Artificial intelligence for facial soft‐tissue landmark localization, measurement, and classification in orthodontics: A systematic review

This systematic review provides a broad overview of three distinct applications of artificial intelligence (AI) in facial soft‐tissue assessment: landmark localization, automated facial measurement, and classification of soft‐tissue conditions in orthodontic and dentofacial evaluation. PubMed, Embase, Scopus, Web of Science, Cochrane Library, ClinicalTrials.gov, Google Scholar, and Open Science Framework were searched. Studies using 2D facial photographs, 3D facial scans, or AI‐generated facial models were included. Risk of bias was assessed with an adapted QUADAS‐2 tool and certainty of evidence with GRADE. The protocol was registered in PROSPERO (CRD420261452521). Fifteen studies were included. Landmark‐localization performance varied substantially by imaging modality, landmark definition, data set, reference standard, and error metric; central and anatomically well‐defined landmarks were generally more accurate than peripheral or subjective points. Agreement for AI‐derived measurements ranged from poor to excellent, with ICCs of approximately 0.01–0.984. Classification accuracy ranged from approximately 0.83 to 0.98, with AUCs up to 0.98. Given only two narrowly scoped classification studies, these findings should be considered exploratory and hypothesis‐generating rather than broadly generalizable. Because of substantial methodological heterogeneity, findings were synthesized qualitatively. Current evidence supports AI as an adjunctive tool, but routine clinical implementation requires robust external validation and clinician oversight. PROSPERO registration: CRD420261452521