A deep learning-based myocardial infarction classification based on single-lead electrocardiogram signal

Abstract

Acute myocardial infarction (AMI) carries a significant risk, emphasizing the critical need for precise diagnosis and prompt treatment of the responsible lesion. Consequently, we devised a neural network algorithm in this investigation to identify myocardial infarction (MI) from electrocardiograms (ECGs) autonomously. An ECG is a standard diagnostic tool for identifying acute MI due to its affordability, safety, and rapid reporting. Manual analysis of ECG results by cardiologists is both time-consuming and prone to errors. This paper proposes a deep learning algorithm that can capture and automatically classify multiple features of an ECG signal. We propose a hybrid convolutional neural network (CNN) and long short-term memory (LSTM) for automatically diagnosing MI. To generate the hybrid CNN-LSTM model, we proposed 39 models with hyperparameter tuning. As a result, the best model is model 35, with 86.86% accuracy, 75.28% sensitivity and specificity, and 83.56% precision. The algorithm based on a hybrid CNN-LSTM demonstrates notable efficacy in autonomously diagnosing AMI and determining the location of MI from ECGs.