RELIABILITY AND ACCURACY OF ARTIFICIAL INTELLIGENCE–BASED SOFTWARE FOR CEPHALOMETRIC DIAGNOSIS: A DIAGNOSTIC STUDY
Abstract
Artificial intelligence (AI) is transforming cephalometric diagnostics in orthodontics through platforms such as CephX and WebCeph. These AI-based tools utilize machine learning methods and neural networks for automatic identification of cephalometric landmarks and for performing angular and linear measurements, thereby improving diagnostic accuracy, consistency, and efficiency. CephX demonstrated higher reliability and accuracy compared to WebCeph, particularly for FMA and L1–MP measurements, while WebCeph showed greater discrepancies in SNA, ANB, U1–NA, and L1–MP measurements. Both platforms allow manual adjustment of landmarks, which is essential for ensuring clinical accuracy. Key challenges in AI implementation include variability in results across skeletal classes, data privacy concerns, and the need for clinician supervision to verify AI-generated outputs. The expanding use of AI in orthodontics offers opportunities to optimize cephalometric analysis, improve treatment planning, and standardize diagnostic procedures.