Clinical Reliability of AI-Based Cephalometric Analysis Using WebCeph: A Comparative Agreement Study

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Artificial intelligence (AI) offers rapid, standardized cephalometric analysis.
• Clinical interchangeability of AI measurements with expert manual tracing is uncertain.

Purpose of the Study:

• To compare automated AI cephalometric analysis (WebCeph) with expert manual tracing.
• To evaluate the clinical agreement of AI-generated Steiner cephalometric variables.

Main Methods:

• Comparative study of 100 lateral cephalometric radiographs.
• Analysis using expert manual tracing and automated WebCeph platform.
• Evaluation of seven Steiner variables using paired t-tests, ICC, and Bland-Altman analysis.

Main Results:

• Six of seven variables showed statistically significant differences between AI and manual methods.
• Automated measurements tended to overestimate skeletal and dental parameters.
• Intraclass correlation coefficients (ICC) were poor for clinically relevant variables (e.g., ANB, FMA), indicating low agreement.

Conclusions:

• WebCeph AI analysis lacks clinically acceptable agreement with expert manual tracing for key cephalometric variables.
• Discrepancies in sagittal and vertical diagnostic parameters may impact clinical interpretation and treatment planning.
• AI cephalometric analysis is best utilized as an adjunctive tool, not a replacement for clinician evaluation.