From eye to algorithm: Assessing artificial intelligence performance in cephalometric landmark identification – a cross-sectional study

Introduction: Cephalometrics plays an essential role in orthodontic diagnosis and treatment planning. It requires accurate identification of anatomical landmarks. Manual tracing is time-consuming and varies per observer.Aim and Objectives: The study aimed to compare the accuracy and reliability of AI-driven software WebCeph Premium against traditional manual tracing in detecting anatomical landmarks in a clinical setup.Materials and Methods: A comparative study was done on 80 pre-treatment lateral cephalograms. Cephalometric analyzes - Steiner's and Tweed's were performed using WebCeph and manual methods. Wilcoxon signed-rank test and Intraclass correlation coefficients (ICC) were used for comparison and reliability assessment.Results: Significant differences were observed in key cephalometric parameters (SNA, SNB, and ANB). Vertical angles showed moderate agreement, while skeletal and dental landmarks revealed poor reliability.Conclusion: AI software should be viewed as a supporting tool rather than a complete replacement, requiring an experienced orthodontist for accurate diagnosis.