Computational MEG Epileptic Spike Network Modeling for Prediction of Surgical Outcome in Intractable Epilepsy: Retrospective Analysis of 26 Children
Abstract number :
2.028
Submission category :
3. Neurophysiology / 3D. MEG
Year :
2021
Submission ID :
1826725
Source :
www.aesnet.org
Presentation date :
12/5/2021 12:00:00 PM
Published date :
Nov 22, 2021, 06:56 AM
Authors :
Ekta Shah, DO, MS - University of Texas McGovern Medical School; Pablo Cuesta – University Complutense of Madrid; RIcardo Bruna – University Complutense of Madrid; Michael Funke – University of Texas McGovern Medical School; Fernando Maestu – University Complutense of Madrid; Gretchen Von Allmen – University of Texas McGovern Medical School
Rationale: Personalized computational epileptic spike network modeling using data from magnetoencephalography (MEG) can be a valuable tool for prediction of outcome in epilepsy surgery.
Methods: In this retrospective study, MEG pre-surgical data from 26 children who had undergone resective surgery for intractable epilepsy was analyzed using an automated functional connectivity algorithm for interictal spikes. We developed a model using transitions-locked networks, instead of single state networks, to identify brain areas whose epileptic activity would depict a specific transition dynamic between a pre-spike state and the spike-onset state that remain stable for several consecutive frequency bands. This model was used to predict the brain regions containing node clusters most likely to destabilize the epileptic network if removed. These regions were then compared with the post-operative MRI for each individual to determine overlap with their respective resections. Clinical data pertaining to surgical outcome was used to classify each case as favorable (seizure-free) or unfavorable (not seizure-free). In our cohort of patients, 15 were favorable and 11 were unfavorable.
Results: The performance of the model was very high in 12 of the 15 favorable cases (80%), meaning that the brain area inferred by our model was included within the resected tissue cavity on post-operative MRI. In the group of subjects classified as surgically unfavorable, the model displayed candidate clusters outside the corresponding resected regions.
Conclusions: This is the first personalized data-driven computational neurosurgical model for epilepsy surgery using MEG data. The method was validated against the epilepsy surgery outcome. Although the model was tested in a limited number of patients, it provided a reliable result in a cohort containing variable epilepsy localization. We view that a future prospective study using this MEG-based model can further optimize and validate this methodology and positively impact surgical outcomes.
Funding: Please list any funding that was received in support of this abstract.: Intramural funding from University of Texas McGovern Medical School, Department of Pediatrics.
Neurophysiology