Diagnosis of Acute Coronary Syndrome with CNN and LSTM Based Deep Learning Model

Abstract

Early and accurate diagnosis of acute coronary syndrome (ACS) and its subtypes is essential for patient health. The aim of this study is to develop a deep learning approach utilizing electrocardiography (ECG) signals to classify ACS and its different types. The model was constructed using a combination of convolutional neural network and long short-term memory structures to categorize ECG signals representing acute myocardial infarction with ST-elevation (STEMI), myocardial infarction without ST-elevation (NSTEMI), and healthy individuals. The dataset comprises 12-lead ECG signals collected from patients who presented with chest pain at the Erciyes University Hospital Emergency Department. ECG data were processed to remove noise using notch, low-pass, and high-pass filters, and then standardized using z-score normalization. Model performance was assessed through k-fold cross-validation, calculating metrics such as accuracy, sensitivity, specificity, precision, F1 score, and classification rate. With 5-fold cross-validation, classification accuracy was observed to be 0.928 ± 0.0172 for the ACS-Normal group, 0.891 ± 0.0083 for the NSTEMI-Normal group, and 0.886 ± 0.02275 for the STEMI-Normal group. These findings suggest that the proposed deep learning model is effective in distinguishing ACS and its subtypes, showing promise for future integration into clinical applications.