A novel approach for cardiac pathology detection using phonocardiogram signal multifractal detrended fluctuation analysis and support vector machine classification

The aim of this study is to develop a reliable method for assisting doctors in the early detection and diagnosis of heart disease by analyzing normal and abnormal phonocardiogram signals (PCG) using multifractal detrended fluctuation analysis (MFDFA).

The MFDFA technique is a model-independent method for uncovering the self-similarity of a stochastic process or autoregressive model, which allows for the extraction of the most important characteristics of the PCG signal.

These characteristics include time evolution of the local Hurst exponent (Ht), q-order mass exponent (tq), root mean square (RMS), q-order Hurst Exponent (Hq), q-order singularity exponent (hq), and q-order dimension exponent (Dq) also proved its effectiveness by 98.5075% when classifying its results in support vector machine (SVM).

The proposed method was applied using MATLAB R2022b with record signals from PhysioNet and Michigan websites. The MFDFA technique appears to be promising in heart disease study.