Favorable outcomes were found when artificial intelligence (AI) was used in mammography screenings compared with the standard double reading in a new study published in The Lancet.
AI Improves Breast Cancer Detection in Swedish Trial
A large trial in Sweden demonstrates that artificial intelligence can improve breast cancer screening. Researchers found AI assistance led to fewer interval cancers – cancers diagnosed between scheduled screenings – without increasing false positives. The findings, based on the Mammography screening with Artificial Intelligence (MASAI) trial (NCT04838756), could significantly impact how mammograms are read and interpreted.
The study involved over 105,000 women aged 40 to 74 in southwest Sweden. Participants were invited for screening every 1.5 to 2 years, with those at moderate risk due to family history screened annually. All mammograms were randomly assigned to either AI-assisted review or standard double reading by radiologists. The AI model, trained on over 200,000 prior examinations, provided a risk score and highlighted specific areas of concern.
Researchers focused on the interval cancer rate – the number of cancers detected between screenings or within two years of a negative result - as the primary measure of success. They also analyzed characteristics of interval cancers, as well as the AI’s sensitivity and specificity.
The trial included 16 radiologists, most with over five years of experience, who assessed the mammograms. the screenings were randomized to ensure unbiased evaluation.
