Tools and Guardrails for Sharing AI Resources at Scale with ATLAS
ATLAS enhances accessibility for radiologists, researchers, and developers to find essential AI models and datasets. Charles E. Kahn Jr., MD, emphasizes the platform’s role in improving the discoverability and interoperability of AI resources. ATLAS features an innovative Card Creator for effortless sharing of information, which includes an AI extractor to auto-fill submissions. Each model card undergoes validation, ensuring accurate metadata through JSON schema checks and live URL verification. Users benefit from a searchable interface and API integration, with ontology-driven indexing using RadLex and RSNA codes for precise categorization. The platform also employs a Digital Object Identifier (DOI) for traceability. The Radiology Ontology of AI Datasets, Models, and Projects (ROADMAP) provides standardized terminology. The imaging AI community is urged to publish ATLAS cards to enhance global discoverability and facilitate collaboration. For more information, explore additional RSNA AI resources, peer-reviewed research, and innovation initiatives.
