Abstracts

Rationale: Here we describe a method for separating electric field sources (EFSS) based on cortico-cortical evoked potentials (CCEPs) to improve our analysis for surgical planning. We previously described a novel metric of correlation between the pattern of connectivity determined by CCEPs and early seizure spread, and showed that we were able to predict the area targeted for ablation in 10 patients. 6 of 10 patients had significant CCEP/ictal correlations in an area targeted for ablation. However, one limitation of our prior method was that we collapsed CCEPs across all contacts of an electrode, even when the electrode was known to traverse to two anatomically distinct areas. Here we improve our method by developing EFSS, thus dividing contacts into physiologically-independent cortical units before analysis of CCEPs. Methods: Single pulse electrical stimulation was performed at 1 Hz in trains of 20-30 seconds at multiple sites in 10 patients with refractory epilepsy undergoing stereo-EEG monitoring. The resulting CCEPs were averaged and the amplitude was defined as the root mean squared error (RMSE) from 10 ms to 50 ms after the onset of the stimulus artifact. To perform EFSS, we examined the amplitude of the CCEPs on a given contact for all stimulation runs triggered from a different electrode. We ranked the CCEPs amplitude from largest to smallest and computed pair-wise Spearman correlation coefficients between all contacts of a given depth electrode. We then visually separated the resulting correlation matrix into two clusters when the intra-cluster correlation was high (typically R>0.75) but the inter-cluster correlation was lower (typically R<0.5). We then chose the largest amplitude CCEP for each cluster, and computed CCEP-ictal correlations as in our previous methodology.  Results: We performed a EFSS for the 4 patients which had no significant area of CCEP/ictal correlation. This changed the classification for Patient 1, who now has significant (p=0.034) CCEP/ictal correlation inside the ablation zone after segmentation increased the independent units from 7 electrodes to 11 clusters. We analyzed preliminary results for surgical outcome (follow-up ranged from 3 to 17 months after surgery). Overall, 7 out of 7 (100%) of patients which had significant /CCEP ictal correlation inside the ablation zone had reduction in seizures frequency after surgery (3 were seizure free). In contrast, only 1 out of 3 (33%) patients without an area of CCEP/ictal correlation inside the ablation zone had an improvement in seizure frequency after surgery. This difference between in surgical outcomes between patients with or without CCEP/ictal overlap inside an ablation zone was not statistically significant (p=0.067, Fisher exact test). Conclusions: We were able to separate contacts of depth electrodes into independent sources by examining a correlation matrix computed from observed CCEPs. EFSS increased the sensitivity for detecting significant CCEP/ictal overlap, including in a patient seizure-free after surgery who was a false negative using our prior methodology. However, application of this method in our sample of 10 patients was underpowered to detect a difference of surgical outcome. There was a non-significant trend towards improvement in seizure frequency if an area of CCEP/ictal overlap was ablated (7/7 vs. 1/3, p=0.067) which we plan to follow-up in a larger patient cohort. We are developing a validated means for automated EFSS for more general use in other SEEG-based analyses. Funding: NPP was supported by the AAN and American Brain Foundation (CRTF).