Surface-based analysis of cortical FLAIR intensity. Clinical application in epilepsy surgery and its translational potential for basic neuroscience research.

Background: Surface-based analysis of cortical FLAIR intensity can reveal subtle abnormalities that escape conventional inspection. We developed a workflow that projects normalized FLAIR signal onto cortical surfaces (SUPR-FLAIR) and applies vertex-wise statistical testing (SUPR-FLAIR-sa) to support presurgical planning in epilepsy and to probe translational applications in basic neuroscience. Methods: Using a FreeSurfer/FSL pipeline, individual scans were registered to fsaverage, FLAIR was normalized to white-matter median, and signal was sampled onto the cortical surface for vertex-wise GLM analysis (QDEC), visualized as signed –log10(P). Clinical evaluation comprised illustrative MRI-positive and MRI-negative cases and a retrospective cohort of MRI-negative patients undergoing temporal surgery (July 2013–July 2021; ≥24 months’ follow-up). Outcomes (Engel class), histopathology, and—in a performance subset—sensitivity/specificity relative to the resection zone (RZ) were assessed. We also examined extra-lesional patterns in hippocampal sclerosis (HS) and conducted a single-subject, paired rest–task pilot (finger tapping) to test for hyperacute FLAIR modulations. Results: Among MRI-negative surgical patients (n=29), outcomes were Engel I 62%, II 24%, III 14%; histology showed nonspecific gliosis 76%, malformations of cortical development 17%, and HS 7%. In the performance subset (21 patients + 30 controls), SUPR-FLAIR-sa yielded sensitivity 53.8% and specificity 44.7%. When the peak significant cluster (P<0.05) fell within the RZ, 70% achieved seizure freedom versus 38% when the peak lay outside. Concordance with FDG-PET statistics (SUPR-PET-sa) and colocalization within the RZ correlated with the highest seizure-freedom rates (up to 89%). In five HS cases, the most significant SUPR-FLAIR-sa cluster consistently localized to the ipsilateral temporal pole despite no visually appreciable FLAIR hyperintensity. In the pilot task–rest experiment, small but reproducible increases in cortical FLAIR were observed over sensorimotor areas (hand knob, SMA/pre-SMA). Conclusions: SUPR-FLAIR-sa provides actionable, though imperfect, localization in MRI-negative epilepsy, particularly when integrated with FDG-PET and SEEG. Key limitations include scanner specificity and sample size. Prospective, multi-scanner studies with harmonized normative cohorts are needed to refine thresholds, validate performance, and clarify whether cortical FLAIR can index transient functional states.