Advancing Heart Beat Classification in Alignment with AAMI EC57 Standards for Enhanced Medical Instrumentation

Abstract

An electrocardiogram, commonly known as an EKG or ECG, captures a visual representation of the heart's electrical conduction. Physicians use this diagnostic tool to detect a wide range of cardiovascular activities by identifying deviations from the normal pattern. Tracking the performance of the cardiovascular system is one of the applications of an electrocardiogram (ECG). Recently, there has been increased attention to reliable heartbeat classification due to greater similarities among various ECGs. In this study, we propose a method for classifying heartbeats based on a generalized linear model, which, in accordance with the AAMI EC57 standard, accurately classifies five different arrhythmias. To facilitate portable representation, we utilize a dataset in our research. Additionally, we introduce a transfer method to apply the knowledge gained from classifying arrhythmias to the challenge of identifying myocardial infarctions (MIs). We evaluated our approach using the PTB Diagnostics and MIT-BIH datasets from Physion Net. The results show that, on average, our proposed method achieves an accuracy of 95.96% in predicting arrhythmias.