It’s been a hot topic at every breast imaging conference for the last few years, but 2026 has seen some of the strongest evidence yet for its applications within UK breast screening practice. The Royal College of Radiologists’ fifth edition of its breast imaging guidance (October 2025) even included a dedicated section on AI for the first time.

The Swedish MASAI trial, involving more than 105,000 women, found that AI-supported screening had higher sensitivity than standard double reading, at 80.5% versus 73.8%, while specificity remained almost identical at 98.5%. The AI-supported pathway was also associated with fewer invasive and larger interval cancers.¹

Closer to home, the GEMINI study is particularly relevant. Published in Nature Cancer in March, this prospective study involved 10,889 women screened within NHS Grampian and evaluated 17 different ways of incorporating AI into the screening workflow.² In the primary workflow, AI was associated with a 10.4% increase in cancer detection, while recall rates remained broadly unchanged.

The study also found the potential to reduce radiologist workload by up to 31%. This means that the value of AI in mammography isn’t just limited to improved detection of breast cancer, but also alleviating some of the pressure associated with the current NHS limited radiology capacity. The aim is not to replace radiologists, but to allow them to focus their attention on other aspects of their role that are equally as important as reading scans.

NHS preparations to evaluate these technologies at scale are currently underway, with NHS England having committed £6 million to AIR-SP, a national programme designed to support AI research in screening, with a projected patient cohort of almost 700,000 women across 30 hospitals.³

So what does all of this mean for UK practice? Well, following evaluations of its safety and efficacy in our screening pathways, AI has the potential to increase screening capacity, reduce radiology workloads and give radiologists more time for complex cases and time with patients while also hopefully improving detection rates.

In December, NHS England updated its clinical guidance for breast screening assessment to include the use of CEM as an optional problem-solving tool in some cases including downgrading false positives, reducing unnecessary biopsies and identifying multiple foci.

As evidenced by the RACER trial, CEM can also provide a more comprehensive assessment of disease extent during the initial diagnostic work up which potentially reduces the need for sequential imaging and enable appropriate treatment planning earlier in the pathway.⁴ It’s particularly useful because it facilitates detailed morphological assessment and provides additional information about vascularity and lesion conspicuity.

The future for breast imaging teams may be less about which modality is the best overall, but instead about asking which modality provides the most information for each individual patient’s needs.

NHS England data shows that 1.94 million women between the ages of 50 and 70 attended breast screening in the 2024/25 report – that’s an increase of almost 194,000 on the previous year, and attendance after the first invitation reached its highest level in a decade (63.6%).⁵ An additional 2,614 cancers were detected via the screening route compared to the year before.

While this shows a move in the right direction, 3 in 10 women still did not take up their invitation for breast screening, so there is evidently still work to be done. NHS England continues to drive attendance using targeted outreach, reminder messaging and expanding the use of mobile screening units as well as various national awareness incentives.

This one has been a bit quieter than AI developments but it could be really impactful. NHS England is developing Rubie; a new suite of digital services designed to replace to National Breast Screening System with the goal of implementing it by 2029.⁶ This year, appointment management and image-reading functionalities for the system are being tested.

With a growing need to connect large volumes of imaging, clinical and outcome data and the looming implementation of AI, modern breast screening will increasingly depend on the system’s ability to facilitate these needs. Future breast screening environments may require AI algorithms to access images, screening teams to monitor outcomes, and researchers to evaluate performance across large populations; Rubie will make all of these easier. In simpler terms, the modernisation of NHS screening IT could be an important catalyst for implementing further innovations within breast imaging.

In August 2026, researchers at The Institute of Cancer Research and the Royal Marsden reported promising results from a new "tissue-free" liquid biopsy approach for detecting early triple-negative breast cancer.⁷

Their liquid biopsies analyse circulating tumour DNA (ctDNA) in the patient’s blood. Conventional tumour-informed approaches require information from the original tumour to develop a personalised test… but the new approach does not require the original tumour tissue at all.

In the study, detection of ctDNA was strongly associated with recurrence, with blood testing predicting recurrence around 7.9 months (average) before it became visible on conventional scans. The tissue-free approach performed comparably to more complex tumour-informed testing.

It’s important however to clarify that this test is not a replacement for imaging or routine follow-up; it’s not just about whether we can molecularly detect recurrence earlier than imaging, but also about assessing whether acting on that information earlier makes a tangible difference to patient outcomes. If future trials demonstrate that it does, then surveillance after breast cancer could change dramatically.

As we know, breast cancer affects people of all ages – not just women aged 50 – 70. The fifth edition of the RCR breast imaging guidance now includes a dedicated section on risk-adapted screening and stresses the increasing importance of this approach.⁸

Research published in August 2026 corroborates this; the University of Cambridge and The Institute of Cancer Research examined women under 50 from the Breast Cancer Now Generations Study and compared current NICE referral criteria with the BOADICEA multifactorial risk model. They found that current NICE criteria identified only 4.4% of women who subsequently developed breast cancer within 10 years, compared with 34.8% identified using the full BOADICEA model. The most stunning takeaway from this research was that following NICE criteria alone could miss up to 95% of women under 50 who subsequently developed breast cancer within 10 years.⁹

Now, this needs careful interpretation. This doesn’t mean that 95% of women with breast cancer are being missed by the NHS screening programme, but rather it raises an important question… could a risk model based on multiple factors identify women who would benefit from additional surveillance more effectively than age/family history alone?