AI in breast cancer screening reduces later diagnoses by 12%, study finds

A Swedish randomised trial of about 100,000 women found AI-supported mammography cut the rate of cancer diagnoses in the years after screening by 12% and raised the share of cancers detected at screening to 81% versus 74% in standard double reading.

A Swedish randomised trial examined the use of artificial intelligence to support mammography screening and reported lower rates of later cancer diagnosis and higher early detection. About 100,000 women who attended screening between April 2021 and December 2022 were assigned to either AI-supported screening or the usual double reading by two radiologists. The AI system triaged images, assigning low-risk cases to a single reading and high-risk cases to double reading, and highlighted suspicious areas for radiologists. The study, described as the largest trial of its kind, was published in The Lancet. Key findings: - The AI-supported group had 1.55 cancers per 1,000 women diagnosed after screening compared with 1.76 per 1,000 in the control group, a reported 12% reduction in later diagnoses. - In the AI group, 81% of cancers were detected at the screening visit versus 74% in the standard-reading group. - The AI group had about 27% fewer aggressive-subtype cancers compared with the control group. - Researchers emphasised that at least one human radiologist remained part of the screening pathway and did not recommend replacing clinicians with AI. - The lead author noted potential benefits for earlier detection and radiologist workload, while urging cautious, monitored rollout of tested AI tools. - External experts said the results are promising but come from a single-centre trial and that further evaluation and assessment across programmes will be important. Summary: The trial reports that AI-supported mammography was associated with fewer cancers diagnosed after screening and a higher proportion of cancers found at the screening visit, which could affect how screening pathways operate. Further evaluation and ongoing trials, including work reported in the UK, are described as important to understand safety, effectiveness and how results might vary between settings.