149 Predictive Value of Magnetoencephalography to Guide Intracranial Implant Strategy. (1st April 2022)

• Record Type: Journal Article
• Authors: Anand, Adrish; Gadot, Ron; Najera, Ricardo A.; Smith, David; Hegazy, Mohamed; Gavvala, Jay R.; Shofty, Ben; Sheth, Sameer A.
• Is Part Of: Neurosurgery. Volume 68(2022)Supplement 1
• Page Start: 44
• Page End: 44
• Publication Date: 2022-04-01
• Subjects: Nervous system -- Surgery -- Periodicals; 617.48005
• DOI: 10.1227/NEU.0000000000001880_149
• Languages: English
• ISSNs: 0148-396X

Abstract

Abstract : INTRODUCTION: Magnetoencephalography (MEG) is a useful component of a pre-surgical evaluation. Due to its high spatiotemporal resolution, MEG often provides nonredundant information to the clinician when forming hypotheses about the epileptogenic zone (EZ). With the increasing utilization of stereo-EEG (sEEG), MEG clusters are more commonly used as an sEEG electrode target. However, there are no pre-defined features of an MEG cluster that predict whether it is representative of intracranial EEG interictal or ictal activity, which limits optimal utilization of MEG in surgical planning. METHODS: We retrospectively analyzed patients who had an MEG study since it became available at our center (2017-2021). Patients were included if they had a positive MEG prior to an sEEG evaluation. MEG dipoles and sEEG electrodes were reconstructed in the same coordinate space to calculate overlap between electrodes and MEG clusters, and to quantify MEG cluster characteristics. MEG cluster features including brain region, stability (degree to which dipoles are parallel), tightness (density of dipole distribution), and number of dipoles were included in a binary classifier to predict ictal and interictal activity. RESULTS: Across 39 included patients, 13% of sEEG electrodes sampled MEG clusters. In these contacts, there were higher rates of ictal (43.22% vs 17.36%, p < 0.001) and interictal activity (39.63% vs 18.93%, p < 0.001) compared to electrodes not sampling MEG clusters. For contacts sampling the MEG cluster, binary classification predicted ictal activity with 76.7% accuracy compared to 54.4% in shuffled data (c-statistic = 0.816), while interictal activity was predicted accurately at 68.2% compared to 57.8% in shuffled data (c-statistic = 0.672) . Further analysis of individual characteristics showed that cluster stability contributed most to the model's accuracy (c-statistic = 0.773), whereas tightness (c-statistic = 0.701) and number of spikes (c-statistic = 0.692) contributed to a lesser extent. Brain region (c-statistic = 0.553) was not predictive of ictal activity. CONCLUSION: MEG cluster stability, tightness, and number of dipoles can be used to predict ictal activity. Quantitative analysis of these features may be useful for prospective planning of intracranial diagnostic implants.