Cardiac Arrhythmia Classification based on the RMS Signal and Cyclostationarity

Abstract

The goal of this paper is to present a method of classification for cardiac arrhythmias. This method is based on the analysis of the (RMS) signal extracted from the electrocardiogram (ECG) leads and the heart beat cyclostationarity nature. The coefficients of spectral correlation fuction are useful in order to differentiate among the five types of beats: Normal (N), Premature Ventricular Contraction (PVC), Premature Atrial Contraction (APC), Left Bundle Branch Block (LBBB) and Right Bundle Branch Block (RBBB). Some simple statistical measures for spectral correlation coefficients are extracted while Principal Component Analysis (PCA) is used to reduce its features. Finally, the classification using Artificial Neural Networks (ANNs) is implemented. Two experiments were performed, one that used a multi-lead ECG signal, and the other that used the RMS signal as input. These experiments showed that when using the RMS signal better results are obtained. This last method led to some specific results: sensitivity, specificity and accuracy indexes of 97.68%, 99.08% and 98.72%, respectively.