Abstracts

Rationale: The efficacy of epilepsy surgery depends upon successful localization of the epileptogenic zone. Magnetoencephalography (MEG) and high-density electroencephalography (HD-EEG) enable the non-invasive localization of interictal epileptic discharges (IEDs) by solving the inverse problem. However, to date, it is still unclear the sensitivity and specificity of MEG and HD-EEG in the localization of the irritative zone (IZ). The goal of this study is to examine the significance of MEG and HD-EEG in the non-invasive localization of the IZ by evaluating the concordance of MEG and HD-EEG findings to invasive EEG (iEEG) recordings in pediatric patients undergoing surgery. We hypothesize that both modalities localize the IZ with high localization accuracy, but MEG is superior to EEG since it provides an undistorted view of brain activity due to the negligible effect of conductivity differences on the magnetic flux. Methods: We retrospectively analyzed interictal whole-head MEG (306 sensors) and HD-EEG (70 channels) recordings from 8 children (13.4 ± 2.7 years) who underwent epilepsy surgery after intraoperative electro-corticography. IEDs with varying morphology and topography were independently visually marked in MEG and EEG signals. The IEDs were localized using the MEG or EEG signals with an Equivalent Current Dipole (ECD) scanning method. Cluster analysis was further used to classify these discharges on the basis of their spatial distribution. ECDs with goodness-of-fit >75% were considered. For each patient, boundary element models were estimated from the pre-operative MRIs. We co-registered patient’s MRI with the post-implantation computerized tomography to determine the location of iEEG electrodes on patient’s cortex. Results for non-invasive localization were compared with the IZ defined by iEEG findings (Fig. 1 a & b). Each ECD and cluster was compared to the iEEG findings and classified as: (i) concordant when the ECD/cluster was localized in the vicinity of an IZ-iEEG electrode; (ii) non-concordant when the ECD/cluster was localized in a brain area covered by iEEG electrodes but outside the IZ; and (iii) not covered, when the ECD/cluster was localized in a brain area not covered by iEEG.  Results: IEDs were identified in seven patients in both MEG and EEG signals. For one patient, only MEG data were available. All ECDs localized with MEG were inside the iEEG coverage, while 13.2% of ECDs localized with EEG were outside the iEEG coverage (Fig. 1c). MEG presented a significantly higher proportion of ECDs which were concordant with the IZ compared to EEG [χ²(1, N=460) = 53.37, p < 0.001]. The mean distance of ECDs to the IZ was 2.25 (±3.23) cm for the MEG and 3.05 (±2.82) cm for the HD-EEG (Fig. 1d; p=0.003). Clustering method (Fig. 2) improved MEG but not EEG source localization: the mean distance of clusters to the IZ was 1.35 cm (±0.42) for the MEG compared to 3.02 cm (±3.28) for the EEG. Conclusions: MEG and HD-EEG are both valuable tools for the non-invasive mapping of interictal activity in epilepsy surgery. MEG can localize the irritative zone with higher localization accuracy and specificity compared to EEG. Clustering method of dipoles in MEG data provides an accurate non-invasive assessment of the IZ in pediatric patients undergoing epilepsy surgery.  Funding: ...