Investigation of Valvular Heart Defects Through Phonocardiogram Signals

Abstract

Cardiovascular disease (CVD) is a serious illness that affects the world. Early detection and prevention of CVD is thought to help reduce CVD mortality rates. Valvular heart defects will be challenging to diagnose without echocardiogram. Despite the fact that this method is relatively reliable, both the device and the process are time-consuming where this can be dangerous for those who require rapid medical attention. Therefore, this study would concentrate on the use of phonocardiogram (PCG) signals to provide early screening assessment for valvular heart defects such as aortic stenosis (AS), mitral stenosis (MS), mitral regurgitation (MR) and mitral valve prolapse (MVP). Signal processing techniques which involves pre-processing, segmentation, feature extraction and classification are applied to analyze PCG signal. Daubechies wavelet with 5th order (Db5) with 7th level of decomposition is used to remove undesirable signal in PCG signal and reconstructed back from 1 to level 6 Daubechies wavelet with 5th order. The signal was then segmented using average Shannon energy. The segmented signal is then entered into feature extraction process. Then, feature extraction was done in time-frequency analysis by using Stockwell transform which is also known as S-transform. Finally, the classification process was done using K Nearest Neighbor (KNN), Support Vector Machine (SVM), and Ensemble classifier on all dataset with an overall best accuracy of 96.32%, 94.88% and 98.02%, respectively. This study’s outcome would be an advancement diagnostic tool that capable of analyzing PCG signal data and helps physicians by providing early detection for any valvular heart defects.