Abstracts

This abstract has been invited to present during the Neuroimaging platform session
This abstract has been invited to present during the Broadening Representation Inclusion and Diversity by Growing Equity (BRIDGE) poster session

Rationale: A single unprovoked seizure (SUS) occurs in up to 10% of patients but does not necessarily develop into epilepsy. It is unclear what determines the susceptibility to develop epilepsy. Although brain network changes have been ascertained in people with epilepsy, this has not been studied in SUS patients.

Methods: Using 7T MRI, fMRI scanning and T1 anatomical scan with resting state and a movie for naturalistic analysis of functional connectivity (FC) was co-registered. The images were pre-processed with fmriprep, denoised with nilearn, and had connectivity matrices extracted using Pearson correlations between the 300 regions in the Schaffer parcellation (Figure 1). We next applied graph theory measures of centrality to identify regions with the highest difference between healthy control (HC) and SUS patients using sensitivity index. We then used these regions to fit a logistic regression model to classify SUS patients against HC.

Results: We recruited 12 patients with a SUS and 17 HC with similar demographic characteristics. The data was analyzed utilizing eigenvector centrality (EC) measures with the highest average accuracy with 5-fold cross-validation when compared to other common centrality measures (Figures 2A, 2B). Using EC, we achieved an average accuracy of 87% with 5-fold cross-validation in resting state recording and 97% in the movie-driven recording. Patients with a SUS had alteration in FC involved in the prefrontal, insular and temporoparietal regions versus HC in the movie-driven data. Resting state analysis showed differences in the limbic system,  default mode network versus HC (Figure 2C). 

Conclusions: This is the first study to show that alterations in  brain network connectivity can be seen early and after a single seizure with normal EEG and normal MRI. Epilepsy as a network disease happens early and causes an alteration in connectivity, regardless of the seizure type. These changes in connectivity may signify a future degree of alteration in patients that may develop epilepsy after a single seizure. 
_x000D_ Figure 1. Schematic representation of data acquisition and preprocessing. The FMRI recording is collected during movie and resting states from 17 healthy and 12 first-seizure participants. The recordings are preprocessed using Schaefer2018.
_x000D_ Figure 2.Healthy vs. first seizure classification. (a) Average accuracy (5-fold cross-validation) of classification models developed to detect first-seizure participants plotted as a function of the number of regions included with the highest sensitivity index. The sensitivity index is computed using degree centrality, closeness centrality, betweenness centrality and EC. The set of regions selected based on the EC outperforms the rest of centrality measures. (b) The EC used to detect first seizure participants across all recording types including movie-driven, resting state recordings. Similarly, the average accuracy is plotted as a function of the number of regions with the highest sensitivity index. (c) The regions correspond to the highest average accuracy for each recording type and their corresponding values in healthy and first-seizure participants.

Funding: No funding was received in support of this abstract.