Classifying heart conditions based on class probability output networks

Abstract

This paper presents a novel method of classifying heart conditions from an electrocardiography (ECG) signal. For this purpose, the R-R intervals of ECG signal are analyzed by Gamma distribution parameters and classified into normal (NR) or abnormal (AN) ECG waves. For the normal ECG waves, the heart condition is further investigated by analyzing the dynamic behavior of heart activity based on the correlation between successive R-R intervals and long-term analysis. The classification of heart conditions is made by estimating the conditional class probabilities using class probability output networks (CPONs). The simulation for classifying heart conditions using the MIT-BIH data sets reveals that the proposed approach is effective for classifying heart conditions and allows more accurate classification than the existing classifiers such as the k-NN and SVM. (C) 2019 Published by Elsevier B.V.