Despite these attempts, drug errors still occur with alarming regularity.
“I had read certain studies that said essentially 90% of anesthesiologists Admit to have a medication error at some point in their career, “said Dr. Kelly Michaelsen, colleague from Wiederspan to Uw Medicine and assistant professor of anesthesiology and pain at Washington University. She started wondering if emerging technologies could help.
As a healthcare professional and trained engineer, he struck him to identify an error about to be made and alert anesthesiologists in real time, should be in AI capabilities. “I said to myself:” It seems to be something that shouldn’t be too difficult to do for AI “,” she said. “Ninety-nine percent of the drugs we use are these same 10 to 20 drugs, and therefore my idea was that we could train an AI to recognize them and act as a second set of eyes.”
The study
Michaelsen focused on the vial exchange errors, which explain about 20% of all drug errors.
All injectable drugs are available in labeled bottles, which are then transferred to a labeled syringe on a drug truck in the operating room. But in some cases, someone selects the bad bottle, or the syringe is poorly labeled and the patient is injected with the bad drug.
In a particularly notorious bottle exchange error, a 75 -year -old woman treated at the Vanderbilt University Medical Center in Tennessee was injected with a deadly dose of paralyzing drug Vecuronium instead of the sedative payment, resulting in her death and a high profile criminal trial.
Michaelsen thought that such tragedies could be prevented by “smart glasses” – adding a portable camera powered by AI to the protective glasses carried by all the staff during operations. By working with her colleagues from the IT department of the University of Washington, she designed a system that can scan the immediate environment for syringe and bottle labels, read and detect them if they correspond.
“He zooms in on the label and detects, let’s say, propofol inside the syringe, but wave inside the bottle, and therefore produces a warning,” she said. “Or the two labels are the same, so everything is fine, move on.”
The construction of the aircraft has taken Michaelsen and his team for more than three years, half of which has been devoted to approval to use pre -recorded video flows of anesthesists preparing the drugs properly inside the operating room. Once given the green light, it was able to form AI on this data, as well as additional images – this time in the laboratory – in progress errors.
“There are a lot of problems with the fatigue of the alarm in the operating room, so we had to make sure that it works very well, it can do an almost perfect job to detect errors, and therefore [if used for real] This would not give false alarms, “she said.” For obvious ethical reasons, we could not make express mistakes with the patients involved, so we did it in a simulated operating room. “”
In a study Published at the end of last year, Michaelsen said that the device detected of bottle exchange errors with 99.6%accuracy. Everything is left is to decide in the best way to warn the messages to relay and it could be ready for real use, awaiting administration of food and drugs. The study was not funded by AI technology companies.
