An Intelligent Systems for Heart Sound Signal Analysis

Globally, heart failure is the leading cause of death. One of the first diagnostic techniques for examining the state of the heart is cardiac auscultation. The heart's mechanical activity, which is triggered by biopotentials produced by the Sino-Atrium (SA) node, results in non-stationary and non-deterministic signals known as Heart Sound Signals (HSS). When captured by an electronic stethoscope, the graphical representation of HSS is known as phonocardiogram (PCG). PCG can provide very helpful diagnostic information when evaluating a range of cardiac abnormalities, like valvular heart disease, irregular heartbeat, cardiac decompensation, and anatomical defects in the heart. PCG is composed of the first and second (S1, S2) heart sounds, which are frequently the two loudest heart sounds. In both diseased and non-pathological situations, the PCG signal may comprise murmurs, the third and fourth cardiac sounds (S3), and other odd sounds like clicks, snaps, and rubs.PCG analysis is effective for non-invasive cardiovascular disease diagnosis. The thesis is on creating an Automatic Computerised Auscultation (ACA) for effective cardiac state monitoring. ACA provides diagnostic and prognostic data in a non-invasive, affordable, fast, and precise manner, helping clinicians detect cardiac sickness early and prevent cardiovascular disease deaths.Following the same steps for classifying heart sounds, the suggested RNN+LSTM approach deals with classifying heart murmurs as harmless (normal) and pathological (abnormal). The classification AUC for the optimal murmur classification combination was $\color {Orange}{\text{93}}{\text{.69}}\% $.