Segmentation and identification of some pathological phonocardiogram signals using time-frequency analysis

Abstract

Heart sounds that are multicomponent non-stationary signals characterise the normal phonocardiogram (PCG) signals and the pathological PCG signals. The time-frequency analysis is a powerful tool in the analysis of non-stationary signals especially for PCG signals. It permits detecting and characterising abnormal murmurs in the diagnosis of heart disease. In this study, the authors introduce a novel method based on time‐frequency analysis in conjunction with a threshold evaluated on Rényi entropy for the segmentation and the analysis of PCG signals. The method was applied to different sets of PCG signals: early aortic stenosis, late systolic aortic stenosis, pulmonary stenosis and mitral regurgitation. The analysis has been conducted on real biomedical data. Tests performed proved the ability of the method for segmentation between the main components and the pathological murmurs of the PCG signal. Also, the method permits elucidating and extracting useful features for diagnosis and pathological recognition.