Applications of statistical equivalence to newborn EEG recordings

Epochs were selected from sleep EEGs of newborns during both quiet and REM phases. A first order parameter analysis was done in order to test the homogeneity (run test and one way variance analysis) of the data within the epochs, follwed by a study of spectral equivalence between epochs. Comparisons were made between average spectra in the frequency range 0.5--4 c/sec; and between their band-by-band components, to test the feasibility of automatic classification of sleep stages by means of frequency analysis. On the basis of the statistical parameters we used, spectral analysis appears to be of limited value for the purpose of automatic classification of newborn EEG records; due to a lack of complete homogeneity of the spectra within each sleep state, it is not possible to identify a characteristic spectrum for active sleep and quiet sleep.