Progressive Orthodontic Motion Planning based on Hierarchical Diffusion Transformer.

Area of Science:

• Biomedical Engineering
• Computer Science
• Dental Medicine

Background:

• Orthodontic motion planning is vital for digital orthodontics, aiding treatment plan formulation.
• Current methods require both initial and target tooth alignments to generate motion sequences.
• A practical limitation is the unavailability of target tooth alignment at the outset of planning.

Purpose of the Study:

• To introduce OrthoDiff, a novel target-free framework for orthodontic motion planning.
• To enable prediction of complete tooth motion sequences using solely the initial tooth alignment.
• To address the limitations of existing methods in predicting tooth motion without a defined target.

Main Methods:

• Developed OrthoDiff, a target-free framework employing a progressive generation strategy.
• Decomposed tooth motion sequences into multi-level motions for progressive inference space constraint.
• Utilized a hierarchical diffusion transformer backbone, treating tooth alignment as sequential tokens and incorporating dental model topology.

Main Results:

• OrthoDiff significantly outperforms state-of-the-art methods in target-free tooth motion generation.
• Ablation studies validated the effectiveness of the framework's core components.
• Achieved state-of-the-art results in tooth target alignment prediction as a secondary outcome.

Conclusions:

• OrthoDiff offers a robust solution for orthodontic motion planning using only initial alignment.
• The progressive generation strategy and hierarchical diffusion transformer are key to the framework's success.
• The framework demonstrates potential for advancing digital orthodontics and treatment planning efficiency.