Artificial Intelligence-Assisted Cephalometric Analysis: A Comparative Study Between WebCeph and NemoCeph

The incorporation of artificial intelligence (AI) into orthodontics has facilitated automated cephalometric assessments, offering greater efficiency and minimizing human-related inaccuracies. Nevertheless, the precision of AI-driven systems, when compared with conventional digital analysis software, requires thorough validation. To evaluate the accuracy and consistency of AI-assisted cephalometric analysis using WebCeph in relation to manual tracings performed using NemoCeph, which is regarded as the reference standard. Cephalometric radiographs were obtained from 50 adult patients. Linear and angular parameters were measured using NemoCeph and WebCeph, and statistical comparisons were performed. Data normality was verified using the Shapiro-Wilk test, while paired t-tests identified significant discrepancies. For the linear variables, the most notable differences occurred in S-Ar, AO-BO, overbite, and A-Na-Pog, with WebCeph occasionally overestimating or underestimating the results relative to NemoCeph. Regarding angular parameters, significant variations were identified in the sella angle, gonial angle, upper gonial angle, lower gonial angle, SNA, SNB, ANB, lower facial height, mandibular plane to S-Na angle, occlusal plane to S-Na angle, upper incisor to Na-A angle, lower incisor to Na-B angle, and lower incisor inclination. Differences were also statistically significant in determining the facial typology. WebCeph yields dependable outcomes for linear dental evaluations but demonstrates marked discrepancies in angular and vertical skeletal dimensions compared with NemoCeph. These findings underscore the importance of cautious clinical interpretation and suggest that AI-based cephalometric tools should augment, rather than replace, expert orthodontic assessments.