Abstract / Summary
This study aimed to investigate the value of T2-weighted fluid-attenuated inversion recovery (T2W-FLAIR)-based histogram radiomics for differentiating multiple sclerosis (MS) lesions from non-specific and migraine-related white matter lesions. A total of 56 patients were classified into three groups: non-specific white matter lesions (Group 1, n=20), migraine-related lesions (Group 2, n=16), and MS lesions (Group 3, n=20). Lesions were manually segmented by two independent observers using 3D Slicer, and histogram-based first-order features were extracted. Group comparisons were performed using the Kruskal–Wallis test, followed by pairwise Mann–Whitney U tests. Binary logistic regression assessed the independent predictive value of significant parameters, diagnostic performance was evaluated by receiver operating characteristic (ROC) analysis, and inter-observer reliability by the intraclass correlation coefficient (ICC). Significant differences were found among the three groups for entropy, variance, uniformity, mean, and median values (all adjusted p<0.05), mainly driven by the MS group. MS lesions showed higher entropy, variance, mean, and median values, and lower uniformity. After adjustment for age and sex, all five parameters remained significant predictors of MS, and logistic regression confirmed their independent predictive value (all adjusted p<0.05). Entropy, variance, and uniformity showed high discrimination (univariate AUC: 0.832–0.838); the adjusted entropy model reached an AUC of 0.879. Inter-observer reliability was good to excellent (ICC 0.818–0.999). These findings suggest that T2W-FLAIR-based histogram radiomics provides useful quantitative information for differentiating MS lesions from other white matter lesions, particularly through heterogeneity-related features that may serve as supportive imaging biomarkers in the characterization of MS lesions.