Neural Network Crown Pose Estimation for 3D Dental Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining crown position in restorative dentistry are imprecise and prone to errors due to inaccuracies in segmentation and interpolation of 3D dental models, particularly when natural teeth neighbors are far from the crown position or when invalid inputs are used.
Innovation Solution
A computer-aided dental restoration system utilizing a trained neural network to estimate crown pose by encoding 3D model representations with surface normal data and dental notations, applying registration optimization to determine a local coordinate system for precise crown placement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If PCA model is used to improve interpolation accuracy, then crown position estimation is improved, but the system becomes sensitive to invalid inputs and segmentation errors
Solution Approach 1:
The patent replaces the PCA model (statistical mechanical approach) with a neural network model that processes encoded 3D dental models directly. The neural network learns from training data to predict crown positions without relying on statistical interpolation, thereby eliminating sensitivity to invalid inputs and segmentation errors while maintaining high accuracy.
Solution Approach 2:
The patent changes the input representation from raw 3D models to encoded representations that include surface normal data and dental notation information. This parameter transformation enables the neural network to process and interpret dental structures more effectively, improving both accuracy and robustness simultaneously.
2Productivity
If interpolation method is used to determine crown position, then processing speed is improved, but measurement precision deteriorates due to inaccuracies in interpolation
Solution Approach 1:
The patent substitutes the traditional interpolation method with a neural network-based prediction system. The neural network, trained on labeled dental data, directly predicts accurate crown positions without relying on geometric interpolation, thereby achieving both high processing speed and high measurement precision.
Solution Approach 2:
The patent performs preliminary encoding of the 3D dental model into a representation that includes surface normal data and spatial relationships before feeding it to the neural network. This pre-processing step prepares the data in a format that enables fast and accurate prediction, eliminating the need for slow and imprecise interpolation operations.
3Ease of operation
If segmentation and interpolation of 3D model is used, then crown position can be determined, but the process is time-consuming and error-prone
Solution Approach 1:
The patent replaces the manual and computationally intensive segmentation and interpolation process with an automated neural network system. The neural network directly processes 3D dental models to predict crown positions, eliminating time-consuming manual segmentation and interpolation steps while providing automated, accurate results.
Solution Approach 2:
The patent performs preliminary encoding of the 3D dental model into a compressed representation that captures essential geometric and topological features. This pre-encoding step reduces the computational complexity and processing time required for crown position prediction while maintaining accuracy.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
The present disclosure relates to computer-aided dental restoration system (102) that is configured to estimate crown pose and representing virtual crown in 3D model (110). The system obtains 3D model (110) of dentition of the patient. The system segments 3D model to obtain segmented tooth data and generates encoded 3D model representation suitable for processing by trained neural network (112). The method for generating encoded 3D model representation comprises subsampling point cloud (404) representation based on segmented tooth data, determining surface normal representation, retrieving dental notation of restorative site, encoding each of plurality of points with corresponding surface normal representation and with value relative to restorative site to produce encoded 3D model representation, inputting encoded 3D model representation into trained neural network, and producing output, using trained neural network, wherein output comprises prediction of translation of each point of encoded 3D model representation into positions corresponding to crown pose.