Photograph-Based Skeletal Malocclusion Screening with Geometric Machine Learning: External Validation Across Turkish and Non-European Cohorts

A side-profile smartphone photograph classifier for skeletal malocclusion achieved AUCs of 0.902 in the Turkish external cohort and 0.835 in a non-European cohort for the Class III task.

• Why it matters: Radiographic imaging limits scalable orthodontic triage, and prior performance may not generalize to new populations, so external validation on target cohorts is needed.
• What they did: The method mapped 100 engineered geometric feature vectors from smartphone side-profile photographs to orthodontic class labels, using separate models for Class III vs non-Class III and Class I vs Class II.
• The result: For Class III, sensitivity stayed similar (0.801, 0.836, 0.848; all adjusted p>0.7) while specificity dropped (0.766 to 0.697) and AUC differed by 0.066 (p=0.029), with lower sensitivity in children than adults.

AI summary of the abstract · may contain errors · not medical advice · read the paper.