Abstracts

Rationale: In the evaluation of patients with medically refractory epilepsy for resective surgery, localization of ictal onset zone (IOZ) using invasive EEG with subdural grids is conventionally considered the gold standard. However, this procedure is associated with not insignificant morbidity. In contrast, scalp EEG though much safer, has localization only to head regions and may have a poor co-registration with underlying cortex especially in patients with distorted anatomy. Further, a single electrode on scalp EEG carries dipolar source information from about 30 cm2 of cortical surface, which has insufficient specificity for surgical planning. We used several different algorithms for EEG source localization and co-registration with patient's segmented cortex to study the internal validity of these algorithms and agreement with localization of IOZ on invasive EEG monitoring. Methods: All patients with medically refractory epilepsy admitted for non-invasive pre-surgical evaluation were eligible for inclusion. Inability to co-operate with electrode digitization was an exclusion criterion. After scalp EEG electrode placement, the electrode position was digitized using a Polhemus digitizer. Sections of the EEG showing seizure onset were snipped and uploaded as ‘functional data' in the Curry 7 Neuroscan software, the electrode positions being obtained from the digitizer file. Patient's brain MRI was uploaded as ‘image data' and boundary element head model was constructed with segmented cortex. EEG was filtered at 1-70 Hz, with additional notch filter at 50 Hz and its harmonics. Following source localization algorithms were then carried out: rotating equivalent current dipole (ECD), multiple signal classification (MUSIC), current density methods including minimum norm estimate (MNE), standardized low resolution electro-tomography (sLORETA) and sLORETA weighted minimum norm (SWARM). Source localizations were obtained at the point of ictal onset determined by visual analysis of scalp EEG (+/-5 msec). The agreement between lobar localization thus obtained was compared with that of scalp EEG using Fleiss kappa statistic. Results: Five patients with a total of 9 seizures have been analyzed so far. Each seizure was counted separately as an observation. All patients had extra-temporal epilepsy. In this small sample, sLORETA algorithm was found to have best agreement with visual analysis of scalp EEG at a lobar level (κ=0.67, 95% CI 0.54, 0.80). In addition, all 3 current density/ distributed dipole methods (MNE, sLORETA, SWARM) had high internal consistency (κ= 0.73, 95% CI 0.59, 0.87). Conclusions: EEG source imaging may potentially offer a safe, non-invasive and relatively specific method to increase accuracy of localization of ictal onset zone for surgical candidates with medically refractory epilepsy. For better clinical utility we intend to compare EEG source localization with IOZ defined on prolonged video-ECoG monitoring.