Comparison between spectral and fractal EEG analyses of sleeping newborns

Spectral parameters (power spectrum in the delta, theta, alpha, beta1 and beta2 bands) and the fractal dimension were estimated for each two seconds frame of the EEG sleep time series at the awake condition or during one of the four EEG sleep states: active sleep (two stages: mixed and low voltage irregular) and quiet sleep (two stages: quiet sleep high voltage and trace alternant). Six EEG channels were selected from 11 channels of polygraphic EEG-sleep recordings of 10 healthy 111 term newborns (5 males and 5 females), with normal psychomotor development evaluated at 24 and 36 months of age. Both methods of analysis, each with its peculiarity, proved to be effective in discriminating the various behavioural states: the spectral analysis showing that the values of the theta and delta band, and to a lesser degree, the alpha band, have different patterns according to the state of sleep, particularly in TA, during which there is a large standard deviation; the calculation of the fractal dimension showing specific variations in the morphology and complexity of the EEG traces according to the state of sleep analyzed.