Aims: We investigate RR interval (RRI) correlations in subjects with obstructive sleep apnea (OSA) and congestive heart failure (CHF), two conditions with differing impacts on autonomic regulation. Our aim is to identify distinct nonlinear heart rate variability (HRV) features that differentiate these pathologies, with healthy subjects representing an intermediate baseline. Previous studies have reported HRV alterations in both conditions, but a systematic comparisons across these groups remains lacking.
Methods: We analyze nighttime Holter and polysomnography ECG recordings, corresponding to typical sleep hours. We include three groups: healthy controls (N=93), OSA patients (N=42), and CHF patients (N=44). We apply second-order scale-dependent detrended fluctuation analysis (SDFA-2), focusing on the scaling exponents a(s) and their gradients. Patients are classified with XGBoost ensemble method and compared into combined classifier of mean RR and conventional HRV measures (RMMSD, SD1/SD2 and LF/HF).
Results: SDFA-2 a(s) shows significant differences between all the groups (Fig.~1). The CHF population has significantly lower a(s) values in the shortest scales, whereas OSA patients have significantly higher a(s) values at scales 20-30 compared to healthy controls. The differences in the mean behavior between groups are evident, although individual variability is substantial, as indicated by the wide distributions of a(s) values. The confusion matrix in Fig.~1 illustrates the classification performance of the SDFA-2, with the highest accuracy observed for the healthy group (87%), followed by OSA (72%) and CHF (72%). It outperforms the combined classifier of the conventional HRV measures with 85%, 72% and 54% for healthy, CHF and OSA respectively. SDFA-2 a(s) reveal distinct autonomic regulation patterns among healthy, OSA, and CHF subjects, offering potential improvements in early diagnostic.