Medical Image Registration via Local ROI Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image registration methods in medical imaging lack precision, especially when dealing with low-contrast or moving structures, and require expert knowledge for modifying registration parameters.
Innovation Solution
An image processing device with a global registration module, a selection module, and a local registration module that uses a different parameter vector for refining registrations within a selected Region of Interest (ROI), based on automatic analysis of intensity ranges, image edges, entropy measurements, and anatomical position, allowing for improved registration accuracy with minimal user input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a global registration is performed using a given registration algorithm with a first parameter vector, then the overall alignment of images is achieved, but the registration precision in specific regions of interest is insufficient
Solution Approach 1:
The patent divides the image registration process into two segments: a global registration step that aligns the overall images using a first parameter vector, and a local registration step that refines specific regions of interest using a second parameter vector. This segmentation allows different parameter sets to be applied to different spatial scales, improving local precision without sacrificing global alignment.
Solution Approach 2:
The patent implements local quality by allowing different parameter vectors to be applied to different regions of the image. The second parameter vector is specifically tailored for regions of interest, enabling optimized registration precision for critical areas while maintaining efficient global registration with the first parameter vector.
2Measurement precision
If registration parameters are manually adjusted to improve local registration accuracy, then registration precision in specific regions is improved, but the ease of operation and user accessibility deteriorates due to requiring expert knowledge
Solution Approach 1:
The system performs self-service by automatically determining and applying the second parameter vector for local registration without requiring manual expert adjustment. The automated selection and application of region-specific parameters enables high-precision local registration while maintaining ease of operation for non-expert users.
Solution Approach 2:
The patent automatically changes registration parameters based on the specific region of interest. Different parameter vectors are selected and applied according to the characteristics of different image regions, enabling optimized local registration accuracy without requiring users to manually adjust complex parameters.
3Device complexity
If a single parameter vector is used for the entire image, then the device complexity is low, but the registration accuracy in regions with different characteristics deteriorates
Solution Approach 1:
The patent segments the parameter application into global and local levels. The first parameter vector is applied globally to maintain simplicity, while the second parameter vector is applied locally to regions of interest to improve accuracy. This dual-level segmentation balances parameter management simplicity with registration precision.
Solution Approach 2:
The patent implements local quality by allowing different parameter vectors to be applied to different regions. The second parameter vector is specifically optimized for regions of interest, enabling higher registration accuracy in critical areas while the first parameter vector maintains simple global parameter management.
Data Source
AI summary
The invention relates to a method and an image processing device (50) for the registration of two images (I1, I2) that may for example be provided by a CT scanner (10) and/or an MRI scanner (20). According to one embodiment of the invention, the images are first globally registered (GR) with a given registration algorithm using a first parameter vector (p). A user may then select a region of interest ROI, and a plurality of local registrations (LR1, . . . LRs, . . . LRn) are calculated for this ROI using the same registration algorithm but different parameter vectors (p, p, . . . p). The results of the local registrations (LR1, . . . LRs, . . . LRn) are displayed and the user can select the best local registration(s). In a final step, the selected local registration(s) (LRs) and the global registration (GR) may be merged. Additionally or alternatively, a parameter vector for a local registration in the ROI may be determined by an automatic analysis of the ROI.