Abstracts

This abstract has been invited to present during the Translational Research platform session.
This abstract has been invited to present during the Basic Science Poster Highlights poster session.

Rationale: Medically intractable epilepsy patients undergo multiple neuroimaging tests, which in turn generate multimodal preoperative data. To facilitate multimodal integration and structure-function coupling, we developed an anatomical brain atlas – Yale Brain Atlas (YBA) – specialized for epilepsy surgery evaluation. Here, we first applied YBA to integrate seizure spread data from iEEG and pairwise connectivity matrices from rsfMRI. We assessed if rsfMRI could be used to localize seizure onset and spread areas. Second, we applied YBA to test if seizure spread features are associated with surgical outcome. Predicting surgical outcome could prevent or modify or modify unsuccessful resective surgeries.

Methods: To localize seizure onset and spread areas, rsfMRI connectivity matrices in YBA space labeled according to onset and spread parcels from iEEG were analyzed for 23 patients and 33 healthy controls. Each index in the labeled rsfMRI matrices reflected one of three possible pairwise correlation types: epileptogenic vs. epileptogenic (E|E), epileptogenic vs. nonepileptogenic (E|N), or nonepileptogenic vs. nonepileptogenic parcel (N|N). Two main statistical workflows were executed. The first workflow assessed if the pairwise correlations of E|E, E|N, and N|N had any significant differences across the 23 patients. The second workflow used shared response modelling (SRM) to assess if the pairwise correlations of E|E, E|N, and N|N had any significant differences for the 23 patients compared to 33 controls. As for predicting surgical outcome, six features were extracted from 56 seizure spread profiles in YBA space: total parcel number, first spread distance, total spread distance, first spread velocity, average spread velocity, and white matter tract count. Statistical tests were performed to evaluate the association of these features with surgical outcome.

Results: We created a brain atlas (YBA) specialized for epilepsy surgery evaluation to serve as a substrate for multimodal analysis. For predicting surgical outcome from spread patterns, we found that total spread distance is a significant predictor of surgical outcome for temporal neocortical onset seizures and that first and average spread velocities are significant predictors of surgical outcome for frontal onset seizures. For localizing epileptogenic networks from rsfMRI, we found that rsfMRI pairwise correlations were greatest for E|E pairs, followed by N|N and E|N. Further analysis using SRM revealed that the difference between patient and healthy SRM-inverted rsfMRI data was greatest for E|E pairs, followed by N|N and E|N.

Conclusions: Our results suggest that epileptogenic parcels have rsfMRI connectivity values greater than those of nonepileptogenic parcels. Moreover, our results suggest that resective surgical outcome is significantly poorer for temporal neocortical intractable patients with extensive seizure spread and frontal intractable patients with fast seizure spread.

Funding: Department of Neurosurgery, Yale School of Medicine