Deep learning screening of otoscopic images for ENT referral: development, validation and deployment of NHREF OtoScan
Source check: Source checked. Not independent scientific confirmation.
Finding: A MobileNetV2 screening model for ENT referral achieved 97.0% sensitivity and 75.6% specificity on 718 held-out otoscopic images. The retrospective evaluation used near-duplicate grouping and found source-dependent performance. The authors report no prospective clinical validation or clinician comparison; the software is described as research use with mandatory specialist review.
Technology: MobileNetV2 convolutional neural network
Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A calibrated MobileNetV2 model estimates abnormal-ear probability to support an ENT-referral screening decision.
- Test and result
- On a held-out set of 718 otoscopic images, the frozen referral threshold yielded 97.0% sensitivity and 75.6% specificity. Images were grouped to reduce near-duplicate leakage; performance varied by source dataset.
- Limitation / still unknown
- Retrospective public-image validation only; no prospective clinical validation or comparison against clinicians. The paper says the system is research software, not a medical device, and requires specialist review.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://www.medrxiv.org/content/10.64898/2026.11.04.26367286v1.full.pdf · https://doi.org/10.64898/2026.11.04.26367286
Source checked · editorial details
The source page was checked against this entry. 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: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
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 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.