An international team of researchers has developed an AI tool that can predict which breast cancer patients may be at risk of side-effects after surgery and radiotherapy.

The technology, developed by a research team led by the University of Leicester, will be tested later this year in a clinical trial in the UK, France and Netherlands and is hoped to help patients access more personalised care.

Two million women globally are diagnosed with breast cancer every year, which is the most common cancer in females in most countries.

Greater awareness, earlier detection and a wider range of treatment options have improved survival rates, but many patients experience side-effects after treatment.

Some of the factors that increase the risk of side-effects are already know, but the research project aims to make more accurate predictions for each individual patient, as well as providing understandable explanations for doctors and patients.

“The explainable AI tool shows the reasoning behind its decision-making so it’s easier not only for doctors to make decisions, but also to provide data-backed explanations to their patients,” explained Dr Tim Rattay, a consultant breast surgeon and associate professor at the University’s Leicester Cancer Research Centre.

“Thankfully, long-term survival rates from breast cancer continue to increase, but for some patients, this means having to live with the side-effects of their treatment, including skin changes, scarring, lymphoedema, which is a painful swelling of the arm, and even heart damage from radiation treatment.

“That’s why we’ve developed an AI tool to inform doctors and patients about the risk of chronic arm swelling after surgery and radiotherapy for breast cancer. We hope this will assist doctors and patients in choosing options for radiation treatment and reduce side effects for all patients.”

The researchers used information from European datasets on 6,361 breast cancer patients to train different machine learning algorithms to predict arm swelling up to three years after surgery and radiotherapy.

The AI tool correctly predicted lymphoedema in an average of 81.6 per cent of cases and correctly identified patients who would not develop it in an average of 72.9 per cent of cases. The overall predictive accuracy of the model was 73.4 per cent.

Dr Rattay said patients identified at higher risk of arm swelling could be offered additional supportive measures, such as wearing an arm compression sleeve during treatment, which has been shown to reduce arm swelling in the long-term.

“Clinicians may also use this information to discuss options for lymph node irradiation in patients, where its benefit may be fairly borderline,” he said.

“We will test the effect of the prediction model on clinician and patient behaviour and use of the prophylactic arm sleeve in the proposed clinical trial.”

The team, Rattay said, will incorporate the current AI model into software that can provide evaluations and predictions to doctors and patients. This will be tested when the clinical trial starts later this year.

They hope to enrol 780 patients as part of a clinical trial, called the Pre-Act project, who will be followed up for a period of two years. They are also developing the tool further so that it can predict other side effects, such as skin and heart damage.

Dr Guido Bologna, associate professor at the University of Applied Sciences and Arts of Western Switzerland in Geneva, and co-investigator on the project, said: “The final, best-performing model makes predictions using 32 different patient and treatment features, including whether or not patients had chemotherapy, whether sentinel lymph node biopsy under the armpit was carried out, and the type of radiotherapy given.”