A scientific team at Yokohama City University Hospital has used facial recognition technology to develop an automated system that can predict with moderate accuracy (75%) when patients in the intensive care unit (ICU) are at high risk of unsafe behavior. The automated risk detection tool has the potential to continuously monitor patients’ safety when limited capacity makes it difficult for staff to continuously observe critically ill patients at the bedside. Dr. Akane Sato, who led the research, said that the researchers used images of a patient’s face and eyes to train computer systems to recognize high-risk arm movement that could lead to behavior such as accidentally removing a breathing tube. A proof-of-concept model was created using pictures taken by a camera mounted on the ceiling above patients’ beds. Around 300 hours of data were analyzed to find daytime images of patients facing the camera in a body position that showed their face and eyes clearly. Ninety-nine images were subject to machine learning. Ultimately, the model was able to alert against high-risk behavior, especially around the subject’s face, with high accuracy. The study included 24 postoperative patients. “Various situations can put patients at risk, so our next step is to include additional high-risk situations in our analysis, and to develop an alert function to warn health care professionals of risky behavior,” Sato said. “Our end goal is to combine various sensing data such as vital signs with our images to develop a fully automated risk prediction system.” Although more images of patients in different positions are needed, the facial recognition tool is the first step toward a smart ICU that is being planned for the hospital. The research was presented at Euroanaesthesia 2019, June 1-3, 2019, in Vienna.
