Abstracts

Rationale: The spatial sampling bias of sEEG, which limits the ability to completely map the epileptic network, remains a major challenge for sEEG target planning. Magnetoencephalography (MEG) provides excellent spatiotemporal resolution and can be used to assist sEEG target planning. However, the most common MEG source localization approach, equivalent current dipole (ECD), often fails to fully localize the epileptogenic zone when complex epileptic networks exist. To overcome this limitation, connectivity-based MEG analyses to map crucial ‘hub’ regions; the highly connected cluster nodes within the network level, have been studied in patients with epilepsy. It is crucial to validate these novel methods and determine how to best apply them for clinical decision-making. However, prospective validation of these methods is lacking. We hypothesized that MEG network analyses more accurately define the seizure onset zone (SOZ) compared to ECD.

Methods: MEG recordings obtained as part of the pre-surgical evaluation were used for functional connectivity analysis and compared to ECD clusters. Whole brain estimates of source activity were obtained using a linearly constrained minimum variance beamformer, comparing 3-second baseline (no spikes) and 3-second spike phase. Second, functional connectivity was estimated using weighted phase lag index (wPLI). Finally, highly connected nodes were quantified using eigenvector centrality. Distribution of top 95% nodes was defined as the maximum distance over all pairs of nodes of the distances between each hub. Post-operative CT scans with sEEG positions were co-registered to the patients’ presurgical MRI in order to measure the Euclidean distance between the SOZ defined by sEEG to both the ECD and area of maximum hub. Surgical margins were defined using post-operative MRI or CT.

Results: Thirty-one patients were identified (F/M = 16/15, age at seizure onset = 4.57 ± 4.8 yrs, age at surgery = 10.83 ± 5.7 yrs) who underwent MEG, sEEG, and focal resective or Laser Interstitial Thermal Therapy. Both the centroid of ECD cluster and maximum hub of connectivity were lobar concordant with SOZ defined by sEEG in all cases. Connectivity and network-based mapping identified a maximum hub that was significantly closer to the sEEG-defined SOZ compared to ECD (13.42 ± 9.28 mm vs 22.25± 17.22 mm) (t-test, p=0.015). 16.2% of patients received complete resection of ECD cluster, whereas 45.2% received complete hub resection. Fifteen out of 31 patients were seizure free (ILAE 1) at the most recent follow up >1 year. The sensitivity and specificity in seizure freedom outcome prediction was 73.3% (95% CI: 47.5-89.3) and 81.3% (56.0-94.0) in patients with complete resection of the hub region, compared to 26.7% (10.7-52.5) and 93.8% (69.4-100) in patients with complete resection of the ECD.

Conclusions: MEG connectivity analysis exhibited better accuracy for localization of SOZ compared to conventional ECD source modeling. Therefore, MEG connectivity may offer better guidance than dipole clusters for sEEG in clinical practice.

Funding: Please list any funding that was received in support of this abstract.: Postdoctoral Research Fellowship Award.