AI in the ICU? University of Florida researchers develop model to promote better patient outcomes


Ethan Kispert


Improved patient care and better patient outcomes are the goals of one team at the University of Florida College of Medicine as they analyze an artificial intelligence (AI)-driven solution to monitoring critical care patients. 


Stay one step ahead. Join our email list for the latest news.



UF Health researchers are currently working on developing an “intelligent ICU” where patients are monitored with sensors that record movements of their head and other body parts. Following the study, researchers hope physicians will have the ability to more effectively monitor patient’s vital signs and recommend a proper course of treatment. 

According to UFHealth, in this “intensive care unit of the future” nurses will continue to come and go as needed to monitor the patient’s condition and to check on their physiological signs. The AI systems will continue to run in the background as it collects data from various cameras, wearable sensors, light sensors, noise meters, and other equipment in the room. 

Dr. Parisa Rashidi, an associate professor at the J. Crayton Pruitt Family Department of Biomedical Engineering at UF, emphasized the ongoing collection of this data and how it would help providers better assess the acuity of patients in ICUs. 

Dr. Rashidi specifically highlighted the system’s ability to monitor patients’ vital signs and its potential to improve recovery outcomes — even if this means monitoring the slightest change such as the twitching of an arm or the movement of a leg.

“We know that mobility of the patient is really important in the critical care unit. If the patients are able to take a couple of sips or if they are able to sit upright in their bed for somebody who is critically ill, that means that they are actually taking one step towards recovery.”

She points to the system’s ability to provide patient data which could help assess how the patient will recover long term. 

“You also know what the trajectory of the patient is, what type of intervention we can introduce in the environment, or what kind of recommendation we can make to the physicians, the nurses, [and] the caregiver, so we can help the patient in terms of their health outcomes.”

Implementing an AI-based system for monitoring patients is necessary, according to Dr. Rashidi, because the ICU patient’s status can fluctuate rapidly, oftentimes too quickly for nurses to take note. The system can then process this data into an electronic health records (EHR) system for health providers to access later on. 

Initial tests showed that the model has some disadvantages.

Dr. Rashidi pointed to the sensor pads used for monitoring patients’ vital signs. 

“In our studies, we have found that depending on where you place wearable sensors on a patient’s body, you can get different measurements.”

The system is collecting patient data constantly and disturbances, such as nurses taking the pads off or forgetting to reapply them, creates inconsistencies in the EHR reports.

The study, which is funded by a $2.4 million National Institutes of Health grant, is part of two projects focused on intelligent data collection for better patient outcomes. According to Dr. Rashidi, UFHealth has recruited roughly 170 patients for analysis and plans to bump that number to 400 in the future.