Radiogenomics Analysis of DCE-MRI Textures for Tracking Tumor Response and Uncovering Biological Processes in HER2-Negative Breast Cancer

Motivation: DCE-MRI provides non-invasive insights into vascularization and the extracellular matrix, rendering it particularly suitable for assessing tumor response. Goal(s): To identify the optimal DCE-MRI timing and texture features for predicting pathologic complete response during neoadjuvant chemotherapy in breast cancer patients. Approach: DCE-MRI scans were taken before, early, and mid-treatment. A sample size of 488 was used for discovery and 250 for validation. Logistic regression compared predictions using benchmark FTV-based analysis and radiomics-based textures with an additional DCE-MRI scan. Results: Mid-treatment DCE-MRI, measuring texture entropy of dependence count, is optimal for predicting pCR. This texture is uniquely linked to angiogenesis and TGF-beta signaling. Impact: DCE-MRI effectively tracks breast tumor response by analyzing vascularization and the extracellular matrix. Specifically, mid-treatment scan measuring texture entropy of dependence count optimally predict pathologic complete response, exhibiting exclusive linkage to biological pathways such as angiogenesis and TGF-beta signaling.