Identification of deep sleep and awake with computational EEG measures.

The objective of the present work was to examine identification of deep sleep and awake with computational analysis of sleep EEG traces from central brain regions. All-night EEG traces from a total of 56 male subjects, 22 healthy control subjects and 34 age-matched apnea patients, were examined. A spectral mean frequency measure, a Hilbert transform based EEG amplitude and a correlation coefficient method were compared. The EEG amplitude provided a good identification of deep sleep, reaching 86.25% but was relatively poor in the identification of wakefulness, reaching 39.06%. Mean frequency provided a relatively good identification of deep sleep and awake, reaching 84.66% and 77.67%, respectively, while the correlation coefficient produced the lowest results of 37.89% and 44.43%. Optimal threshold values for deep sleep and awake identification were determined as 4.20 and 9.76 Hz, respectively, for the mean frequency measure. Mean frequency measure can be used to provide overall context information about sleep depth for automated sleep EEG analysis methods.