Automatic analysis overcomes limitations of sleep stage scoring

A computer programme for the automatic analysis of the sleep EEG and EMG is presented. An EEG parameter is derived which is based on the joint frequency-amplitude distribution of the EEG activity. This newly developed parameter stresses the dynamic development of sleep, composed of alternating phases of EEG synchronization and desynchronization. While synchronization develops slowly, the opposite phase of EEG activity, desynchronization, is more rapid. A particular advantage of the EEG parameter is its continuous scale which results in high resolution. Thus, the parameter reflects gradual EEG changes which a visual analyser would have to classify into the same sleep stage. In addition to the EEG, the EMG is analysed and two parameters are extracted, representing the mean muscle tone and transient EMG activation, respectively. There is a close temporal relationship between the EEG and the EMG with a maximum of transient EMG activity during the phase of EEG desynchronization. The properties of the automatic analysis were compared to the visual analysis on a sample of 12 all-night sleep records. The results show that the EEG parameter also agrees sufficiently with the traditional sleep scoring method and therefore is a valid descriptor of the time course of sleep.