Validation of an MRI-based classification of peroneus brevis tendon morphology: a four-type system with high inter-rater reliability

Objectives Split tears of the peroneus brevis tendon are a common and often underrecognized cause of lateral ankle pain and instability in athletes. Although MRI is widely used for ankle assessment, no validated system exists for classifying normal morphological variation of the peroneus brevis tendon. This study aimed to validate an MRI-based classification system. Methods We analyzed 130 normal peroneus brevis tendons (power > 0.8, α = 0.05, effect size = 0.31) on consecutive 3T ankle MRI scans. Persons with fractures, ligament injuries, or peroneal tendon pathology were excluded. Each tendon was evaluated on a single standardized axial PD-weighted slice, selected just proximal to the lateral malleolus apex. Seven independent raters classified tendons into four types: general flat, flat with lateral bulge, flat with medial bulge, and oval. Inter-rater reliability was assessed using Cohen’s kappa, Gwet’s AC1, and Fleiss’ kappa. Classification robustness was evaluated using F1 scores, ROC analysis, and precision-recall metrics. Results The most common forms were general flat (37.7%) and oval (22.3%). Fleiss’ kappa was 0.691 (95% CI 0.629–0.752), with strong pairwise agreement (Cohen’s kappa and Gwet’s AC1: 0.58–0.87). ROC AUC was 0.89; majority voting improved agreement (F1 = 0.93). Conclusions This classification system is robust and reliable, supporting consistent MRI-based assessment of tendon morphology in clinical and research settings.