An SNN-Inspired Area- and Power-Efficient VLSI Architecture of Myocardial Infarction Classifier for Wearable Devices

Myocardial infarction (MI) is a cardiac abnormality in which the coronary artery gets blocked, causing millions of fatalities every year. MI has a very high mortality and disability rate; therefore, with the detection of MI, it is also imperative to determine the location of the blockage to provide on−time treatment to avoid any fatality. In this brief, for the first time, a VLSI architecture is proposed that can determine the location of the infarction in real-time. The proposed architecture classifies the electrocardiogram (ECG) into twelve classes and achieves an average accuracy, sensitivity, and specificity of 99.90%, 99.49%, and 99.94%, respectively. Its area utilization is 1.69 $mm^{2}$ at SCL $180~nm$ Bulk CMOS technology node, and the power consumption is $268.9 ~\mu W$ at 250 KHz. The low area and power requirements and real-time classification capability make the classifier suitable for wearable devices.

I. Introduction

Myocardial infarction (MI) [1], also known as heart attack, denotes atherosclerotic changes in the coronary arteries of the nutritional myocardium. A thrombus, generated by the cholesterol plaque deposited in the inner wall of the blood vessel, blocks a coronary artery or one of its smaller branches. This blockage decreases the blood supply to the myocardium, causing oxygen deprivation to the heart muscles and making them necrotic. During MI episodes, patients can experience symptoms, such as loss of consciousness, shortness of breath, and chest pain [2]. However, many patients experience mild or no symptoms, which lead to silent heart attacks. Around 22-64% of all MI cases are estimated to be silent heart attacks [3], increasing MI disability and mortality rate.