MSFNet: Multi-Stage Fusion Network for Myocardial Infarction Classification

Electrocardiogram (ECG) is a critical diagnostic tool in healthcare, and numerous deep learning models have been explored and applied to map ECG patterns to detect heart abnormalities. In this paper, we propose a novel network called Multi-stage Fusion Network (MSFNet), a multi-stage architecture network capable of accurately localizing abnormal segments within ECG sequences and performing myocardial infarction (MI) type diagnosis based on these segments. From the experiments on PTB-XL datasets, we observed that MSFNet achieves competitive performance compared to other methods.