Accuracy and reliability of 2 artificial intelligence platforms for cephalometric analysis compared with a...

Web-based platforms offer cephalometric tracing using artificial intelligence (AI) with varying performance levels. This study assessed the accuracy, reliability, and time efficiency of cephalometric tracings performed with the AI Web-based platforms WebCeph (Assemble Circle, Seoul, South Korea) and CephX (ORCA Dental AI, Las Vegas, Nev) in both their automated and corrected forms.

Fifty pretreatment lateral cephalograms of patients were randomly selected and traced using AI platforms WebCeph and CephX in both their automated and landmark-corrected forms, along with the Dolphin Imaging software (version 13.01; Dolphin Imaging and Management Solutions, Chatsworth, Calif) as the "gold standard." Twelve parameters involving sagittal, vertical, dental, and soft-tissue dimensions were selected. The time required for each analysis was measured using a stopwatch. Intersystem comparisons were performed using ordinary least squares linear regression models, with Dolphin Imaging software as the reference. The intraclass correlation coefficient was used to determine the agreement among systems. A significance level of P <0.05 was applied, and 95% confidence intervals were calculated for all outcomes. Clinically relevant differences were defined as angular discrepancies greater than 2° or linear discrepancies exceeding 2 mm.

The AI systems in their corrected form showed similar results to those of Dolphin Imaging software. If a 14% error is accepted, they were accurate and reliable in 11 of 12 parameters. Moreover, it was possible to reduce the tracing time by 46% compared with Dolphin Imaging software. The automated systems demonstrated low reliability and accuracy for cephalometric analysis. CephX and WebCeph are still not suitable for assessing soft-tissue parameters.

CephX and WebCeph platforms for cephalometric tracing are valuable diagnosis tools only when landmark correction is applied.