TL;DR

The Institute of Cancer Research (ICR) in London announced on 20 October that Breast Cancer Now is funding two of its projects with £800,000 between them. Both will use AI to investigate why certain breast cancers that are oestrogen receptor (ER) positive come back many years after treatment. The aim is to predict which patients face a higher risk of relapse, so follow-up care can be tailored to them.

The problem

ER-positive disease accounts for around 80% of UK breast cancer diagnoses, roughly 44,000 women a year, according to the ICR. Most of these cancers do not return. In some people, though, cancer cells survive treatment and sit inactive in the body, sometimes for decades, before starting to grow again. When that happens, the diagnosis is often metastatic breast cancer that can no longer be cured.

The two projects

Professor Luca Magnani is working with Dr Maria Secrier of University College London on how ER-positive cells slip into dormancy and stay there. The team is training AI on large volumes of data gathered from cancer cells during treatment, so it can pick up the rules cells follow as they go quiet and as they reactivate. They are also studying how dormant cells interact with the tissue, blood vessels and immune cells around them.

Dr Maggie Cheang’s team will draw on tumour samples from POETIC, a clinical trial that tracked patients with ER-positive breast cancer for over a decade after surgery. Samples were taken before treatment, during therapy and from cancers that later recurred. Combining imaging and molecular analysis with AI, the team hopes to find signs, as early as two weeks after diagnosis, that a cancer is resisting treatment or likely to come back.

“Late relapse can feel unpredictable and frightening. We want to change that,” Cheang said.

Dr Simon Vincent, Breast Cancer Now’s chief scientific officer, said the projects “could help develop tools to identify who is most at risk of recurrence, helping doctors to tailor follow-up care”.

Looking forward

Much of the UK’s breast cancer AI coverage, including Google’s NHS screening results, concerns finding cancer earlier. In our view, this funding targets a different and harder question: what happens to the cells that survive. The POETIC samples, with more than ten years of follow-up, are the kind of long-run dataset AI needs for that work. Any clinical tool is some way off: the ICR frames practical risk estimates for doctors as something that could follow in the future.