Four-Year Data Generation Project Aims to Use AI for Critical Care

After receiving a large grant from the National Institutes of Health's Bridge to Artificial Intelligence (AI), also known as Bridge2AI, researchers from the University of Florida (UF) aim to test AI use in critical care practices to assist patients recovering from life-threatening illnesses.

In recent years, the use of AI in healthcare has been a growing practice. Using  $3.6 million of a $23.5 million multicenter grant, UF researchers aim to develop an AI infrastructure in critical care to assist the recovery process.

“This project is a huge win for UF AI research and will put us on the map for biomedical AI,” said Azra Bihorac, MD, the senior associate dean for research affairs at the UF College of Medicine and co-director of UF’s Intelligent Critical Care Center, or IC3, in a press release. “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.”

To build the AI infrastructure, researchers will create a repository of data for AI research from more than 100,000 critically ill patients. A network of health systems will supply the data, including UPMC, Massachusetts General Hospital, and Mayo Clinic.

In addition to the 100,000-patient data set, the project will include AI workforce training events, standards for ethical AI use, AI tutorials that are publicly available, and guidelines for collaborative AI approaches.

Known as “A Patient-Focused Collaborative Hospital Repository Uniting Standards for Equitable AI,” or CHoRUS, the project also aims to generate and expand biomedical data for monitoring patients with critical illnesses.

Researchers also noted that previous efforts were often limited in terms of diversity. CHoRUS, however, aims to widen the range of AI-ready data sets to include more geographically and demographically diverse medical 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 Parisa Rashidi, PhD, the J. Crayton Pruitt Family term fellow and an associate professor at the UF Herbert Wertheim College of Engineering, who co-directs the IC3, in a press release.

Further, UF plans to collaborate with AI computing company NVIDIA to assist industries aiming to create AI algorithms for medical practice, alongside highlighting the CHoRUS data set.

Researchers also intend to have a team of legal, ethics, and communication scholars to ensure privacy surrounding the data that is used for AI.

This is the latest effort to use AI to enhance critical care. Last May, researchers from the University of Washington developed an AI framework to reduce the time, effort, and resources needed to predict patient outcomes in critical care settings.

The framework, known as the Cost-Aware Artificial Intelligence (CoAI) model, lowered the cost of data acquisition by about 90 percent for predictions related to trauma response for patients on their way to the hospital and the in-hospital mortality risk of critical care patients in the intensive care unit.

More recently, Johns Hopkins University researchers developed machine-learning (ML) algorithms to detect the early warning signs of delirium and predict which patients will be at high risk of delirium at any point while receiving care in the ICU.

They developed two models, which accurately identified delirium-prone patients 78.5 percent and 90 percent of the time, respectively.