Abstracts

Rationale: Non-invasive investigations of interictal epileptic discharges (IED) using electric source imaging (ESI), magnetic source imaging (MSI) and EEG-fMRI aim at localizing the epileptogenic focus during presurgical investigations. Exploiting complementarities between EEG and MEG data when recorded simultaneously, the fusion of EEG and MEG within the source imaging framework of Maximum Entropy on the Mean (MEM) improves the accuracy and the reliability of IED source localizations (Chowdhury et al., HBM 2018). On the other hand, whereas fMRI hemodynamic responses to IED are often extended and could contain propagated activity, the fMRI cluster exhibiting the most significant response was suggested as a good predictor of the seizure onset zone. The objective of this study was to retrospectively evaluate, for similar IEDs, the concordance between EEG-MEG fusion results using MEM (MEM-fusion) and the fMRI cluster exhibiting the highest absolute t-value, and then to compare those results with available clinical reference and postsurgical outcomes. Methods: We retrospectively studied ten consecutive neocortical epilepsy patients who underwent simultaneous EEG-MEG recordings and EEG-fMRI investigations. For each patient, similar IEDs were selected for both EEG-MEG and EEG-fMRI analysis. For source localization, we used the consensus MEM-fusion map which consists in applying MEM fusion source reconstruction for every single IED, followed by hierarchical clustering of all source maps. The average map from the cluster involving the largest number of IEDs was then defined as the consensus MEM-fusion map. This map was then compared with the fMRI cluster exhibiting the most significant response.  Concordance was qualitatively evaluated at a sublobar level. Results from this comparison between MEM-fusion and fMRI response were then further compared to the underlying clinical reference and to postsurgical outcomes.The clinical reference was defined combining information from intracranial EEG findings, the resected region, and an eventual epileptic lesion detected on MRI. Results: Consensus MEM-fusion maps were found to be sublobar concordant with the peak of the most significant fMRI cluster in 7 patients, including 5 activations (positive BOLD responses) and 2 deactivations (negative BOLD responses). Two situations occurred: a) when the most significant fMRI cluster was spatially extended, covering several lobes (2 patients), consensus MEM-fusion was usually more focal and source results were concordant with the clinical reference. Among these two patients, only one was operated and is not seizure free. b) When results from both consensus MEM-fusion and the fMRI most significant cluster were more localized (5 patients), these two investigations were concordant with the clinical reference for all patients. All these patients were operated and 4/5 are seizure-free with a follow-up of more than 12 months. The consensus MEM-fusion map and the most significant fMRI cluster were discordant for 3 patients, often exhibiting extended results and propagation patterns, including either concordance at the lobar level (1 patient) or complete discordance. Two of them were operated and are not seizure-free. Conclusions: These preliminary results suggest that a concordance between the consensus MEM-fusion map and the circumscribed fMRI cluster exhibiting the most significant hemodynamic response was usually associated with a good postsurgical outcome, whereas a discordance or a partial concordance but involving extended regions was not, suggesting limited localization efficiency. Funding: Canadian Institutes of Health Research: MOP-93614; FRQS; Savoy Foundation