TEANet: Automated tooth extraction and arrangement with tooth-level graph spatial transformation network.

Area of Science:

• Orthodontics
• Artificial Intelligence in Dentistry
• 3D Medical Imaging

Background:

• Current automated orthodontic treatment planning methods struggle with cases requiring tooth extraction.
• Existing techniques using point clouds and Graph Convolutional Networks (GCN) are limited in accuracy for complex extraction scenarios.

Purpose of the Study:

• To develop an automated orthodontic treatment planning method, TEANet, capable of handling tooth extraction and arrangement.
• To improve accuracy and efficiency in orthodontic planning for complex dentition cases.

Main Methods:

• Proposed Tooth Extraction & Arrangement Network (TEANet) featuring a double-coordinate classifier for extraction and an adaptive graph for arrangement.
• Utilized a node-erasable adaptive graph for feature propagation in GCN and a learning-based 3D spatial transformer model.

Main Results:

• TEANet demonstrated superior performance on three orthodontic tasks: tooth classification, extraction classification, and tooth arrangement regression.
• Evaluated on 304 clinical cases, the method consistently outperformed existing state-of-the-art approaches.

Conclusions:

• TEANet provides a novel and robust framework for orthodontic treatment planning, especially for extraction cases.
• The method's innovative components, including the adaptive graph and spatial transformer, offer an efficient solution for complex orthodontic cases.