Comparative Evaluation for Two and Five Classes ECG Signal Classification: Applied Deep Learning

Abstract

ECG can be using to reliably monitor the health of a cardiovascular system.The proper classification of heartbeats has received a lot of attention recently. While there are numerous similarities among ECG situations, instead of learning and to use transferable knowledge across tasks, most research has concentrated on classify a group of situations using a dataset labeled for that task.This research suggested a technique for heartbeat classification based using deep CNN modelsthat can accurately categorize distinct arrhythmias for two and five classes ECG signals in line to AAMI EC57 standards. The authors also propose a strategy for transferring this job's knowledge to the myocardial infarction (MI) categorization problem. Physician Net's MIT-BIH and PTB Diagnostic databases were used to evaluate the suggested technique. The ECG data was gathered from various Physio Bank databases, which provide clinical research data freely available. Images of ECG signals with time-frequency encoding were fed into architecture such as CNN, LSTM, Alex Net, VGG-16, Resnet50, and Inception. The categorization of ECGs was completed, as well as the performance of CNN, LSTM, Alex Net, VGG-16, Resnet50, and Inception architectures were evaluated using a transfer learning technique and modifications in particular output layers for fivedesigns.