Multi-Channel Seismic Network-Inspired Myocardial Infarction Detection and Localization Models

Cardiovascular diseases (CVDs) are one of the most significant public health issues globally, Among CVDs, myocardial infarction (MI) has a particularly high mortality rate. Early prediction of MI can reduce related complications and mortality rates. Inspired by the multi-channel seismic network, named Universitat Politecnica de Catalunya- Universidad Central de Venezuela (UPC-UCV), this study improves the moving step and order of filters of UPC-UCV to construct a MI detection model and name it UPC-UCV-MI. Meanwhile, we combine UPC-UCV with a Transformer Encoder to form a hybrid network (named MI-Hybrid) for locating the occurrence site of MI. We test the robustness of the proposed framework by using intra-patient and inter- patient experimental protocols. Under the intra-patient protocol, the average accuracies of MI detection and location have reached 99.96% and 99.67% respectively. Under the inter-patient protocol, the average accuracies of MI detection and location have reached 95.69% and 73.33%, which are better than other related models. Our models can quickly and accurately detect and locate MI, with good clinical application prospect.