Abstracts

Rationale: Seizures may originate in neuronal networks of neocortex, subcortical grey matter, or mesial temporal structures. Localization of the seizure onset zone often requires recordings from a combination of intracranial depth and grid electrodes. The choice of reference may impact physiological interpretations of EEG studies based on such measurements. EEG measures of functional interrelations among neuronal networks (correlation or coherence based) are known to be particularly sensitive to the choice of reference.Methods: Intracranial EEG recordings obtained during multiple seizures (3-7) were remontaged to an average reference or a distant silent reference, parameterized using multivariate autoregressive (MVAR) models, and analyzed with short-time direct directed transfer function (SdDTF). SdDTF estimates interrelations among sites of EEG recordings, i.e. the direction and intensity of effective connectivity between the sites. EEG activity remontaged to both average or distant silent references were analyzed across a wide range of frequencies (0-175 Hz) in order to identify effective connectivity and seizure foci among epileptic networks. Seizure foci were identified independently by visual inspection of the EEG recordings by experienced epileptologists. Results of SdDTF analysis were compared to the clinical evaluation.Results: When intracranial grid electrodes were used, patterns of effective connectivity correctly identified the locations of ictal onset zones with an average reference montage, but not with distant referential grid electrodes as references. For the combination of intracranial depth and grid electrodes, patterns of effective connectivity identified seizure onset zones (localized near a depth contact) using distant referential montages, but not average reference montage.Conclusions: The choice of references has a significant impact on the determination of patterns of epileptic effective connectivity and identification of the seizure onset zone. For intracranial grid recordings, the average reference provides accurate results because the contribution of active electrodes is minimized by larger numbers of not active electrodes. Combining recordings of intracranial grid and depth electrodes, especially when depth contacts are near seizure foci make average referencing unsuitable due to the contribution of active depth electrodes. In this case, a distant silent grid contact is more suitable as a reference. This project was funded by NINDS R01 NS40596 and NS48222.