Radiomics Analysis of Contrast-Enhanced Breast MRI for Optimized Modelling of Virtual Prognostic Biomarkers in Breast...

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Breast cancer management relies on clinical stage, nodal status, and biomarkers like ER, HER2, and grade.
• Accurate prediction of these factors can prevent unnecessary interventions, such as surgery.

Purpose of the Study:

• To evaluate the utility of MRI radiomics in predicting prognostic biomarkers for breast cancer.
• Investigate the potential of radiomics to yield virtual biomarkers including ER, HER2 expression, tumor grade, molecular subtype, and T-stage.

Main Methods:

• Retrospective review of 209 invasive breast cancer patients who underwent dynamic contrast-enhanced (DCE) MRI.
• Extraction of Haralick texture features from DCE images and selection using Bootstrap Lasso.
• Performance assessment using area under the receiver operating characteristic curve (AUC).

Main Results:

• Radiomics models showed moderate performance in differentiating nodal status (AUC=0.78 for N0 vs N1-N3) and predicting HER2 status (AUC=0.64).
• The model demonstrated potential in distinguishing high nuclear grade (AUC=0.71) and ER status (AUC=0.67).
• Performance was also assessed for molecular subtypes, with AUCs of 0.60 for triple-negative and 0.66 for Luminal A.

Conclusions:

• Quantitative MRI radiomics texture analysis shows potential for identifying aggressive, node-positive triple-negative breast cancer.
• Radiomic features correlated well with higher nuclear grades, T-stages, and N-positive stages.