Autoencoder and Machine Learning Method for Myocardial Infarction (MI) Detection Application

Cardiac events may occur intermittently in daily life, in this light, this paper proposes a standalone application to diagnose myocardial infarction (MI) based on machine learning and autoencoder method. To capture these heart abnormalities an initial prototype data used from a well-known platform, physionet platform. The novelty of the work is in how it would localize the heart abnormalities as well as the final results show that our model is not only more efficient in its efficiency compared with related work in terms of accuracy, but also competitive in terms of MI detected stage whether its acute or subacute. It is envisioned that the proposed research and resulting application will assist frontline workers (i.e., doctors, nurses, technicians, cardiologists, etc.) to screen patient's ECG recording, and ultimately, diagnose and support the patient with an accurate and rapid response.