Medical AI

Artificial intelligence can significantly improve the detection of breast cancer during routine screening, according to findings from a world-first clinical trial released on Friday. The results point to AI as a potential solution to both diagnostic gaps and rising workload pressures faced by radiologists globally. Published in The Lancet, the study is the first completed randomised controlled trial to rigorously evaluate AI-assisted breast cancer screening. Conducted in Sweden, the trial followed more than 100,000 women who underwent routine mammography in 2021 and 2022. Participants were randomly divided into two groups. One group had their scans reviewed by a single radiologist supported by an AI system, while the other followed the standard European protocol of double reading by two radiologists. The outcomes showed that the AI-assisted approach identified 9% more cancer cases than the conventional method. Importantly, over a two-year follow-up period, women in the AI-supported group had a 12% lower incidence of “interval cancers” — cancers detected between regular screening rounds, which are often more aggressive. The benefits were consistent across age groups and breast density levels, and the rate of false positives remained comparable between both groups. Kristina Lang, senior author of the study and a researcher at Lund University, said the findings indicate that large-scale adoption of AI-supported mammography could ease staffing pressures in radiology departments while improving early cancer detection. However, she emphasised that any rollout must be done carefully, with ongoing evaluation and oversight. Experts cautioned that AI should complement, not replace, human expertise. Jean-Philippe Masson, head of the French National Federation of Radiologists, noted that radiologists must validate AI-generated findings, as the technology can sometimes flag benign tissue changes as cancer. He also pointed out that high costs and concerns around overdiagnosis have slowed AI adoption in countries like France. Stephen Duffy, emeritus professor of cancer screening at Queen Mary University of London, who was not involved in the research, said the trial adds to growing evidence that AI-assisted screening is safe. However, he flagged that the reduction in interval cancers was not statistically significant and called for longer follow-up to assess whether outcomes between the two groups eventually converge. Earlier interim results from the trial, published in 2023, showed that AI nearly halved the time radiologists spent reviewing mammograms. The AI system used in the study, Transpara, was trained on over 200,000 mammography exams from 10 countries. Breast cancer remains the most commonly diagnosed cancer among women worldwide. According to the World Health Organization, more than 2.3 million women were diagnosed with the disease and around 670,000 died from it in 2022. Source: PTI