An Efficient Computational Approach for Phonocardiogram Signals Analysis and Normal/Abnormal heart sounds diagnosis

Abstract

In the present work, the authors proposedan intelligent approach for the examination and classification of cardiac sound signals{''}phonocardiogram (PCG){''}. In this approach, an artificial neural network (ANN) is executed as indicator and classifier of PCG abnormalities using the features extracted from PCG acoustic signals via the discrete wavelet transform (DWT). To develop and validate the proposed approach, the PASCAL CHSC 2011 dataset was utilized. Thek-fold cross validation was utilized to assessthe efficiency of the proposed intelligent approach. The results demonstrate that the approachachievesa high performance compared to other classification techniques for PCG datasets. The obtained results showed an overall accuracy of 99.89%. Moreover, the proposed approachresults are compared with the recently published ones that achieved utilizing different machine learning (ML) approaches. The achieved results showed thatthe proposed system has the ability to efficientlydiagnose and classify the PCG acoustic signals. It can also assist the clinicians to take accurate decisions in detecting cardiovascular abnormalities.