Image Registration Using Fixed Topology Model Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image registration methods are inadequate for registering images of objects with complex properties, such as the heart, which can lead to reduced registration quality due to difficulties in modeling sliding surfaces and highly flexible structures like papillary muscles, as well as the absence of anatomical correspondence in the blood pool.
Innovation Solution
An image registration apparatus and method that adapts a model with a fixed topology to both images, allowing for reliable registration by determining corresponding image elements based on spatial positions, and includes an adaptation unit that deforms the outer surface structure and minimizes internal energy of the inner structure, enabling accurate alignment even in complex anatomical structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image registration methods are used on complex anatomical structures like the heart, then the registration process can be performed, but the registration quality deteriorates due to difficulties in modeling sliding surfaces and highly flexible structures
Solution Approach 1:
The patent segments the complex anatomical structure into a parametric model with fixed topology (e.g., heart model with defined chambers, walls, and surfaces) and the image data. This segmentation allows the system to handle complex structures by breaking them into manageable components with known topological relationships, thereby improving registration quality without overwhelming complexity.
Solution Approach 2:
The patent transforms the complex anatomical structure into a parametric representation where geometric parameters are optimized to match the image data while maintaining fixed topological constraints. This parameter transformation approach enables accurate modeling of flexible structures like papillary muscles and sliding surfaces by adjusting parameters within a constrained topological framework.
2Adaptability or versatility
If parametric spatial transformations are used to model complex anatomical structures, then flexibility in adapting to different shapes is improved, but reliability of corresponding element identification deteriorates due to topology changes
Solution Approach 1:
The patent establishes a parametric model with predetermined fixed topology before performing registration. This preliminary structuring of the model with known topological relationships allows for reliable identification of corresponding elements across different images, as the topological framework is established in advance and maintains consistency throughout the transformation process.
Solution Approach 2:
The patent separates topological parameters (which remain fixed) from geometric parameters (which are optimized). This separation allows the model to adapt to different shapes through geometric parameter optimization while maintaining reliable corresponding element identification through fixed topological relationships, effectively decoupling adaptability from reliability concerns.
3Measurement precision
If iterative warping methods are used to align volumetric images, then registration can be achieved, but the process becomes computationally intensive and time-consuming
Solution Approach 1:
The patent transforms the image registration problem into a parameter optimization problem where a limited set of parametric model variables are optimized to match the image data. This parameter-based approach reduces the computational complexity compared to voxel-by-voxel iterative warping, as it optimizes a smaller number of parameters while maintaining alignment accuracy through the constrained parametric model structure.
Solution Approach 2:
The patent segments the registration process into model adaptation (optimizing parametric model parameters to match image data) and corresponding element determination (using the adapted model to identify correspondences). This segmentation allows for more efficient computation by separating the optimization phase from the correspondence identification phase, reducing overall registration time while maintaining accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach improves registration quality by reliably aligning complex anatomical structures, such as the heart, by adapting the model to the images and determining corresponding elements based on spatial positions, facilitating accurate comparison and analysis of anatomical features across different images or modalities.
Implementation Method 1
The method includes extracting first and second corresponding surfaces from the respective first and second images, which surfaces delineate the same feature, such as a bone/tissue interface. This is done by extracting a stack of contours for each surface. The first surface is iteratively warped toward alignment with the second surface to arrive at a global translation vector and a set of residual surface distortion vectors.
Implementation Method 2
The first surface is iteratively warped toward alignment with the second surface to arrive at a global translation vector and a set of residual surface distortion vectors. Volume distortion vectors, which are determined by applying a weighting function to the residual surface distortion vectors, are used to indicate the locations in the second volumetric image of voxel centers whose interpolated intensities are to be moved to lattice points.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to an image registration apparatus for registering a first image and a second image with respect to each other. A model, which has a fixed topology, is adapted to the first image for generating a first adapted model and to the second image for generating a second adapted model,and corresponding image elements (40, 48, 49; 50, 58, 9) are determined in the first image and in the second image based on spatial positions of first image elements in the first image with respect to the first adapted model and spatial positions of second image elements in the second image with respect to the second adapted model. Since the model has a fixed topology, corresponding image elements can relatively reliably be found based on the adapted models, even if the first and second images show objects having complex properties like a heart, thereby improving the registration quality.