Abstracts

RATIONALE: Dipole modeling of temporal lobe epileptiform (spike and seizure) activity has been reported to add quantitative localizing ability above that provided by visual inspection. Few studies have reported on its use on frontal lobe patients, which are less common and often present a more difficult localization problem than temporal lobe patients. Our objective was to begin to determine if spatiotemporal dipole modeling could increase our localizing ability in frontal lobe patients.
METHODS: Since the installation of our completely digital Grass-Telefactor video-EEG systems, we have identified seven patients at our center with presumed frontal lobe epilepsy. For each of these we have manually selected representative spikes, sharp waves, and seizures for analysis. All were imported into BESA 2000 [MEGIS, Germany] and bandpass filtered at 1.6-50 Hz. Obviously bad channels were excluded from further analysis. One second epochs were chosen for modeling; spikes and sharp waves were centered in the epoch. Seizure epochs were chosen as close as possible to EEG onset. Rhythmic seizure activity was sometimes bandpass filtered in a tighter range to remove more noise. We used one-three regional dipoles optimized with a genetic algorithm set to default parameters. This is an objective method that requires no user intervention. After modeling the entire one second epoch we selected spike components (50-200 ms) based on the global field power, and single peaks of rhythmic seizure activity, for further modeling.
RESULTS: For all seven patients, the dipoles localized to frontal cortex, generally to the side of the presumed epileptigenic zone. In some fits for some patients they were located very near the midline, making it difficult to lateralize. In some other fits, dipole locations were close to the temporal pole and were thus difficult to ascribe entirely to the frontal lobe. The use of two or three dipoles often was able to account for other brain activity (alpha), eye movements, or electrode noise, thus apparently making the location of the epileptiform activity more robust and precise. The results of modeling only the spike components were almost identical to modeling the entire one second epoch. Confirmation of the epileptogenic zone was accomplished in three patients so far by the use of implanted electrodes. For two of these, the correlation and lateralization was complete. For the third, the dipole localization was deep mesial frontal, but difficult to lateralize. Depth electrodes recorded spikes mostly in left orbital frontal and deep mid-frontal cortex.
CONCLUSIONS: These initial results demonstrate the potential utility of spatiotemporal dipole modeling for frontal lobe seizure localization. As these patients proceed to intracranial studies and surgery we will be able to better confirm lateralizing and localizing results. To improve our ability to lateralize midline frontal activity, and separate lateral frontal and anterior temporal activity, we are beginning to record true electrode postions, which should impove localization accuracy over using generic electrode locations. We are also acquiring high resolution MRI scans in order to build more realistic head models, which should also improve accuracy. Two year post-surgical evaluations will provide the final determination of the epileptogenic zone.
[Supported by: Legacy Health System]