Combining 1D CNN and LSTM for Automated Myocardial Infarction Detection from ECG Signals

Myocardial infarction (MI) is a life-threatening condition that requires rapid and accurate diagnosis. In this study, we propose a novel approach to detect and localize MI using a combination of 1D convolutional neural network (CNN) and long short-term memory (LSTM) algorithm from digital ECG records from the PTB-XL dataset. The ECG signals were retrieved from the dataset and denoised. The Q wave amplitude, T wave amplitude, and ST deviation were extracted as features of 12 lead ECG to give a 36-dimensional feature vector. This feature vector was then passed to the combination of 1D CNN and LSTM architecture to give the output. Our results showed that the proposed approach achieved an accuracy of 95.1%. Our study demonstrates the potential of hybrid algorithm by combining the 1D CNN and LSTM algorithms for accurate detection and localization of MI from digital ECG records.