The abstract is read at the publisher; the summary is JClub's.
Algorithm D achieved the highest AUROC (0.849) and the top sensitivity across matched AI-positive rates in 707,922 mammography screening exams from 2022–2023.
- Why it matters: Independent, head-to-head evaluation of commercial mammography AI on large, diverse real-world cohorts is limited, leaving uncertainty about which algorithms perform best at clinical operating points.
- What they did: Using electronic medical records, four AI CAD models (three FDA-cleared commercial algorithms A, B, D and open-source Mirai) were applied to standard screening views, with cancer within 12 months identified via registry linkage and compared by AUROC and sensitivity/specificity/PPV at AI-positive rates of 5%–25%.
- The result: Algorithm D led sensitivity at every AI-positive rate from 55.3% (5%) to 77.9% (25%), with non-overlapping confidence intervals versus other algorithms, while no algorithm matched radiologist sensitivity at AI-positive rates of 10% or less.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.