Back

Anatomical variations of the peroneus brevis tendon on axial MRI: A validated cross-sectional classification at the lateral malleolus

Zych, R.; Mocanu, D.; Hagberg, Y.; Bokwa-Dabrowska, K.; Huuskonen, M.; Romanus, I.; Dziedzic, D.; Szaro, P.

2025-05-19 orthopedics
10.1101/2025.05.18.25327845 medRxiv
Show abstract

PurposeThe shape of the peroneus brevis tendon may contribute to the development of split tears, but a classification of its anatomical variants on transverse section in vivo is lacking. This study aimed to validate a proposed classification system for peroneus brevis tendon morphology on the transverse cross section on magnetic resonance imaging (MRI). MethodsWe analysed 130 normal peroneus brevis tendons (power >0.8, = 0.05, effect size = 0.31) on consecutively selected ankle MRI performed at 3T. We excluded patients with fractures, ligament injuries, or peroneal tendon pathology. Seven independent raters classified tendons into four morphological categories: general flat, flat with a lateral bulge, flat with a medial bulge, and oval. Inter-rater reliability was assessed using Cohens kappa, Gwets AC1, and Fleiss kappa. We evaluated classification robustness using F1 score, receiver operating characteristic analysis, and precision-recall metrics. ResultsConsensus classification showed 37.7% of tendons were general flat, 20.8% flat with a lateral bulge, 19.2% flat with a medial bulge, and 22.3% oval. Inter-rater agreement was substantial, with Fleiss kappa of 0.69 (95% CI 0.63-0.75). Cohens kappa ranged from 0.59 to 0.87, and Gwets AC1 from 0.58 to 0.87, confirming strong pairwise agreement. Classification performance was high (area under the curve = 0.89), and majority voting further improved agreement (F1 score = 0.93). The most common misclassification occurred between oval and flat with a medial bulge. ConclusionThis proposed classification is robust and reliable, offering a validated framework for consistent tendon evaluation in both clinical and research settings.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.