AI is beginning to narrow long-standing gaps in breast cancer care, promising earlier detection and more personalised treatment — but only if new tools are rigorously evaluated, safeguarded and used alongside human expertise.
Breast cancer remains the most common cancer in women worldwide. The World Health Organization recorded about 2.3 million new cases and 670,000 deaths in 2022. Outcomes vary sharply depending on how quickly disease is found and treated: American Cancer Society data show five-year survival is about 99 per cent for localised disease, falling to roughly 31 per cent for distant metastases.
Conventional mammography saves lives but has limits. Dense breast tissue can cut sensitivity from over 80 per cent in fatty breasts to about 57–71 per cent in extremely dense breasts, meaning some cancers are missed. Supplemental methods such as tomosynthesis, ultrasound and MRI each have trade-offs, making personalised screening strategies important.
AI tools are emerging as a “second set of eyes”. In Germany’s PRAIM study, radiologists using AI achieved a cancer detection rate of 6.7 per 1,000 — a 17.6 per cent relative increase on standard double reading — while recall rates stayed non-inferior and predictive values improved. An NHS pilot of the Mia AI tool analysed over 10,000 mammograms and flagged 11 additional cancers missed in the first human read.
On 4 February 2025, the NHS launched the EDITH trial, inviting nearly 700,000 women across around 30 breast clinics to take part in a multi-platform AI study, backed by £11 million from the National Institute for Health Research. Professor Lucy Chappell, NIHR chief executive, called it a “landmark trial” that could test whether AI speeds diagnosis and enables single-reader workflows without losing accuracy.
Commercial developments are also integrating AI into imaging and workflow. SimonMed Imaging’s Mammogram+ combines 3D mammography with FDA-cleared AI to generate hundreds of high-resolution images per scan, delivering same-day patient reports on breast density, personal risk and action plans. In treatment support, platforms such as OncoGenomX’s PredictionStar™ aim to predict the most effective drug combinations for individual tumours, reducing over-treatment and the risk of recurrence.
AI could yield wider health benefits. SimonMed has added a module to detect breast arterial calcifications during mammography, flagging cardiovascular risk without extra radiation or procedure time.
Yet evidence gaps remain. Studies have found inconsistent performance across settings, limited generalisability and concerns over bias, overdiagnosis and workflow disruption. Experts urge a “comprehensive AI governance framework” with adaptable, transparent systems that prioritise patient safety. Responsible adoption will depend on large, real-world trials; retaining clinicians as decision-makers; transparent reporting of training data and subgroup performance; independent validation; and clear patient consent.
For the UK, EDITH offers a chance to show how a national health system can evaluate and scale AI responsibly. Success will depend not just on innovation but on commissioning pathways, workforce training, interoperability and regulatory oversight.
AI is already improving detection and personalising care in real-world pilots. With rigorous trials, strong governance and close collaboration between policymakers, clinicians, researchers, developers and patients, the UK could set a model for ethical, scalable use of AI in cancer care.
Created by Amplify: AI-augmented, human-curated content.
