Related Experiment Video
Updated: Jun 4, 2026
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Assessment of the accuracy of automated tooth segmentation using different orthodontic platforms and models with a
Serena Moschitto1, Maria Giovanna Pantalena1, Umberto Gulisano2
1Department of General Surgery and Medical-Surgical Specialties, University of Catania, Via Santa Sofia 78, Catania 95123, Italy.
Background/Objectives:
Tooth segmentation remains the most time-consuming task during model preparation for digital orthodontic setup. This study aimed to assess the accuracy of AI-based tooth segmentation tools of four widely used orthodontic platforms, namely Medit LINK (©Medit Corp, South Korea), DentOne (©DIORCO Co, South Korea), BlueSkyPlan (Blu Sky Bio, USA), and ArchForm, using models categorized by specific clinical characteristics.
Materials/Methods:
A total of 270 digital dental models were retrospectively selected and assigned to nine groups: aligned dentition, crowding, spacing, ectopia, agenesis, gingival recession, tooth wear, orthodontic brackets, and attachments. Automated labeling success and segmentation quality were evaluated for each software. Mesio-distal (MD) tooth widths were calculated using Meshmixer software (Autodesk, San Rafael, Calif) and compared with unsegmented models (ground truth). All data were statistically analyzed.
Results:
BlueSkyPlan consistently outperformed the other tested software in tooth labeling, surface segmentation, and mesio-distal measurements accuracy across all model groups (P < 0.05). Labeling errors were most frequent in models with ectopic teeth (≤12.2% success) and agenesis (≤53.1% success). Surface-specific errors increased with crowding (buccal/lingual errors), spacing (mesio-distal errors), gingival recession (buccal errors), and tooth wear (occlusal errors). All software overestimated MD widths compared with ground truth with the limits of agreement considered clinically relevant (>0.30 mm) in specific software-model combinations).
Limitations:
The software evaluated does not provide transparency regarding the AI technologies implemented.
Conclusions/Implications:
The accuracy of automatic segmentation of intra-oral scans is influenced by clinical and anatomical characteristics of the dental arches and still requires clinician supervision. BlueSkyPlan currently appears to provide the most reliable performance among the tested platforms.
