Nonlinear-analysis of human sleep EEG using detrended fluctuation analysis

Abstract

Quantification of the fractal scaling properties of human sleep EEG dynamics was sought and each normal sleep stage was compared with that of sleep apnea. The fractal scaling exponents that quantify power-law correlations were computed using detrended fluctuation analysis. Six healthy subjects, aged 30-35 years, participated and six recordings of the apnea were acquired from MIT/BIH polysomnography database. The data were 8-h baseline recordings (23:00-07:00 h). The EEG signals from the C4-A1 derivation were acquired with a resolution of 250 Hz. The sleep stages were visually scored for 30 s epochs, according to the criteria of Rechtschaffen and Kales. The mean scaling exponents increased from the awake stage to stages 1, 2 and 3-4, but decreased during rapid eye movement (REM) sleep. The scaling exponents of the apnea were lower than those of the healthy subject for all the stages. The scaling exponents could be attributed to the fractal nature of EEG, which would be more appropriate for describing the complexity of EEG due to its assumption of non-stationarity.