Deformable Image Registration Using Region-Weighted Force Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deformable image registration techniques treat all image regions equally, which can be a disadvantage when the primary purpose is to match specific organs, as it equally weights all regions, potentially leading to inefficient alignment processes.
Innovation Solution
The proposed system uses a deformation engine with different logic components to compute and apply forces based on distinct regions within the images, allowing for time-dependent weighting and iterative refinement, where the identity and weighting of regions can change between iterations, using external and internal force models to align deformable images with reference images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all image regions are treated equally in deformation algorithms, then the registration process is simple to implement, but the alignment accuracy for specific organs of interest is reduced
Solution Approach 1:
The patent applies local quality by assigning different weights to different image regions based on their importance. Specifically, regions containing organs of interest are assigned higher weights while less important regions receive lower weights. This allows the deformation algorithm to focus computational effort on achieving accurate alignment for critical anatomical structures while maintaining a relatively simple overall algorithm framework.
2Measurement precision
If region weighting is adjusted iteratively during registration, then the registration accuracy for regions of interest is improved, but the computation time increases
Solution Approach 1:
The patent implements dynamics by making the region weights adaptive rather than static. The weights are updated iteratively during the registration process based on the current alignment quality and the importance of different regions. This dynamic adjustment allows the algorithm to progressively improve registration accuracy for regions of interest while controlling computation time through the iterative refinement approach.
Solution Approach 2:
The patent maintains continuity of useful action by performing registration iterations that continuously refine the alignment. Each iteration builds upon the previous one, progressively improving the registration accuracy. The process continues until convergence criteria are met, ensuring that the useful action of aligning images is maintained throughout the computation without unnecessary interruptions or restarts.
Data Source
AI summary
Disclosed are systems for and methods of registering (i.e., aligning) a deformable image with a reference image subject to a plurality of regions within the deformable and reference images. Different members of the plurality of regions may be used in different phases of a deformation algorithm and the identity of these regions may change between different iterations of the deformation algorithm. In some embodiments, most of an image is used for calculation of the internal force of the demons algorithm while a smaller subset of the image is used for calculating the opposing external force.