Dental Prosthesis Generation via Blended Dataset and Dimensionality Change
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current deep learning-based methods for generating dental prostheses face challenges in achieving accurate reconstruction, surface roughness, and anatomical shape reproduction, particularly due to limitations in resolution, noise handling, and the separation of natural teeth and technician-designed datasets, which affect the quality and accuracy of generated dental restorations.
Innovation Solution
The proposed solution combines a blended dataset of natural teeth, technician-designed teeth, and clinically adjusted teeth with deep implicit representation (DIR) and Denoising Diffusion Probabilistic Model (DDPM) to enhance the generation algorithm, incorporating edge-preservation and noise-reducing filters for preprocessing, and uses a special loss function to emphasize anatomical features, along with digital geometry processing for post-processing to ensure accurate and smooth surfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If deep learning methods are used to reconstruct missing tooth surfaces from 3D intraoral scans represented as 2D depth images, then the automation of dental CAD is improved, but the reconstruction accuracy deteriorates due to down sampling and limited resolution
Solution Approach 1:
The patent transforms the 2D depth image representation back into a 3D point cloud representation, allowing the deep learning model to operate in three dimensions. This dimensional transformation enables the model to reconstruct missing tooth surfaces with higher accuracy while maintaining automation, as the 3D point cloud preserves spatial information that is lost in 2D down-sampled representations
Solution Approach 2:
The patent applies preprocessing steps including noise removal and completion of discontinuous steps before the deep learning reconstruction process. By preparing the input data in advance with these preliminary actions, the model can achieve better reconstruction accuracy without sacrificing automation, as the preprocessing enhances the quality of input data for the automated deep learning system
2Manufacturing precision
If higher resolution is used to improve reconstruction accuracy, then the manufacturing precision is improved, but the memory consumption increases which may render the program inoperable
Solution Approach 1:
The patent segments the tooth surface reconstruction into different regions of interest, focusing computational resources on critical areas such as occlusal surfaces and contact points. This segmentation allows the system to achieve high reconstruction accuracy in important regions while using lower resolution in less critical areas, thereby reducing overall memory consumption while maintaining operational feasibility
Solution Approach 2:
The patent applies different resolution levels to different parts of the tooth model, using high resolution only where anatomical features are most critical (such as grooves, pits, fossa and ridge areas) and lower resolution in other regions. This local quality approach ensures manufacturing precision where needed while minimizing memory consumption across the entire model
3Quantity of substance
If separate datasets for natural teeth and technician-designed prostheses are used, then the training data availability is improved, but the quality of generated restorations deteriorates due to lack of integration between anatomical accuracy and design expertise
Solution Approach 1:
The patent merges separate datasets for natural teeth and technician-designed prostheses into a unified training dataset. This combination allows the deep learning model to simultaneously learn from anatomical features of natural teeth and design principles from technician-created restorations, producing high-quality generated restorations that satisfy both anatomical accuracy and design expertise requirements
Data Source
AI summary
A computer-implemented geometric processing method generates the design for a model of a dental prosthesis beginning with obtaining a blended prosthesis dataset for dental prostheses including natural tooth data and prosthesis tooth data designed by a technician. This dataset is preprocessed by generating a depth map with preprocessing so that the data is more suitable for deep learning (DL) so as to ensure generation of a smooth surface. An artificial intelligence neural network generation model with tooth feature loss is trained on the preprocessed dataset and is used to form a model of the prosthesis. Then the dental information associated with the dental model of dentition is used to generate a 3D dental prosthesis surface with a post-processing method to meet the requirement of the dental prosthesis. Finally, the rest of the dental prosthesis is completed to meet full function.