Heartbeat classification using a deep neural network

Abstract

As we age the cardiovascular system weakens and is more prone to cardiovascular diseases. Those cause arrhythmia, which is an abnormal heartbeat rhythm that can be life-threatening. An electrocardiogram (ECG) is the principal diagnostic tool used to record and interpret heart activity. Therefore, ECG are complex signals contain information about the each heartbeat, but are difficult to manually analyze. Hence, a computer-aided diagnosis (CAD) system is proposed to classify the different types of heartbeat and ensure that the assessment of ECG signals is objective and accurate. The method used for feature extraction of the heartbeats, once they are segmented form the original ECG signal, is the independent component analysis (ICA) of discrete cosine transform (DCT) coefficients. A deep neural network classifier is used to cluster heartbeats into one of 13 or 5 classes, corresponding to class-based or subject-based assessment strategies, by using those two kinds of features (one for each lead). The method acquires an overall accuracy of 97.68%, in the class-based assessment strategy and 97.87% in the subject-based assessment strategy, based on the MIT-BIH arrhythmia database. These results show that the proposed automated diagnosis system provides high reliability to be used by clinicians. The method can be extended for detection of other abnormalities of heart and to other physiological signals.