A Dual Discriminator Adversarial Learning Approach for Dental Occlusal Surface Reconstruction.

Area of Science:

• Biomedical Engineering
• Computer Science
• Dental Restoration

Background:

• Restoring masticatory function in partially edentulous patients is complex due to individual tooth morphology.
• Existing deep learning methods for dental restorations often neglect crucial dental biological characteristics in occlusal surface reconstruction.

Purpose of the Study:

• To propose a novel dual discriminator adversarial learning network for accurate occlusal surface reconstruction in dental restorations.
• To address the limitations of current methods by integrating dental biological characteristics into the reconstruction process.

Main Methods:

• A novel dual discriminator adversarial learning network integrating a dilated convolutional generative model and a dual global-local discriminative model.
• The generative model uses dilated convolutions for feature representation preserving tissue structure.
• The dual discriminative model employs two discriminators for real/fake assessment, focusing on global coherence and local consistency.

Main Results:

• Experiments on 1000 real-world patient dental samples validated the method's effectiveness.
• Quantitative analysis showed a root mean square error of 0.114 mm between generated and target crowns.
• Qualitative analysis confirmed the generation of more reasonable dental biological morphology.

Conclusions:

• The proposed method significantly outperforms state-of-the-art techniques in occlusal surface reconstruction.
• The generated occlusal surfaces possess natural tooth anatomical morphology and high clinical application value.