Abstracts

Rationale:
Depending on MRI results, ~40%-70% of people with drug-resistant temporal lobe epilepsy (TLE) experience seizure recurrence following anterior temporal lobectomy (ATL) (Wiebe et al N Engl J Med 2001; Bell et al Epilepsia 2009). This highlights the need for tools to improve the surgical selection of patients with TLE (Jehi et al Lancet Neurol 2015). Recent studies suggest that resting-state functional connectivity (rs-FC) and network properties may improve patient stratification, presurgical planning, and surgical outcomes (He et al Neurology 2017; Corona et al Brain 2023). However, the profile of normative rs-FC patterns and deviations in TLE cohorts remains understudied. Here we extract normative scalp EEG rs-FC patterns using population-level data and investigate whether connectivity within the posterior default mode network regions (temporal, parietal, occipital) can lateralize epilepsy and prognosticate post-ATL outcomes.

Methods:
Our retrospective study included resting-state eyes closed awake interictal epochs from controls without evidence of neurologic disease and people with TLE. These scalp EEG datasets were collected at Mayo Clinic, Rochester (MCR). A per-subject static estimate of rs-FC was computed using spectral coherence (1-45Hz). A population set of routine EEGs expertly reviewed as normal (MCR-pop; N=6,737) was decomposed in a domain-guided fashion to establish normative whole-brain rs-FC patterns with distinct spectral and connectivity signatures (Fig 1). Next, twelve intra-hemispheric temporal-to-posterior (T-to-P) channel pairs (6 on each side) were used to analyze manually identified focal temporal epilepsy cases (MCR-focal; N=167) and presurgical EEGs of TLE-ATL subjects (MCR-tle; N=69). MCR-focal and MCR-tle data were assessed for the strength of normative patterns. The resulting “weights” revealed group-level differences in epilepsy lateralization and seizure-free outcomes as defined using the Engel score at one-year post-ATL (Fig 2).

Results:
Two of the six rs-FC patterns extracted from MCR-pop (Fig 1a, #1 #6) are likely of artifactual origin (> 25Hz; ~40% var. exp.), while the remaining four (#2-5) correspond to a mixture of potentially physiologic processes (1/f-slowing, theta, alpha, beta bands; ~60% var. exp.). These interpretations were supported by trends with increasing age (more 1/f-slowing, less alpha; Fig 1b). In T-to-P intra-hemispheric analysis of MCR-focal, higher 1/f-slowing connectivity (#2) was observed in the hemisphere contralateral to the seizure side (Fig 2a; red). Subjects with seizure-free outcomes versus poor outcomes did not show a difference in ipsilateral (ipsi) or contralateral (contra) T-to-P connectivity (Fig 2b).

Conclusions:
Our results suggest that intra-hemispheric differences in normative EEG rs-FC patterns may serve as non-invasive presurgical biomarkers for lateralizing TLE from normal EEG epochs. Further investigation is needed, including extensions to non-lesional MRI cases, larger age-matched TLE datasets, and out-of-sample validations.

Funding:
This research was supported by the Mayo Clinic & Illinois Alliance Fellowship (N. Wagh). Scalp EEG data was provided through the Mayo Clinic Neurology Artificial Intelligence Program.