An ECG Heartbeat Classification Strategy using Deep Learning for Automated Cardiocare Application

Recent Studies show exponential growth in heart diseases and cardiac ailments cases endangering the patient’s life. An automated smartphone application can ease symptoms monitoring, medical aids and treatment compliance that usually are challenging and frustrating tasks. In this work, we proposed a stepwise procedure used for the functional development of a smartphone app. To identify a cardiac situation, reliably and accurately understanding the ECG heartbeats is a complex and crucial task. In this process, precise categorization of heartbeats is critically essential. We developed a 26 layered Deep CNN-based model that correctly categorized ECG heartbeats in ‘Normal Beat(Non-Ectopic),’ ‘Supraventricular Ectopic Beat,’ ‘Ventricular Ectopic Beat,’ ‘Fusion Beat’ and ‘Unknown Beat.’ The proposed model is tested on a publicly available MIT-BIH ECG Heartbeat categorization dataset. The test results affirm that the developed method classifies the ECG heartbeats in mentioned five rhythms (classes) by an average accuracy of 98.80 percent. Furthermore, the developed classifier model accomplishes better results in terms of average accuracy while comparing it with recent similar works.