Artificial intelligence has helped doctors detect significantly more cases of breast cancer in routine screening, according to a landmark clinical trial in Sweden that researchers say could reshape national screening programmes.

The study, published in The Lancet, is the first completed randomised controlled trial to test AI-supported mammography at scale. More than 100,000 women undergoing regular breast cancer screening at multiple Swedish centres were randomly assigned to two groups. In one group, a single radiologist interpreted each mammogram with the support of an AI system. In the other, scans were read in the traditional way by two radiologists, the current standard in many European countries.

The AI-assisted approach detected 9 percent more cancers than the conventional method, without increasing the rate of false positives. Researchers also reported a 12 percent reduction in interval cancers in the AI group, meaning fewer cancers were diagnosed between scheduled screening rounds, when tumours are often more advanced and harder to treat.

Lead author Kristina Lang of Lund University said the findings indicate that integrating AI into screening could both improve early detection and ease pressure on overstretched radiology services. Earlier interim data from the same trial showed that AI support nearly halved the time radiologists spent reading scans, suggesting substantial potential for efficiency gains.

The AI system used in the trial, Transpara, was trained on more than 200,000 mammograms from 10 countries. It assigns a risk score to each image, flagging those most likely to contain cancer for closer human review. The performance gains were consistent across age groups and breast density levels, both important factors in screening accuracy.

Specialists caution, however, that AI should augment rather than replace human expertise. Jean-Philippe Masson, head of the French National Federation of Radiologists, said the radiologist’s judgement remains essential to interpret AI findings and avoid unnecessary biopsies or treatment. He and others warn that overdiagnosis and cost remain concerns, particularly in health systems where AI tools are not yet widely deployed.

Independent experts such as Stephen Duffy of Queen Mary University of London say the trial adds to evidence that AI-assisted screening is safe, but argue that longer follow-up is needed to confirm whether the reduction in interval cancers translates into fewer deaths from breast cancer.

Globally, breast cancer remains the most commonly diagnosed cancer in women, with millions of new cases each year. Researchers say that if validated in other settings, AI-supported screening could become a critical tool in catching more cancers earlier, when treatment is most effective.