MRI-based prediction of malignancy in peripheral nerve tumors: model development and internal validation in a biopsy-confirmed cohort.

PURPOSE: Differentiating malignant from benign peripheral nerve sheath tumors (PNSTs) is challenging due to overlapping imaging features. To develop and internally validate an MRI-based model to differentiate malignant from benign PNSTs and guide biopsy decisions.

MATERIALS AND

METHODS

We retrospectively analyzed 139 consecutive adults with histologically proven PNSTs (121 benign, 18 malignant) discussed at a tertiary referral tumor board (2015-2024). Pre-treatment MRI examinations were independently assessed by two experienced musculoskeletal radiologists blinded to clinical and histopathological data. Based on prior knowledge and sample size limitations, a parsimonious logistic regression model was prespecified, including tumor volume, presence of a well-defined margin, and presence of cystic degeneration. Model performance was evaluated for discrimination, calibration, and clinical utility, with internal validation using 500 bootstrap resamples and uniform shrinkage.

RESULTS

Malignant tumors were substantially larger, less often well-circumscribed, and more frequently cystic than benign lesions. The final three-predictor model demonstrated good discrimination (area under the curve [AUC] 0.944; optimism-corrected 0.936). At a 20% biopsy threshold, the model would prevent 68 unnecessary biopsies per 100 patients while missing three malignancies. Findings were robust in sensitivity analyses.

CONCLUSION

A simple MRI-based model using three available features - tumor volume, margin definition, and cystic change - distinguished malignant from benign PNSTs with high accuracy. This proof-of-concept provides a practical framework for biopsy decision-making and merits external validation in multicenter cohorts.

KEY POINTS: · A simple MRI-based model estimates malignancy risk in peripheral nerve tumors.. · The model combines tumor volume, margin irregularity, and cystic degeneration.. · An internally validated model may help prioritize biopsy decisions in specialized centers..

CITATION FORMAT: · Ghali Gataa K, Dabrowska K, Kader M et al. MRI-based prediction of malignancy in peripheral nerve tumors: model development and internal validation in a biopsy-confirmed cohort. Rofo 2026; DOI 10.1055/a-2852-8884.