Abstracts

Rationale: Epilepsy surgery fails to effect a cure in 30-40% of children despite optimal evaluation. Visual interpretation of subdural electrodes may be confounded by complex fields and the overpowering influence of source components close to certain electrodes. Although computer simulations have demonstrated that subdural ESL provides more precise localization of deep sources than scalp ESL, no study has evaluated the relationship between subdural ESL and the epileptogenic zone as defined by successful epilepsy surgery. We aimed to develop a methodology for evaluating ESL with respect to the epileptogenic zone, hypothesizing that in successful epilepsy surgery, interictal spikes localized using subdural ESL should be within the surgical resection zone. Methods: Retrospective data on surgical patients were reviewed, preoperative and postoperative MRIs were coregistered with intraoperative CTs of subdural electrodes, and coregistration quality was compared with surgical outcome using Spearman's rho. Patients with accurate coregistration and good surgical outcome were selected for ESL. Representative interictal spikes from subdural EEG were selected, and sources were reconstructed using a boundary element model. Five reconstruction methods were applied: one single dipole, one scanning model (Multiple Signal Classification, or MUSIC), and three distributed source models (minimum norm estimation [MNE], standardized low-resolution electromagnetic tomography [sLORETA], and sLORETA-weighted accurate minimum norm [SWARM]). Dipoles or maximal regions of charge density were categorized as within or outside the surgical resection zone (SRZ) using a 3-dimensional volume resection mask and compared using Chi-square, and euclidean distances to surgical mask epicenter were compared among the reconstruction methods using ANOVA, and location within the resection zone was compared to patients. Results: Coregistration on 19 individuals trended toward association with seizure freedom (p=0.07). Coregistration was adequate in 3, all of whom were seizure free after surgery. In each patient 14 to 15 spikes were reconstructed (total 215 solutions), of which 59 dipoles (27%) lay within the SRZ. The single dipole method performed best (40%) and SWARM worst (19%). Considering subcategories, 41% were localized for the case with left temporal corticectomy, 37% for right parietal corticectomy, and 5% for right frontal/interhemispheric corticectomy. Considering model, dipole performed better in the parietal case (86%), and MNE and sLORETA performed better in the temporal case (50%, p<.001 for all). Conclusions: Spikes were localized within the resection zone best using dipole model in parietal epilepsy. Success in image coregistration was weakly associated with outcome, suggesting that improved coregistration modalities should be investigated to allow ESL in patients with worse prognosis. These data support the use of subdural ESL, and the prospective evaluation of ESL to identify models suited to specific regions for improved planning of surgical resection.