Repeatability and Agreement of CephX AI-Driven Cephalometric Analysis in Patients with Cleft Palate Only: A Comparative Study with Non-Cleft Controls

Cephalometric analysis is essential in patients with cleft palate only (CPO) for the assessment of craniofacial morphology and treatment planning. This study evaluated the test–retest repeatability of CephX AI-driven cephalometric analysis and its agreement with expert manual measurements in patients with CPO versus non-cleft controls. Cephalometric radiographs of 50 patients with CPO (25 males and 25 females; aged 14–23 years) and 50 non-cleft controls matched for sex and ethnic origin were traced manually by two experienced orthodontists using CephX. All analyses were repeated after a two-week interval. Repeatability was assessed using Lin’s concordance correlation coefficient (CCC), and agreement between AI and manual measurements was evaluated using Bland–Altman analysis. CephX showed high test–retest repeatability for both groups, with CCC values of 0.97 or higher for all cephalometric variables evaluated. However, the agreement with expert manual measurements varied among cephalometric variables. In the CPO group, 34 of 50 radiographs required manual scale calibration and 31 of 50 generated a landmark-review warning; nevertheless, no skeletal or dental landmarks included in the analysis were manually repositioned by the investigators. CephX provided highly reproducible measurements, but the need for calibration and landmark review indicates that clinician supervision remains important, particularly in patients with altered craniofacial anatomy.