Abstracts

Rationale: Before surgery, the delineation of the seizure onset zone (SOZ) is necessary to estimate the epileptogenic focus and tailor the surgical resection. Intracranial electroencephalography (iEEG) is the tool of choice to define the SOZ but carries risks related to its invasiveness. The possibility of estimating the SOZ non-invasively through electric source imaging (ESI) of ictal data recorded using conventional 20-channel scalp EEG (convEEG) would largely benefit the patient’s presurgical evaluation procedure. The goal of this study is to assess the accuracy of convEEG to localize the SOZ non-invasively, by comparing it with the iEEG-defined SOZ, and to estimate its predictive value in terms of postsurgical outcome, by comparing it with the surgical resection. Methods: Thirty-five children who underwent epilepsy surgery after extraoperative iEEG and convEEG at Boston Children’s Hospital were studied (age: 11.9 ± 5.7 years). The electrographic discharges at ictal onset captured during the long-term monitoring with convEEG and iEEG were identified by two epileptologists blind to clinical data. Patient's presurgical MRI was coregistered with post-implantation computerized tomography (CT). Resected volume was determined by coregistering preop- and postop-MRIs. Surgical outcome was evaluated using the Engel scale and dichotomized into good (Engel 1) and poor (Engel ≥2).The predominant ictal frequency (Fig. 1) was identified via Morlet time-frequency decomposition. Ictal data were filtered around this frequency. Single ictal events were marked, averaged and localized using the Equivalent Current Dipole (ECD). Boundary Element Method (BEM) was used to obtain a realistic head model (three-layer for convEEG; one-layer for iEEG). The distance of the convEEG-SOZ and iEEG-SOZ (i.e. localization error) was estimated by calculating the Euclidian distance of each convEEG ECD from the closest iEEG ECD. The distance of both convEEG-SOZ and iEEG-SOZ from the resection margin was also calculated (Dres) and compared between good- and poor-outcome patients. Finally, for each patient, we estimated the percentage of resection (proportion of dipoles with Dres <15 mm) of both conv-EEG and iEEG-SOZ and tested whether it predicted outcome. Results: A total of 378 seizures were analyzed for all patients (198 iEEG, 180 convEEG). ConvEEG ECDs approximate iEEG ECDs with a localization error of 32.2 ± 13.3 mm. Dres was significantly lower for good outcome patients (Fig. 2A, convEEG: 13.9 ± 9.3 mm; iEEG: 7.6 ± 8.3 mm) compared to poor outcome patients (convEEG: 27.2 ± 17.6 mm; iEEG: 22.7 ± 12.7 mm), independently from clinical confounding factors (i.e. age, gender, duration of epilepsy, seizure frequency, resection volume and presence of MRI lesion; multivariate regression analysis: p < 0.05). The optimal threshold on the percentage of resection to predict outcome was assessed using the receiver operating characteristic (ROC) curve analysis for convEEG and iEEG separately (Fig. 2B). A threshold of 77% convEEG and 50% iEEG resected dipoles was able to predict good outcome with a positive predictive value (PPV) of 85% and negative predictive value (NPV) of 54% for convEEG, and 80% PPV and 90% NPV for iEEG (Fig 2C, Fisher exact test < 0.05 for both). Conclusions: The proximity of the SOZ localized using non-invasive convEEG to the surgical resection is predictive of the patient’s outcome. The non-invasive source localization of the SOZ with 20-channel convEEG can facilitate pre-surgical planning, optimize iEEG placement prior to surgery and possibly improve outcome in pediatric epilepsy surgery. Funding: No funding