Computer characterization of tracé alternant and REM sleep patterns in the neonatal EEG by adaptive segmentation--an...

The possible utility of the computer technique of adaptive segmentation in the comparative quantitative characterization of tracé alternant (TA) and REM sleep in the neonatal EEG has been explored in a pilot study of the EEGs of 3 full-term infants of ages 1, 13 and 23 days. The technique of adaptive segmentation, originally developed by its authors for automatically delimiting and characterizing different types of patterns within the same EEG recording, had previously been found effective for this purpose for normal and abnormal adult EEGs. The same computer program parameter values that had been found to be optimal for segmentation of adult EEGs were also found to be optimal for the neonatal ones, as typified by segmentation or demarcation of the bursts and interburst periods in tracé alternant. Adaptive segmentation, in conjunction with clustering of the resulting segments and computation of temporal profiles showing the times of occurrence of the different types of activity in a given phase of sleep together with the mean amplitude and mean frequency of each, supplemented by the respective power spectra, was found to be an effective way of characterizing these EEGs, including certain types of artifact. Further characterization was afforded by statistical summaries of segment durations.