A Machine Learning Approach for STEMI and NSTEMI Diagnosis Using Empirical Wavelet Transform of Multi-Lead ECG Signals

Today, with technological advancements, numerous computer- aided systems are being developed for diagnosing heart diseases using Electrocardiography (ECG) signals. Myocardial infarction (MI), a cardiovascular disease, occurs due to insufficient blood supply or blockage of blood flow to the heart. Analyzing ECG signals is crucial for the early diagnosis of MI. In this study, a novel approach involving feature extraction and machine learning algorithms is proposed to classify different MI conditions (STEMI, NSTEMI) with a healthy control (HC) group using 12-lead ECG signals. In the proposed method, after preprocessing the 12-channel ECG signals, a classification study was conducted using features extracted from sub-band signals obtained via Empirical Wavelet Transform, classifying the groups into HC-NSTEMI, HC-STEMI, HC-NSTEMI-STEMI, and NSTEMI-STEMI. According to the results, the HC-STEMI group was classified with the highest performance, achieving an AUC of 0.996 using Artificial Neural Networks through 5-fold cross-validation. In conclusion, this study demonstrates the significant potential of multi-channel ECG signal analysis for MI diagnosis, contributing to future studies in this field..