An artificial intelligence–driven predictive model for EPclin risk assessment (ENDOPREDICT test) in HER2-negative luminal breast cancer

Accurate prognostic stratification is essential for guiding adjuvant treatment decisions in HER2-negative luminal breast cancer. Artificial intelligence applied to digital histopathology offers an opportunity to extract complementary information for molecular risk assessment. We developed a multimodal deep learning model combining whole-slide histopathological images, tumor size, and nodal status to predict ENDOPREDICT EPclin risk groups in 768 patients, including an Internal cohort (n = 256) and a multicenter External cohort (n = 512). Cross-cohort testing yielded AUCs of 0.74 for the Internal-trained model tested on the External cohort and 0.77 for the External-trained model tested on the Internal cohort, compared with 0.715 and 0.702 for the clinical-only model, respectively. Improvement was significant in the Internal test cohort (paired DeLong p = 0.0079), but not the External test cohort (p = 0.145). Pooled 10-fold cross-validation yielded AUCs of 0.787 and 0.789 for patients originating from the Internal and External cohorts, respectively, demonstrating moderate and relatively consistent discrimination. High-risk predictions had PPVs of 88.8–93.4%, sensitivities of 56.7–64.9%, and NPVs of 37.3–47.8%. These findings demonstrate the feasibility of EPclin risk prediction from digitized histopathological slides and clinical variables. The potential incremental contribution of image-derived features warrants further investigation and prospective multicenter validation.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-20
DOI
https://doi.org/10.1038/s41598-026-80438-9
Primary Topic
AI in cancer detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An artificial intelligence–driven predictive model for EPclin risk assessment (ENDOPREDICT test) in HER2-negative luminal breast cancer

Laura Haik, Valérie Velasco, Céline Blaye, Gaëtan Marie MacGrogan et al.
9 links
Scientific Reports
AI in cancer detection
article

An artificial intelligence–driven predictive model for EPclin risk assessment (ENDOPREDICT test) in HER2-negative luminal breast cancer

Laura Haik, Valérie Velasco, Céline Blaye, Gaëtan Marie MacGrogan, Julie Sentis, Maxime Brunet, Claire B. Larmonier, Léonie Alran, Nicolas Lembege, Olivier Saut, Nathalène Truffaux, Sophie AURIOL, Camille Chakiba, Margaux Andrieu, Dorothée Ulliel, Van-Linh Le
article en

Abstract

Accurate prognostic stratification is essential for guiding adjuvant treatment decisions in HER2-negative luminal breast cancer. Artificial intelligence applied to digital histopathology offers an opportunity to extract complementary information for molecular risk assessment. We developed a multimodal deep learning model combining whole-slide histopathological images, tumor size, and nodal status to predict ENDOPREDICT EPclin risk groups in 768 patients, including an Internal cohort (n = 256) and a multicenter External cohort (n = 512). Cross-cohort testing yielded AUCs of 0.74 for the Internal-trained model tested on the External cohort and 0.77 for the External-trained model tested on the Internal cohort, compared with 0.715 and 0.702 for the clinical-only model, respectively. Improvement was significant in the Internal test cohort (paired DeLong p = 0.0079), but not the External test cohort (p = 0.145). Pooled 10-fold cross-validation yielded AUCs of 0.787 and 0.789 for patients originating from the Internal and External cohorts, respectively, demonstrating moderate and relatively consistent discrimination. High-risk predictions had PPVs of 88.8–93.4%, sensitivities of 56.7–64.9%, and NPVs of 37.3–47.8%. These findings demonstrate the feasibility of EPclin risk prediction from digitized histopathological slides and clinical variables. The potential incremental contribution of image-derived features warrants further investigation and prospective multicenter validation.

Scientific Reports
Centre National de la Recherche Scientifique (FR), Institut national de recherche en sciences et technologies du numérique (FR), Université de Bordeaux (FR), Institut Bergonié (FR), Institut de Mathématiques de Bordeaux (FR)
Good health and well-being
AI in cancer detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon