ECG Classification using Deep Transfer Learning

The state-of-the-art deep neural networks trained on a large amount of data can better diagnose cardiac arrhythmias than cardiologists. However, the requirement of the high-volume training data is not pragmatic. In this research, the identification and classification of three ECG patterns are analyzed from a transfer learning prospect. The features learned from the general image classification are transferred to the time-series signal (ECG) classification using transfer learning. In this research, various modern deep networks trained on the ImageNet database are re-utilized for classifying scalograms (2D representation) of ECG signals. The performance of these deep transfers on the classification of ECG time-series data is then assessed.