ECG Heartbeat Classification of Myocardial Infarction and Arrhythmia using CNN

The most common cause of death all over the world is cardiovascular disease (CVD), with Coronary Artery Disease (CAD) being the most common. To lower the risk of CAD consequences such as myocardial infarction, it is critical to diagnose and treat CAD early. The most popular diagnostic screening tool for detecting CAD through its symptoms like arrhythmia, is the electrocardiogram (ECG) readings. Current procedures for diagnosing CVDs, such as heart catheterization or blood tests, are invasive. Other non-invasive cardiac methods of testing have drawbacks, such as uncertainty about the best order, type, and frequency of cardiac imaging tests for various medical settings performance. In this paper we propose an approach to this problem using ECG values to train our respective models and focus on vertical improvement i.e. higher accuracy and better results for existing techniques rather than horizontal improvement i.e. building new techniques. We tested our techniques on PhysioNet’s MIT-BIH and PTB Diagnostics Datasets, resampling certain data wherever necessary. As per our results, our models are able to predict the categories with average accuracies of 99% and 98% on different sub-categories of Myocardial Infarction and Arrhythmia ECG signals, respectively.