Transfer Learning in Deep Neural Network Model of ECG Signal Classification

An electrocardiogram (ECG) is a simple test that can be used to evaluate a patient's cardiac rhythm and electrical activity. Cardiologists and medical practitioners now frequently employ electrocardiograms (ECGs). Currently, measuring the proper classification of heartbeats is challenging research work in which conventional machine learning algorithms are widely used to fulfill the objectives. However, more research focuses on effectively identifying a collection of circumstances on a dataset instead of learning and applying transferable knowledge across tasks. This work introduces a convolutional neural network based on a ResNet framework with 1-dimensional (1-D) convolution layers to classify five distinct heart arrhythmias according to the AAMI EC57 standard. Finally, the model is adapted for transferring the learned knowledge to the myocardial infarction (MI) task classification. The MIT-BITH and PTB Diagnostic datasets from physioNet are employed to test the suggested technique. Experimental outcomes exhibit that the accuracy of arrhythmia classification on the MIT-BITH dataset is 98 percent, while MI classification on the PTB dataset is 97 percent, according to our proposed method.