RSNA Expands ATLAS AI Data Hub
Growing catalog broadens what users can search, compare and cite
RSNA has announced that its Annotated Library of AI Systems (ATLAS) now houses more than 230 model cards and dataset cards in 31 subspecialties, giving research communities the tools to make AI research more transparent, understandable and trustworthy.
In November 2025, RSNA launched ATLAS, a tool that makes it easier to share and find information about AI tools and datasets used in medical imaging.
ATLAS is a free website where AI researchers, developers and health care providers can share detailed descriptions, called “AI index cards,” about their imaging AI models and data.
These index cards are created using a simple and consistent format, which helps users understand and compare different AI tools more easily. Users can search for and compare AI resources, and the system supports meeting regulatory requirements, including those set by the U.S. Food and Drug Administration.
Tools and Guardrails for Sharing AI Resources at Scale
“With ATLAS, radiologists, researchers and developers can easily find AI models and datasets that meet their needs,” said Charles E. Kahn Jr., MD, MS, member of the RSNA AI Committee and the Radiology AI Data Standards Subcommittee, and editor of Radiology: Artificial Intelligence. “The site helps make AI resources more findable and interoperable.”
The platform also offers tools to help create AI index cards, making it simple to share information. The ATLAS Card Creator provides a template for building ATLAS cards. It also features an AI extractor tool that can pull information from existing documents to pre-fill a card for submission.
The ATLAS website and repository validates submissions before making them public:
- Model cards are checked against a JSON schema and verified for live URLs.
- Searchable Interface and API: Users can browse and retrieve model cards via a user-friendly web interface or integrate with their systems using an API.
- Ontology-Driven Indexing: Cards are tagged with RadLex and RSNA content codes for precise categorization.
- Digital Object Identifier (DOI) Assignment: Each published card receives a DOI, ensuring traceability and citation.
The Radiology Ontology of AI Datasets, Models and Projects (ROADMAP) provides a controlled terminology for the metadata describing AI models and datasets. The ATLAS data schema and ROADMAP ontology are maintained by a panel of imaging AI experts.
“ATLAS uses widely used vocabularies, such as SNOMED and RadLex, to index its content,” Dr. Kahn said. “The site incorporates language from ROADMAP, a newly released set of descriptors from RSNA’s Radiology AI Data Standards Committee, to help standardize information about AI resources.”
The imaging AI community is encouraged to publish ATLAS cards for AI models and datasets they wish to share. Submitting a card helps ensure the work is discoverable by the global medical imaging and radiology community, easy to understand and evaluate, and well positioned for broader collaboration and real‑world use.
For More Information
Explore additional RSNA AI resources, including datasets and tools for model development.
Discover peer-reviewed research on imaging AI in Radiology: Artificial Intelligence.
See how RSNA fosters innovation and benchmarking through AI Challenges.
Read previous RSNA News stories on data standards and interoperability: