Deep Learning Identifies BMI-Associated Morphologic Features in Breast Cancer Tissue

In a cohort of 195 breast cancer patients, AI embeddings from non-adipose stromal histology classify high BMI (≥30 kg/m2) vs low BMI (<25 kg/m2) with average AUC 0.78 (95% CI 0.73–0.83).

• Why it matters: Obesity is linked to worse breast cancer incidence and outcomes, but the tissue-level morphologic mechanisms behind this association are not well defined.
• What they did: They used an AI pathology foundation model to extract vector embeddings and train machine learning classifiers focused on tumor, stromal, or combined regions, then compared performance and embedding dissimilarity.
• The result: Stromal signatures tied to higher BMI also appear across hormone-receptor and age subgroups and are associated with reduced inter-patient heterogeneity in both tumor and stromal tissues.

AI summary of the abstract · may contain errors · not medical advice · read the paper.