Abstracts

Rationale: Our goal is to determine if distinguishable patterns of spatial phase clustering are present in high frequency bands of EEG derived from interictal 256-channel high density scalp EEG (dEEG), and if these patterns correspond to the proven epileptogenic zone, as documented by subdural grid electrocorticographic (ECoG) recordings. Methods: We studied an patient with medically refractory frontal lobe epilepsy who underwent intracranial subdural grid ECoG recordings to establish the localization of ictal onsets. Seizure proved to originate from the left parasagittal frontal region near motor cortex. Prior to invasive EEG studies the subject underwent dense array dEEG recordings. Ten minutes of interictal dEEG data, sampled at 1000 Hz, was randomly selected for analysis. The selected segment was at least two hours away from an epileptic seizure and, based on visual analysis, free of interictal epileptiform patterns. Data were imported into MATLAB for analysis. The EEG data was filtered in the appropriate EEG band. The analysis was performed in low gamma (30-50 Hz), high gamma (50-80 Hz) and in high frequency (80-160 Hz) bands. Hilbert transform was applied to compute the analytic phase and then it was unwrapped. From this, the instantaneous phase frequency was computed. Spatiotemporal contour plots with 1.0 ms intervals were constructed using a montage layout of 256 electrode positions. The rate of change in phase with distance (rad/mm) was computed from spatial location of the electrodes on the scalp. Several criterions were applied to select stable phase cone patters. These included: (1) phase frequency was within the temporal band, e.g., 30-50 Hz for low gamma band, (2) signs of spatial gradient and maximum or the minimum did not change for at least 3 time samples, and (3) The frame velocity was within the range of condition velocities of cortical axons, 1-10 meters/sec. Stable clusters of frames in each second of dEEG data were computed and an average rate of phase clusters over a period of 10 min period was computed. Spatial plots of averaged rate of phase clusters were constructed in different EEG bands. Results: Spatiotemporal plots showed formation of cone-like structures which varied in spatial shapes from one frame to the next. We found that the peak intensity also varied from one frame to the next. In general, stronger and more stable patterns were observed in the seizure-onset regions as compared with surrounding brain. Clustering of spatial patterns was observed with greatest density in the seizure-onset areas in the low gamma (30-50Hz) and the high frequency (80-160 Hz) bands. Conclusions: These preliminary results show that the spatiotemporal dynamics and clustering of phase patterns have the potential to assist in localizing the seizure-onset zone from scalp dEEG data. Studies from additional patients are currently underway to confirm these findings.