Abstract PS3-04-01: Prediction of pathologic complete response from histopathology images of HER2+ breast cancer using an AI foundation model

Abstract Background: Neoadjuvant taxane-based chemotherapy in combination with anti-HER2 therapy is standard in patients with locally advanced HER2-overexpressing (HER2+) breast cancer. Pathologic complete response (pCR) after combination therapy is associated with improved long-term outcomes and occurs less frequently in patients whose disease is also estrogen and/or progesterone receptor positive (HR+) compared to patients with estrogen and progesterone receptor negative (HR-) disease. In patients with HER2+, HR+ disease, improved patient selection for standard neoadjuvant therapy is needed to optimize outcomes and minimize toxicity in patients unlikely to achieve pCR. Here we sought to utilize digital pathology and artificial intelligence/machine learning (AI/ML) algorithms on baseline biopsy tissue of newly diagnosed patients with operable HER2+ breast cancer to predict likelihood of pCR after standard neoadjuvant therapy, with a particular focus on HR status-specific differences. Methods: Digitized H 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-04-01.