Abstracts

RATIONALE: Classical spectral analysis is a well established method for the characterization of statistical properties in EEG time series. Extended information about the underlying dynamical processes can be obtained by applying higher order spectral methods. We have evaluated the capability of these techniques for the characterization of spatio-temporal dynamics of the epileptogenic process. METHODS: Bicoherence and skewness were estimated from cross- and auto-bispectra of long-lasting multi-channel stereoelectroencephalographic recordings using a moving-window technique. The temporal evolution of these measures was derived after appropriate averaging in the frequency domain within the bands 0-20 Hz and 20-40 Hz. RESULTS: Interpreting the bispectral functions as a measure for the complexity of phase relationships in the EEG, a less complex phase behavior was found at recording sites ipsilateral to the side of the primary epileptogenic area. Moreover, this phenomenon showed a high temporal stability and appeared to be independent of the analyzed frequency bands. CONCLUSIONS: Complexity of phase relationships as measured by bispectral functions can help to gather more insight into the spatio-temporal dynamics of the epileptogenic process. Thus, these methods can be regarded as a valuable addition to other linear and nonlinear EEG analysis techniques. [Supported by the Deutsche Forschungsgemeinschaft]