Clinical validation of magnetoencephalography network analysis for presurgical epilepsy evaluation

• Published in: Clinical neurophysiology Vol. 142; pp. 199 - 208
• Main Authors: Fujiwara, Hisako, Kadis, Darren S., Greiner, Hansel M., Holland, Katherine D., Arya, Ravindra, Aungaroon, Gewalin, Fong, Susan L., Arthur, Todd M., Kremer, Kelly M., Lin, Nan, Liu, Wei, Mangano, Francesco T., Skoch, Jesse, Horn, Paul S., Tenney, Jeffrey R.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2022
• Subjects: Drug-resistant epilepsy; Eigenvector centrality; Functional connectivity; Magnetoencephalography; Surgical outcome; Functional connectivity; sEEG; Drug-resistant epilepsy; Magnetoencephalography; Eigenvector centrality; ECD; Surgical outcome; LITT; MEG; SOZ
• ISSN: 1388-2457; 1872-8952
• DOI: 10.1016/j.clinph.2022.07.506

•Primary hub was significantly closer to the stereotactic EEG-defined seizure onset zone compared to the center of ECD cluster.•Both ECD cluster and network analyses suggested that spatially restricted region correlate with better seizure outcomes.•Our findings suggest that magnetoencephalography network analysis could be a valuable complement to the conventional ECD method. To clinically validate the connectivity-based magnetoencephalography (MEG) analyses to identify seizure onset zone (SOZ) with comparing to equivalent current dipole (ECD). The ECD cluster was quantitatively analyzed by calculating the centroid of the cluster and maximum distance (the largest distance between all dipoles). The “primary hub” was determined by the highest eigencentrality. The distribution of nodes in the top 5% of eigenvector centrality values was quantified by generating the convex hull between each node. Thirty-one patients who underwent MEG, stereotactic-EEG, and focal surgery were included. The primary hub was significantly closer to the sEEG-defined SOZ compared to ECD (p = 0.009). The seizure freedom positive and negative predictive values of complete ECD cluster and primary hub resections did not significantly differ, although complete resection of the primary hub showed slightly better negative predictive value (ECD: 50.0% NPV, hub: 64.7% NPV). Both quantitative ECD and functional connectivity analyses suggested that spatially restricted dipole distributions and higher connectivity in a smaller region correlate with better seizure outcomes. Our findings suggest that MEG network analysis could be a valuable complement to the ECD methods. The results of this study are an important step towards using non-invasive neurophysiologic recordings to accurately define the epileptic network.