Beyond pCR Prediction: Subtype-Specific Artificial Intelligence for Treatment Tailoring in Breast Cancer Neoadjuvant Therapy

Abstract

Simple Summary Breast cancer comprises biologically distinct subtypes rather than a single disease, and each is treated with a different pre-surgery strategy. Clinicians want to know early whether a patient is likely to respond, because that could guide safer treatment adjustment. Artificial intelligence has therefore been used to estimate response from routinely collected imaging, pathology slides, and laboratory data, using models that range from statistical scores to deep-learning systems and multimodal designs. This review finds such estimates are not interchangeable across subtypes and must be read in their original treatment context. Most models were tested only where they were built, few in independent patients, and fewer still for reliable probabilities or improved decisions. Current evidence is therefore not strong enough for AI to change an individual treatment or surgery decision safely. We map how mature the evidence is, how it differs across subtypes, and what is still required.