Graduate Research Assistant The University of Texas at Arlington, TX, USA Arlington, Texas
This abstract will be presented during a Pediatric Highlights platform session
Rationale: Epilepsy is a disorder of brain networks that can be assessed using functional connectivity measures. An increase of functional connectivity is linked to the epileptogenic zone (EZ), while a functional connectivity decrease is seen in distal brain regions. Connectivity alterations during interictal periods can guide intracranial EEG (icEEG) placement and EZ resection without the occurrence of a seizure. Here, we propose an innovative method that estimates non-invasively functional connectivity with magnetoencephalography (MEG) and high-density electroencephalography (HD-EEG) in children with drug resistant epilepsy (DRE). The proposed method uses virtual sensors (VSs), built at the same locations where implanted icEEG electrodes are placed during Phase II, to estimate functional connectivity aiming to compare MEG and HD-EEG connectivity estimates with icEEG estimates. Methods: We analyzed HD-EEG, MEG, and icEEG data from 37 patients with DRE who underwent surgery, and performed source localization (beamformer) to build VSs (for HD-EEG and MEG) at the icEEG electrode locations (Figure 1). We selected data with Interictal Epileptiform Discharges (IEDs) and data of resting-state activity (20 epochs of 3 s each), and computed per patient three connectivity matrices [Amplitude Envelope Correlation (AEC), Correlation (CORR), and Phase Lock Value (PLV)] (Figure 1). Each matrix was used to generate a network (graph) using the Minimum Spanning Tree (MST). For each node (i.e. each sensor) of the network, we estimated their mean connectivity (AEC, CORRE, PLV) plus four centrality measures from graph theory: degree, betweenness, closeness, and eigenvector. We tested the reliability of VSs measures via linear correlation with icEEG measures, and compared values in resected and non-resected areas for seizure-free (good outcome) and non-seizure-free (poor outcome) patients. Results: Analysis on icEEG showed higher AEC in resected than non-resected areas (p=0.04) in good outcome patients and no differences in poor outcome patients, as well as for all the centrality measures (Figure 2). These connectivity metrics correlated with the same ones obtained through VSs (high correlation, 0.8-0.9). In good outcome patients, we found differences between resected and non-resected areas (p< 0.05) from HD-EEG-VEs analysis for the three connectivity measures and the betweenness centrality measure, and from MEG-VEs analysis for all the centrality measures (Figure 2), AEC and CORR connectivity measures. No connectivity differences were found for poor outcome patients with any modality. Conclusions: We developed an innovative method that estimates functional connectivity non-invasively and reconstructs epilepsy networks in children with DRE. Our results show that non-invasive functional connectivity measures estimated via VSs localize the EZ, as invasive recordings do. Source functional connectivity on non-invasive data can help in the presurgical evaluation of children with DRE, overcoming the spatial limitations of icEEG, and potentially improving the outcome for these patients. Funding: Please list any funding that was received in support of this abstract.: RO1NS104116-01A1 and R21NS101373-01A1 by NINDS.