On the analysis and classification of heart sounds based on segmental Bayesian networks and time analysis

This paper describes a Computer-aided Heart Sound Analysis and Classification System (CHACS) to automatically detect and identify the heart sound components of the phonocardiogram based on segmental Bayesian networks and time analysis. In this system, two subsystems in both time and frequency domains are proposed. In the frequency-domain subsystem, a segmental Bayesian network is proposed to model each heart sound template. The generalized probabilistic descent (GPD) algorithm is adopted for discriminative training. In the time-domain subsystem, based on doctor's auscultation manner, the onset time difference between each heart sound and reference heart sound is statistically estimated and used for heart sound detection. The individual classification results of these two subsystems are combined to yield the final results. For experimental evaluation, twenty-two patients with heart disease and thirty-six healthy individuals are being requested to provide a 20-second recording of heart sounds in a silent environment. Experimental results shall demonstrate that these two subsystems complement each other and a classification rate of 92.9% is obtained.