Deep Learning Predicts Breast Cancer Chemotherapy Response via MRI
September 23, 2026
A deep learning model using EfficientNet-B0 encoders achieves an AUC of 0.79 for predicting pathological complete response in breast cancer patients. The architecture fuses pretreatment diffusion and contrast-enhanced MRI data with clinical variables, finding HR/HER2 subtype to be the most significant clinical predictor.
HOW THIS AFFECTS YOU
●
researcherThe late-fusion architecture demonstrates the utility of combining multimodal imaging with specific clinical subtypes.
●
healthThis approach may enable more personalized therapeutic planning for neoadjuvant chemotherapy.