A Multicenter Prospective Observational Cohort Study: Predicting Neoadjuvant Chemotherapy Response Using Pre-Treatment DCE-MRI-Based AI Models in HR+/HER2- Breast Cancer
This study is a multicenter, prospective, observational cohort study to evaluate the predictive performance of pre-treatment DCE-MRI-based artificial intelligence (AI) models for neoadjuvant chemotherapy benefit in HR+/HER2- breast cancer. The study plans to enroll eligible HR+/HER2- breast cancer patients receiving routine standard neoadjuvant chemotherapy and stratify participants into high-benefit and low-benefit subgroups via the established AI model based on baseline breast DCE-MRI images. All enrolled patients will undergo systematic collection of baseline clinical-pathological data, pre-treatment DCE-MRI scans, neoadjuvant chemotherapy regimens, postoperative residual cancer burden (RCB) classification, objective response rate (ORR), and long-term survival endpoints including disease-free survival (DFS) and overall survival (OS). The primary objective compares the rate of RCB 0-1 between AI-defined high-benefit patients and published historical control data; secondary analyses compare ORR, RCB 0-1 proportion, DFS and OS between AI-stratified high-benefit and low-benefit subgroups to comprehensively verify the clinical value of this imaging AI model for individualized neoadjuvant chemotherapy selection.
• Female patients aged ≥ 18 years old.
• Histopathologically confirmed invasive breast carcinoma.
• Hormone receptor positive (ER and/or PR ≥1%), HER2-negative status (IHC 0-1+, or IHC 2+ with negative FISH result).
• Clinical stage II-III breast cancer per the 8th AJCC staging system, with clinical indication for neoadjuvant chemotherapy or primary surgery.
• Standard pre-treatment breast DCE-MRI performed before neoadjuvant chemotherapy, with image quality eligible for AI model analysis.
• ECOG performance status 0 or 1; adequate function of major vital organs to tolerate planned clinical treatment.
• Voluntary participation with written informed consent obtained.