The AIR·MS data platform for artificial intelligence in healthcare

Area of Science:

• Clinical Informatics
• Artificial Intelligence in Healthcare
• Data Science

Background:

• Healthcare systems generate vast amounts of diverse clinical data.
• Integrating disparate data sources and computational infrastructure is crucial for advanced AI research.
• Existing platforms often lack the scale and flexibility required for complex AI model development.

Purpose of the Study:

• To introduce the Artificial Intelligence-Ready Mount Sinai (AIR·MS) platform.
• To showcase the integration of diverse clinical datasets and computational infrastructure.
• To demonstrate the platform's utility through three AI research projects.

Main Methods:

• AIR·MS integrates structured EHR data and unstructured pathology/radiology data using the OMOP Common Data Model.
• Data is stored in an in-memory columnar database, with metadata linking raw source data.
• HIPAA-compliant cloud and on-premises HPC environments support data access and analytics.

Main Results:

• The platform provides access to over 12 million patient records, including EHRs, clinical notes, and imaging metadata.
• AIR·MS supports interactive cohort building and AI model training with high system performance.
• Three use cases demonstrated risk-factor discovery and federated cardiovascular risk modeling.

Conclusions:

• AIR·MS successfully integrates clinical data and infrastructure for large-scale AI research.
• The platform's scalability, security, and collaborative design serve as a model for future initiatives.
• AIR·MS enables AI-driven healthcare research on multimodal clinical data.