Foundation AI pathology models predicted breast cancer TP53 mutations (AUROC 0.70), CDH1 mutations (0.71), and breast cancer-specific survival (C-index 0.67) accurately.
17 cohorts covering 12 cancer types, including breast cancer (CPTAC-BRCA, PLCO-Breast), bladder (PLCO), GBM, LUAD, SARC, STAD, and UCEC. Training data included 250M patches extracted from 260k whole slide images (WSIs).
Digital pathology foundation model trained using self-supervised learning method DINOv2 on hematoxylin and eosin (H&E)-stained slide patches, with regressors or classifiers trained on top of mean-pooled embeddings.
Biomarker classification (measured by AUROC) and survival analysis (measured by C-index)
A digital pathology foundation model trained on H&E slides can accurately predict biomarkers and survival across multiple cancer types, including breast cancer, without requiring clinical data.
Abstract Background: The application of machine learning methods to oncology has historically been challenging due to the high resolution of medical imaging modalities and scarcity of downstream supervised data. Over the past years, however, foundation models have enabled the field to bypass these constraints by leveraging self-supervised learning on large quantities of unsupervised imaging data. These foundation models learn effective latent representations for the various morphologies seen throughout the training data. These representations can then be used downstream for supervised machine learning tasks with minimal additional training. Methods: A digital pathology foundation model was trained using self-supervised learning method DINOv2 on 250M patches extracted from 260k whole slide images (WSIs). We evaluated the model on biomarker classification and survival analysis tasks across a total of 17 cohorts covering 12 cancer types. For all evaluations, we kept the foundation model frozen and trained regressors or classifiers on top of mean-pooled embeddings obtained by passing hematoxylin and eosin (H STAD: 0.68±0.06; UCEC: 0.67±0.07), and microsatellite instability (MSI) in TCGA-STAD (0.69±0.06). See Table 1 for results for all tasks and cohorts. Conclusions: Pathology foundation models are applicable to a wide variety of tasks across a range of cancer subtypes, even in spite of sparse data and naive model architectures. With the right data, similar methodology could be used to train predictors of recurrence risk, metastasis risk, treatment benefit, and more. Citation Format: J. Cappadona, J. Witowski, K. Zeng, J. Park, B. Machura, K. Geras. Pan-cancer ai foundation models yield accurate biomarker and survival predictions in breast cancer abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-10.
Building similarity graph...
Analyzing shared references across papers
Loading...
Cappadona et al. (Tue,) reported a other. Foundation AI pathology models predicted breast cancer TP53 mutations (AUROC 0.70), CDH1 mutations (0.71), and breast cancer-specific survival (C-index 0.67) accurately.
www.synapsesocial.com/papers/699a9dcd482488d673cd3f2a — DOI: https://doi.org/10.1158/1557-3265.sabcs25-ps3-06-10
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:
J. Cappadona
J. Witowski
K. Zeng
Clinical Cancer Research
Aims Community College
Building similarity graph...
Analyzing shared references across papers
Loading...