Related Experiment Video
Updated: May 5, 2026
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Artificial Intelligence-enabled Automatic Computed Tomography Segmentation for Craniomaxillofacial Surgery Using a
Songying Wu1, Pui Hang Leung1, Andy Wai Kan Yeung2
1From the Division of Oral and Maxillofacial Surgery, Faculty of Dentistry, The University of Hong Kong, Hong Kong.
Background:
Skull computed tomography (CT) segmentation is the cornerstone of computer-assisted craniomaxillofacial surgery. Clinical routine threshold-based segmentation is time-consuming and yields suboptimal results in critical anatomical regions. This study evaluated the performance of an artificial intelligence (AI)-enabled automatic segmentation of skull CT scans and investigated clinical factors affecting its performance.
Methods:
The segmentation outcomes of 44 preoperative skull CT scans from a surgery database by AI-enabled and clinical routine methods were evaluated using quantitative metrics and qualitative assessment. The impact of occlusal contact, metallic artifact, bone involvement, slice increment, and pixel size of the CT scan was analyzed.
Results:
The mean Dice coefficient (DICE) of AI-enabled segmentation of the upper skull was 92.19% ± 2.59%, comparable to the clinical routine segmentation at 91.72% ± 3.77% (P = 0.228). The mean DICE of AI-enabled segmentation of the mandible was 94.81% ± 3.24%, outperforming clinical routine segmentation at 91.77% ± 5.21% (P < 0.001). AI yielded superior segmentation at the anterior maxillary wall and the temporomandibular joint. The presence of occlusal contact adversely affected AI segmentation of the mandible. Smaller slice increments and pixel sizes were associated with improved AI accuracy, whereas metallic artifacts and bone involvement had no significant effect.
Conclusions:
AI yielded comparable accuracy to the clinical routine method for skull CT segmentation, with better performance in critical anatomical regions and elimination of metallic artifacts. This study served as an external validation cohort to support future application of this AI-enabled segmentation model in the workflow of computer-assisted craniomaxillofacial surgery.
