2D/3D Medical Image Registration Across Multiple Imaging Geometries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 2D/3D registration methods using a point-to-plane correspondence model are limited to a single imaging geometry, leading to information loss and accuracy issues when using multiple imaging geometries for tracking three-dimensional motion in image-guided interventional procedures.
Innovation Solution
A method that selects contour points in a 3D image for each 2D image acquired using different imaging geometries, determines their 2D positions based on initial registration parameters, and updates these parameters by solving equations or optimization problems to define the registration parameters, which describe the position and orientation of the region of interest relative to the imaging geometry, allowing for registration across multiple imaging geometries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a point-to-plane correspondence model is used for 2D/3D registration, then the registration can be achieved with a single 2D image, but information loss occurs and accuracy deteriorates when tracking three-dimensional motion
Solution Approach 1:
The patent transitions from using a single 2D image (two-dimensional information) to utilizing multiple 2D images acquired from different imaging geometries (multiple two-dimensional perspectives). This dimensional expansion provides comprehensive three-dimensional motion tracking capability by combining information from various viewing angles, thereby resolving the information loss inherent in single-view methods.
Solution Approach 2:
The patent creates a registration method that works universally across multiple imaging geometries. By formulating a correspondence model that can handle N different 2D images with different imaging geometries, the system achieves multi-functional capability for tracking three-dimensional motion from various perspectives, rather than being limited to a single imaging geometry.
2Measurement precision
If multiple imaging geometries are used for 2D/3D registration, then three-dimensional motion tracking accuracy is improved, but the complexity of the registration method increases
Solution Approach 1:
The patent extends the correspondence model by incorporating parameters for multiple imaging geometries. The model uses parameters including three-dimensional coordinates of contour points, two-dimensional coordinates in multiple images, and imaging geometry parameters (rotation and translation). By systematically organizing these parameters, the method handles multiple imaging geometries without excessive complexity.
Solution Approach 2:
The patent segments the registration problem into discrete correspondence relationships between contour points in 3D space and their projections in multiple 2D images. By establishing independent correspondence models for each imaging geometry and then integrating them, the complex problem of multi-geometry registration is divided into manageable segments that can be solved systematically.
3Ease of manufacture
If a single 2D image is used for registration, then the method is simple to implement, but accuracy deteriorates when patient movement occurs
Solution Approach 1:
The patent adds multiple imaging geometry dimensions to the registration system. By acquiring and processing multiple 2D images from different angles and positions, the system gains the ability to reliably track patient movement in three-dimensional space, overcoming the limitations of single-image methods when patient motion occurs.
Solution Approach 2:
The patent establishes correspondence relationships between 3D contour points and their 2D projections across multiple images, creating a feedback mechanism. The registration parameters are optimized to minimize differences between actual 2D image contours and projected 3D contours, providing continuous feedback that maintains registration accuracy even when patient movement occurs between imaging sessions.
Data Source
Figure 1
Figure 2
AI summary
Method for determining or updating a registration between a 3D image (1) of a region of interest (26) and a set of at least two 2D images (2, 3) of the same region of interest (26), wherein the 2D images (2, 3) are acquired using different imaging geometries, wherein the registration is described by a set of registration parameters, comprising the steps: - Selecting at least one respective contour point (4 - 7) in the 3D image (1) for each 2D image (2, 3), - Determining a 2D position (8 - 11) of a respective depiction of each contour point (4 - 7) in the respective 2D image (2, 3), - For each contour point (4 - 7) determining at least one condition for the registration parameters that needs to be fulfilled to map the respective contour point (4 - 7) to the respective 2D position (8 - 11) in the respective 2D image (2, 3), - Determining an updated set of registration parameters by solving a set of equations or an optimization problem that depends on the conditions.