Abstract

Cardiovascular diseases (CVDs) are one of the leading causes of death each year. Early diagnosis of CVDs can help to control and prevent the complication of heart diseases. Although auscultation is one of the conventional methods of CVDs diagnosis, it is not accurate enough because of the human hearing restrictions and nonstationary nature of the heart sounds. Because the heart sound or phonocardiogram (PCG) signal contains heart functional information, it can be employed to diagnose various types of CVDs. The goal of this study is to detect Mitral valve Prolapse (PMV) using PCGs. To reach the goal, first, the PCGs were denoised using the Chebyshev filter along with the Wavelet Transform (WT). Then, using the Shannon Energy Envelope (SEE) along with adaptive thresholding, the denoised PCGs were divided into the cardiac cycles. Fractional Fourier Transform (FrFT) was performed to extract the desired features in the time-frequency space. Based on the Mahalanobis distance criterion, the optimal features were selected. The results of the proposed algorithm on the 15 prolapsed and 5 non-prolapsed patients show 95.65 accuracy using the SVM classifier. © 2020 IEEE.