[Articles] Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: a randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial

Back to news list

Source: The Lancet

Original: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(25)02464-X/fullt...

Published: 2026-01-31

The MASAI trial compared AI-assisted mammography screening with standard double-reading without AI in a randomized, controlled, non-inferiority, single-blind, population-based accuracy screening trial.[1] AI-assisted screening showed a non-inferior frequency of interval cancers and a smaller number of interval cancers with unfavorable characteristics.[1] The sensitivity was higher, the specificity the same as with the standard procedure.[1] Cancer detection increased by 29% (6.4 versus 5.0 per 1,000 screened), while maintaining a similar false-positive rate.[1][2] The workload of radiologists decreased by 44%.[1][2] AI has contributed to the detection of small invasive cancers without lymphatic involvement, including triple-negative forms.[1] Findings suggest that AI can improve screening performance and reduce reading burden.[1]