Intraoral Scan Registration Using Tissue Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dental scanning technologies face challenges in accurately registering intraoral measurements due to deformation of soft tissues, which introduces errors and interruptions, particularly when relying on rigid registration processes that fail to account for non-rigid parts.
Innovation Solution
A deep learning-based system that semantically segments intraoral measurements, identifying rigid and non-rigid parts, and applies corrective measures to ensure accurate registration by weighing the contributions of hard and soft tissues differently based on their proportion in the image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If feature based techniques such as Fast Point Feature Histograms are used to compute transformations for registering scans, then registration can be performed without prior knowledge of relative orientation, but regions that are deformable must be avoided which limits the scanning area
Solution Approach 1:
The patent segments the intraoral scan data into different tissue types (hard tissue, soft tissue, air) using machine learning classification. This segmentation allows the system to selectively use only rigid hard tissue regions for registration computations, while excluding deformable soft tissue regions. The segmentation enables comprehensive scanning of all oral regions while maintaining registration accuracy by using only suitable regions for the registration process.
2Productivity
If conventional registration algorithms are used, then processing speed may be maintained, but registration errors increase due to soft tissue deformation
Solution Approach 1:
The patent performs preliminary classification and identification of tissue types before the registration process. By pre-segmenting the scan data into hard tissue, soft tissue, and air regions using machine learning, the system prepares the data in advance so that only appropriate hard tissue regions are used for registration. This preliminary action eliminates the need for complex error correction during registration, maintaining processing speed while significantly improving registration accuracy.
3Measurement precision
If deep learning methods are used to identify rigid and non-rigid parts, then registration accuracy improves, but computational complexity increases
Solution Approach 1:
The patent replaces complex mechanical registration processes with machine learning-based classification. Instead of using traditional iterative alignment algorithms that require complex computational geometry operations, the system uses trained neural networks to automatically classify tissue types and identify rigid regions. This substitution simplifies the overall system architecture while improving accuracy, as the machine learning model performs the complex differentiation task more efficiently than traditional geometric methods.
Data Source
Figure 1(i)~1(iv)
Figure 2A~2B
Figure 3A
AI summary
A method, system and computer readable storage media for segmenting individual intra-oral measurements and registering said individual intraoral measurements to eliminate or reduce registration errors. An operator may use a dental camera to scan teeth and a trained deep neural network may automatically detect portions of the input images that can cause registration errors and reduce or eliminate the effect of these sources of registration errors.