Abstracts

Rationale: Previous studies provided evidence that seizure prediction algorithms that are based on a quantification of relationships between recording sites appear to perform significantly better than random prediction, even if rigorous statistical validation is applied. Some studies reported on seizure precursors that were not in close vicinity to the seizure onset zone but could be located in remote or even contralateral brain structures. Although this seemingly counterintuitive finding may indicate the importance of brain regions outside the ictal onset zone but within the epileptic network for the generation of seizures, it renders an a priori selection of EEG channels for seizure prediction studies problematic. While previous studies considered the strength of interactions only, we here investigated whether estimates for both strength and direction of interactions can help to identify “optimal” EEG channels for seizure prediction studies. Methods: We retrospectively analyzed strength and direction of interactions in multi-day, multi-channel intracranial EEG recordings from 6 patients (5 patients with unilateral and 1 patient with bilateral mesial temporal lobe epilepsy (TLE); total number of seizures: 43; total recording duration: 44 days; 20 bilateral intrahippocampal recording sites) using bivariate measures that are based on the concept of phase synchronization. From the temporal average of the direction of interactions we identified asymmetric driving brain regions and evaluated their relationships to the seizure onset zone. For each channel combination, we estimated the seizure prediction performance via ROC statistics using the strength of synchronization. Results: In 4 patients with unilateral TLE the mean intra-hemispheric asymmetry was more pronounced in the focal hemisphere and the channel exhibiting maximum asymmetry indexed the seizure onset zone. Interestingly, in all patients with unilateral TLE the recording sites with maximum prediction performance were not located ipsilateral to the seizure onset zone but were found in contralateral brain regions, and the overlap between these regions was significantly less pronounced than expected a priori. The patient with bilateral TLE showed no clear driving structure and we observed only one channel that exhibited maximum prediction performance. Conclusions: Our results suggest the existence of an epileptic network with functionally different and spatially distributed structures involved in the generation of focal seizures. Estimating both strength and direction of interactions within the epileptic network can help to identify location and extent of precursor structures without relying on performance estimates. This advantage can be regarded helpful for the design of prospective seizure prediction studies. (Supported by Deutsche Forschungsgemeinschaft)