UF to use $23.5 million grant to build AI infrastructure to improve critical care

By Cody Hawley

The University of Florida has been awarded $3.6 million of a $23.5 million multicenter grant for a four-year data-generation project that is unprecedented in its scope, aimed at building an infrastructure for artificial intelligence in critical care and advancing artificial intelligence in ways that improve patients’ ability to recover from life-threatening illnesses.

Funded by the National Institutes of Health’s Bridge to Artificial Intelligence, or Bridge2AI program, this project creates a network of university health systems that will support a comprehensive repository of data for AI research from more than 100,000 critically ill patients. The patients’ data will be made anonymous.

Although the project’s highlight will be the 100,000-patient data set, key aspects of the project include AI workforce training events, a set of standards for ethical use of AI in critical care, publicly available AI tutorials and guidelines for a collaborative approach to medical AI research.

A team of eight principal investigators, including three from UF — Azra Bihorac, M.D., M.S., FCCM, FASNParisa Rashidi, Ph.D.; and Yulia Levites Strekalova, Ph.D., M.B.A. — will lead the network of connected intensive care units. UF Health will be a vital contributor to the data repository, along with other major health systems, including Massachusetts General Hospital at Harvard; Emory University; Duke University; the University of California, Los Angeles; Nationwide Children’s Hospital; Columbia University and the Mayo Clinic.

“This project is a huge win for UF AI research and will put us on the map for biomedical AI,” said Bihorac, the senior associate dean for research affairs at the College of Medicine and co-director of UF’s Intelligent Critical Care Center, or IC3. “The success of our UF team builds on the investment of UF Health and the UF College of Medicine in the digitization of clinical infrastructure and the generation, integration and standardization of medical data for both clinical and research use.”

The program, called “A Patient-Focused Collaborative Hospital Repository Uniting Standards for Equitable AI,” or CHoRUS, will expand and generate biomedical data that can be used for monitoring, diagnosing and treating critically ill patients, as well as augmenting doctor’s rapid decision-making.

Previous data-generation efforts have lacked geographical and demographic diversity. CHoRUS will greatly increase the scope and scale of AI-ready data sets by creating the repository of anonymized data, composed of structured electronic health records and other biomedical information.

“The CHoRUS data set will be the largest and most comprehensive critical care data set. Providing this data set to the scientific community will accelerate advances in the development of AI algorithms in the critical care domain and in the medical AI domain in general,” said Rashidi, the J. Crayton Pruitt Family term fellow and an associate professor at the UF Herbert Wertheim College of Engineering, who co-directs the IC3.

Earlier large-scale data sets used by ICUs have been insufficient for general use, Bihorac said.

“So far, the large, high-resolution data sets needed to utilize AI technology in ICUs have been limited to single sites,” Bihorac said. “CHoRUS will lead to more research advances and more ICUs being able to utilize AI advances and give critically ill patients the best possible care. UF’s involvement will yield benefits for patient care that will extend beyond the duration of our project.”

Other Bridge2AI components — involving UF faculty from the colleges of medicine, communication, pharmacy, law and engineering — include expanding access to AI knowledge and resources by involving the Gainesville community in AI training through UF’s Citizen Scientist program and a fellowship initiative for local high school teachers.

The Citizen Scientist program, which enables researchers to receive feedback from the community about their work, will offer a publicly available educational module about medical AI ethics, and the UF Center for Precollegiate Training will offer summer fellowships for two local high school math teachers who will develop an AI and programming curriculum for high school students that can be shared with teachers across the country.

“By design, this project brings together experts from multiple disciplines,” said Levites Strekalova, an assistant professor in the College of Public Health and Health Professions department of health services research, management and policy. “We’re creating skills and workforce development programs for medical professionals, K-12 teachers and citizen scientists. Our efforts will have a national impact, which is both humbling and exciting.”

Furthermore, UF will build an industry innovation collaborative with NVIDIA and others to help industries seeking to develop AI algorithms for medical practice and to raise awareness about CHoRUS data. NVIDIA provides the technological power behind HiPerGator, one of the nation’s fastest supercomputers, and partners with UF to advance the university as a national leader in the application of AI.

“By leveraging our open-source technology stacks like MONAI, a framework for developing medical imaging AI, and NVIDIA FLARE, a systems development kit for ensuring security and data privacy, we can quickly equip a new generation of AI builders and accelerate their journey to AI proficiency,” said Mona G. Flores, M.D., the global head of medical AI at NVIDIA.

To fortify the privacy of personal data used in medical AI, a team of legal, ethics and communication scholars will recommend privacy policies beyond the current legal standards to strengthen protections and enhance public trust.

The team will consult with up to 10 CHoRUS citizen scientists to identify and address any ethical questions or weaknesses in the public’s perception of medical AI.

“CHoRUS is building data resources to support a wide array of socially beneficial goals — not just to enhance research but also to improve health care, to make medical products safer and to benefit public health,” said Barbara J. Evans, J.D., Ph.D., the Stephen C. O’Connell Chair and a professor of law and engineering at UF. “Each such use poses distinct legal and ethical issues, so we’re examining ways to give people confidence that their data will be safe in medical AI systems.”

UF will help this multicenter group provide better critical care for all patients through more effective AI monitoring and diagnostics.

“This exciting project will leverage the investments made by the University of Florida to build the AI University, as well as the donation of the HiPerGator AI supercomputer by UF alumnus and NVIDIA co-founder Chris Malakowsky,” said Erik Deumens, Ph.D., UF’s senior director of research computing.

Overall, the CHoRUS program expands upon UF’s campuswide AI initiative and UF Health’s focus on advancing trustworthy AI and machine learning in the clinical realm.

Principal investigators for the UF multidisciplinary team:

  • Azra Bihorac, M.D., M.S., FCCM, FASN, a professor and senior associate dean of Research, College of Medicine
  • Parisa Rashidi, Ph.D., an associate professor of biomedical engineering and J. Crayton Pruitt Family term fellow, Herbert Wertheim College of Engineering
  • Yulia Levites Strekalova, Ph.D., MBA, an assistant professor in the College of Public Health and Health Professions department of health services research, management and policy

Members of the UF multidisciplinary team:

  • Barbara Evans, J.D., Ph.D., Stephen C. O’Connell Chair and a professor, Levin College of Law and professor, Herbert Wertheim College of Engineering
  • Benjamin Shickel, Ph.D., an assistant professor of nephrology, hypertension & renal transplantation, College of Medicine
  • Tezcan Ozrazgat-Baslanti, Ph.D., a research assistant professor, College of Medicine
  • Elizabeth Shenkman, Ph.D., chair of the department of health outcomes and biomedical informatics, College of Medicine; co-director, the UF Clinical and Translational Science Institute; and associate director for community outreach and engagement, UF Health Cancer Center
  • Yi Guo, Ph.D., an associate professor of health outcomes & biomedical informatics, College of Medicine
  • Serena Guo, M.D., Ph.D., an assistant professor of pharmaceutical outcomes & policy, College of Pharmacy

Data acquisition sites of the CHoRUS program:

  • University of Florida
  • Massachusetts General Hospital at Harvard
  • Beth Israel Deaconess Medical Center at Harvard (through Massachusetts Institute of Technology)
  • Columbia University
  • University of Pittsburgh Medical Center
  • Duke University
  • University of Virginia
  • University of California, Los Angeles
  • Emory University
  • Nationwide Children’s Hospital
  • Mayo Clinic
  • Seattle Children’s Hospital
  • University of New Mexico
  • University of California, San Francisco

Additional partners in the CHoRUS program:

  • Johns Hopkins Medical Center
  • Tufts University Medical Center
  • University of Oregon
  • University of Texas Health Science Center, Houston