Accuracy of Cephalometric Analysis in Cleft Lip and Palate: Comparison of Manual, Artificial Intelligence-Based, and Examiner-Corrected Artificial Intelligence Approaches.
Journal:
The Cleft palate-craniofacial journal : official publication of the American Cleft Palate-Craniofacial Association
Published Date:
Dec 2, 2025
Abstract
ObjectiveTo evaluate the accuracy of the artificial intelligence (AI)-powered orthodontic imaging system OrthoDx™ for cephalometric analysis in patients with cleft lip and/or palate (CLP), compared with manual semi-automated measurements obtained using Dolphin Imaging software.DesignRetrospective study with anonymized lateral cephalometric radiographs.SettingDepartment of Orthodontics, Faculty of Dentistry.Patients, ParticipantsThe study included 188 patients with CLP (mean age, 9.18 ± 4.73 years).InterventionsManual cephalometric analysis was performed using Dolphin Imaging software used as reference, while AI-based and examiner-corrected analyses were conducted using OrthoDx™. Seventeen angular and eight linear cephalometric parameters were analyzed.Main Outcome Measure(s)Primary outcome was the agreement between manual, AI, and examiner-corrected AI cephalometric measurements, assessed using intraclass correlation coefficients (ICCs), one-sample t-tests, and Bland-Altman analyses. Primary outcome measures were defined prior to data collection.ResultsIntraobserver reliability for the manual method showed good to excellent reliability with no significant differences between repeated measurements. SNA, saddle, articular, and U1-FH angles differed significantly between manual and AI methods but not after examiner correction. Significant differences were observed between manual and AI, and between manual and corrected AI, for several other parameters. ICCs ranged from moderate (0.70-0.75) to excellent (>0.90), indicating variable agreement across parameters.ConclusionsAI-based cephalometric analysis using OrthoDx™ demonstrated limited accuracy in patients with CLP. Examiner intervention reduced the variability of certain cephalometric measurements, making the results closer to the manual group, supporting the role of clinician-supervised AI as a complementary rather than replacement tool.
Authors
Keywords
No keywords available for this article.
