AI Shows Promise in Improving Breast Cancer Screening, Global Studies Find
Recent research from the United Kingdom, Australia, and Sweden indicates that artificial intelligence (AI) tools are being evaluated for their potential to improve breast cancer detection and streamline screening processes. The studies, conducted in real-world and simulated clinical settings, report findings on increased cancer detection rates, reduced radiologist workload, and new capabilities in predicting patient risk.
Key Findings from UK Research
Two major UK studies, published in the journal Nature Cancer, evaluated the integration of AI into National Health Service (NHS) breast screening programs.
The NHS AI Study
Involving approximately 175,000 women, this research compared standard double reading by two radiologists against a combination of one human reader and one AI reader. The AI software was developed by Google.
Key reported outcomes include an increase in the cancer detection rate from 7.54 to 9.33 per 1,000 women when AI acted as the second reader.
The study also noted a reduction in false positives and a 39.3% decrease in patient recalls for first screening mammograms. Radiologist reading time was reduced by 32.1%. In a separate arbitration scenario, where a third reader resolves disagreements, AI performed comparably to human arbitrators.
The GEMINI Study
This prospective study analyzed mammograms from 10,889 women in the NHS Grampian region using an AI system named Mia.
Researchers reported a 10.4% increase in cancer detection when AI was used as a second reader.
The recall rate for further investigations decreased by 0.8%. Radiologist workload was reduced by up to 31%, with simulated models showing potential cost savings of 36%. The time to notify patients of results was reduced from an average of 14 days to 3 days.
Both UK studies note a national shortage of clinical radiologists and suggest AI could help address workload pressures. The findings are intended to inform the larger, prospective EDITH trial, which will further evaluate AI tools across multiple NHS sites.
Developments in Australian AI for Risk Prediction
Separate research from Australia focuses on an AI tool named BRAIx, developed using millions of mammograms from the Australian population screening program.
The tool is designed for two functions: assisting in cancer detection and predicting a woman's risk of developing breast cancer within the next four years.
A study published in The Lancet Digital Health reported that the AI identified that 1 in 10 women in the top 2% of risk scores developed breast cancer within four years, despite having received clear results from initial mammograms.
Researchers state the tool analyzes mammogram pixels to identify patterns of early cancer and risk not typically perceptible to human readers. They report it offers greater accuracy in risk prediction than methods based on age, family history, and breast density alone.
The cancer detection component of BRAIx is undergoing a randomized controlled trial in Australia. Researchers project a potential rollout within five years.
Swedish Trial on Screening Outcomes
A large-scale trial in Sweden, published in The Lancet, involved 100,000 women randomly assigned to either AI-supported screening or standard screening by two radiologists.
In the AI-supported group, mammograms were analyzed by an AI system, which then directed low-risk cases to a single radiologist and high-risk cases to a double review.
The study reported that the AI-supported group had 1.55 cancer diagnoses per 1,000 women in the period following screening, compared to 1.76 per 1,000 in the standard screening group—a difference of 12%.
A higher proportion of cancers were detected at the initial screening stage in the AI group (81%) compared to the control group (74%). The study also noted a nearly 27% reduction in diagnoses of aggressive cancer sub-types in the AI-supported group.
Researcher Perspectives and Future Directions
Researchers involved across the studies consistently describe AI as a tool to support, not replace, radiologists.
Experts highlight potential benefits including earlier cancer detection, reduced patient anxiety from fewer false-positive recalls, and allowing radiologists to focus on more complex tasks.
Several researchers have emphasized the need for continued evaluation, caution in implementation, and monitoring of AI's long-term impact across different healthcare systems.
The upcoming EDITH trial in the UK and planned prospective studies in Australia aim to generate further evidence on the effectiveness and integration of AI in routine breast screening programs.