Identification of chronic heart failure using linear and nonlinear analysis of heart sound

Abstract

Chronic heart failure (CHF) is a cardiac condition caused by various of cardiac diseases in the end stage. This paper employed the linear and nonlinear approaches to analyze the heart sound (HS) signals from the patients with CHF. The linear approaches include the time and frequency domain analysis. The nonlinear parameters include largest Lyapunov exponent, correlation dimension, sample entropy and the width of multifractal spectrum, which describe the chaos, fractal characteristics and complexity of the HS signals. Statistical test and receiver operating characteristic (ROC) curve analysis have been applied to the characteristic parameters extracted from the HS signals of the healthy subjects and CHF patients. The results show that the statistically significant differences of linear and nonlinear features between the healthy and CHF groups can be observed. Compared to the healthy people, the cardiac mechanical activity of the patients with CHF has a decreased chaotic characteristic, complexity and randomness, and it indicates the HS features could be the measure to distinguish the CHF patients from the healthy subjects. Hence, our study suggests the proposed features could be as supplementary indexes or efficient clues for the diagnosis of CHF.