Novel approach to safeguard patient data included among NSF-led National AI Research Resource Pilot

The U.S. National Science Foundation and the Department of Energy recently announced that a team comprised of Vanderbilt’s newly created ADVANCE center and VALIANT lab is among the first round of 35 projects that will be supported with computational time through the National Artificial Intelligence Research Resource (NAIRR) Pilot. 

The NAIRR Pilot — the result of a recent Executive Order on the Safe, Secure and Trustworthy Development and Use of AI — will provide AI researchers and students access to key AI resources and data.

Projects granted computing allocations in this initial round encompass a diverse range of AI-related areas, including investigations into language model safety and security, privacy and federated models, and privacy-preserving synthetic data generation.

The Vanderbilt team, which also includes Oak Ridge National Laboratory (ORNL), has proposed an innovative secure and trustworthy AI approach to safeguard healthcare data from unauthorized use in AI model training through the development of unlearnable examples (UE). By embedding invisible noise in over 2 million radiological images from MRI and CT scans, the project will generate label-agnostic UE models that protect medical imaging data against misuse in future AI training, without compromising their clinical utility.

“This initiative will enhance the collaboration between the Vanderbilt Lab for Immersive AI Translation (VALIANT) and AI Discovery and Vigilance to Accelerate Innovation and Clinical Excellence (ADVANCE), further strengthening our ties with Oak Ridge National Lab,” said  Yuankai Huo, assistant professor of computer science at Vanderbilt and a team member.

“We are thankful to the NSF and DOE for the NAIRR pilot, which provides researchers and students like us with key AI resources and data. This access is crucial in unlocking AI’s full potential to improve humanity and society.”

Bennett Landman, director of VALIANT and chair of the Department of Electrical and Computer Engineering at Vanderbilt, said the project highlights the radical collaboration the university continues to foster.

“This project brings together unique expertise across Vanderbilt University, Vanderbilt University Medical Center (VUMC), and Oak Ridge National Lab,” said Landman, a preeminent scholar who holds the Stevenson Chair of Electrical and Computer Engineering and has joint appointments in computer science, biomedical engineering, radiology and radiological sciences, psychiatry and behavioral sciences, biomedical informatics, and neurology.

“I am excited about our potential to explore previously inaccessible challenges intersecting privacy, security, and medical imaging.”

ADVANCE co-director Bradley Malin said he’s pleased to see Vanderbilt’s collaborative approach to AI being recognized.

“This project will enable us to showcase that socially responsible AI can be developed and achieved at scale,” said Malin, VUMC’s Accenture Professor of Biomedical Informatics, Biostatistics and Computer Science and vice chair for Research Affairs in DBMI.

In tandem with the announcement of initial awards, the NAIRR Pilot opened the next opportunity for researchers and educators to apply for access to resources that support AI research, including advanced computing systems; cloud computing platforms; access to foundation models, software and privacy enhancing technology tools, collaborations to train models; and education platforms.

 

Contact: Lucas Johnson, lucas.l.johnson@vanderbilt.edu