Stacked Variational Autoencoder in the Classification of Cardiac Arrhythmia using ECG Signals with 2D-ECG Images

Abstract

Cardiac Arrhythmia is an endangered signal to human life. Most arrhythmias are shortfalls of symptoms. Electrocardiogram (ECG) is a non-invasive, low-priced, and powerful tool to record the electrical signals of the heart and detect severe cardiovascular diseases. ECG interpretation is generally done by the cardiologist which is time-consuming and sometimes may lead to the wrong diagnosis. Moreover, numerous arrhythmia heartbeats remain unexplored. The arrhythmia record has an expeditious and divergent ECG. Prompt diagnosis of arrhythmia can reduce the mortality rate. In this paper, we have investigated the application of a Stacked Variational Autoencoder (SVAE) for the automatic diagnosis of arrhythmia from ECG signals. Furthermore, the augmented dat aset is used for training the model, to resolve the imbalance in the classes. The proposed model reached an overall accuracy of 98.96% and sensitivity of 97.32%. SVAE classified twelve classes of cardiac arrhythmia including normal sinus rhythm.