Automated recognition of EEG changes accompanying arousal in respiratory sleep disorders.

Daytime sleepiness and impaired cognitive function can be a consequence of recurrent transient arousal from sleep, associated with abrupt changes in the electroencephalogram (EEG). EEG is normally assessed by trained observers from paper records, but automation offers the advantages of speed and objectivity. We assessed 10 automated indices of EEG activity as potential indicators of arousal. Arousals from light, slow wave and rapid eye movement sleep were studied in 30 subjects. Segments of EEG recorded immediately before and after each arousal were analyzed by automated measurement of 10 EEG indices using a personal computer. We investigated the ability of each index to recognize arousal while rejecting change due to variability during sleep. Nine of the 10 indices showed significant changes with arousal (p < 0.001); the better indices were related to EEG frequency, and 3 were chosen for further study. In these indices, the mean changes with arousal were 3.8 Hz (ZeroCross), 1.7 Hz (Hjorth's Mobility) and 1.2 Hz (FrqMean, an index of central EEG frequency). With none of these three indices were significant differences in performance due to base sleep stage or subject group found. We conclude that detection of arousal is feasible using automated methods that measure simple indices related to the frequency of the EEG waveform.