Temporal lobe epilepsy surgical outcomes can be inferred based on structural connectome hubs: a machine learning study
Abstract number :
537
Submission category :
5. Neuro Imaging / 5A. Structural Imaging
Year :
2020
Submission ID :
2422878
Source :
www.aesnet.org
Presentation date :
12/6/2020 5:16:48 PM
Published date :
Nov 21, 2020, 02:24 AM
Authors :
Ezequiel Gleichgerrcht, Medical University of South Carolina; Simon Keller - University of Liverpool; Daniel Drane - Emory University School of Medicine; Brent Munsell - University of North Carolina; Kathryn Davis - University of Pennsylvania; Carrie McDo
Rationale:
Medial temporal lobe epilepsy (TLE) is one of the most common forms of medication-resistant focal epilepsy in adults. Despite the removal of medial temporal structures, over a third of patients continue to have disabling seizures after surgery. Post-operative seizure refractoriness implies that extra-medial and/or extra-temporal networks are capable of influencing the brain networks and generating seizures. We tested whether the abnormalities of structural network integration of regional nodes could be associated with surgical outcomes.
Method:
Magnetic resonance diffusion images from the presurgical evaluation of 121 patients with drug-resistant TLE across three independent centers were used to reconstruct personalized whole-brain structural connectomes. Node-based graph theory measures were measured and used for testing out-of-sample prediction of surgical outcomes using support vector machine (SVM) and deep neural networks. An independent cohort of 47 patients from three different referral centers was used to probe the independent validity of our predictive models.
Results:
Machine learning approaches achieved classification accuracies > 90% when employing betweenness centrality relative to other topological measures. Nodes most strongly contributing to the predictive models involved extra-medial limbic and limbic projection zones in the temporal regions, contralateral medial temporal regions, ipsilateral thalamus, and bilateral frontal and parieto-occipital structures, in line with prior studies showing aberrant grey and white matter changes in these circuits. We found excellent validity of this model in the independent cohort.
Conclusion:
Nodal features mainly in extra-medial temporal, thalamic, frontal, and parieto-occipital regions prior to surgery are related to surgical outcomes in TLE. Patients with abnormally integrated structural network nodes are less likely to achieve seizure freedom. These findings provide additional information on the mechanisms of surgical refractoriness and may help predict chances of surgical success
Funding:
:This study was supported by an NIH/NINDS R01 grant (1R01NS110347-01A1)
Neuro Imaging