Smartphone profile photograph screening for skeletal malocclusion, externally validated in Turkey and Belgium
A preprint reports that a classifier using 100 geometric features from a single side-profile smartphone photograph, with no radiograph, kept its sensitivity for skeletal Class III malocclusion across two external cohorts but lost specificity in a non-European cohort from a Belgian clinic, where its AUC fell from 0.902 to 0.835, and fell further in a second task separating Class I from Class II (0.875 to 0.723).
- Model
- Geometric machine learning classifiers on 100 engineered facial-profile features
- Field
- Medicine
- Date
- 2026-10-29
What was found
Two binary tasks were evaluated, Class III against the rest and Class I against Class II, each with its own development set and model, applied unchanged at the default threshold. The reference standard was the ANB angle on pre-treatment lateral cephalograms traced by orthodontists, never shown to the model. One external cohort came from a Turkish university orthodontic department outside the development set (256 patients in the Class III task) and the other from a private orthodontic clinic in Brussels described as non-European (282 patients). For Class III, sensitivity was similar across the evaluation sets (0.801, 0.836 and 0.848), while specificity fell from 0.766 to 0.697. Sensitivity was lower in children than in adults in both cohorts (0.71 to 0.80 against 0.89 to 0.90).
Novelty check
Europe PMC, 2026-11-05: TITLE:"malocclusion" AND photograph* AND ("machine learning" OR "deep learning" OR "artificial intelligence") returned 28 records, including cephalogram-based deep learning for orthognathic surgery need (2021), intraoral-image classification of Angle's classes (2026) and a 2024 concept paper on AI-assisted self-monitoring for skeletal malocclusion; none reported external validation of a profile-photograph classifier across populations.
Caveats and known objections
A medRxiv preprint, not peer-reviewed, so capped at claimed. The developer is among the authors: both corresponding authors list Kedi Mobil Uygulama A.Ş., an Istanbul company, as an affiliation, and Europe PMC lists funding from the TÜBİTAK 1507 SME R&D programme and the EU's Horizon Europe Women TechEU scheme; a competing-interest statement was not found in the text read for this entry. The authors attribute the specificity drop to the reference standard rather than the model, since the ANB threshold comes from Caucasian normative data, which is a fair point but also means the tool's accuracy depends on which population it is used in. The reported ancestry of the Belgian cohort is the paper's own label. Retrospective, with development-set sizes not recorded here; the endpoint is classification accuracy against a cephalometric criterion, a surrogate, and no screening decision was studied. Autonomy is ai-assisted: the tool is proposed for triage ahead of an orthodontist, who makes the diagnosis.
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 Claimed 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 claimed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Smartphone profile photograph screening for skeletal malocclusion, externally validated in Turkey and Belgium. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-29-photo-malocclusion-screeningBibTeX
@misc{whataifound-bezmialemvakifuniversitymulastkkomanuniversityandkedimobiluygulama-2026-screening,
title = {Smartphone profile photograph screening for skeletal malocclusion, externally validated in Turkey and Belgium},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Bezmialem Vakif University, Muğla Sıtkı Koçman University and Kedi Mobil Uygulama. Verification: Claimed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-10-29-photo-malocclusion-screening}
}