Optimization of heart sound feature combination for HCM analysis based on BPSO

An optimization method of heart sound feature combination based on binary particle swarm optimization (BPSO) is proposed to analyze hypertrophic cardiomyopathy (HCM), which is a typical cardiovascular disease. The analysis procedure mainly includes feature extraction, feature combination optimization and HCM analysis. Firstly, 13-dimension energy features are extracted by the wavelet transformation. Secondly, feature combinations are selected by the proposed optimization method for classifying the normal and HCM heart sounds as well as the different types of HCM heart sounds. Lastly, 362 clinical cases including normal and HCM heart sounds are used to verify the validity of the proposed optimization method for heart sound feature combination, compared with principle component analysis (PCA), genetic algorithm (GA) and the analysis method without optimization of heart sound features. The average accuracy by our proposed optimization method, considering the number of features selected out, reaches to 20.91%, higher than the results utilizing feature combinations obtained by the other methods. In addition, the proposed optimization method of feature combination has potential for analyzing different kinds of HCM cases.