Related Experiment Video
Updated: Apr 27, 2026
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
Beyond Pathology: MRI-derived Metrics Unlock Preoperative Prognostic Risk Stratification in Breast Cancer
Mehrad Zare1, Alisa Mohebbi1, Afshin Mohammadi2
1School of Medicine, Tehran University of Medical Sciences, Tehran, Iran; Association of Nuclear Medicine and Molecular Imaging (ANMMI), Universal Scientific Education and Research Network (USERN), Tehran, Iran.
Background:
The Nottingham Prognostic Index (NPI) remains the gold-standard for breast cancer (BC) prognostic risk stratification, yet its reliance on postoperative pathology delays critical treatment decisions. We hypothesized that MRI could unlock prognostic insights before surgery, enabling immediate preoperative risk classification without waiting for histology.
Methods:
In a retrospective cohort of 791 patients with invasive BC, we leveraged quantitative (tumor size, volume, and enhancement metrics) and categorical (presence or absence of axillary lymphadenopathy, multicentricity, contralateral disease, chest wall and skin involvement) MRI features to predict high-risk NPI grades (>3.4). Discriminative performance was determined by receiver operating characteristic analysis and multivariable logistic regression, with stratification by molecular subtype, tumor size, and breast density.
Results:
Quantitative MRI metrics demonstrated strong discriminatory power; tumor size and total enhancing volume achieved area under curve (AUC) of 0.855 and 0.867, while washout-enhancing volume provided vascular insights (AUC 0.828). Notably, performance varied by tumor biology; volumetric metrics dominated in HER2-enriched tumors (AUC 0.965) and washout kinetics excelled in triple-negative disease (AUC 0.932). Among categorical features, axillary lymphadenopathy and multicentricity contributed complementary prognostic value. Integration of 3 selected parameters of tumor size, lymphadenopathy, and multicentricity yielded a robust combined model (AUC 0.883) with 89.8% sensitivity and 72.0% specificity. Critically, MRI metrics remained robust across breast density categories.
Conclusion:
MRI serves as a powerful noninvasive proxy for pathological NPI, enabling risk-stratified preoperative counseling and treatment planning. Molecular subtype-tailored MRI interpretation offers a precision medicine pathway toward individualized breast cancer management before surgery.
