The UK’s first comprehensive evaluation of an Artificial Intelligence system in routine breast cancer screening reports that it can increase breast cancer detection by 10.4% and has the potential to reduce the workload of healthcare workers by more than 30% compared to the current clinical process. Conducted by the University of Aberdeen, NHS Grampian and Kheiron Medical Technologies, now part of DeepHealth Inc., and published in Nature Cancer, the study assessed how the Artificial Intelligence software tool Mia could support screening for 10,889 women in NHS Grampian. Researchers found that the technology helped identify more cancers, most of which were invasive and high grade, while also shortening the time to notify affected women from 14 days to just 3 days, which is described as crucial for enabling earlier treatment with a higher chance of success.
The evaluation was designed to reflect real-world conditions in the UK breast screening programme, where all women aged between 50 and 70 years old are invited for mammograms every three years, resulting in over 2 million mammogram examinations being performed annually. At present, two radiologists read each mammogram to minimize missed cancers, yet approximately 20% of cancers are missed using this process. For each five women recalled, approximately one will be diagnosed with breast cancer, leaving many to undergo unnecessary and often invasive tests and associated anxiety. By incorporating Artificial Intelligence into the workflow, the study found that it could reduce the number of women recalled unnecessarily for further assessment, including avoidable biopsies, which is expected to cut patient stress and save healthcare resources and costs.
To understand where Artificial Intelligence could add the most value, the team tested seventeen different scenarios that integrated Mia into the existing screening workflow at various points and with different operating point configurations. The results showed that combining Artificial Intelligence as a second reader substituting one human reader, and as an extra reader serving as a safeguard, delivered the best balance of workload savings and improved early cancer detection without increasing recall rates. Researchers and clinical leaders involved in the project argue that the findings address key evidence gaps highlighted by the UK National Screening Committee, which has so far not recommended Artificial Intelligence for national use due to limited data. They say the work demonstrates that Artificial Intelligence can be tailored to local needs, augment radiologists by picking up cancers that would otherwise be missed, and inform forthcoming studies such as the EDITH trial, which will expand evaluation of Artificial Intelligence in breast screening across sites throughout the UK.
Clinicians and scientists emphasise that Artificial Intelligence can perform tasks similar to human experts, such as examining mammograms and highlighting areas of concern, and that the novel trial design allowed simulation of multiple real-world deployment options in a way not previously attempted in this field. Senior figures from DeepHealth and the National Institute for Health and Care Research describe the project as evidence that Artificial Intelligence-powered solutions can enhance clinical accuracy, reimagine care delivery and ease pressure on an overstretched NHS workforce facing high workloads, a shortfall of clinical radiologists and an ageing population. The study’s backers state that rigorous testing through programmes like GEMINI helps build trust, supports safe adoption of clinical Artificial Intelligence at scale, and sets a foundation for translating cutting-edge technologies into tangible improvements in outcomes and experience for women attending breast screening.
