Personalized dental crown design: A point-to-mesh completion network.

Area of Science:

• Biomedical Engineering
• Computer Science
• Dental Technology

Background:

• Designing dental crowns using traditional computer-aided design (CAD) software is intricate and demands substantial laboratory time.
• Existing methods for automated crown generation often lack personalization and precision.

Purpose of the Study:

• To develop and validate an end-to-end deep learning model for automatic generation of personalized dental crown meshes.
• To enhance the efficiency and accuracy of dental crown design in digital dentistry.

Main Methods:

• A novel deep learning architecture was created, incorporating a feature extractor and a transformer-based model to predict crown geometry.
• A point-to-mesh module and differentiable Poisson surface reconstruction were employed to generate accurate crown meshes.
• Training utilized a customized margin line loss, contrastive Chamfer distance loss, and mean square error (MSE) loss for mesh quality control.

Main Results:

• The developed model demonstrated superior performance compared to the previous Dental Mesh Completion (DMC) method.
• Achieved a 12.32% reduction in Chamfer distance and a 46.43% reduction in MSE compared to DMC.
• The margin line loss specifically improved Chamfer distance by 5.59%.

Conclusions:

• The proposed deep learning model effectively automates personalized dental crown mesh generation with high accuracy.
• This approach offers a significant advancement over existing methods, streamlining the digital dentistry workflow.
• The model's components and loss functions contribute to improved geometric accuracy and margin line definition.