Abstracts

Rationale: Historically, the localization of interictal and ictal sources has primarily relied on visual inspection of scalp EEG traces. Identification of the source is based mainly on the assumption that the source underlies the most prominent negative activity. With the advent of modern computer-assisted techniques such as dipole source modeling, more quantitative methods are now available to localize EEG activities. The objective of this study was to compare the results of dipole source analysis and visual analysis of scalp EEG in surgical epilepsy patients Methods: Non-invasive scalp EEG's from 24 surgical epilepsy candidates were analyzed using the Brain Electrical Source Analysis (BESA) software, developed by MEGIS Software. Scalp EEG was recorded using 10-20 international system with placement of additional electrodes at F9/F10 and P9/P10 for more accurate dipole source localization. EEG activity was digitally recorded referentially to midline electrodes. Continuous monitoring with digital video and EEG was performed using the Nicolet digital video/EEG system. A single neurophysiologist performed dipole source analysis on interictal and ictal waveforms using BESA. Both individual and averaged waveforms were analyzed. Principle Component Analysis was used to separate potential concurrent sources. EEG localization by visual inspection was done independently by two neurophysiologists. The BESA results were compared to the two EEG results for agreement of localization, and were categorized as being in agreement ( concordant ); somewhat in agreement, but with dipole analysis providing further localizing information, ( concordant plus ); or in complete disagreement ( discordant ). Results: Up to 15 interictal spikes and 3 seizures were analyzed for each patient. In total, 41 seizures were analyzed. Compared to the first set of EEG localizations, concordance rate with dipole source modeling results was 34.4%, concordance plus rate was 29.5%, and discordance rate was 36.1%. Compared to the second set of EEG localizations, concordance rate was 27.9%, concordance plus rate was 41.0%, and discordance rate was 31.0%. A Kappa test for inter-rater reliability was done for the two EEG data sets and showed substantial agreement for interictal spikes (kappa 0.612, p < 0.001), moderate agreement for seizure sets 1 and 2 (kappa 0.599, p < 0.001; and kappa 0.53, p = 0.003, respectively), and substantial agreement for seizure set 3 (kappa 0.63, p = 0.005). Conclusions: The results of this study suggest that dipole source modeling may have further localizing value over visual inspection of EEG alone. There was a surprisingly high rate of discordance between dipole source analysis and visual EEG localizations (approximately one-third with both data sets). Furthermore, in the remainder when there was concordance, in a significant percentage, the dipole source analysis provided further localizing information that was not readily apparent by visual inspection alone. Overall, this study suggests that dipole source modeling may play a valuable role in the seizure localization.