But it’s difficult to assess how much sedative to use on patients – with too much posing added health risks and too little meaning they wake up and are then at risk of hurting themselves by accidentally removing medical devices after an operation.
Dr Sato said the “proof of concept” AI model was created using 24 postoperative patients, aged 67 on average, who were admitted to his hospital’s ICU between June and October 2018.
A ceiling-mounted camera above patients’ heads was used to analyse 300 hours of data when patients showed their faces and eyes clearly.
A total of 99 images were fed into the AI’s machine-learning algorithm to create alerts when a patients’ facial expression suggested they were about to move their arms in a risky way.
The research has been presented at this year’s Euroanaesthesia congress in Vienna, Austria and not yet submitted to a medical journal.
Dr Sato said: “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 healthcare professionals of risky behaviour.
“Our end goal is to combine various sensing data such as vital signs with our images to develop a fully automated risk prediction system.
The researchers said the AI needed to learn using more images of patients in different positions to make it useful in real life.
Dr Sato added that monitoring of a patient’s consciousness could improve the system’s ability to tell between between high-risk behaviour and voluntary movements.
Source: Read Full Article