AI-assisted mammograms help radiologists detect more aggressive breast cancers.

A large Swedish trial published in The Lancet reported that AI-supported mammogram readings reduced interval cancers by 12 percent among more than 100,000 screened women.

A large randomized trial in Sweden examined whether artificial intelligence could help radiologists detect breast cancers that appear between routine screenings. More than 100,000 women were screened over two years and mammograms were either supported by an AI system or read by two radiologists, a common practice in Sweden. The trial measured interval cancers, which are invasive tumours that arise between regular mammograms and are often harder to detect. Key findings: - The trial divided screenings roughly in half: one group used AI-supported reads and the other used standard double reading by two radiologists. - The rate of interval cancers was reported to be 12% lower in the AI-supported group (82 versus 93 cases), a numerical difference of 11 fewer interval cancers. - AI-supported readings triaged images into low-risk cases reviewed by a single radiologist and higher-risk cases reviewed by two. - The study was carried out at a single centre and did not record participants' race or ethnicity, which the authors noted as limitations. - Researchers said longer-term outcomes such as mortality remain undetermined and that a cost-effectiveness analysis is planned. Summary: The reported effect was a reduction in interval cancers in the AI-supported screening arm, based on the trial's screenings and measurements. Researchers noted limitations including the single-centre setting and missing race/ethnicity data, and they described further study on mortality impact and cost-effectiveness as next steps.