Artificial intelligence has been studied in radiology for years. Since 2023, large trials have started to show whether it can safely share the reading of screening mammograms, and in 2025 the NHS began its own trial. The question is no longer whether AI can analyse a mammogram, but whether it makes screening better for patients.
What the MASAI and EDITH trials show
The MASAI trial in Sweden randomised more than 100,000 women to standard double reading or to AI-supported reading, where AI sorted mammograms into low-risk scans read by one radiologist and higher-risk scans read by two. AI-supported screening found about 29% more cancers without raising false-positive recalls, and cut radiologists' reading workload by around 44%. The final results, published in The Lancet in 2026, found it was not worse at preventing interval cancers (cancers diagnosed between screens). In the UK, the EDITH trial, announced in February 2025 with £11 million of government funding, involves nearly 700,000 women at around 30 sites and tests whether AI can safely replace one of the two specialists who read each mammogram.
What this means for the NHS screening programme
The NHS breast screening programme invites women from age 50 up to their 71st birthday every three years. Every mammogram is read by two radiologists independently — a process designed to catch cancers one reader might miss. AI is being evaluated as a replacement for the second reader, which would release significant radiologist capacity. With radiologist shortages a known challenge for NHS radiology, this has real implications for waiting times and programme coverage.
What AI cannot yet do
- AI tools are currently validated for mammography only — not for MRI or ultrasound breast screening
- AI performance varies by breast density and by cancer subtype — it remains better at detecting some cancers than others
- Clinical decision-making (biopsy, referral, treatment) remains firmly with human clinicians
- AI tools require ongoing validation as the screening population and imaging technology evolves
- AI is not part of routine NHS screening yet: EDITH is testing whether it can be, site by site
The global picture: AI in low-resource settings
For high-income countries, AI offers efficiency gains in an already-functioning system. The more transformative potential may be in low-income communities, where radiologist shortages mean mammograms are simply not read at all. Researchers are exploring whether AI could enable community screening programmes — operating with portable digital mammography units and cloud-based AI reading — in settings where a specialist radiologist is not available. This remains early-stage, but the direction is significant.
Sources
Share this article