AI additional reader found 0.7-1.6 extra cancers per 1,000 in prospective Hungarian mammography screening rollout
Source check: Needs source check. Not independent scientific confirmation.
Finding: Ng et al. prospectively assessed a commercial AI system as an additional reader beside standard double reading in three rollout phases (about 25,000 women, four sites), finding 0.7-1.6 additional cancers per 1,000 cases, 0.16-0.30 percent additional recalls and a 0.1-1.9 percent gain in positive predictive value after 7-11 percent additional human reads. Most extra cancers were invasive and small. Authors are from the vendor, Kheiron Medical Technologies; there is no randomization or mortality outcome, and the extra reading workload is part of the result. Abstract-level review.
Technology: Commercial deep-learning mammography reader (vendor system)
Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.
- AI contribution
- The AI flagged cases for arbitration review in addition to two human readers.
- Test and result
- Performance was assessed prospectively across a pilot, a multicentre pilot and a live rollout (per abstract).
- Limitation / still unknown
- Long-term outcomes, interval cancers and independent evaluation were not assessed.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://europepmc.org/article/MED/37973948
Needs source check · editorial details
The source has not been fully checked or contains conflicting information. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 17 Nov 2026. Review scope: abstract and bibliographic metadata (Europe PMC) only; Methods and full text not assessed.
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 4/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.