Overview
- The completed randomized controlled trial, published in The Lancet, compared AI-supported screening with standard double reading across four Swedish sites from 2021 to 2022.
- More cancers were detected with AI support without an increase in false positives, with screening false-positive rates of 1.5% in the AI group versus 1.4% in the control group.
- The AI group had a 12% lower rate of interval cancers over two years and saw fewer severe disease measures, including 16% fewer invasive, 21% fewer large, and 27% fewer aggressive subtype cancers.
- In the AI-supported arm, 81% of cancers were found at screening compared with 74% under standard double reading, using an AI system trained on more than 200,000 exams from over 10 countries.
- Study authors say AI could ease radiologist workloads and shorten waits, while experts emphasize radiologist oversight, vigilance for overdiagnosis, and the need for further validation including the ongoing UK EDITH trial of roughly 700,000 women.