"Effectiveness of Morphometric Analyses for Detecting Focal Cortical Dysplasia in Epilepsy Surgery Candidates."

Drug-resistant epilepsy (DRE) affects approximately 30% of patients and requires accurate identification of the epileptogenic zone (EZ) to enable curative surgery, particularly in MRI-negative or subtle-lesion cases. Focal cortical dysplasia (FCD) is a common cause of DRE but remains challenging to detect radiologically. FlaT1 is a voxel-based morphometric approach based on the FLAIR/T1 ratio that enhances grey–white matter contrast to improve FCD detection. We validated its performance against MAP18. We retrospectively included 47 patients with histopathologically confirmed FCD who underwent epilepsy surgery. Surgical outcome was classified using the Engel scale. Preoperative 3D T1-weighted and FLAIR images were processed with FlaT1 and MAP18 to generate junction and extension z-score maps. Postoperative MRI was segmented to define the resection cavity, used as spatial reference. Qualitative correspondence between morphometric clusters and the resection cavity was classified as full, partial, or absent by two independent raters. Inter-rater agreement was assessed using Cohen’s κ and PABAK. Quantitative validation compared mean z-scores inside versus outside the resection cavity using paired t-tests. Diagnostic performance metrics were calculated, and Fisher’s exact test assessed association with surgical outcome. Subgroup analyses were performed by lobar location and histopathology. Inter-rater agreement was fair (κ = 0.389) but substantial after prevalence adjustment (PABAK = 0.75). For FlaT1, junction maps showed 74.5% full and 21.3% partial correspondence (95.7% overall), while extension maps showed 63.8% full and 29.8% partial correspondence (93.6% overall). Joint full correspondence was observed in 61.7% of cases. MAP18 demonstrated comparable junction performance but lower extension concordance. FlaT1 showed significantly higher mean z-scores inside versus outside the resection cavity for both junction (Δ = 0.36, p = 0.012) and extension maps (Δ = 0.27, p = 0.001), whereas MAP18 reached significance for junction maps only (Δ = 0.20, p = 0.005). For overall full correspondence, FlaT1 achieved sensitivity 80.6%, specificity 100%, PPV 100%, and accuracy 85.1%, outperforming MAP18 (accuracy 80.9%). Fully corresponding clusters were strongly associated with favourable surgical outcome for both pipelines (FlaT1 overall p = 1.83 × 10⁻⁶). High detection rates were maintained across lobar locations and FCD subtypes. FlaT1 demonstrated robust spatial concordance with resected dysplastic cortex and strong association with postoperative outcome. Its high specificity and predictive value support its use as a reliable adjunct tool in the presurgical evaluation of FCD.