AI system tested in Belgian breast clinics to personalize cancer treatment

Four breast clinics in Belgium's Flanders region are testing an artificial intelligence system designed to help doctors predict which treatment may be most suitable for individual breast cancer patients, Dutch-language public broadcaster VRT reported on Friday.

Developed by RADar, the learning and innovation center of AZ Delta, the system seeks to provide doctors with additional information at the time of diagnosis, potentially enabling more targeted treatment and helping avoid unnecessary surgical procedures.

The AI system attempts to predict three key factors from the outset of a patient's treatment: how well the patient is likely to respond to chemotherapy, whether the cancer has spread to the lymph nodes in the armpit, and the size of the tumor.

"These are things we usually only find out later in the care trajectory today," Barbara Bussels, coordinator of the breast clinic at AZ Delta, said.

"If we already have that information at the time of diagnosis, it can help to better assess which treatment is most suitable for which patient," she added.

According to Philip Poortmans, a radiation oncologist at ZAS clinic, the system can identify patterns that may be difficult for doctors to detect by examining individual sources of medical information.

"Through this approach, artificial intelligence can recognize patterns that are difficult or even impossible to detect with the human eye," Poortmans said.

One potential application is predicting the likelihood of cancer spreading to the axillary lymph nodes.

More accurate predictions could allow surgeons to tailor procedures involving the armpit more precisely to individual patients and, in some cases, potentially avoid surgery altogether.

This could reduce the risk of complications including lymphedema, pain and restricted mobility.

The AI system is not yet being used to make treatment decisions for current patients.

Since early June, cancer specialists at the four participating hospitals have been independently evaluating the system using historical patient files to determine whether it can improve and personalize treatment recommendations.

If the results are positive, the most promising AI models will move to a subsequent phase involving testing in routine clinical practice.

Bussels added that once the system receives European certification, the AI tool could potentially be deployed in other breast clinics across Europe.