YOUR GO-TO INTELLIGENCE RESOURCE FOR THE SCHOLARLY PUBLISHING INDUSTRY

The Radiological Society of North America (RSNA) has expanded the Annotated Library of AI Systems (ATLAS), which now includes more than 230 model cards and dataset cards across 31 subspecialties, giving research communities the tools to make AI research more transparent, understandable and trustworthy.

Launched in November 2025, ATLAS is a platform designed to enable researchers, developers, and healthcare professionals to share and access structured information about AI tools and datasets used in medical imaging. The platform uses standardized “AI index cards” to describe models and datasets, allowing users to compare resources and assess their applicability more effectively.

ATLAS supports regulatory compliance requirements, including those established by the U.S. Food and Drug Administration, by providing consistent documentation and metadata for AI systems. Users can search and retrieve model cards through a web-based interface or integrate access through an API.

The platform includes tools such as the ATLAS Card Creator, which provides templates for submitting structured entries, and an AI extractor tool that pre-fills cards by extracting information from existing documents. Submissions undergo validation processes, including checks against JSON schemas and verification of live URLs, before publication.

Additional features include ontology-driven indexing using RadLex and RSNA content codes, and the assignment of Digital Object Identifiers (DOIs) to published cards to ensure traceability and citation.

The ATLAS framework is supported by the Radiology Ontology of AI Datasets, Models and Projects (ROADMAP), which provides standardized terminology for describing AI resources. The ontology and data schema are maintained by a panel of imaging AI experts and incorporate widely used vocabularies such as SNOMED and RadLex to improve interoperability and consistency.

The platform is intended to support discoverability, evaluation, and reuse of AI models and datasets across the global medical imaging and radiology community.