Does an AI orthodontic diagnosis tool work equally across ethnic groups?

Detection of Racial Bias in an Artificial Intelligence Based Orthodontic Diagnosis System

Observational Bezmialem Vakif University · NCT07162753

This project tests whether an AI trained on Turkish profile photos and lateral X‑rays can correctly identify Class I, Class II (Div 1 & 2), and Class III skeletal malocclusions in Turkish patients and patients from other ethnic backgrounds.

Quick facts

Study typeObservational
Enrollment5000 (estimated)
Ages4 Years and up
SexAll
SponsorBezmialem Vakif University Academic / other
Locations1 site (Istanbul, Fatih)
Trial IDNCT07162753 on ClinicalTrials.gov

What this trial studies

Researchers trained a deep‑learning system on 3,280 Turkish patients using pre‑treatment profile photographs and lateral cephalometric radiographs to recognize four skeletal malocclusion classes. The trained AI was then tested on two independent groups: Turkish patients whose records were not used for training and patients of other ethnic backgrounds treated at a collaborating clinic in Belgium. Only anonymized pre‑treatment images meeting strict photo quality and positioning criteria are included, and cases with prior orthodontic treatment or major facial soft‑tissue issues are excluded. Comparing performance between the two test groups will reveal whether the model generalizes across ethnicities or shows bias related to its Turkish training data.

Who should consider this trial

Good fit: Ideal candidates have pre‑treatment lateral profile photographs with a plain background taken in natural head position plus matching lateral cephalometric radiographs, and no prior orthodontic treatment or facial hair that obscures the profile.

Not a fit: Patients with blurry, missing or nonstandard profile photos, significant facial deformities (including cleft lip/palate), or prior orthodontic treatment are unlikely to benefit from this AI evaluation.

Why it matters

Potential benefit: If successful, the AI could help clinicians make faster, more consistent skeletal diagnoses and highlight where additional training data are needed to ensure fair performance across populations.

How similar studies have performed: Prior AI work in dental imaging and cephalometric analysis has shown good accuracy for classification tasks, but rigorous cross‑ethnic validation is limited in the literature.

Eligibility criteria

Show full inclusion / exclusion criteria
* Patients with lateral profile photographs taken before starting orthodontic treatment
* Profile photographs with a plain background
* Patient in natural head position facing forward
* No beard or mustache that could camouflage the profile
* Availability of lateral cephalometric radiographs taken before treatment together with the profile photograph

Exclusion criteria:

* Blurry missing or non standard profile photographs
* Presence of beard mustache scars cleft lip palate or soft tissue deformities that may significantly affect facial appearance
* Previous orthodontic treatment history

Where this trial is running

Istanbul, Fatih

Study contacts

How to participate

  1. Review the eligibility criteria above with your treating physician.
  2. Visit the official trial page on ClinicalTrials.gov for the most current contact information and recruitment status.
  3. Contact the listed study coordinator or principal investigator to request pre-screening. Pre-screening is free and never obligates you to enroll.
Conditions Class I MalocclusionClass II Div 1 MalocclusionClass II Division 2 MalocclusionClass III MalocclusionArtificial IntelligenceDeep LearningEthnic BiasMachine Learning
Last reviewed 2026-06-15 by the Find a Trial editorial team. Information on this page is for educational purposes and is not medical advice. Always consult qualified healthcare professionals about clinical trial participation.