Foundation models for cancer: advancing diagnosis, prognosis, and treatment response-an NCI workshop report.
Authors
Affiliations (5)
Affiliations (5)
- Division of Cancer Treatment and Diagnosis, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
- Division of Cancer Biology, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
- Office of Data Sharing, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
- Office of the Director, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
- Division of Cancer Control and Population Sciences, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
Abstract
From March 24-26, 2026, the National Cancer Institute convened the Foundation Models for Cancer Workshop, bringing together more than thirty speakers from academia, government, and industry and attracting more than 2,000 registrants worldwide. The workshop examined the current state and future potential of foundation models and related artificial intelligence (AI) technologies in oncology. Presentations spanned single-cell biology, computational pathology, medical imaging, electronic health records, drug discovery, federated learning, and clinical decision support. Several themes emerged across the three-day meeting. First, the development of robust foundation models depends on access to high-quality, multimodal cancer datasets that are representative of the patient populations, disease stages, and clinical settings encountered in routine care. Second, prospective validation and evaluation in real-world clinical environments are essential for establishing the reliability, generalizability, and utility of AI-enabled tools. Third, while technical advances continue at a rapid pace, successful translation into cancer research and patient care will also require attention to implementation, workflow integration, interpretability, and data governance. Collectively, the workshop highlighted both the opportunities and challenges associated with applying foundation models to cancer research and clinical care and provided a forum for discussing future directions for the field.