Lancet
AI-assisted mammography matches standard double reading in major screening trial
A Swedish randomized, controlled, non-inferiority screening-accuracy trial (NCT04838756; N = 105,915) compared AI-supported mammography screening with standard double reading. Interval cancer rates were 1.55 vs. 1.76 per 1,000 participants, respectively (proportion ratio 0.88; 95% confidence interval, 0.65–1.18), meeting the non-inferiority margin. The AI-supported group had fewer invasive, T2+, and non‑luminal A interval cancers. Sensitivity was higher with AI support (80.5% vs. 73.8%; p=0.031), with consistency across age groups and breast density, and for invasive cancers. Specificity was identical between groups at 98.5%. The AI workflow also reduced screen reading workload.
Clinical takeaway: AI-enabled single reading may be a viable alternative to traditional double reading, potentially preserving diagnostic quality while improving efficiency in high-volume breast screening programs.
Source:
Gommers J, et al. (2026, January 31). Lancet. 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. https://pubmed.ncbi.nlm.nih.gov/41620232/