Leiden University Medical Center (LUMC) in the Netherlands is starting to use a new AI tool called PERISCOPE. This tool helps doctors know which patients might get an infection after surgery—either within 7 days or within 30 days.
The AI was developed and tested by PhD researcher Siri van der Meijden. She also studied how to use the tool in real hospital settings and defended her research on May 6.
The idea behind PERISCOPE came from a need to better predict infections after surgery. Right now, about 5% to 20% of patients get infections after an operation. These include wound infections, lung infections, urinary infections, and sometimes serious bloodstream infections (sepsis). These infections can make patients stay in the hospital longer, require more treatments, or even be readmitted. PERISCOPE helps doctors find high-risk patients early so they can give them extra care.
Van der Meijden and her team used 10 years of hospital records to train the AI. This data was made anonymous and included patient history, other illnesses like diabetes, and health signs like heart rate and blood pressure. The AI learned to match this data with who did or didn’t get an infection after surgery.
The tool was tested using over 250,000 past surgeries from three hospitals: LUMC, Radboudumc, and a hospital in Genk, Belgium. At LUMC, they compared PERISCOPE’s predictions with those made by doctors. The AI performed just as well as experienced doctors—and even better when doctors were unsure or had less experience.
Doctors, nurses, and surgeons in areas like general surgery, orthopedics, and brain surgery will use PERISCOPE. They’ll see a risk percentage on their screen, along with a risk level (low, medium, or high), and other useful patient info. This makes it faster and easier to decide if a patient should stay in the hospital or come back for a check-up.
PERISCOPE doesn’t replace doctors—it helps them make better decisions. Doctors’ own judgment and medical guidelines will always come first.
Even though the tool is ready, it won’t be used in real-life care until mid-2026. Right now, the team is working to connect the tool with the hospital’s digital patient system. It took 5 years to build PERISCOPE, partly because of all the safety checks required for medical tools.
Van der Meijden also had to work with many departments—like orthopedics and infection specialists—to decide how infections are defined and what data to include.
She is still improving PERISCOPE. Right now, the tool needs to be updated manually with new data to get smarter. In the future, they hope the tool can predict the type of infection and even risks before surgery, not just after. It may also predict other problems like bleeding or readmission.
Only time will show how helpful PERISCOPE will be in the long run—but early signs are promising.
Source: healthcare-in-europe
