Prediction of Heart Disease using ECG


Team

Table of Contents

  1. Introduction
  2. Our Project
  3. Requirement Analysis
  4. Software Models
  5. System Overview
  6. Machine Learning Aspects
  7. Links

Introduction

A Myocardial Infarction or a Heart Attack occurs due to the lack of blood floor to heart muscles. This is an emergency medical condition where your heart muscles begin to die which could lead to permanent heart damage and even death. The main cause of a heart attack has been identified as Coronery Artery Disease i.e. blockages in the tubes of the circulation system. Coronery Artery Disease The sure and the well-known way of knowing the risk of a MI is by performing an invasive angiogram. This method detects blokages using x-rays that are taken during injection of a contrast iodine dye.

What are the issues associated with this procedure?

What is Angiogram?

Our Project

Our focus is to develop a solution to predict a potential myocardial infarction using a more accessible, less complicated and a more affordable method. Our approach is to develop an ECG based method to address this.

What is ECG?

An ECG (ElctroCardioGram) is used to analyze arrythmias i.e.irregularities in heart rhythms. There’s an electrical system in the heart that conducts electric signal impulses that produces the pqrs waves on the ECG strip. These arrythmias occur when the electrical signals aren’t working properly. Because injured heart muscles aren’t able t0 conduct electrical impulses normally, ECGs usually identify a heart attack has occured or is in progress.

How can we incoporate ECG data in our solution?

A point to note in this particular domain is that there is no straight-forward way of detecting potential MIs directly from an ECG. Usually it is used to identify whether a heart attack has occured which produces abnormal heart rhythms on the ECG. We are developing an approach to identify potential MIs through minor arrythmias that are not very visible at early stages to the naked eye, through an automated computerized system.

What is the benefit of our solution?

Since ECGs are more accessible, affordable and less-complicated, people can get predictions on potential MIs through our system which helps them to take actions beforehand without having to wait for much more advanced tests. Then they can proceed with further actions and tests.

Existing solutions

Various reaseaches have been done on this area. These projects have focused on automatic extraction of relavant and reliable information from ECG signals that has not been as easy task for a computerized system and classifying into heart disease classes. Read more about this research Our approach is extending these reasearches to predict a risk of a MI.

Requirement Analysis

Functional Requirements

Non-functional Requirements

Software Models

Main epics of the system

Use Case Diagrams

High level software system Auth System Predictor system Tracker

UML Class Diagram

UML Class Diagram

System Overview

Machine Learning Aspects

We are developing a machine learning based approach to predict potential heart diseases. The dataset we have used for this is ECG Heartbeat Categorization Dataset which has been derived from the PTB Diagnostic ECG Database

High-level implementation process

  1. Develop various models to predict heart diseases and compare accuracies.
  2. ECG pre-processing
    • Analyze the ECG under 16 channels and extract features at a frequency. A signal processor has to be developed for this
    • Pad for missing values ML workflow
  3. Predict heart diseases using the developed ML algorithm

Current progress

Currently binary classifier models have been developed as the initial step using the following machine learning algorithms.

progress progress

…..