Tooth position prediction method based on adaptive geometry optimization.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Current tooth position prediction methods struggle with high-precision medical data access.
• Existing approaches have limitations in capturing hierarchical features and spatial relationships among teeth.

Purpose of the Study:

• To develop an advanced Transformer-based method for multi-layer feature optimization in tooth position prediction.
• To overcome challenges related to data acquisition and feature representation in dental AI.

Main Methods:

• Implemented geometric and physiological adaptive optimization strategies for pose invariance.
• Designed a hierarchical feature tooth position prediction network utilizing Transformer architecture.
• Constructed a jointly supervised loss function to manage tooth spatial relationships and prevent misalignments.

Main Results:

• Achieved a 2.87% improvement in prediction accuracy compared to existing methods.
• Reduced rotation and translation errors by 28.28% and 37.53%, respectively.
• Demonstrated effective supervision of tooth spatial structure and prevention of collisions.

Conclusions:

• The proposed multi-layer feature optimization Transformer-based method significantly enhances tooth position prediction accuracy and robustness.
• The adaptive strategies and hierarchical network effectively address limitations in current dental AI approaches.
• The jointly supervised loss function ensures precise spatial relationships and minimizes alignment errors.