Multimodality MRI radiomics analysis for predicting TP53 mutations in triple negative breast cancer

Motivation: TP53 mutations in triple-negative breast cancer(TNBC) are linked to aggressive behavior and treatment resistance. A non-invasive detection method could improve treatment planning and prognosis. Goal(s): This study evaluates the potential of MRI-based radiomic features combined with a machine learning classifier to predict TP53 status in TNBC preoperatively. Approach: In a retrospective study of 105 TNBC patients, 20 features were selected from 3,411 MRI radiomic features using Pearson correlation and Recursive Feature Elimination(RFE). A Support Vector Machine(SVM) was then trained and evaluated. Results: The SVM classifier achieved an AUC of 0.79 and an accuracy of 0.82 in predicting TP53 mutations in the validation cohort. Impact: This machine learning-based MRI radiomics model, trained on multi-center, multi-vendor data, demonstrated strong predictive performance, enhancing reliability, generalizability, and patient convenience. It reduces costs compared to invasive methods and offers broad clinical applicability across diverse fields.