Fractal Dimension of EEG in Sleep Onset

Sleep onset is typically monitored using EEG, EMG and EOG. Scoring rules of Rechtschaffen and Kales (RK) are normally used in 30 s epochs for scoring wakefulness (S0) and sleep stages (S1, S2, S3, S4 and REM). One daytime sleepiness test is Maintenance of Wakefulness Test (MWT) where the subject is instructed to stay awake under soporific circumstances. We analyzed 17 of 40 minute MWT recordings using linear and non linear analysis to study vigilance fluctuations beyond 30 s RK scoring. Using adaptive epoch lengths with minimum epoch duration of 2 seconds a total of 768 wake-sleep and sleepwake transitions were used to study change in fractal dimension (FD) in sleep onset and in following awakenings. Time domain fractal dimension was calculated using Higuchi algorithm (kmax=8, 16) for 2 seconds before and 2 seconds after each sleep onset and awakenings. Ratios of dimensions were calculated. Using kmax=8: For sleep onset (S0-S1) transitions fractal dimension changed from 1.28±0.08 to 1.32±0.09 and fractal dimension ratio was 1.03±0.07 and for sleep-wake (S1S0) transitions ratio was 1.00±0.06. Using kmax=16: For sleep onset (S0-S1) transitions fractal dimension changed from 1.52±0.09 to 1.52±0.10 and fractal dimension ratio was 1.00±0.05 and for sleep-wake (S1-S0) transitions ratio was 1.04±0.06. Findings using kmax=8 are in contradiction of lower frequencies (theta activity of S1) resulting in lower FD than higher frequencies (alpha activity of S0). Our results indicate that fractal dimension could be used as an assisting parameter in computer assisted sleep onset detection. Key-Words: EEG, Fractal Dimension, Sleep Onset, Adaptive Scoring, MWT