A comparison of EEG signal dynamics in waking, after anesthesia induction and during electroconvulsive therapy seizures.

Journal Article (Journal Article)

Evidence suggests that quantitative dynamical measures of electroencephalogram (EEG) signals are more appropriate for characterizing the differences between states in an individual rather than as absolute indices. One such measure, the largest Lyapunov exponent (lambda 1), appears to have potential for identifying seizure activity and for being of clinical utility for characterizing electroconvulsive therapy (ECT) seizures. As a result, we compared lambda 1 for the EEG recorded in 8 depressed subjects in 3 states: (1) during right unilateral ECT seizures, (2) during the pre-ECT waking state, and (3) following anesthesia administration but prior to ECT. Spectral amplitude and autocorrelation were also calculated in these states, allowing a comparison of these measures with lambda 1. We hypothesized that lambda 1 would be lowest during the ECT seizures, suggestive of greater EEG signal predictability over time during the seizures. We found that during the seizures lambda 1 was smaller, while spectral amplitude was larger. Significant inter-state differences were not found for the left temporal and occipital regions suggesting that these measures might serve as markers of the degree of seizure involvement of specific brain regions. Spectral amplitude and lambda 1 were uncorrelated and varied independently in some cases. The autocorrelation time was shortest in the waking EEG, and longest for the post-anesthesia EEG, and did not account for the differences seen in lambda 1. In contrast, the persistence of oscillations in the autocorrelation functions was greater for the ictal EEG than the other two states and may relate to lambda 1.

Full Text

Duke Authors

Cited Authors

  • Krystal, AD; Greenside, HS; Weiner, RD; Gassert, D

Published Date

  • August 1996

Published In

Volume / Issue

  • 99 / 2

Start / End Page

  • 129 - 140

PubMed ID

  • 8761049

International Standard Serial Number (ISSN)

  • 0013-4694

Digital Object Identifier (DOI)

  • 10.1016/0013-4694(96)95090-7


  • eng

Conference Location

  • Ireland