Evaluation of monofractal and multifractal properties of inter-beat (R-R) intervals in cardiac signals for differentiation between the normal and pathology classes

Abstract

In this study, the monofractal and multifractal properties of inter-beat (R-R) intervals in cardiac signals for normal and pathology classes were studied and applied to a dataset in PhysioNet which consists of 24 h R-R intervals from 54 healthy subjects (hs) and 92 patients with various diagnoses [44 with congestive heart failure (chf), 25 with atrial fibrillation (af) and 23 diagnosed with sudden death syndrome (sd)]. The results in this study indicate that the most suitable method for estimating the monofractal properties of R-R intervals is detrending moving average (DMA). The Hurst exponents (H) of the healthy and pathological groups, calculated using the DMA method, are shown to be statistically different by the Kruskal–Wallis test [p-value < 0.01; healthy $H_{{\rm hs}}$Hhs = 0.23(0.18–0.27); pathological group $H_{{\rm chf}}$Hchf = 0.06(0.03–0.10), $H_{{\rm af}}$Haf = 0.05(0.03–0.07), $H_{{\rm sd}}$Hsd = 0.05(0.03–0.08)]. To study the multifractal properties of R-R intervals, the multifractal detrended fluctuation analysis was used. Formal statistical tests indicate statistically significant differences (p-value < 0.05, ANOVA and pairwise testing) between the width of the spectrum of the chf group $W_{{\rm chf}}$Wchf = 1.11 ± 0.12 and the width of the spectra of the af group $W_{{\rm af}}$Waf = 0.58 ± 0.15, the sd group $W_{{\rm sd}}$Wsd = 0.60 ± 0.20, and the hs group $W_{{\rm hs}}$Whs = 0.61 ± 0.08. Using these results, a logistic regression model was developed for differentiating chf from other pathologies (af and sd).