Study: AI camera helps doctors avoid medication errors

Together with hospital staff, researchers have tested an AI-supported head camera that is designed to detect medication errors in everyday clinical practice. The scientists from the University of Washington, Carnegie Mellon University and Makerere University used video data from hospitals and operating theaters to train a deep learning algorithm to recognize the labels on syringes and vials. The application checks in real time whether the medication matches the patient. If this is not the case, the system warns the medical staff acoustically or visually of errors.

In a training phase lasting just under two months, head-mounted cameras filmed the hands of doctors and certified hospital staff administering medication during operations. "Our algorithms can detect vial mix-ups in real time using videos that are wirelessly transmitted to a local edge server with a graphics processing unit (GPU)," the researchers write in their study.

Wrong medication in 5 to 10 percent of all cases

According to the research group, incorrectly administered medication and syringes pose the greatest risk to patient safety in hospitals and operating theaters. This affects around one in 20 patients. In the USA, the "error rate in the administration of medication in hospitals is estimated at 5 to 10 percent", the researchers write. They go on to argue that anesthesia staff are already used to wearing protective eyewear and that a move towards smart glasses would only mean a minor change in their daily routine.

A survey of 109 anesthesiologists found that a large majority would be willing to use such a lightweight and certified system if it "was proven to lead to fewer medication errors, make charting easier or if it was required by their employer."

The use of artificial intelligence has long been established in other areas of medicine, as Felix Nensa, Professor of Radiology at Essen University Hospital, explains in an interview.

(hoh)