Abstract

Abstract Cardiovascular disorders (CVDs) have attracted a great deal of attention given that they are one of the most common causes of mortality every year. With early diagnoses, CVDs can be controlled before reaching the critical stage. Listening to heart sounds has been a common way of evaluating the cardiovascular system. The sound signal generated by the heart’s mechanical activity provides useful information on the operation of the heart valves. Due to human hearing limitations as well as transient and non-stationary nature of heart sounds, diagnosis based on auscultation is difficult and requires training, repetition, and expertise. The purpose of this study is to diagnose and assess the severity of tricuspid regurgitation (TR), a disease that most adults suffer from without being aware of it, by using phonocardiogram (PCG) and without the employment of auxiliary tools such as the electrocardiogram (ECG). To this end, PCG signal is first preprocessed using Shannon energy envelope (SEE) and Hilbert-transform (HT). The Mel Frequency Cepstral Coefficient (MFCC) is used along with the wavelet transform method for feature extraction; superior features are then selected with the genetic algorithm (GA). The accuracy of the proposed method was 98.78 ± 0.95% and kappa was 98.12 ± 1.48% with k-nearest neighbors (KNN) classifier.