Multimodal MRI and clinicopathologic features predict BRCA mutation status with AUC up to 0.77 in high-risk breast cancer

A model combining clinicopathologic factors and radiologist-assessed MRI semantic features predicted germline BRCA mutations with AUC 0.77 in external validation (492 high-risk patients, multicenter). Radiomics did not improve performance in this retrospective study.

• Retrospective multicenter study of 492 high-risk breast cancer patients (Center A n=270, Center B n=222) who underwent preoperative MRI and BRCA testing.
• In external validation (train Center A, test Center B), the combined clinicopathologic + MRI semantic RF model achieved AUC 0.77, while clinicopathologic-only RF reached AUC 0.73; multimodal radiomics + clinicopathologic + semantic model (LR) had AUC 0.72.
• Internal validation (Center A) showed clinicopathologic-based LR AUC 0.73 and radiomics gradient-boosting models AUC 0.71–0.72; the primary limitation is the retrospective design with only two centers, and modest AUCs may not suffice for clinical triage.

RadPigeon summaries are original and for information only. They are not clinical advice.