LETA: Tooth Alignment Prediction Based on Dual-branch Latent Encoding.

Area of Science:

• Biomedical Engineering
• Computer Science
• Orthodontics

Background:

• Accurate tooth positioning is crucial in orthodontics.
• Current manual methods for tooth alignment are time-consuming and inefficient.
• Digital dentistry requires automated solutions for complex dental tasks.

Purpose of the Study:

• To present LETA, a novel deep learning system for automated 3D tooth alignment.
• To develop an efficient and accurate method for predicting tooth pose transformations.
• To reduce the manual workload for orthodontists in digital dental workflows.

Main Methods:

• LETA utilizes a dual-branch Latent Encoding approach for 3D tooth alignment.
• The system processes segmented individual 3D tooth meshes from intraoral scanners (IOS).
• Key components include an Encoder for latent code learning, a Projector for transformation, and a Solver for pose estimation, guided by ground truth features.

Main Results:

• LETA achieves state-of-the-art performance on a large-scale dataset of 9,868 IOS surfaces.
• The system accurately predicts 3D pose transformations for individual teeth.
• A clinical study showed LETA reduces orthodontists' workload by over 60%.

Conclusions:

• LETA offers a highly effective automated solution for 3D tooth alignment.
• The deep learning approach demonstrates significant potential for advancing digital dentistry.
• This method promises to streamline orthodontic treatment planning and execution.