Related Experiment Video
Updated: Aug 6, 2026
10:37
A Multimodal Imaging Framework to Advance Phenotyping of Living Label-free Breast Cancer Cells
Published on: August 22, 2025
Development and Validation of a Multimodal-Multitask Deep Learning Approach for Estimating Late Distant Recurrence
Eleftherios P Mamounas1,2, Ming Chen3, Joseph A Sparano4
1NSABP Foundation , NRG Oncology, Pittsburgh, Pennsylvania.
Cancer Research Communications
|July 17, 2026
Summary
An artificial intelligence (AI) model accurately predicts late distant recurrence (DR) risk in hormone receptor-positive (HR+) early breast cancer. This AI tool helps identify patients who may benefit most from extended endocrine therapy (ET).
Area of Science:
- Oncology
- Artificial Intelligence
- Pathology
Background:
- Late distant recurrence (DR) is a significant risk for hormone receptor-positive (HR+) early breast cancer patients after initial endocrine therapy (ET).
- Current methods for risk stratification and predicting benefit from extended endocrine therapy (ELT) have limitations.
- Accurate identification of patients at high risk for late recurrence is crucial for personalized treatment decisions.
Purpose of the Study:
- To develop and validate a multimodal artificial intelligence (AI) model for improved long-term risk stratification of late distant recurrence (DR) in HR+ early breast cancer.
- To assess the heterogeneity in benefit from extended letrozole therapy (ELT) across different risk groups identified by the AI model.
- To provide a scalable and generalizable tool that complements existing genomic assays for guiding treatment decisions.
Main Methods:
- A deep learning model was developed integrating digitized hematoxylin and eosin whole-slide images with clinicopathologic variables.
- The model was trained and cross-validated using data from 2,271 patients in the NSABP B-42 trial.
- External validation was performed on 4,300 patients from the TAILORx trial, evaluating prognostic performance using hazard ratios (HR) and absolute risk differences.
Main Results:
- The AI model effectively stratified patients into distinct risk groups with significantly different 10-year absolute DR risk in the NSABP B-42 cohort (7.95% difference between high- and low-risk groups; HR, 5.71).
- High-risk patients demonstrated a greater absolute benefit from ELT (4.09%) compared to low-risk patients (0.49%).
- External validation in the TAILORx cohort confirmed the model's prognostic performance in identifying patients with significantly different late DR outcomes (HR, 1.893).
Conclusions:
- This multimodal AI approach, utilizing routine pathology and clinical data, provides robust and generalizable stratification of late DR risk in HR+ breast cancer.
- The AI model can identify patients who may benefit more from extended endocrine therapy, supporting individualized treatment strategies.
- This scalable AI strategy offers a promising complement to existing genomic assays for optimizing patient management.