Related Experiment Video
Updated: May 18, 2026
A Finite Element Approach for Locating the Center of Resistance of Maxillary Teeth
Published on: April 8, 2020
AI-driven CBCT segmentation and 3D modeling of the anterior surface of maxilla for computer-assisted surgery: a
Hei Yuet Lo1, Pui Hang Leung1, Yu-Xiong Su1
1Division of Oral and Maxillofacial Surgery, Faculty of Dentistry, University of Hong Kong, Prince Philip Dental Hospital, 34th Hospital Road, Hong Kong Special Administrative Region.
Aims:
The maxilla is frequently involved in virtual surgical planning (VSP), serving as a base for osteotomies and designing patient-specific devices. However, segmenting thin bone structures like the anterior surface of the maxilla is challenging, often resulting in defects that compromise VSP. Our study aimed to compare various segmentation and 3D modeling algorithms for the anterior surface of the maxilla in CBCT, to serve as a reference for clinical practice.
Materials And Methods:
The study included 20 patients preparing for orthognathic joint surgery. Various segmentation and 3D modeling algorithms were compared, including manual segmentation, threshold segmentation, 3D hole repairing, and AI segmentation, using Mimics Viewer and Blue Sky Plan software. The accuracy of each segmentation method was evaluated using the Dice Similarity Coefficient (DSC) and 95 % Hausdorff distance (HD95). Additionally, the clinical applicability of the 3D models was qualitatively evaluated using questionnaires focused on surface consistency, structural completeness, surface smoothness and noise, accuracy of anatomical features, and overall suitability for virtual surgical planning. Statistical analysis was performed using non-parametric tests.
Results:
For segmentation accuracy, threshold segmentation and 3D hole repairing achieved significantly higher DSC and HD95 compared with other algorithms. AI segmentation in Mimics Viewer achieved a DSC of 0.90 ± 0.03 and an HD95 of 0.72 ± 0.50 mm, less than 1 mm.With regard to the qualitative assessment of clinical applicability, the score for Mimics Viewer was 3 (IQR: 3-4; p < 0.001), which outperformed the other algorithms for accuracy of anatomical features. Blue Sky Plan had the lowest median DSC (0.78 ± 0.12) and highest HD95 (2.02 ± 0.80 mm).
Conclusion:
3D hole repairing using 3-matic gave the best performance in terms of both accuracy and quality assessment for the anterior surface of the maxilla. AI-driven segmentation using Mimics Viewer, designed specifically for craniomaxillofacial surgery, provides optimal performance and could be a valuable tool in VSP.
