Using AI in breast cancer screening reduces late diagnosis rate by 12%

The study found that the use of artificial intelligence in breast cancer screening was associated with a 12% reduction in the number of cancer cases detected in subsequent years, as well as a higher rate of early diagnosis. According to scientists, this is the largest study to date on the use of AI in mammography screening. It involved 100,000 women in Sweden who were screened between April 2021 and December 2022. Participants were randomly divided into two groups: in one, mammograms were analyzed using an AI system, in the other, the images were evaluated by two radiologists according to a standard procedure.

An artificial intelligence algorithm analyzed the images and assigned low-risk cases to be read once by a specialist, and high-risk cases to be reviewed twice. In addition, the system highlighted suspicious areas, helping doctors focus on possible pathologies. According to a publication in The Lancet, the use of AI in mammography screening has reduced cancer detection rates by 12% within a few years of screening. In the AI ​​group, there were 1.55 cases of cancer per 1,000 women, while in the control group this figure was 1.76 per 1,000.

In the group where mammography was performed using AI, 81% of cancer cases were detected at the screening stage, while in the control group this figure was 74%. In addition, in the “AI group” there were 27% fewer cases of aggressive forms of the disease compared to participants whose images were assessed according to the standard scheme. Lead author of the study, Dr. Christina Lang from Lund University, noted that the use of artificial intelligence in mammography contributes to earlier detection of cancer, but this approach also has certain limitations.

Breast cancer remains the leading cause of death among women aged 35 to 50 years, with over 2 million new cases diagnosed each year worldwide. Despite the demonstrated benefits of AI in the screening process, the researchers stress that the technology is not intended to replace medical professionals. The examination still requires at least one radiologist to interpret the results with the support of an algorithm.