Locked AI model for p16/Ki-67 dual-stain triage validated on 9,006 cervical screening cases
Cytoreader-Fusion, an AI system that classifies p16/Ki-67 dual-stain cytology slides to triage HPV-positive screening results, was locked before validation and then detected biopsy-confirmed CIN 3+ with 88.3% sensitivity and 54.4% specificity in a held-out cohort of 4,996 cases.
- Model
- Cytoreader-Fusion
- Field
- Medicine
- Date
- 2026-09-24
- Human collaborators
- Bernd Lahrmann, Nicolas Wentzensen, Niels Grabe
What was found
Dual-stain cytology is the guideline-endorsed triage for HPV-positive screening results but depends on manual reading. The ensemble model was trained on 1,919 cases and evaluated under blinded honest-broker oversight on 9,006 cases in four cohorts, with all preprocessing, aggregation rules and decision parameters fixed before validation. The endpoint was histologically confirmed CIN 3+, or AIN 3+ in an anal cohort, rather than agreement with cytologists, and one classifier served both ThinPrep and SurePath slides without retraining. In the post-lock cohort sensitivity was 88.3% (90% interval 85.3 to 91.3) and specificity 54.4% (52.9 to 55.9); across cohorts the authors report both above ASC-US+ Pap cytology and comparable to or exceeding manual dual-stain reading.
Novelty check
Europe PMC, 2026-10-29: TITLE_ABS:"dual stain" AND TITLE_ABS:"artificial intelligence" AND TITLE_ABS:"cervical" returned 3 records: this study, a 2024 review, and a 2021 JNCI paper on deep-learning automation of dual-stain cytology by Wentzensen, Lahrmann, Clarke and colleagues, several of the same investigators. Automated dual-stain reading is therefore not new; what this adds is a model locked in advance and evaluated on a held-out cohort against histological endpoints.
Caveats and known objections
External in data, not in investigators: the title calls the validation independent, and the post-lock cohort was held out under an honest broker, but the authors include the group behind the 2021 automation work and staff of H-Labs GmbH, so it is not independent in this registry's sense. Funded by the US National Cancer Institute and others, with Kaiser Permanente Northern California among the participating institutions. Retrospective, on archived slides; the endpoint is a hard histological one, but no patient was managed on the model's output. Specificity of about 54% in the post-lock cohort implies many referrals, and the abstract does not give the manual dual-stain figures for that cohort. Whether the model is available outside the developers is not stated. Autonomy is ai-assisted: the system classifies slides for a triage decision clinicians make.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Locked AI model for p16/Ki-67 dual-stain triage validated on 9,006 cervical screening cases. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-09-24-cytoreader-fusion-dual-stain-triageBibTeX
@misc{whataifound-universitymedicinegttingen-2026-triage,
title = {Locked AI model for p16/Ki-67 dual-stain triage validated on 9,006 cervical screening cases},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by University Medicine Göttingen (with H-Labs and the US National Cancer Institute). Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-09-24-cytoreader-fusion-dual-stain-triage}
}