Related Experiment Video
Updated: Jul 14, 2026
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Exploration of predominant error sources in intraoral scanning and cone-beam computed tomography semiautomated
Lei Chi1, Yueran Fu1, Yan Bao1
1State Key Laboratory of Oral Diseases, National Center for Stomatology, National Clinical Research Center for Oral Diseases, Department of Oral Implantology, West China Hospital of Stomatology, Sichuan University, Chengdu, China.
Purpose:
Registration accuracy is vital for digital workflows because misalignment increases errors. This study explores the effects of registration methods, cone-beam computed tomography (CBCT) image quality, and number of registration points (NRP).
Methods:
All participants underwent intraoral and CBCT scanning. In Section 1, registration included four groups: three points, two points and one surface, one point and two surfaces, and three surfaces. In Section 2, after grouping the registrations on the basis of image quality (high, medium, or low), clinicians performed calibration trials. In Section 3, the NRP was increased from two to seven points. Deviations were measured by evaluating the anatomical landmarks (including the cusp tips, central fossae, and ends of the developmental grooves).
Results:
Significant differences were observed in the registration methodologies (P < 0.001), with deviations increasing from 257.0 μm (group 3 points) to 479.3 μm (group 3 surfaces). Calibration precision improved proportionally with enhanced CBCT image quality. Deviations significantly decreased from 314.4 μm (low quality) to 237.6 μm (high quality; P < 0.001). Increasing the NRP from two to four points reduced the deviations from 375.4 to 228.2 μm; however, further increasing the NRP beyond four points yielded no significant improvement (P > 0.999, P > 0.999, P = 0.152).
Conclusions:
Clinicians should prioritize point-based registration to improve accuracy. High CBCT image quality significantly improves accuracy. Increasing the NRP beyond four points demonstrates no statistically significant reduction in deviation.
