable imaging system along with an efficient algorithm for predicting IMT anomalies from
ultrasound images are proposed. Finally, we describe a custom-designed prototype of the
proposed system and demonstrate its capability in acquiring ultrasound images.[6]

7
Paper 7:Design of a Wearable Device for ECG Continuous Monitoring Using Wireless
Technology. Santiago Led, Jorge Fernández and Luis Serrano.2004

This project focuses on the design and implementation of an intelligent wearable device for
ECG continuous acquisition and transmission to some remote gateway using Bluetooth
technology. The acquisition device has been designed for having very low power consumption
and reduced size. The Analog Devices’ ADuC831 Micro-Converter for achieving the analog
to digital conversion and the CSR’s BlueCore2 chip for the Bluetooth transmission are the
core of the device. The designed device is an important component of a complete prototype
for remote ECG continuous.[7]

8
 2.3 Literature Summary
A literature review is an objective, critical summary of published research literature relevant
to a topic under consideration for research. The summary is presented here.

Table 2.1 Summary of literature survey

Sr Name of paper with Year Technique/ Success Conclusion

no Author Algorithm ratio

1 Artificial intelligence 2022 k-nearest 97.99% Sensors in today's wearables can

based early detection of neighbours measure a plethora of vital signs, such
cardiac diseases using (KNN) as blood pressure, temperature, heart
smart wearables rate, HRV, respiration rate, arterial
support
vector oxygen saturation, and thoracic fluid
Authors: Arti Rana, content. It has been hypothesized that in
Vineet machines
the future, these wearables could take
Kishore Shrivastava (SVM)
the place of implanted devices in
convolutiona determining when a heart failure patient
l neural will need to be readmitted to the
networks hospital.It may also help those who are
(CNNs) afflicted by viral infections and prevent
the spread of epidemics.

2 Cardiac monitoring with 2022 - - In this work, a novel sensor based on

novel low power sensors measuring the electrostatic charge
measuring upper changes during cardiac activity was
thoracic electrostatic designed, developed and evaluated
charge variation for long for cardiac monitoring.
lasting wearable devices.

Authors: Kanika
Dheman,David
Werder.Michele Mango

3 Design of Low Power 2021 CNN 95.82% This paper proposes a power and
VLSI Architecture model . area efficient end-to-end arrhythmia
for Classification of classification system employing
Arrhythmic Beats ECG signals. The high multi-class
Using DNN for classification accuracy, low area and
Wearable Device power requirements make the
Applications. proposed architecture suitable for

9
 Authors: Meenali low power wearable device
Janveja, Mayank applications. Their endeavor is to
Tantuway, Ketan test the proposed classifier on a large
Chaudhari, Gaurav number of healthy and unhealthy
Trivedi ECG excerpts, so that it can aid
cardiologists in diagnosing
arrhythmia with an improved
efficiency in future.

4 Severe Analysis of 2020 Machine 91.8% After the brief explanation, we

Cardiac Disease learning conclude that, with the help of
Detection using the algorithms AIOT, doctors can take decisions to
Wearable Device by include take care of patients, more
Artificial Intelligence Regression efficiently by the access to data.
, Decision Some better decisions can be taken
Authors: Maryam Khan, trees, by using Artificial Intelligence. The
Qareena Yaseen, Asia Support purpose of this study is to develop
Mumtaz, Ayesha Vector the cardiac vascular device that
Saleem, Seemab Ishaq, Machines, monitors the patients also in the
Haseeb Udeen Neural medical field and provides
networks, efficiency and more enhancement.
Deep It works to eliminate working
learning all repetitively, increasing effort in the
connected treatment process, reducing time
to and money wastage. Creating
ensemble preventive care by Artificial
Methods. Intelligence application the model
proposed an AI-based calculation
for arrhythmia classification and is
more upgraded for all arrhythmia
detection. AI-based model is
possible on a chip. The model is
simpler more flexible as in the
future, patients can use it to meet
the real AIOT based cardiac disease
detection more smoothly

5 Artificial Intelligence of CNN 94.96% This study proposes an artificial

Things Wearable intelligence of things (AIoT) system
System for Cardiac 2019 for electrocardiogram (ECG)
Disease Detection
analysis and cardiac disease
detection. The system includes a
Authors:Yu-Jin Lin,
front-end IoT-based hardware, a
Chen-Wei Chhuang,
user interface on smart device’s
Chun-Yueh Yen,
application (APP), a cloud database,
Sheng-Hsin
and an AI platform for cardiac
Huang,Peng-Wei
disease detection.
Huang

10
6 Early Detection of 2019 Image 91.00% In this study a novel wearable
Cardiovascular processing ultrasonic imaging assembly is
Diseases Using algorithm proposed for routine monitoring of
Wearable Ultrasound to analyze the carotid arteries in an easy-to-use
Device the and economical way , using
acquired standard B-mode ultrasound which
Authors: Sumaiya images and is suitable for wearable form
Shomaji, Domenic enable factors, the device monitors
Forte, Alex automatic intima–media thickness (IMT),
Roman,Swarup Bhunia detection of which is a proven indicator of
plaque in cardiovascular disease.
the carotid.

7 Design of a Wearable 2004 The wireless NA The wearable Holter has been
Device for ECG transmission designed in order to improve the
Continuous Monitoring technologies present ECG continuous monitoring
Using Wireless (GPRS, devices. The design and selection of
Technology. GSM, the electronic components and an
Bluetooth). efficient software implementation
Authors: Santiago Led, provide a device with great
Jorge Fernández and PIN code
authentication autonomy, reduced dimensions and
Luis Serrano minimum weight. These features
for bluetooth
connection. lead to a device easy to use
improving a patient's quality of life.

11
 Chapter 3
Implemented System

3.1 Overview
3.1.1 Existing System Architecture
The Content Based Information Filtering (IF) systems need proper techniques for representing
the items and producing the user profile, and some strategies for comparing the user profile
with the item representation.

Fig. 3.1 Existing system architecture used for AIOT wearable system for cardiac diseases

The high level architecture of a content­based recommender system is depicted in Figure 3.2.
The recommendation process is performed in three steps, each of which is handled by a
separate component:
● Wearable ECG monitoring device: This study proposes a sensing hardware structure,
as shown, which includes an analog front-end circuit with low-power consumption, a
commercial power management integrated circuit (IC), and a commercial Bluetooth
module. The analog front-end circuit is a self- designed system on chip (SOC), which
includes a 10-bit sigma—delta analog to digital converter, a level shifter and digital
signal processing units. The commercial Bluetooth module applies Bluetooth Low
Energy 4.0 to transmit the ECG signal collected by the front-end SOC to the APP
instantly. The wearable ECG monitoring device with a single lead is attached with two
silver chloride wet electrodes on the chest, and it can be used up to 24 h under normal
usage.[4]

12
● User interface on smart device APP: A user interface on APP is also proposed, and
the structure is shown in Fig. 3, which includes three main parts: an ECG-displaying
function, an AI-based arrhythmia- analyzing function, and data-storing and
data-transmitting function. The real-time ECG signal will be shown on the screen, and
the AI algorithm is used to classify the user’s ECG signal into different cardiac
arrhythmias at the same time. By taking the computing power of modern mobile into
account, the classification on smart devices shows only two categories: normal and
abnormal. For further classification, the classification will be completed on the cloud
server to obtain a more precise arrhythmia type from the user’s ECG signal. The
collected ECG data will not only be stored in local mobiles but also uploaded to the
cloud database. For the sake of data safety and correctness, all the data will be
encoded and added with time stamps.[4]

● Cloud server and database: The structure of the cloud server and database is shown.
This server contains a big-data database, which includes three segments: data storage,
web user interface, and AI-based algorithm for arrhythmia analysis. First, the data
storage is in charge of receiving the data packages from the front-end smart devices,
and the data packages are decoded as ECG signals. Furthermore, the ECG signals will
be stored separately according to the measured objects and the measuring time stamps.
Second, the web user interface provides doctors, patients, and patients’ families a
clear information platform. Doctors can diagnose patients’ condition more specifically
with the stored ECG data, and patients and their families can realize more about their
daily ECG signal. Third, the AI-based algorithm can detect unusual signals from a
large amount of data in several minutes. In general, a normal human produces
approximately one hundred thousand heartbeats per day, and most of them are normal
ECG signals; just a few are abnormal. Given this reason, doctors face a great
challenge of diagnosing correctly with long-term ECG data. Through this cloud
platform, the AI-based algorithm can quickly detect unusual signals, which will be
displayed on the web user interface.[4]

13
 ● AI-based algorithm: The AI-based algorithm for arrhythmia classification has four
categories: normal ECG, atrial fibrillation, atrial flutter, and ventricular fibrillation, as
shown. The structure of this algorithm has two segments: data pre-processing and
CNN model. To make the CNN model have better feature learning, traditional ECG
signal processing, such as time–frequency analysis, feature extraction, and R-peak and
QRS-complex detection, are not carried out. The pre-processing structure proposed in
this study includes three steps: noise removal, baseline removal, and image
generation.[4]

3.1.2 Proposed System Architecture

As the leading cause of illness and death in the current world, cardiovascular disease
(CVD) can be significantly decreased thanks to significant advancements in diagnosis and
treatment. Unfortunately, due to the limitations of the skills of extracted characteristics,
standard ECG problem identification algorithms exhibit high rates of misdiagnosis.
Therefore, initiatives and approaches for awareness and care must be developed to prevent
the untimely death of people with coronary heart disorders. This chapter discusses how to
manage cardiovascular disorders by prompt diagnosis, treatment, and surveillance. It has
been found that wearable technology is much more effective at managing heart disease
scenarios. The study suggests using artificial intelligence (AI) in the Internet of Things
(IoT) to recognize heart disorders as well as ECGs. In this article, we described a recently
employed cloud-based artificial intelligence framework for atrial fibrillation and briefly
characterized the results of the experiments.

14
Fig. 3.2 Proposed system architecture

15
An AIOT wearable device for heart disease detection could be designed with the following
architecture:

Sensor module: The sensor module will be responsible for collecting data from the wearer's
body. It can include a heart rate sensor, an ECG sensor, and an activity sensor. These sensors
will measure the wearer's heart rate, rhythm, and activity level.

Data processing module: The data processing module will be responsible for processing the
data collected by the sensor module. It will use machine learning algorithms to analyze the
data and detect abnormalities in the wearer's heart rate, rhythm, and activity level. This
module will be powered by an embedded system, which can be a microcontroller or a
microprocessor.

Connectivity module: The connectivity module will be responsible for transmitting the data
processed by the data processing module to the cloud. It can use Wi-Fi or Bluetooth Low
Energy (BLE) for data transmission. This module can also receive data from the cloud, such
as software updates and machine learning model updates.

Cloud computing module: The cloud computing module will receive the data from the
connectivity module and process it further using more powerful machine learning algorithms.
It will then generate reports on the wearer's heart health status and transmit them back to the
wearable device. This module can be hosted on a cloud server, such as Amazon Web Services
or Google Cloud.

User interface module: The user interface module will provide a graphical user interface
(GUI) for the wearer to view their heart health status. It can be a simple display showing the
wearer's heart rate, rhythm, and activity level, or it can provide more detailed reports on the
wearer's heart health status.

Power management module: The power management module will be responsible for
managing the power consumption of the wearable device. It can use a rechargeable battery or
energy harvesting techniques to extend the battery life of the device.

16
3.2 Implementation Details

3.2.1 Technique CNN

Convolutional Neural Networks (CNN) can be used in an AIOT wearable device for heart
disease detection to extract meaningful features from the raw sensor data collected from the
wearer's body. Here's a high-level overview of how CNNs can be used in such a device:
Data collection: The wearable device collects data from various sensors, including heart rate
and ECG sensors, which capture heart rate and rhythm data.

Data preprocessing: The collected data is preprocessed to remove noise and artifacts that can
affect the performance of the CNN model.

Data augmentation: The data can be augmented to increase the size of the dataset, which can
improve the performance of the CNN model. For example, the data can be randomly shifted in
time or scaled in amplitude.

CNN model design: A CNN model is designed to process the preprocessed and augmented
data. The CNN model typically consists of convolutional layers, pooling layers, and fully
connected layers. The convolutional layers extract features from the input data, the pooling
layers reduce the spatial dimensions of the output from the convolutional layers, and the fully
connected layers classify the input data into different classes.

Training: The CNN model is trained on the preprocessed and augmented data using a suitable
optimization algorithm such as stochastic gradient descent (SGD). The training process
involves adjusting the weights of the CNN model to minimize the difference between the
predicted output and the actual output.

Validation: The trained CNN model is validated on a separate dataset to evaluate its
performance. The validation dataset should be different from the training dataset to ensure that
the model can generalize well to unseen data.

17
Deployment: Once the CNN model is trained and validated, it can be deployed on the
wearable device to classify the heart rate and rhythm data in real-time.

By using CNNs in an AIOT wearable device for heart disease detection, the device can
achieve high accuracy in classifying heart rate and rhythm data, which can aid in early
detection of heart diseases.

3.2.2 Use Case Diagram / Activity Diagram

fig 3.3 Use case diagram

18
fig 3.4. Activity diagram

19
 3.2.3 Sample Dataset Used

3.2.3.1. Dataset 1
Table 4.1 Sample Dataset Used for Experiment

age sex cp trest chol fbs reste thal exan oldp slop ca thal target
bps cg ach g eak e

52 1 0 125 212 0 1 168 0 1 2 2 3 0

53 1 0 140 203 1 0 155 1 3.1 0 0 3 0

70 1 0 145 174 0 1 125 1 2.6 0 0 3 0

61 1 0 148 203 0 1 161 0 0 2 1 3 0

62 0 0 138 294 1 1 106 0 1.9 1 3 2 0

58 0 0 100 248 0 0 122 0 1 1 0 2 1

Parameters used in the dataset:

● Age: Age of the individual in years.

● Sex: Gender of the individual (1 = male, 0 = female).
● CP (Chest Pain Type): The type of chest pain experienced by the individual (0, 1, 2, 3, 4).
This is a categorical variable representing different types of chest pain.
● Trestbps: Resting blood pressure of the individual in mm Hg.
● Chol: Serum cholesterol level in mg/dl.
● FBS (Fasting Blood Sugar): Fasting blood sugar level greater than 120 mg/dl (1 = true, 0
= false).
● Restecg: Resting electrocardiographic results (0, 1, 2). This is a categorical variable
representing different electrocardiographic results.
● Thalach: Maximum heart rate achieved.
● Exang (Exercise Induced Angina): Angina induced by exercise (1 = yes, 0 = no).

● Oldpeak: Depression induced by exercise relative to rest, a measure on an ECG reading.

● Slope: Slope of the peak exercise ST segment (0, 1, 2). This is a categorical variable
representing different slopes.

20
● CA (Number of Major Vessels Colored by Fluoroscopy): Number of major vessels
colored by fluoroscopy (0-3). This indicates the number of major blood vessels supplying blood
to the heart that are visible under fluoroscopy.
● Thal (Thalassemia): Thalassemia is a blood disorder. 3 = normal, 6 = fixed defect, 7 =
reversible defect.
● Target: The presence of heart disease (1 = presence, 0 = absence).

It seems like this dataset contains information about different individuals, their demographics,
and various medical measurements. The goal is likely to use this data to build a machine learning
model that can predict the presence or absence of heart disease (the "target" variable) based on
these features.

3.2.3.2. Dataset 2

ECG Arrhythmia Image Dataset

This dataset is composed of two collections of heartbeat signals derived from two famous
datasets in heartbeat classification, the MIT-BIH Arrhythmia Dataset and The PTB Diagnostic
ECG Database. The number of samples in both collections is large enough for training a deep
neural network.

This dataset has been used in exploring heartbeat classification using deep neural network
architectures, and observing some of the capabilities of transfer learning on it. The signals
correspond to electrocardiogram (ECG) shapes of heartbeats for the normal case and the cases
affected by different arrhythmias and myocardial infarction. These signals are preprocessed and
segmented, with each segment corresponding to a heartbeat.

Content

● Arrhythmia Dataset
● Number of Samples: 109446
● Number of Categories: 5
● Data Source: Physionet's MIT-BIH Arrhythmia Dataset
● Classes: ['N': 0, 'S': 1, 'V': 2, 'F': 3, 'Q': 4]

The PTB Diagnostic ECG Database

21
● Number of Samples: 14552
● Number of Categories: 2
● Sampling Frequency: 125Hz
● Data Source: Physionet's PTB Diagnostic Database

This dataset is created by saving the each ECG arrhythmia into the image form. Then total
images from each classes are divided into train and test data where training samples are a80% of
the total data and test samples are 20%.

Figure 3.2.3.1.Sample Ecg input image

3.2.4 Hardware and Software Specifications

The experiment setup is carried out on a computer system which has the different hardware
and software specifications as given in Table 3.2 and Table 3.3 respectively.

Table 3.2 Hardware details

Processor 2 GHz Intel

HDD 180 GB

RAM 2 GB

22
 Table 3.3 Software details

Operating System Windows XP Professional With Service pack 2

Programming Language JDK 1.8

Database Oracle 9

23
 Chapter 4

Result and Discussion

4.1.Sample of Inputs, Outputs and GUI Screenshots.

4.1.1.ECG Sensing Device

4.1.2.Wire Connections

24
 4.1.3.Home page 4.1.4.Login Page

4.1.5.Menu Option page 4.1.6.ECG Graph (Result)

25
 4.1.7. Health Tips 4.1.8.Graph image upload page

4.1.9.Analyzed Result with treatment 4.1.10. Manually detection of disease

26
4.2 Evaluation Parameters

The quality of a prediction system can be evaluated by comparing predictions to a test set of
known user ratings. These systems are typical measured using predictive accuracy metrics,
where the predicted ratings are directly compared to actual user ratings. The most commonly
used metric in the literature is Mean Absolute Error (MAE)- defined as the average absolute
difference between predicted ratings and actual ratings, given by:

(9)

Where pu,i is the predicted rating for user u on item i, ru,i is the actual rating, and N is the total
number of ratings in the test set.

(10)

A related commonly-used metric, Root Mean Squared Error (RMSE), puts more emphasis on
larger absolute errors, and is given by:

Predictive accuracy metrics treat all items equally. However, for most recommender systems we
are primarily concerned with accurately predicting the items a user will like. As such, researchers
often view recommending as predicting good, i.e. items with high ratings versus bad or
poorly-rated items. In the context of Information Retrieval (IR), identifying the good from the
background of bad items can be viewed as discriminating between “relevant” and “irrelevant”
items; and as such, standard IR measures, like Precision, Recall and Area under the ROC Curve
(AUC) can be utilized.

Precision: a measure of exactness, determines the fraction of relevant items retrieved out of all
items retrieved, e.g. the proportion of recommended movies those are actually good

(11)

27
Recall: a measure of completeness, determines the fraction of relevant items retrieved out of all
relevant items, e.g. the proportion of all good movies recommended

(12)

Hybrid models discussed in the work, two techniques have merged in four different ways. The
first one works by combining the separate recommender ratings using a linear combination or a
voting scheme which basically selects the recommendation that is seen better in terms of quality
and more consistent with past user’s ratings. As for the second method, it adds the content-based
characteristics to collaborative models which can help in overcoming the sparsity problem since
we are not only relying on ratings but also on item profiles for our prediction. The third way is to
add collaborative characteristics to content-based models where latent factors are introduced to
describe the user preferences. The fourth mode is to develop a single unifying recommendation
model based on content-based and collaborative characteristics using probabilistic approaches
such as rule-based classifier or Bayesian regression models.

To select the optimum values, the weights α and β have to be selected to maximize the accuracy
of the resulting MAE evaluated with the combined rating. Alternatively, the choice of the
weights needs to minimize the error resulting from the difference between predicted ratings and
actual ratings available in training data. For each rating-prediction pair <pu,ru>, pu being the
predicted value and qu the correct value available in the training data, the absolute error is
computed as |pu-ru|. The MAE is then evaluated by examining N ratings-prediction pairs, and
computing the average error as shown in equation(9) above. The lower the MAE the better the
accuracy is. As a result, the choice of the weights (α,β ) needs to minimize the MAE.

28
4.3. Performance Evaluation

In order to evaluate the proposed system, experiments were conducted on data selected from
kaggle, an application-based research prediction system. The data was collected from thousands
of individuals who have heart diseases and the individuals with normal heartbeat. The data was
stored in img files that were transformed to a user-item matrix. The data was divided into
training sets and corresponding testing sets and thus, a 5-fold cross validation approach was
applied (i.e. 80% training data and 20% test data) to evaluate the system as shown in Figure 4.1.

Figure 4.3.1 Confusion matrix of the predictive model.

29
The accuracy of the proposed technique was compared to user-based collaborative filtering as
stand-alone and item-based collaborative filtering as stand-alone. They searched for the best
weights following the proposed values of α & β using an empirical approach by observing the
MAE for the different combinations of α & β. It was observed that α=1/6 and β=5/6 produced the
lowest MAE compared to the other suggested combinations. Although the combination α=1/8 &
β=7/8 was expected to represent the optimum solution since the weight accorded for item-based
collaborative filtering is higher as shown in Figure 4.2.

Figure. 4.3.2 Accuracy of the model.

30
 Chapter 5

Conclusion

5.1 Conclusion

An artificial intelligence of things (AIoT) wearable system for cardiac disease detection has the
potential to improve patient outcomes and reduce healthcare costs. The system can monitor a
person's heart rate, rhythm, and other vital signs continuously, and use AI algorithms to detect
signs of cardiac disease at an early stage.[5] The system can also provide personalized treatment
plans for patients with cardiac disease, remotely monitor patients, and predict a person's risk of
developing cardiac disease. Additionally, the system can be used in clinical trials to monitor
patients' heart health and track the effectiveness of new treatments.[3] Overall, an AIoT wearable
system for cardiac disease detection can increase access to healthcare and improve patient
outcomes.
.

5.1 Future Scope

The future scope of Artificial Intelligence of Things (AIoT) wearable systems for cardiac disease
detection is promising, with significant potential for advancements in healthcare. Here are some
key areas of growth and development:

● Early Detection and Prevention: AIoT wearable devices can continuously monitor an
individual's vital signs, such as heart rate, ECG, blood pressure, and oxygen saturation. They can
detect abnormal patterns and alert users or healthcare professionals in real-time, allowing for
early intervention and prevention of cardiac diseases.[1]
● Personalized Healthcare: AI algorithms can analyze data from wearable devices to
provide personalized insights and recommendations for individuals to improve their cardiac
health. This includes tailored exercise regimens, dietary advice, and medication management.

● Telemedicine and Remote Monitoring: The integration of AIoT with telemedicine

platforms allows for remote monitoring of patients with cardiac conditions. This reduces the
need for frequent hospital visits, making healthcare more accessible and cost-effective.

● Data Analytics and Research: The data collected from AIoT wearable devices can be

31
used for large-scale research and epidemiological studies. Researchers can gain valuable insights
into cardiac disease trends, contributing to the development of better prevention and treatment
strategies[3].
● Continuous Monitoring of Chronic Conditions: AIoT wearables can be used to monitor
individuals with chronic cardiac conditions, such as arrhythmias or heart failure, providing
healthcare providers with a continuous stream of data to better manage these conditions.
● Improved Diagnostics: AI can assist in the interpretation of cardiac data, potentially
improving the accuracy of diagnostics. AI algorithms can quickly identify anomalies and help in
the early diagnosis of cardiac diseases.[2]
● User-Friendly Design: The development of more user-friendly and comfortable wearable
devices will encourage broader adoption among individuals of all ages. This will lead to a more
extensive data pool for analysis and research.
● Integration with Smart Healthcare Ecosystems: AIoT wearables can be integrated with
other healthcare devices and systems, such as electronic health records (EHRs) and hospital
information systems, creating a seamless healthcare ecosystem for cardiac disease
management.[7]

32
 References

[1]Arti Rana, Vineet Kishore Shrivastava.”Artificial intelligence based early detection of

cardiac diseases using smart wearables”.2022

[2]Kanika Dheman,David Werder.Michele Mango.”Cardiac monitoring with novel low power

sensors measuring upper thoracic electrostatic charge variation for long lasting wearable
devices.”.2022

[3].Meenali Janveja, Mayank.”Design of Low Power VLSI Architecture for Classification of

Arrhythmic Beats Using DNN for Wearable Device Applications”.2021

[4]Maryam Khan, Qareena Yaseen, Asia Mumtaz, Ayesha Saleem, Seemab Ishaq, Haseeb
Udeen.”Severe Analysis of Cardiac Disease Detection using the Wearable Device by Artificial
Intelligence”2020

[5]Yu-Jin Lin, Chen-Wei Chhuang, Chun-Yueh Yen, Sheng-Hsin Huang,Peng-Wei

Huang.”Artificial Intelligence of Things Wearable System for Cardiac Disease
Detection”.2019

[6]Sumaiya Shomaji, Domenic Forte, Alex Roman,Swarup Bhunia.”Early Detection of

Cardiovascular Diseases Using Wearable Ultrasound Device”.2019

[7]Santiago Led, Jorge Fernández and Luis Serrano.”Design of a Wearable Device for ECG
Continuous Monitoring Using Wireless Technology”2004

[8]https://www.mdpi.com/1424-8220/23/2/828

[9]https://www.xenonstack.com/artificial-intelligence-solutions/healthcare/

33
 Journal Paper

Artificial Intelligence of Things System

for Cardiac Disease Detection
Prasanna Pawar Pranav Bandal Anit Roy Vedant Phatangare
Prof. Krishnendu Nair
Department of Department of Department of Department of
Department of
Information Technology Information Technology Information Technology Information Technology
Information Technology
Pillai College of Pillai College of Pillai College of Pillai College of
Pillai College of
Engineering, New Panvel Engineering, New Panvel Engineering, New Panvel Engineering, New Panvel
Engineering, New Panvel
Navi Mumbai, India Navi Mumbai, India Navi Mumbai, India Navi Mumbai, India
Navi Mumbai, India
ppawr20it@student.mes.a bandalpsh20ite@student. apayyappilly20it@student vphata20ce@student.mes.
knair@mes.ac.in
c.in mes.ac.in .mes.ac.in ac.in

Abstract— As the leading cause of illness and death in the The artificial intelligence of things (AloT) platform
current world, cardiovascular disease (CVD) can be proposed in this study aims at analyzing real- time ECG
significantly decreased thanks to significant advancements in signals to reduce the risks of severe arrhythmias. For
diagnosis and treatment. Unfortunately, due to the limitations real-time detection, low-power consumption, and long
of the skills of extracted characteristics, standard ECG
problem identification algorithms exhibit high rates of
duration of use, a complete system structure, including a
misdiagnosis. Therefore, initiatives and approaches for wearable front- end ECG sensing device, a user interface on
awareness and care must be developed to prevent the untimely smart device APP, a cloud database, and an A1-based
death of people with coronary heart disorders. This study algorithm for cardiac disease analysis, is presented and
proposes an artificial intelligence of things (AIoT) system for described in the following sections.
electrocardiogram (ECG) analysis and cardiac disease
detection. The system includes a front-end IoT-based B. Objectives
hardware, a user interface on smart device’s application
(APP), a cloud database, and an AI platform for cardiac
disease detection. The front-end IoT-based hardware, a
Early detection of cardiac disease: The primary objective
wearable ECG patch that includes an analog front- end circuit of the wearable device should be to detect early signs of
and a Bluetooth module, can detect ECG signals. The APP on cardiac disease. This could include monitoring heart rate,
smart devices can not only display users’ real-time ECG blood pressure, and other vital signs to identify
signals but also label unusual signals instantly and reach real- abnormalities that may indicate the presence of a heart
time disease detection. These ECG signals will be uploaded to condition.
the cloud database. The cloud database is used to store each
user’s ECG signals, which forms a big-data database for an AI Continuous monitoring: The wearable device should be
algorithm to detect cardiac disease. The algorithm proposed by
designed to continuously monitor the user's vital signs,
this study is based on convolutional neural networks and the
average accuracy is 94.96%. This device will help individuals providing real-time feedback on any changes or
to monitor their heart condition and will notify when signal abnormalities.
variation happens.
User-friendly design: The wearable device should be
Keywords— Cardiovascular disease, diagnosis, treatment, ECG designed to be comfortable and easy to use, so that users
algorithms, misdiagnosis, awareness, care, AIoT system, will be willing to wear it for extended periods of time.
electrocardiogram, cardiac disease detection, IoT-based
hardware, user interface, smart device application, cloud Accurate and reliable: The device should be accurate and
database reliable in detecting and tracking changes in the user's
vital signs. This is important to ensure that users can trust
I.INTRODUCTION
the information provided by the device.
A. Fundamentals Data collection and analysis: The device should be
34
 capable of collecting and analyzing data over time, of the research topic by exploring previous work and
providing insights into the user's health and any trends or identifying research gaps.
patterns that may indicate the presence of a heart
condition. A. Literature Review

Notification and alert system: The device should be 1. Arti Rana, Vineet Kishore Shrivastava.”:Artificial
equipped with a notification and intelligence based early detection of cardiac diseases using
smart wearables”(2022)[1].This chapter discusses how to
Alert system to notify users of any abnormalities in their manage cardiovascular disorders by prompt diagnosis,
vital signs that may require 1 medical attention. treatment, and surveillance.The study suggests using
artificial intelligence (AI) in the Internet of Things (IoT) to
Integration with healthcare providers: The device should recognize heart disorders as well as ECGs. In this article,
be designed to integrate with healthcare providers, they described a recently employed cloud-based artificial
allowing for remote monitoring and telemedicine intelligence framework for atrial fibrillation and briefly
consultations as needed. This can help to improve patient characterized the results of the experiments.
outcomes and reduce healthcare costs.
2. Kanika Dheman,David Werder.Michele Mango.”Cardiac
C. Scope monitoring with novel low power sensors measuring upper
thoracic electrostatic charge variation for long lasting
The scope of a wearable device for cardiac disease detection wearable devices”(2022)[2].This paper presents the
project will depend on various factors such as the specific evaluation of a novel low-power sensor exploited to
goals of the project, available resources, technological measure the electrostatic charge variation in the upper
capabilities. However, some possible areas of scope for such thorax to provide an energy-efficient and accurate detection
a project could include: of the electrical activity of the heart. The sensor is
investigated for measuring the heart activity in terms of the
Design and development of the wearable device: This QRS complex. The paper presents the design of a wearable
could involve designing the physical form and user sensor device, optimization of electrode positions and
interface of the device, as well as developing the incorporation into a wearable chest strap that can be
necessary sensors and software for monitoring and integrated seamlessly under clothes. Due to the low power
analyzing vital signs. consumption of the sensor, the sensor node consumes only
87.3 μW of power and can provide multiple weeks of
Integration with mobile applications and cloud-based operation using a coin cell battery while providing the same
platforms: The device could be integrated with mobile functionality as that of commercially available sensors such
applications and cloud-based platforms to allow for as photoplethysmography and electrocardiogram (ECG)
remote monitoring and data analysis, as well as for the ICs.
storage and sharing of patient information.
3. Meenali Janveja, Mayank.”Design of Low Power VLSI
Clinical testing and validation: The device would need to Architecture for Classification of Arrhythmic Beats Using
undergo clinical testing to validate its accuracy, reliability, DNN for Wearable Device Applications”(2021)[3].This
and safety for use in detecting cardiac disease. paper proposes a low power VLSI architecture that
facilitates the classification of Electrocardiogram (ECG)
Regulatory compliance: The project would need to ensure into normal and other seven types of arrhythmia beats using
compliance with relevant regulatory requirements and a Deep Neural Network (DNN). Unlike the existing
obtain necessary approvals before the device can be methods for heartbeat classification, in which handcrafted
marketed and sold. ECG features are utilised, the proposed design leverages
DNN for the classification of arrhythmia using a complete
User training and education: Users would need to be ECG beat. This obviates the need to extract ECG features
educated on how to use the device and interpret the data it separately and produces an accurate and power op-timised
provides, as well as on the importance of monitoring their design for arrhythmia classification. Evaluation of the
vital signs for early detection of cardiac disease. proposed methodology on the MIT-BIH dataset exhibits the
accuracy and specificity of 97.01% and 99.09%,
II. LITERATURE SURVEY respectively, which is comparable or better with respect to
other algorithms implemented on software or ASIC based
A literature survey is a critical aspect of any project report. platforms.
It involves reviewing and analyzing existing literature,
research, and other sources related to the project topic.It 4. Maryam Khan, Qareena Yaseen, Asia Mumtaz, Ayesha
Provides a comprehensive understanding of the project Saleem, Seemab Ishaq, Haseeb Udeen.”Severe Analysis of
topic: Literature survey helps to gain a deeper understanding Cardiac Disease Detection using the Wearable Device by
35
Artificial Intelligence”(2020)[4].Although AI technology for predicting IMT anomalies from ultrasound images are
with wearable technology is being used to detect heart proposed. Finally, we describe a custom designed prototype
diseases for quite some time now, sometimes it fails due to of the proposed system and demonstrate its capability in
multiple reasons which include algorithm failure, high cost acquiring ultrasound images.
of treatment, limited battery time wearable device, data
training issues, security and privacy issue in IoT, slow 7.Santiago Led, Jorge Fernández and Luis Serrano.”:Design
working of devices, poor internet or patients don’t reach the of a Wearable Device for ECG Continuous Monitoring
hospital on time. Which gives rise to false results. Security Using Wireless Technology”(2004)[7].This project focuses
and privacy issues in the old devices are the biggest flaws on the design and implementation of an intelligent wearable
due to which old devices work slowly and the internet issues device for ECG continuous acquisition and transmission to
are common, it helps us to check their heart parameters some remote gateway using Bluetooth technology. The
anytime and anywhere in the world which reduces the acquisition device has been designed for having very low
hospital's workload, cost issues and to line onward. power consumption and reduced size. The Analog Devices’
Meanwhile, these problems can be overcome by using ADuC831 Micro-Converter for achieving the analog to
modern models such as ECG assessment, AI-based digital conversion and the CSR’s BlueCore2 chip for the
guidelines, Visy’s model which can recognize five critical Bluetooth transmission are the core of the device. The
diseases. A Wearable ECG patch is a very lightweight designed device is an important component of a complete
model that provides high accuracy and efficiency. These prototype for remote ECG continuous
devices are trained by using a machine learning algorithm,
and AI plays a prime role to detect the diseases. It helps us III. WEARABLE SYSTEM FOR CARDIAC DISEASE
to check their heart parameters anytime and anywhere in the DETECTION
world which reduces the hospital's workload and cost issues,
and the devices provide updated information as real time A. Overview
data is stored online and secured with firebase
authentication. 1) Existing System Architecture:

5. Yu-Jin Lin, Chen-Wei Chhuang, Chun-Yueh Yen, The AIOT wearable system for cardiac diseases is designed
Sheng-Hsin Huang,Peng-Wei Huang.”:Artificial around a content-based recommender system with a
Intelligence of Things Wearable System for Cardiac Disease high-level architecture illustrated in Figure 1. The system
Detection "(2019)[5].This study proposes an artificial operates through three primary components.
intelligence of things (AloT) system for electrocardiogram
(ECG) analysis and cardiac disease detection. The system The first component involves a Wearable ECG Monitoring
includes a front-end IoT-based hardware, a user interface on Device, featuring a hardware structure that includes an
smart device's application (APP), a cloud database, and an analog front-end circuit with low-power consumption. This
A1 platform for cardiac disease detection. The front-end comprises a self-designed System on Chip (SOC) equipped
IoT-bascd hardware, a wearable ECG patch that includes an with a 10-bit sigma–delta analog to digital converter, a level
analog front-end circuit and a Bluetooth module, can detect shifter, and digital signal processing units. The device
ECG signals. The APP on smart devices can not only utilizes a commercial Bluetooth module to transmit
display users' real-time ECG signals but also label unusual real-time ECG signals to the accompanying APP. The
signals instantly and reach realtime disease detection. These wearable device, with a single lead and two silver chloride
ECG signals will be uploaded to the cloud database. The wet electrodes on the chest, can be used for up to 24 hours
cloud database is used to store each user's ECG signals. under normal conditions [4].
which forms a big-data database for A1 algorithm to detect
cardiac disease. The algorithm proposed by this study is The second component focuses on the User Interface on the
based on convolutional neural networks and the average Smart Device APP. This interface encompasses three main
accuracy is 94.96%. functions: ECG-displaying, AI-based arrhythmia analysis,
and data storage and transmission. Real-time ECG signals
6. Sumaiya Shomaji, Domenic Forte, Alex Roman,Swarup are displayed on the screen, and an AI algorithm classifies
Bhunia.”:Early Detection of Cardiovascular Diseases Using them into normal and abnormal categories directly on the
Wearable Ultrasound Device”(2019)[6].In this article, a smart device. Further classification, for a more precise
novel wearable ultrasonic imaging assembly is proposed for arrhythmia type, is conducted on the cloud server. The
routine monitoring of the carotid arteries in an easy-to-use collected ECG data is stored locally and uploaded to a
and economical way. Using standard B-mode ultrasound, secure cloud database, ensuring data safety through
which is suitable for wearable form factors, the device encoding and timestamping [4].
monitors intima–media thickness (IMT), which is a proven
indicator of cardiovascular disease. The design parameters The third component involves the Cloud Server and
for all the essential hardware components of the proposed Database. The server includes a big-data database with
wearable imaging system along with an efficient algorithm segments for data storage, a web user interface, and an
36
 AI-based algorithm for arrhythmia analysis. The data fibrillation, atrial flutter, and ventricular fibrillation. The
storage component decodes data packages from front-end algorithm consists of two segments: data pre-processing
smart devices, storing ECG signals based on measured (including noise removal, baseline removal, and image
objects and timestamps. The web user interface serves as an generation) and a Convolutional Neural Network (CNN)
information platform for doctors, patients, and families, model. Traditional ECG signal processing is bypassed to
allowing detailed diagnosis and insight into daily ECG enhance feature learning in the CNN model [4].
signals. The AI-based algorithm swiftly detects unusual
signals, presenting them on the web user interface [4]. In summary, this integrated system aims to facilitate
efficient ECG monitoring and analysis, seamlessly
Lastly, the AI-Based Algorithm for arrhythmia classification combining wearable technology, smart device applications,
is structured with four categories: normal ECG, atrial cloud servers, and an AI-based algorithm for accurate
arrhythmia classification [4].

Fig.1. Existing system architecture

2) Proposed System Architecture: with coronary heart disorders. This chapter discusses how to
manage cardiovascular disorders by prompt diagnosis,
As the leading cause of illness and death in the current world, treatment, and surveillance. It has been found that wearable
cardiovascular disease (CVD) can be significantly decreased technology is much more effective at managing heart disease
thanks to significant advancements in diagnosis and scenarios. The study suggests using artificial intelligence
treatment. Unfortunately, due to the limitations of the skills (AI) in the Internet of Things (IoT) to recognize heart
of extracted characteristics, standard ECG problem disorders as well as ECGs. In this article, we described a
identification algorithms exhibit high rates of misdiagnosis. recently employed cloud-based artificial intelligence
Therefore, initiatives and approaches for awareness and care framework for atrial fibrillation and briefly characterized the
must be developed to prevent the untimely death of people results of the experiments.

Fig.2 Proposed system architecture

The proposed AIOT wearable device for heart disease collects data, measuring heart rate, rhythm, and activity.
detection comprises essential modules. The sensor module The data processing module analyzes this data using
37
machine learning algorithms. The connectivity module CNN model is trained on the preprocessed and augmented
transmits processed data to the cloud via Wi-Fi or data using a suitable optimization algorithm such as
Bluetooth Low Energy, receiving updates in return. The stochastic gradient descent (SGD). The training process
cloud computing module processes data further, generating involves adjusting the weights of the CNN model to
reports on the wearer's heart health status. The user minimize the difference between the predicted output and
interface module provides a graphical display for the the actual output. Validation: The trained CNN model is
wearer, showcasing heart metrics. The power management validated on a separate dataset to evaluate its performance.
module efficiently handles power consumption, potentially The validation dataset should be different from the training
utilizing a rechargeable battery or energy harvesting dataset to ensure that the model can generalize well to
techniques to extend device battery life. This unseen data
comprehensive system integrates these modules to ensure
effective heart disease monitoring.

B. Implementation Details

1)Technique CNN

Convolutional Neural Networks (CNN) can be used in an 2) Hardware and Software Specification
AIOT wearable device for heart disease detection to
extract meaningful features from the raw sensor data
collected from the wearer's body. Here's a high-level
overview of how CNNs can be used in such a device: Data
collection: The wearable device collects data from various
sensors, including heart rate and ECG sensors, which
capture heart rate and rhythm data. Data preprocessing:
The collected data is preprocessed to remove noise and
artifacts that can affect the performance of the CNN
model. Data augmentation: The data can be augmented to
increase the size of the dataset, which can improve the
performance of the CNN model. For example, the data can
be randomly shifted in time or scaled in amplitude. CNN
model design: A CNN model is designed to process the
preprocessed and augmented data. The CNN model
typically consists of convolutional layers, pooling layers,
and fully connected layers. The convolutional layers
extract features from the input data, the pooling layers
reduce the spatial dimensions of the output from the
convolutional layers, and the fully connected layers
classify the input data into different classes. Training: The

IV. RESULT AND DISCUSSION An experiment is conducted in order to identify the

input/output behavior of the system. Identify inputs. Specify the
A. Standard Dataset Used: sample inputs that would be used in the experiments. The sample
dataset used in the experiment are identified and given in Table III

38
 Table III. Sample Dataset Used for Experiment

fig.4.upload an ecg image to predict the heart condition

fig.5. ECG analysis result high. Precision is the ratio of the number of samples that are
truly classified as positive and the sum of the number of
B. Evaluation Parameters samples truly classified as positive and falsely classified as
positive.
For analyzing the capability of the implemented method, the
following matrices were implemented and used to evaluate our
model.
1. Confusion Matrix: This matrix is used to calculate the degree
of error and correctness in classified objects. In the confusion 4. Recall: Recall is the ratio of the number of positively
matrix, the TP and TN are a total number of samples that are classified samples that are actually positive and the total
truly classified, where TP stands for true positive, and TN is number of correctly classified samples. A recall matrix was
true negative. FP and FN are the total numbers of samples that used to calculate the number of actual positives the model
are incorrectly classified, where FP is false positive, and FN is captured from the testing dataset.
false negative.
2. Accuracy: Accuracy shows the performance of a model in
terms of correctly classified samples. Accuracy can be
determined by the confusion matrix, as is it is the ratio of the
sum of the truly classified samples, both positive and negative, 5. F1 Score: The F1 Score is the weighted average of precision
and the total number of samples in the dataset. and recall. When seeking a balance between the recall and
precision an F1 score is needed.

3. Precision: Precision is mostly used to determine when the

number of samples that are falsely classified as positive is very 6. Error Rate: The error rate is the measure of the percentage of
total samples that are wrongly classified, either as positive or as
39
negative. This matrix was used to calculate the error rate of the monitor patients, and predict a person's risk of developing
CNN model on the given dataset. cardiac disease.Deep learning leverages a suite of techniques,
such as Convolutional Neural Networks (CNN), Recurrent
Neural Networks (RNN), and Long Short-Term Memory
networks (LSTM), to identify and isolate anomalous patterns
within various datasets. Additionally, the system can be used in
C. Applications clinical trials to monitor patients' heart health and track the
effectiveness of new treatments.[3] Overall, an AIoT wearable
1) Social Applications system for cardiac disease detection can increase access to
healthcare and improve patient outcomes.
Early detection of cardiac disease: Wearable devices can
continuously monitor a person's heart rate, rhythm, and B. Future Scope
other vital signs, which can be analyzed using AI algorithms
to detect signs of cardiac disease at an early stage. This can The future scope of Artificial Intelligence of Things (AIoT)
help doctors to intervene before the condition worsens and wearable systems for cardiac disease detection is promising,
improve patient outcomes.[9] with significant potential for advancements in healthcare.
Here are some key areas of growth and development:
Personalized treatment: AI algorithms can analyze the data
collected by wearable devices to create personalized Early Detection and Prevention: AIoT wearable devices can
treatment plans for patients with cardiac disease. This can continuously monitor an individual's vital signs, such as
take into account the patient's unique medical history, heart rate, ECG, blood pressure, and oxygen saturation.
lifestyle, and other factors, leading to more effective and They can detect abnormal patterns and alert users or
tailored treatments.[9] healthcare professionals in real-time, allowing for early
intervention and prevention of cardiac diseases.[1]
. 2) Technical Applications
Personalized Healthcare: AI algorithms can analyze data
from wearable devices to provide personalized insights and
Remote patient monitoring: Wearable devices can be used to recommendations for individuals to improve their cardiac
remotely monitor patients with cardiac disease, allowing health. This includes tailored exercise regimens, dietary
doctors to track their progress and adjust treatment plans as advice, and medication management.
needed. This can be particularly useful for patients who live
far from a medical facility or who have difficulty traveling. Telemedicine and Remote Monitoring: The integration of
AIoT with telemedicine platforms allows for remote
Risk stratification: AI algorithms can analyze data from monitoring of patients with cardiac conditions. This reduces
wearable devices to predict a person's risk of developing the need for frequent hospital visits, making healthcare more
cardiac disease. This information can be used to identify accessible and cost-effective.
high-risk patients who may benefit from early intervention
or preventative measures.[9] Data Analytics and Research: The data collected from AIoT
wearable devices can be used for large-scale research and
Clinical trials: Wearable devices can be used in clinical epidemiological studies. Researchers can gain valuable
trials to monitor patients' heart health and track the insights into cardiac disease trends, contributing to the
effectiveness of new treatments. This can help researchers to development of better prevention and treatment
develop more effective therapies for cardiac disease. strategies[3].
Overall, the use of wearable devices and AI in cardiac
disease detection has the potential to improve patient Continuous Monitoring of Chronic Conditions: AIoT
outcomes, increase access to healthcare, and reduce wearables can be used to monitor individuals with chronic
healthcare costs.[9] cardiac conditions, such as arrhythmias or heart failure,
providing healthcare providers with a continuous stream of
V. CONCLUSION AND FUTURE SCOPE data to better manage these conditions.
A. Conclusion
Improved Diagnostics: AI can assist in the interpretation of
An artificial intelligence of things (AIoT) wearable system for cardiac data, potentially improving the accuracy of
cardiac disease detection has the potential to improve patient diagnostics. AI algorithms can quickly identify anomalies
outcomes and reduce healthcare costs. The system can monitor and help in the early diagnosis of cardiac diseases.[2]
a person's heart rate, rhythm, and other vital signs continuously,
and use AI algorithms to detect signs of cardiac disease at an User-Friendly Design: The development of more
early stage.[5] The system can also provide personalized user-friendly and comfortable wearable devices will
treatment plans for patients with cardiac disease, remotely encourage broader adoption among individuals of all ages.
40
 This will lead to a more extensive data pool for analysis and thoracic electrostatic charge variation for long lasting wearable
research. devices.”.(2022)

Integration with Smart Healthcare Ecosystems: AIoT [3].Meenali Janveja, Mayank.”Design of Low Power VLSI
wearables can be integrated with other healthcare devices Architecture for Classification of Arrhythmic Beats Using
and systems, such as electronic health records (EHRs) and DNN for Wearable Device Applications”.(2021)
hospital information systems, creating a seamless healthcare
ecosystem for cardiac disease management.[7] [4]Maryam Khan, Qareena Yaseen, Asia Mumtaz, Ayesha
Saleem, Seemab Ishaq, Haseeb Udeen.”Severe Analysis of
Cardiac Disease Detection using the Wearable Device by
Artificial Intelligence”(2020)
ACKNOWLEDGEMENTS
[5]Yu-Jin Lin, Chen-Wei Chhuang, Chun-Yueh Yen,
We would like to express our special thanks to Prof. Sheng-Hsin Huang,Peng-Wei Huang.”Artificial Intelligence of
Krishnendu Nair, our major project guide who guided us Things Wearable System for Cardiac Disease Detection”.(2019)
through the project and who helped us in applying the
knowledge that we have acquired during the semester and [6]Sumaiya Shomaji, Domenic Forte, Alex Roman,Swarup
learning new concepts. We would like to express our special Bhunia.”Early Detection of Cardiovascular Diseases Using
thanks to Dr. Satishkumar Varma, Head, Department of Wearable Ultrasound Device”.(2019)
Information Technology, who gave us the opportunity to do this
major project because of which we learned new concepts and [7]Santiago Led, Jorge Fernández and Luis Serrano.”Design of
their application. We are also thankful to our major project a Wearable Device for ECG Continuous Monitoring Using
coordinators Dr. Niteshkumar Agrawal and Prof. Sheetal Wireless Technology”(2004)
Gawande along with other faculties for their encouragement
and support. Finally, we would like to express our special [8]https://www.mdpi.com/1424-8220/23/2/828
thanks to Principal Dr. Sandeep Joshi who gave us the
opportunity and facilities to conduct this major project. Last but [9]https://www.xenonstack.com/artificial-intelligence-solutions/
not the least we would like to thank our parents for financing healthcare/
our education as well as encouraging us to learn engineering is
gratefully acknowledged.

REFERENCES

[1]Arti Rana, Vineet Kishore Shrivastava.”Artificial

intelligence based early detection of cardiac diseases using
smart wearables”.(2022)

[2]Kanika Dheman,David Werder.Michele Mango.”Cardiac

monitoring with novel low power sensors measuring upper

41
Copyright Certificates

42
 Acknowledgement

We would like to express our special thanks to Prof. Krishnendu S Nair , our major
project guide who guided us through the project and who helped us in applying the
knowledge that we have acquired during the semester and learning new concepts.

We would like to express our special thanks to Dr. Satishkumar Varma, Head,
Department of Information Technology, who gave us the opportunity to do this major
project because of which we learned new concepts and their application.

We are also thankful to our major project coordinator Prof. Sheetal Gawande along with
other faculties for their encouragement and support.

Finally we would like to express our special thanks to Principal Dr. Sandeep Joshi who
gave us the opportunity and facilities to conduct this major project.

Prasanna Laxman Pawar

Pranav Shashikant Bandal
Vedant Mohan Phatangare
Anit Roy Payyappilly

43