DCPR-GAN: Dental Crown Prosthesis Restoration Using Two-Stage Generative Adversarial Networks

Area of Science:

• Biomedical Engineering
• Computer Science
• Dental Prosthodontics

Background:

• Restoring masticatory function with dental crowns is crucial but challenging due to complex occlusal surface morphology.
• Existing methods struggle with personalized and functional reconstruction of dental crown surfaces.

Purpose of the Study:

• To develop a data-driven, two-stage generative adversarial network (GAN) for accurate and functional dental crown surface reconstruction.
• To improve the restoration of occlusal relationships and functional characteristics in dental prostheses.

Main Methods:

• A two-stage conditional generative adversarial network (CGAN) approach was designed.
• The first stage learns the relationship between defective teeth and target crowns for occlusal relationship restoration.
• The second stage incorporates an occlusal groove parsing network (GroNet) and occlusal fingerprint constraint for enhanced functional characteristics.

Main Results:

• The proposed GAN framework significantly outperforms existing deep learning methods in functional occlusal surface reconstruction.
• Achieved standard deviation (SD) and root mean square (RMS) errors below 0.161 mm between generated and target surfaces.
• Generated dental crowns exhibit sufficient anatomical morphology and high clinical applicability.

Conclusions:

• The novel two-stage GAN effectively reconstructs functional and anatomically accurate dental crown surfaces.
• This approach offers a significant advancement in data-driven dental prosthodontics.
• The method holds promise for improving dental crown rehabilitation and patient masticatory function.