Detection of a Cardiac Anomaly from a PCG Signal, Based on the Transition from Discrete to Continuous Point Clouds

Atherosclerosis (ATS) is one of the main risk factors for coronary heart disease (CAD). It mainly affects the medium and large arteries, the carotid arteries and the coronary arteries. ATS usually occurs in people over the age of forty and may remain asymptomatic for several years before the first symptoms appear. For the diagnosis of CAD, heart sounds can be used as a simple and non-invasive detection system. For this purpose, the phonocardiogram (PCG), the digital recording of heart sounds, is becoming more and more popular. In recent years, several artificial intelligence (AI) based works have been carried out for the diagnosis of CAD, based on the analysis of PCGs, in particular for the automated segmentation and classification of heart sounds, in order to determine the existence of a predilection in the analyzed patients. This work suffers from a high computational cost and only one study has explored the time-varying frequency characteristics of the systolic and diastolic phases of the PCG. Our method is based on the principle of the discrete-to-continuous DTC approach and attempts to find the superposition between the input signal and the reference signal by seeking a transformation based on the Euclidean metric. We evaluated our approach on the public heart sound dataset provided by the PhysioNet Computing in Cardiology Challenge 2016. In the present work we suggest, a new pattern recognition approach based on discrete to continuous point clouds to analyze and diagnose CAD. Experimental results showed good performance of the proposed approach.