乳腺MRI肿瘤生物学标志物：影像与病理整合指导临床诊疗

BACKGROUND: Breast cancer is biologically heterogeneous, and tissue biomarkers derived from core biopsy may be limited by sampling and spatial heterogeneity. Breast MRI provides multiparametric information on tumor morphology, enhancement behavior, and peritumoral tissue characteristics. We evaluated whether standardized BI-RADS MRI descriptors are independently associated with histopathological features, immunohistochemical markers, and molecular subtypes in invasive breast cancer.

METHODS: This retrospective, single-center study included 340 consecutive women with invasive ductal carcinoma of no special type or invasive lobular carcinoma who underwent pretreatment 3T breast MRI between 2014 and 2019. MRI features were assessed using the BI-RADS MRI lexicon by a single experienced breast radiologist (>20 years), blinded to pathological data. Imaging descriptors were correlated with histological grade, lymphovascular invasion (LVI), ER, PR, HER2, Ki-67, and immunohistochemistry-defined molecular subtypes. Multivariable logistic regression was used to identify independent imaging predictors.

RESULTS: In multivariable models, peritumoral edema independently predicted LVI (OR 2.12, 95% CI 1.18-3.81) and high proliferative activity (Ki-67 ≥ 20%; OR 3.02, 95% CI 1.89-4.83). T2 mixed/hyperintense signal independently predicted high Ki-67 (OR 2.67, 95% CI 1.47-4.85) and HER2 positivity (OR 3.12, 95% CI 1.72-5.66). Rim enhancement was independently associated with the triple-negative subtype (OR 3.83, 95% CI 1.48-9.94) and high Ki-67 (OR 4.12, 95% CI 2.03-8.36). Washout kinetics independently predicted high Ki-67 (OR 3.18, 95% CI 1.82-5.56) and were inversely associated with Luminal A tumors (OR 0.15, 95% CI 0.06-0.37). HER2 positivity was independently associated with peritumoral edema, T2 hyperintensity, plateau kinetics, non-mass enhancement (OR 2.05, 95% CI 1.18-3.56), and segmental non-mass distribution (OR 3.28, 95% CI 1.12-9.61). Across analyses, a consistent imaging pattern combining edema, T2 hyperintensity, rim enhancement, and washout kinetics clustered in biologically aggressive tumors.

CONCLUSIONS: Standardized BI-RADS breast MRI descriptors demonstrate consistent and independent associations with established pathological biomarkers and molecular subtypes. Although MRI does not replace tissue-based assessment, these imaging phenotypes may provide complementary biological context and support radiopathological correlation in multidisciplinary care. Given the retrospective, single-center design with single-reader assessment and absence of external validation, these findings should be interpreted with appropriate caution and require prospective multicenter confirmation before clinical implementation.