Dynamical properties of the sleep EEG in different frequency bands.

The information concerning the dynamic behavior of the sleep process gained by the usual evaluation of sleep EEGs according to the criteria of Rechtschaffen and Kales is limited. Therefore a new methodical approach is presented, which is a special case of spectral analyzed data processing. After digital band-pass filtering of the sleep EEG the root-mean-square (RMS) value of successive 20 s EEG epochs is calculated in defined frequency ranges. This procedure ensures to take into account the influence of the phase relation between different frequency components. The temporal course of these RMS values during the night reveals smooth curves with continuous transitions between different sleep states. In all frequency bands slow oscillations according to the sleep cycles are observable. Whereas the slow frequency bands have a temporal course with local maxima during non-REM and local minima during REM sleep, the fast frequency bands beta and gamma show the opposite behavior revealing higher RMS values during REM sleep. The relationship between the activities in different frequency bands is evaluated calculating the cross correlation coefficient. Taken together the procedure allows an objective and automated quantitative analysis of the sleep EEG. The main advantage of this approach is the characterization of the sleep cycle as a dynamic and continuous process. Compared to the classical analysis it provides a more detailed analysis of the sleep process, especially concerning the dynamics and microstructure of sleep.