Abstracts

Rationale: One third of persons with epilepsy (PWE) are medically refractory and rely on resective surgery for seizure control. Part of pre-surgical planning involves evaluating the integrity of functional networks. Given the discovery of widespread network disruptions in focal epilepsies, it is important to characterize the functionality of these networks during active engagement. We implemented a movie-driven fMRI technique that evokes reliable, time-locked BOLD signal fluctuations that are highly stereotyped across viewers [i.e., produces high inter-subject correlation (ISC)] in order to assess abnormal functional connectivity in PWE. We aimed to (1) examine eight canonical functional networks for idiosyncratic activation among patients during movie-viewing, and (2) assess the predictive utility of these networks to classify patients based on the degree of network asynchrony. Methods: Four pre-surgical patients with temporal lobe epilepsy (age 21-53, one female, all with left or bilateral hippocampal pathology) and 11 controls watched an 8-minute suspenseful movie clip during an fMRI scan in a 3T MRI system (Siemens Prisma). Representative timecourses from eight functional networks in each participant were extracted using independent components analysis and z-transformed. We computed intra-network ISC values for each participant by correlating their timecourse with the median timecourse of the remaining controls. Within each network, we tested for asynchrony using bootstrap confidence intervals (CI; 10 million simulations) computed for each participant. We classified participants as PWE if the upper 95% CI of their ISC did not overlap with the median ISC of the remaining controls. We compared the sensitivity and specificity of this criterion across functional networks. Results: We observed above chance classification across all networks (one-way Wilcox signed rank Z = 2.68, p = .007). Two of our four patients exhibited significantly asynchronous timecourses in every network. The most sensitive networks were auditory, dorsal attention, and right executive control. Movie-driven ISC in these networks classified study subjects as patients with hit rates of 4/4 and false alarm rates of 2/11 [all d's = 2.06, median ISC: Pearson's r controls = 0.83, r patients = 0.66 (auditory), r controls = 0.72, r patients = 0.51 (dorsal attention), r controls = 0.29, r patients = 0.10 (right executive control)]. All patients and controls had significant ISCs compared to their circularly resampled null distributions. Conclusions: The movie paradigm is a rapidly administered, non-invasive tool that is sensitive to alterations in widespread cortical networks at the individual level. We identified three functional networks that exhibit both highly stereotyped movie-driven activation among controls as well as high sensitivity to network asynchrony among PWE. We propose that movie-driven fMRI may aid in diagnosis of equivocal cases and may be valuable in pre-surgical evaluation. Funding: The present research is funded by Canadian Institutes of Health Research.