AI Research Catalyst Fund Awardees Virtual Seminar Series

“Learning Optimal Treatment Strategies for Hypotension During Surgery Using Deep Reinforcement Learning“

by Dr. Tezcan Ozrazgat-Baslanti, Research Assistant Professor, Division of Nephrology, Hypertension and Renal Transplantation, College of Medicine 
and Tianqi Liu, Ph.D. Candidate in Department of Electrical & Computer Engineering

Wednesday, May 11, 2022 Virtual via Zoom

ABSTRACT:

In the United States, where the average American can expect to undergo seven surgical operations during a lifetime, each year 1.5 million patients develop a medical complication. Decision making during surgery has a critical impact on postoperative condition of the patient, as well as the need of healthcare resource allocation, cost of treatment, and duration of postoperative revitalization. An optimized system with potential to generate suggestion based on patient state can be a significant asset in surgical situations. Current practices in the administration of intravenous fluids and vasopressors during surgery as a treatment of hypotension are suboptimal. We developed a Reinforcement Learning model using deep Q-learning and off-policy model free approach to recommend optimum dose of intravenous fluid and vasopressors during surgery to avoid intraoperative hypotension and postoperative acute kidney injury. We divided each surgery duration into 15 minute resampled timestamps, and used demographic variables and intraoperative physiologic time series to identify patient states. Total dose of intravenous fluid and vasopressors were used to determine action space. Our model was successful in separating actions related to higher prevalence of postoperative acute kidney injury and proposed actions to reduce it. The developed model was successful in mimicking physician behavior and showed that higher return was associated with decreased risk of postoperative acute kidney injury.

Bio:

Tezcan Ozrazgat-Baslanti received her PhD degree in Statistics from the University of Florida in 2011. After a four year post doctorate and one year as a Research Assistant Professor in the Department of Anesthesiology, University of Florida College of Medicine, she was recruited as a Research Assistant Professor in the Division of Nephrology, Hypertension and Renal Transplantation, University of Florida College of Medicine in 2016. She has been part of the Precision and Intelligence in Medicine Partnership (PRISMA^P) research group and the University of Florida Sepsis and Critical Illness Research Center (SCIRC) since 2015 and she currently serves as the Associate Director for AI Research in the multidisciplinary Intelligent Critical Care Center (IC3) at UF; a multi-disciplinary center focused on developing and providing sustainable support and leadership for transformative medical AI research, education, and clinical applications to advance patients’ health in critical and acute care medicine. She has been an investigator on many university and federal grants and have been part of multiple grants and various projects on which she served as a statistician, coordinator, and supervisor. Her research focuses on artificial intelligence in nephrology and critical care including phenotyping and prediction models related to kidney health and other hospital complications utilizing large electronic health records data.

Tianqi Liu is a Ph.D. student in the Department of Electrical and Computer Engineering at the University of Florida. He received his B.E. degree in electronic and information engineering from University of Electronic Science and Technology of China, China, and M.S. degree in electrical engineering from Northwestern University, USA. His research interests include reinforcement learning, video coding, and machine learning.

Wednesday, May 11, 2022
12:00PM – 1:00PM
Via Zoom

RSVP below. Instructions to attend talk via Zoom will be emailed to you.