Related Experiment Video
Updated: Jun 16, 2025
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Accuracy and time efficiency of deep learning-based three-dimensional crown segmentation on intraoral scanning: a
Mengfei Xu1, Xinlin Li2, Lingyun Wang1
1Department of Orthodontics, School and Hospital of Stomatology, Cheeloo College of Medicine, Shandong University & Shandong Key Laboratory of Oral Tissue Regeneration & Shandong Engineering Research Center of Dental Materials and Oral Tissue Regeneration & Shandong Provincial Clinical Research Center for Oral Diseases No.44-1 Wenhua Road West 250012, Jinan, Shandong, PR China.
Objectives:
Clinically, accurate tooth segmentation in intraoral scans is vital for clinical diagnosis and treatments. Deep learning is utilized for image segmentation in mesh segmentation such as intraoral scans lately. This review and meta-analysis aimed to present evidence regarding the accuracy of deep learning-based tooth segmentation in intraoral scanning and assess its time-effectiveness.
Data:
The quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2.
Sources:
Search for relevant studies was conducted on 8 February 2025 in PubMed, Web of Science, IEEE Xplore, Scopus, and ScienceDirect.
Study Selection:
Among the 1220 studies, 44 satisfied the inclusion criteria in the systematic review. Heterogeneity was primarily in data selection and index test. No significant publication bias was found based on the funnel plot and Egger analysis. The overall effect for mean intersection over union was 90.15 %, with a 95 % CI of 87.77-92.53 %. Subgroup analysis showed better segmentation results for canines and premolars. Deep learning-based segmentation significantly decreases segmentation time compared to manual methods.
Conclusions:
Deep learning methods for segmenting intraoral scanning models outperform manual human segmentation in speed and efficiency, while reducing reliance on subjective judgments. Although deep learning can currently meet basic clinical needs for tooth segmentation, future research should focus on developing and sharing public datasets and optimizing algorithms to achieve precise gingival boundary segmentation.
Clinical Significance:
Deep learning advances are pivotal for fully automating tooth segmentation, significantly reducing manual labor costs, and ensuring consistent segmentation standards through algorithmic outputs to a certain extent.
