Abstracts

Rationale: Source localisation methods are increasingly applied on scalp EEG signals in clinical practice. However, data on source reconstruction derived from intracranial (subdural) EEG recordings is currently limited to simulation studies and single case reports. We retrospectively implemented novel methods of source imaging in the analysis of ECoG data in an attempt to define the accuracy of inverse solutions, as well as the limitations and feasibility of application in resection planning. The clinical validation of our findings was provided by postoperative outcomes. Methods: Continuous sleep segments of ECoG data obtained during the invasive presurgical workup of 14 patients with refractory frontal lobe epilepsy were retrospectively reviewed. MRI imaging was either unremarkable or showed poorly delineated hyperintense regions that raised suspicion for unilateral frontal lobe dysplasia. Subdural coverage was variable consisting in a lateral grid and additional intrahemispheric strips in nine cases as well as basal frontal strips in four cases. Spikes were visually identified and marked by two independent reviewers according to amplitude, voltage field and disruption of normal background. Consensus spikes were averaged to serve as a substrate for the realistic boundary element model (BEM) based sLORETA and MUSIC (MUltiple SIgnal Classification) algorithms. Results of source localisation were visualised using T1-weighted MRI data sets acquired prior to and after surgical treatment. Results: In ten patients a single spike population was identified, while four patients presented multifocal spikes within the frontal lobe. Inverse solutions localised within the semiologically and electroencephalographically presumed epileptogenic zone in all cases with a distance of 0-7 mm between MUSIC and sLORETA maximums. In seven cases total resection of the epileptogenic zone including inverse solution was possible, followed by Engel Ia outcome. In two cases where epileptogenic zone/solution localised within eloquent cortex areas multiple subpial resections were offered that led to Engel IIIa and Engel IVb outcomes respectively. In another three patients with Engel Ia outcome solutions localised at resection border. Finally, in two patients with multiple spike foci and resulting multiple solutions, partial inclusion of solutions in the resection correlated with Engel IIIb and IVa outcomes. Histology showed focal cortical dysplasia in all but one case. Conclusions: This is the first study evaluating the applicability of source reconstruction methods in subdural recordings and validating results through outcomes of epilepsy surgery. Complementary information derived from the MUSIC and sLORETA algorithms may prove valuable in improving localisation, especially regarding deep sources. This is crucial in the implementation of tailored resections and improvement of clinical outcomes.