Development and Validation of a Pathomics Model for Prognosis Prediction in Neoadjuvant Therapy-Treated Breast...

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate prognosis prediction is crucial for tailoring neoadjuvant therapy (NAT) in breast cancer patients.
• Current methods may not fully leverage integrated clinicopathological and histopathological data.

Purpose of the Study:

• To develop and validate a multimodal artificial intelligence (AI) model, the ClinicHistomics Integrated Outcome Prediction Model (CIOPM).
• To predict prognosis, including overall survival (OS) and disease-free survival (DFS), in breast cancer patients receiving NAT.

Main Methods:

• The CIOPM integrates clinicopathological data with whole-slide images (WSIs) from hematoxylin and eosin (H&E)-stained surgical specimens.
• Model development and external validation were performed using data from 835 patients across four multicenter cohorts (847 WSIs).

Main Results:

• The CIOPM demonstrated robust predictive performance in validation cohorts, with C-indices for OS and DFS ranging from 0.915 to 0.947.
• The model achieved an Area Under the Curve (AUC) of 0.957 for both OS and DFS prediction, enabling accurate risk stratification.
• Ablation studies, subgroup analyses, and comparisons with existing clinical risk models confirmed the CIOPM's superior performance.

Conclusions:

• The CIOPM offers reliable risk stratification for breast cancer patients undergoing NAT.
• This AI-driven approach supports personalized treatment decisions by providing accurate outcome predictions.