Network connectivity separate from the hypothesized irritative zone correlates with impaired cognition and higher rates of seizure recurrence

Introduction: Surgery remains an essential option for the treatment of medically intractable temporal lobe epilepsy (TLE). However, only 66% of patients achieve postoperative seizure freedom, perhaps attributable to an incomplete understanding of brain network alterations in surgical candidates. Here, we applied a novel network modeling algorithm and measured key characteristics of epileptic networks correlated with surgical outcomes and objective measures of cognition. Methods: Twenty-two patients were prospectively included, and relevant demographic information was attained. Resting state functional magnetic resonance imaging (rsfMRI) and electroencephalography (EEG) data were recorded and preprocessed. Using our novel algorithm, patient-specific epileptic networks were mapped preoperatively, and geographic spread was quantified. Global functional connectivity was also determined using a volumetric functional atlas. Neuropsychological pre- and postsurgical raw and standardized scores obtained blinded to epileptic network status. Key demographic data and features of epileptic networks were then correlated with surgical outcome using Pearson's product–moment correlation. Results: At an average follow-up of 18.4 months, 15/22 (68%) patients were seizure-free. Connectivity was measured globally using a functional 3D atlas. Higher mean global connectivity correlated with worse scores in preoperative neuropsychological testing of executive functioning (Ruff Figural Fluency Test [RFFT]-ER; R = 0.943, p = 0.005). A higher ratio of highly correlated connections between regions of interest (ROIs) in the hemisphere contralateral to the seizure onset correlated with impairment in executive functioning (RFFT-ER; R = 0.943, p = 0.005). Higher numbers of highly correlated connections between ROIs in the contralateral hemisphere correlated with impairment in both short- and long-term measures of verbal memory (Rey Auditory Verbal Learning Test Trials 6, 7 [RAVLT6, RAVLT7]; R = − 0.650, p = 0.020, R = − 0.676, p = 0.030). Epilepsy networks were modeled in each patient, and localization of the epilepsy network in the bitemporal lobes correlated with lower scores in neuropsychological tests measuring verbal learning and short-term memory (RAVLT6; R = − 0.671, p = 0.024). Higher rates of seizure recurrence correlated with localization of the epilepsy network bitemporally (R = − 0.542, p = 0.014), with the stronger correlation found with localization to the contralateral temporal lobe from side of surgery (R = − 0.530, p = 0.016). Conclusion: Increased connectivity contralateral to seizure onset and epilepsy network spread in the bitemporal lobes correlated with lower measures of executive functioning and verbal memory. Epilepsy network localization to the bitemporal lobes, in particular, the contralateral temporal lobe, is associated with higher rates of seizure recurrence. These findings may reflect network-level disruption that has infiltrated the contralateral hemisphere and the bitemporal lobes contributing to impaired cognition and relatively worse surgical outcomes. Further identification of network parameters that predict patient outcomes may aid in patient selection, resection planning, and ultimately the efficacy of epilepsy surgery.