ECG Signals- Early detection of Arrhythmia using Machine Learning approaches

Heartbeats can be recorded in the form of electrical signals using a machine called an electrocardiogram. The cardiovascular system's performance can be tracked using an electrocardiogram (ECG), which can also be used to diagnose various cardiac diseases. The paper by Mohammad Kachuee et al., entitled "ECG Heartbeat Classification: A Deep Transferable Representation" classifies heartbeats into five different classes according to AAMI EC57 standards. In this paper, we continue the work on their preprocessed dataset and train the model to detect arrhythmia, i.e., abnormalities in the heartbeat. We also propose a comparative study between different machine learning and deep learning models, namely random forest classification, support vector machines, and convolutional neural networks, for detecting abnormalities in heartbeats. The results we got after the application of all our classifiers showed the best accuracy score of 98.26% for the MIT-BIH dataset and a 100% accuracy score for the PTB Diagnostic dataset.