Analysis of the first heart sound using the matching pursuit method

It is acknowledged that the first heart sound S1 consists of two major, high-frequency components M1 and T1, corresponding, respectively, to the vibrations of the mitral and tricuspid valves and their surrounding tissues following valve closure in early systole. In this study, the matching pursuit (MP) method was used to decompose S1 into a series of time-frequency atoms. M1 and T1 were separated from the parameterised atoms of S1. The first two dominant frequencies of M1 were identified and used as features of a linear classifier to diagnose mitral valve abnormality. This method was applied to two sets of S1 data recorded from 15 patients with normal, and 15 patients with abnormal, bioprosthetic mitral valves, respectively. It was found that the two features exhibit significant differences between the normal and abnormal sets (p< 0.001). Using these two features, a correct classification of 93% was obtained. In addition, when the Wigner distribution of S1 was calculated from the decomposed atoms and compared with a spectrogram, the MP method provided better results. The study demonstrates that the MP method may be a promising technique for heart sound analysis.