Waking and sleeping states in the rat from an EEG data analysis point of view.

This article presents the characteristics of ECoGs of arousal, slow wave sleep and paradoxical sleep in the rat, in terms of analysis of data. In a first part, we have applied four different methods of analysis to the three tracings: the instantaneous amplitude histograms computation, the integrative method of Drohocki, the spectral analysis and the normalized slope descriptor method of Hjorth. Each method provides, after data reduction, characteristic parameters of the tracings. A graph which displays peak spectral frequency versus mean integrated value is enough to discriminate between the 3 quantified tracings. Multivariate discriminant analysis reveals that 3 coefficients altogether allow a good discrimination. In the second part we ask the question: which kind of signal is the paradoxical sleep tracing? After the impossibility to choose between a narrow-band Gaussian process or a sinusoidal wave burried in noise, we propose a third kind of signal found after modulation analysis. This signal is modulated both in amplitude and frequency around a carrier frequency beeing the dominant theta rhythm.