Spatio-temporal EEG measures and their application to human intracranially recorded epileptic seizures

We describe analysis methods which involve the computation of band-integrated, intrachannel spectral measures from the EEG. These measures require only very general assumptions, have a simple interpretation and achieve significant data reduction, while providing a quantitative assessment of significant EEG modification at seizure or onset or in the course of seizure evolution. In developing these measures, we experimented with various measures derived from spatio-temporal power spectra. Bandpass (8-30 Hz) integrated relative energy measures appear to be the most useful since they objectively measure paroxysmal modification. Two such measures are illustrated in this report: the band-integrated power z-score (BIPZ), which is a quantitative measure designed to allow significance testing, and the band-integrated power ratio (BIPR), which is a simply calculated qualitative measure that behaves very similarly to the BIPZ measure and may be useful for real-time data screening. In addition to developing these intrachannel measures, we have also developed methods of displaying these measures in a format appropriate to the analysis of interchannel relationships. In a general sense, the displays provide objective information regarding the spatio-temporal dynamics of the EEG in a condensed format which is still readily understandable to the electroencephalographer. This analysis was applied to 29 spontaneous, depth-recorded seizures in 14 patients with the aim of identifying the initial locus of seizure activity in each seizure episode. The initial locus was defined as that location (or set of locations) to first exhibit statistically significant values of the BIPZ measure. The method was very effective in this regard, as shown by correspondence between the conclusions of this automated analysis and conventional visual analysis.