Related Experiment Video
Updated: May 17, 2026
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Evaluation of Arch Dimensions and Bolton Ratio in Manual and Artificial Intelligence Tooth Segmentation
Sara M Al-Mashhadany1, Dheaa Al-Groosh1, Ammar Sh Al-Ubaydi2
1Orthodontic Department, College of Dentistry, University of Baghdad, Baghdad, Iraq, uobaghdad.edu.iq.
Objective:
A 3D tooth model segmented with the use of deep learning (DL) method program (CephX) as well as MIMICS software (manual segmentation) will be evaluated for reliability in the presented work. In addition, the segmented model was compared with the intraoral scan (IOS)-generated 3D tooth model in terms of the Bolton ratio and arch dimensions.
Methods And Materials:
A total of 30 patients attending the College of Dentistry/University of Baghdad with records of IOSs and CBCT scans were included. CBCT has been transformed into a 3D digital tooth model segmented with the use of MIMICS software and an AI-based program (CephX), and the Bolton ratio and arch dimensions (length and width) were measured utilizing Geomagic Control X software. Statistical analyses, including the mean and standard deviation, were performed, and a paired t-test was used to assess the systematic bias between the three methods. Bland-Altman plots and intraclass correlation (ICC) analysis were used to assess the agreement between the three methods.
Results:
The means of CephX, MIMICS, and IOS of all analyses were mostly similar, and the difference between them was greatest in the anterior Bolton ratio of IOS and CephX images. The systematic bias demonstrated no significant difference (p value > 0.05) between CephX and MIMICS, but CephX versus IOS and MIMICS versus IOS demonstrated significant differences (p value < 0.05) in several measurements, such as Bolton ratios and arch length. Agreement using ICC revealed good to excellent reliability overall, but moderate agreement in overall Bolton between MIMICS-IOS and between CephX-IOS, anterior Bolton and interpremolar distance between CephX-MIMICS, and poor agreement in anterior Bolton between MIMICS-IOS and between CephX-IOS was found. Bland-Altman plots showed that CephX and MIMICS were consistent, implying minimal systematic bias.
Conclusions:
Digital and AI-driven tooth segmentation (MIMICS and CephX), and IOS methods generally provide consistent and reliable measurements in terms of the Bolton ratio and arch dimensions. However, caution is advised when interpreting certain Bolton ratio values because discrepancies and lower agreement may occur.
