Using local scale exponent to characterize heart rate variability in response to postural changes in people with spinal cord injury

Heart rate variability (HRV) is a promising marker for evaluating the remaining autonomic function in people with spinal cord injury (SCI). HRV is commonly assessed by spectral analysis and detrended fluctuation analysis (DFA). This study aimed to investigate whether local scale exponent α(t) can reveal new features of HRV that cannot be reflected by spectral measures and DFA coefficients. We studied 12 participants with SCI and 15 healthy able-bodied controls. ECG signals were continually recorded during 10 min sitting and 10 min prone postures. α(t) was calculated for scales between 4 and 60 s. Because α(t) could be overestimated at small scales, we developed an approach for correcting α(t) based on previous studies. The simulation results on simulated monofractal time series with α between 0.5 and 1.3 showed that the proposed method can yield improved estimation of α(t). We applied the proposed method to raw RR interval series. The results showed that α(t) in healthy controls monotonically decreased with scale at scales between 4 and 12 s (0.083-0.25 Hz) in both the sitting and prone postures, whereas in participants with SCI, α(t) slowly decreased at almost all scales. The sharp decreasing trend in α(t) in controls suggests a more complex dynamics of HRV in controls. α(t) at scales between 4 (0.25 Hz) and around 7 s (0.143 Hz) was lower in subjects with SCI than in controls in the sitting posture; α(t) at a narrow range of scales around 12 s (0.083 Hz) was higher in participants with SCI than in controls in the prone posture. However, none of normalized low frequency (0.04-0.15 Hz) power, the ratio of low frequency power to high frequency (0.15-0.4 Hz) power and long-term (>11 beats) DFA coefficient showed significant difference between healthy controls and subjects with SCI in the prone posture. Our results suggest that α(t) can reveal more detailed information in comparison to spectral measures and the standard DFA parameters.