Angiography Image Registration Using Geometric Pattern Landmarks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital subtracted angiography techniques face challenges in registering images accurately due to patient movement artifacts, requiring manual selection of regions of interest and being computationally costly, especially when dealing with 3D projections on 2D images, which results in suboptimal image quality and increased processing time.
Innovation Solution
A method involving the use of regular geometrical patterns, specifically equilateral triangles, for image registration, where landmarks are defined at the corners of these patterns, and their robustness is tested; if insufficient, the patterns are subdivided into smaller triangles to improve registration precision and reduce computation time by interpolating shift vectors from these landmarks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If rigid registration with manual ROI selection is used, then registration is simple to implement, but registration precision deteriorates due to non-uniform patient movement and rotational motions
Solution Approach 1:
The patent divides the image into multiple overlapping zones, each processed independently with its own shift vector calculation. This segmentation allows different regions with different movement characteristics to be registered separately, improving overall precision while maintaining automated operation.
Solution Approach 2:
The patent extends rigid 2D translation registration to include rotational components by calculating shift vectors that account for both translational and rotational patient movements. This dimensional enhancement allows the system to compensate for complex motion patterns while remaining automated.
2Area of stationary object
If the region of interest size is increased to cover more area, then more structures are included in registration, but artifacts are created within the zone due to excessive surface area
Solution Approach 1:
The patent automatically divides the image into multiple smaller overlapping zones rather than using a single large ROI. Each zone is processed independently, preventing the artifacts that occur in large zones while collectively covering the entire image area through the overlapping design.
Solution Approach 2:
The patent uses overlapping zones where some regions are processed multiple times through the overlap. This excessive action ensures complete coverage and robust registration while maintaining small enough zone sizes to avoid artifacts, with the redundancy providing insurance against registration failures.
3Measurement precision
If manual ROI selection is performed for many images with multiple zones per image, then registration can be customized, but processing time increases significantly
Solution Approach 1:
The patent implements automated zone selection and shift vector calculation that performs registration without manual user input. The system automatically identifies appropriate zones and computes registration parameters, eliminating the time-consuming manual ROI selection process while maintaining registration accuracy.
Solution Approach 2:
The patent pre-defines multiple overlapping zones across the image before processing begins. This preliminary setup eliminates the need for manual zone selection during processing, allowing rapid automated registration while ensuring comprehensive coverage through the pre-planned overlapping zone structure.
4Ease of manufacture
If homogeneous zones are selected for registration, then computation is simplified, but registration fails due to unstable similarity measurement
Solution Approach 1:
The patent uses overlapping zones where registration is performed multiple times on different zones covering the same area. If one zone proves homogeneous and unreliable, the overlapping redundant zones provide alternative registration opportunities, ensuring reliability while maintaining computational efficiency through the use of simpler homogeneous zones.
Data Source
AI summary
In registration of images in digitized subtracted angiography, X-rays irradiate a patient's body. A first digitized mask image is acquired without an injection of the contrast agent into the body, then the contrast agent is injected into the body, and a second digitized contrast image is acquired. The mask image is divided with regular geometrical patterns and a registration is made of the mask image relative to the contrast image through the use of landmarks defined at the corners of the geometrical patterns. These two images are subtracted after registration, and the subtracted image is displayed. The robustness of the landmarks is furthermore tested, and if the robustness is insufficient, a subdivision is made of the regular geometrical patterns.