The use of artificial intelligence (AI) in breast cancer screening has reduced the number of cancers diagnosed later by 12%, according to a major new study from Sweden.
The research, published in The Lancet, involved around 100,000 women who took part in routine mammography screening between April 2021 and December 2022. It is the first large randomised trial to assess how AI performs in real-world breast cancer screening.
Women were randomly divided into two groups. One group received standard screening, where mammograms were read by two radiologists. The other group had AI-supported screening, where an AI system assessed the scans first. Low-risk cases were read by one radiologist, while higher-risk cases were checked by two, with the AI also flagging suspicious areas.
The study found that fewer women in the AI group were diagnosed with breast cancer in the years after screening. There were 1.55 cancers per 1,000 women in the AI-supported group, compared with 1.76 per 1,000 in the standard screening group.
More cancers were detected earlier in the AI group. About 81% of cancers were found during screening, compared with 74% in the standard group. The AI group also had 27% fewer aggressive cancer sub-types.
Dr Kristina Lang from Lund University, who led the study, said AI could help reduce pressure on radiologists while improving early detection. However, she stressed that AI should support doctors, not replace them, and must be carefully monitored.
Experts welcomed the findings but said further studies are needed, especially across different countries and healthcare systems, before AI is widely adopted.
