Personalized design technique for the dental occlusal surface based on conditional generative adversarial networks

Area of Science:

• Dental Restoration
• Artificial Intelligence
• Computer-Aided Design

Background:

• Tooth defects are common, with traditional restoration being time-consuming.
• Current dental computer-aided design and manufacture (CAD/CAM) systems struggle to replicate natural tooth anatomy.
• Existing methods lack personalized anatomical feature restoration for defective teeth.

Purpose of the Study:

• To propose an intelligent network model for designing personalized tooth crown surfaces.
• To overcome the limitations of traditional and CAD/CAM dental restoration methods.
• To improve the accuracy and efficiency of restoring defective teeth.

Main Methods:

• Developed a conditional generative adversarial network (GAN) model for tooth crown surface design.
• Created a training dataset using depth maps from 3D intraoral scans of teeth.
• Employed adversarial training with constraints on occlusal relationships, perceptual loss, and occlusal groove filtering.

Main Results:

• The intelligent network model successfully generated personalized tooth occlusal surfaces.
• The generated surfaces accurately reconstructed natural anatomical features.
• Assessment experiments confirmed the quality of the occlusal surface and its relationship with opposing teeth.

Conclusions:

• The proposed intelligent network model effectively reconstructs personalized anatomical features for defective teeth.
• This AI-driven approach significantly shortens dental treatment time.
• The method restores full functionality to defective teeth with improved accuracy and personalization.