Shell Image Registration for Dental Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current registration techniques in dental implantology, such as those using spin images, are sensitive to noise, scatter, and occlusion, and require manual intervention or feature extraction, leading to inaccuracies and increased computational time, especially when registering high-resolution dental models with noisy CBCT data.
Innovation Solution
The method employs shell images, calculated for points on 3D models using concentric spherically symmetric shells, which are less sensitive to noise and surface normal errors, allowing for automated registration by comparing normalized volume values and calculating similarity measures like Euclidean distance or linear correlation, reducing the need for resampling and improving registration accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If spin images are used for registration, then partial occlusion and clutter can be handled, but noise sensitivity and computational time increase
Solution Approach 1:
The patent segments the 3D surface into multiple local regions around selected points, creating a hierarchical representation where global registration is achieved through coordinated local transformations. This segmentation allows the method to focus on local geometric features that are less sensitive to noise while maintaining adaptability to partial occlusions through selective region processing.
Solution Approach 2:
The patent transforms the 3D surface registration problem into a 2D image matching problem by creating local 2D projections around selected points. This dimensionality reduction simplifies the registration task, making it more computationally efficient and less sensitive to noise while preserving the essential geometric relationships needed for accurate alignment.
2Measurement precision
If ICP algorithm is used with manual starting positions, then registration can be performed, but time consumption increases and automation is reduced
Solution Approach 1:
The patent performs preliminary actions by automatically selecting key points and generating initial local transformations before the final registration step. This preliminary processing establishes a good starting configuration that guides the optimization algorithm toward the correct solution more efficiently, reducing both computational time and the need for manual intervention.
Solution Approach 2:
The patent implements self-service by enabling the system to automatically select points, generate initial transformations, and perform registration without manual intervention. The algorithm autonomously identifies corresponding points and adjusts transformations to minimize distance errors, achieving both automation and accuracy simultaneously.
3Manufacturing precision
If high-resolution digital models are registered with CBCT data, then detailed alignment is achieved, but noise and scatter from metallic materials increase registration errors
Solution Approach 1:
The patent extracts and removes harmful metallic materials and their associated noise from the CBCT data through segmentation and filtering operations. By separating the metallic components from the bone and soft tissue structures, the method eliminates the primary source of noise and scatter while preserving the essential anatomical features needed for accurate registration.
Solution Approach 2:
The patent applies local quality by processing different regions of the data with different techniques. Metallic regions are handled through special filtering and removal, while bone and soft tissue regions undergo standard registration processing. This localized approach optimizes the handling of each material type's specific characteristics, improving overall registration accuracy despite the presence of noisy metallic materials.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Systems and methods, devices and software are described for use in image comparison such as required for medical planning such as in dental implantology or other applications that require registration of digital images. A method and system is described for automatically finding correspondences between two or more digital representations such as images of one or more 3D objects with an identical, partially identical or similar geometry. The method and system, devices and software have the advantage that although the different digital representations of the object may be influenced by either noise, or scatter, or occlusion, or clutter, or any combination of these, correspondence can be found.