Resting-state fMRI biomarkers of surgical outcome in temporal lobe epilepsy: Network reorganization, rsfMRI-based prediction models, and cognition - A systematic review.

Temporal lobe epilepsy (TLE) is a network disorder extending beyond the structurally defined epileptogenic zone. Resting-state functional MRI (rsfMRI) may provide information on brain network alterations relevant to seizure outcome and cognitive risk after epilepsy surgery. We reviewed rsfMRI studies in TLE surgical cohorts to evaluate postoperative network changes, preoperative predictors of seizure outcome, and associations between functional connectivity and cognition. This systematic review followed PRISMA guidelines and was registered in PROSPERO. Fifteen studies were included, comprising 474 TLE patients who underwent resective epilepsy surgery and had at least 1 year of follow-up. Studies were reviewed for patient population, surgery, rsfMRI methodology, networks or regions, outcome definition, postoperative network changes, and cognitive assessments. Despite heterogeneity in patient selection, surgery, rsfMRI analysis, and outcome definition, studies consistently reported reduced DMN-related connectivity in TLE compared with healthy controls. Postoperative functional connectivity in seizure-free patients shifted toward healthy controls patterns in few studies. Preoperative rsfMRI-based prediction models showed promising accuracy for predicting seizure outcome but were generally developed in small cohorts and lacked external validation. Evidence relating rsfMRI to cognition was limited, but implicated thalamic, ascending reticular activating system, and thalamo-occipital connectivity in attention, language, and executive functions. RsfMRI is a promising research biomarker for network-level risk stratification in TLE surgery, particularly for seizure outcome prediction and postoperative network reorganization. Current evidence remains insufficient for clinical implementation. Future studies require larger homogeneous multicenter cohorts, standardized acquisition and preprocessing, consistent outcome definitions, systematic postoperative rsfMRI, neuropsychological assessment, and external validation of prediction models.