Abstracts

Rationale: Accurate localization of epileptogenic focus can be challenging, especially when electroencephalography (EEG) and/or imaging studies (e.g. MRI) are inconclusive. Magnetoencephalography (MEG) is an attractive tool for non-invasive localization of the epileptic focus and can aid the presurgical evaluation of patients with focal epilepsy. The absence of interictal abnormalities in typical routine short-term (60 minutes) MEG studies reduces the yield of MEG, since focus localization relies directly on the presence of epileptiform transients in the recorded data. We herein propose a focus localization algorithm (FLA), which performs connectivity analysis of signals in the source space, to identify the epileptogenic focus. We tested FLA on a consecutive series of 22 patients selected from the Cleveland Clinic epilepsy database who had preoperative MEG, underwent resective surgery, and were seizure free postoperatively for at least 6 months. Methods: MEGs were recorded using a 306 channel Elekta vectorview system of 204 plano-gradiometers and 102 magnetometer channels. Generalized Partial Directed Coherence (GPDC) was estimated from consecutive 1 min artifact free MEG data using the methodology previously described in [1]. An independent MEG reviewer marked data segments with epileptiform discharges and awake/drowsy states. Eigenvector Centrality was estimated from the GPDC matrix. The centrality score of each dipole was statistically compared with its counterpart in the contralateral hemisphere. Corrections for multiple comparisons were performed using FDR (q=0.05). Regions with high centrality score and statistically significant (p < 0.05) when compared to the homologous contralateral ones were identified as candidate epileptogenic foci. For every patient, the post-operative MRI was coregistered with the pre-operative MRI. FLA results were deemed to be concordant if FLA localized region was within the area of resection. The performance of FLA was evaluated using MEG data segments with or without the presence of epileptiform transients, during awake or drowsy states, and was also compared to the one of the conventional Single Equivalent Current Dipole (SECD) methodology (SECD modeling was performed just before or at the peak of the global field power of an interictal transient). Results: FLA-based analysis localized the epileptogenic focus successfully in 14 out of 22 patients. Concordant results were observed in 11 out of the 22 patients using SECD-based analysis (See Table 1). FLA based analysis localized the epileptic focus in 6 patients in whom SECD analysis was indeterminate. Three patients had SECD positive results with negative FLA results. In 5 patients both SECD and FLA analysis produced discordant /indeterminate results. There was no statistically significant difference between FLA results in segments containing interictal spikes vs segments without spikes (Fig.1). Similarly, there was no statistical difference in FLA results between segments of sleep and awake. Conclusions: The results from this study show that novel analysis of interictal MEG can contribute to accurate localization of the epileptogenic focus even in the absence of interictal spikes in the recorded data. Since the performance of FLA does not depend on the presence or absence of intertical spikes and is overall better than SECD's, the yield and cost-effectiveness of MEG studies can thus be enhanced. The accurate identification of the epileptogenic focus and from short, interictal-only, noninvasive electromagnetic recordings could pave the way for a paradigm shift in the diagnosis and treatment of epilepsy. [1] Krishnan, B., et al. Clin Neurophysiol 126.4 (2015) 667-674. Funding: Epilepsy Foundation Postdoctoral Fellowship (01/2015-12/2015)