ImTooth: Neural Implicit Tooth for Dental Augmented Reality.

Area of Science:

• Computer Vision
• Medical Imaging
• Augmented Reality

Background:

• Augmented Reality (AR) in medicine is a growing field, with dental AR showing significant potential.
• Existing dental AR solutions often require complex equipment, markers, and are not optimized for wearable devices like AR glasses.
• This limits the accessibility and practicality of AR in clinical dental settings.

Purpose of the Study:

• To introduce ImTooth, a novel neural-implicit model-driven dental AR system designed for AR glasses.
• To simplify existing dental AR solutions by integrating reconstruction and registration into a single network.
• To enable accurate 3D tooth reconstruction, registration, and interaction using a single wearable device.

Main Methods:

• Developed a scale-preserving, voxel-based neural implicit model trained on multi-view images of a textureless tooth model.
• Integrated reconstruction and registration within a single neural network, leveraging learned edge features and depth information.
• Utilized a Microsoft HoloLens 2 as the sole sensor and display device for real-time application.

Main Results:

• Achieved high-precision 3D tooth model reconstruction and accurate registration without auxiliary markers or complex scanners.
• Demonstrated robustness to challenging visual conditions, including weak, repeating, and inconsistent textures.
• Successfully integrated the system into dental procedures, showcasing its utility in applications like bracket placement guidance.

Conclusions:

• ImTooth offers a simplified, accurate, and cost-effective solution for dental AR applications on wearable devices.
• The system's ability to perform reconstruction and registration in real-time enhances the potential for AR integration in clinical dentistry.
• This work paves the way for more accessible and advanced AR-assisted dental diagnostics and therapeutics.