MADCrowner: Margin Aware Dental Crown design with template deformation and refinement

Area of Science:

• Biomedical Engineering
• Computer-Aided Design
• Digital Dentistry

Background:

• Dental crown restoration is crucial for tooth defects, with computer-aided design (CAD) enhancing efficiency but still requiring manual adjustments.
• Existing learning-based methods for automated dental crown generation face challenges like poor spatial resolution and surface reconstruction issues.

Purpose of the Study:

• To introduce MADCrowner, a novel margin-aware mesh generation framework for automated dental crown design.
• To address limitations of existing methods, including inadequate spatial resolution, noisy outputs, and overextension of surface reconstruction.

Main Methods:

• Developed CrownDeformR, a deformable template model guided by anatomical context from a multi-scale intraoral scan encoder.
• Introduced CrownSegger, a segmentation network for precise cervical margin extraction, used as a constraint and boundary condition.
• Implemented a tailored post-processing method to refine surface reconstruction and remove overextended areas.

Main Results:

• The proposed MADCrowner framework significantly improved geometric accuracy and clinical feasibility compared to existing approaches.
• Integrating cervical margin information enhanced CrownDeformR's performance and enabled effective post-processing.
• Extensive experiments on a large-scale intraoral scan dataset validated the method's effectiveness.

Conclusions:

• MADCrowner offers a robust solution for automated dental crown design, enhancing precision and reducing manual intervention.
• The margin-aware approach and refined reconstruction techniques represent a significant advancement in digital dentistry.
• The framework's performance suggests strong potential for clinical adoption in restorative dental workflows.