In August 2023 The Lancet Oncology published the first interim safety results of MASAI (Mammography Screening with Artificial Intelligence), a randomized controlled trial of AI-supported breast-cancer screening run in Sweden and led by Kristina Lang of Lund University. It was the first randomized trial of AI in breast screening, an important step beyond the retrospective studies that had dominated the field.
The trial enrolled 80,033 women, randomly assigning roughly half to AI-supported screening and half to standard double reading by two radiologists. In the AI group, a commercial system flagged suspicious cases and triaged the reading workload; in the control group, two human readers reviewed every mammogram in the usual way.
AI-supported screening detected cancer at a rate of 6.1 per 1,000 participants, against 5.1 per 1,000 with standard double reading, while reducing the radiologists’ screen-reading workload by 44.3 percent. The authors concluded that AI support produced a similar cancer-detection rate with substantially less reading effort, and judged it safe to continue the trial.
MASAI matters because it moved AI in medical imaging from accuracy claims on stored datasets to a prospective, randomized test in a live national screening program, with measured effects on both detection and clinician workload.