Related Experiment Video
Updated: Sep 14, 2025
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Automatic Point Cloud Patching of Intraoral Three-Dimensional Scanning Based on Deep Learning.
Qianhan Zheng1, Yimin Wang2, Mengqi Zhou1
1Stomatology Hospital, School of Stomatology, Zhejiang University School of Medicine, Clinical Research Center for Oral Diseases of Zhejiang Province, Key Laboratory of Oral Biomedical Research of Zhejiang Province, Cancer Center of Zhejiang University, Hangzhou, Zhejiang, China.
This study introduces a deep learning method to automatically restore missing data in intraoral scans (IOS). The AI model accurately reconstructs incomplete 3D point clouds, enhancing digital dentistry workflows.
Area of Science:
- Digital dentistry
- Artificial intelligence in healthcare
- 3D imaging and reconstruction
Background:
- Intraoral scanning (IOS) is crucial for digital dentistry but often suffers from data loss due to the complex oral environment.
- Incomplete 3D point clouds from IOS hinder the accuracy and efficiency of digital orthodontic workflows.
Purpose of the Study:
- To develop and evaluate a deep learning-based method for automatic restoration of missing regions in intraoral 3D point clouds.
- To improve the accuracy and efficiency of digital orthodontic workflows by addressing data loss in IOS.
Main Methods:
- A Point Fractal Network architecture was utilized for reconstructing incomplete IOS data.
- A dataset of 314 IOS scans (4162 teeth) was used, with simulated data loss (5-20%) for training and validation.
- Model performance was evaluated using Chamfer distance (CD) to quantify point cloud completion accuracy.
Main Results:
- The deep learning model demonstrated robust performance, achieving average CD values below 0.01 across various data loss levels.
- Visual assessment confirmed high geometric fidelity between completed and original 3D point clouds.
- The model processed each point cloud in approximately 0.5 seconds, enabling near real-time restoration.
Conclusions:
- The developed deep learning model accurately restores missing IOS data, significantly enhancing the precision and efficiency of digital dental workflows.
- The method's speed and accuracy support real-time clinical applications, reducing manual corrections and improving treatment outcomes.
- This AI-driven approach has the potential to minimize human error, increase dental restoration precision, and facilitate broader AI integration in clinical practice.
