A method for assessing alertness fluctuations from EEG spectra.

The object of the study was to develop a method for automatic evaluation of the vigilance level using the EEG frequency pattern. Forty-one EEG records were selected as training material, being considered representative of either a fully alert EEG, or sleep stage I. In the same epochs, 22 different variables based on EEG spectral analysis were calculated. Multiple regression analysis was applied to determine the correlation between the spectral values and the assigned levels of vigilance. The coefficients obtained were applied to the spectral values of 265 epochs belonging to another set of tracings, for estimation of the vigilance level. The correlation between the automatic and visual finding was 0.89 in the training material and 0.86 in the test material. The method can be used to evaluate the mean vigilance level and its variance, rhythmicity and trend during the examination. The preliminary results suggest that objective measurement of alertness can be of value in many applications, such as in pharmacological research, in evaluation of certain psychic symptoms and in working environment studies.