ACR AI-LAB™ Demonstrations

Working with the ACR AI-LAB™ platform, seven institutions will demonstrate the process of using transfer learning to share an Artificial Intelligence (AI) model between sites, permitting an AI model developed at one location to be shared and tuned at other locations. This pilot represents a major milestone in AI-LAB’s effort to allow institutions to develop high-quality algorithms that address clinical needs at local levels.

Distributed solutions for local AI model training and testing — which do not require patient information to leave an institution — have not yet been used successfully in radiology at this scale. The group of pilot sites will use ACR AI-LAB™ to apply transfer learning directly within the platform. An institution can download an AI model from other institutions to test, evaluate and improve the model at their site.

For the pilot, a consortium of seven health care institutions was formed including Massachusetts General Hospital, Lahey Hospital and Medical Center, Emory University, The University of Washington, the University of California San Francisco, Brigham and Women’s Hospital, and Ohio State University. NVIDIA and Nuance will support the project by contributing key technology infrastructure necessary to enable the end-to-end workflow.