Effect of epoch length on power spectrum analysis of the EEG

To study the effect of epoch length on the variability of power spectrum analysis of the EEG, 22 64-s segments of EEG were analyzed using epoch lengths of 2, 4, 8, 16, and 32 s. Nine of these segments exemplified EEG changes during transient anesthetic states or surgical conditions. Epoch-to-epoch variability was computed within frequency bins for all segments, and ANOVA with hierarchal classification was used to determine the length of the EEG segment necessary to identify a statistically significant change in those EEG segments recorded during changing conditions. In 16 segments, the epoch-to-epoch variability with power spectra were computed using 2-s epochs was significantly less than the variability when power spectra were computed using longer epoch lengths. In five segments, no significant difference existed between the variance at 2-s epochs and longer (4-s) epochs. In one case, an EEG containing a burst-suppression pattern, the variability was significantly increased when 2-s epochs were used. Analysis using 2-s epochs also identified changes more rapidly than analysis using any longer epoch length in eight of nine segments, and the differences were clinically significant as well (over 30 s faster when 2-s epochs were used instead of 16-s epochs). These findings suggest the preferability of short epoch lengths when power spectrum analysis is used for intraoperative EEG monitoring.