Anatomical Atlas Refinement via Region-Specific Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging technologies face challenges in accurately determining and improving anatomical atlas elements, particularly in identifying and refining anatomical structures within images, due to limitations in image matching and transformation processes.
Innovation Solution
A computer-implemented method that acquires model and patient image data, performs elastic fusion to determine transformations, and iteratively refines atlas element data by superimposing patient images within defined regions, using both automatic and manual processes to enhance anatomical atlas accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If elastic fusion is performed to match patient image to model image, then transformation accuracy is improved, but matching precision within the region deteriorates due to homogeneous grey value assignment
Solution Approach 1:
The image space is segmented into two distinct regions: a first region with homogeneous grey values used for obtaining transformations, and a second region with original image values used for determining atlas elements. This segmentation allows the matching process to operate differently in different spatial zones, resolving the contradiction between needing transformations and needing precise matching.
Solution Approach 2:
Different quality characteristics are applied to different regions: the first region uses simplified homogeneous grey values optimized for transformation calculation, while the second region preserves original image complexity optimized for atlas element identification. This local differentiation resolves the contradiction by allowing each region to serve its specific purpose without interference.
2Measurement precision
If multiple patient images are superimposed to determine atlas element, then atlas accuracy is improved, but processing time increases
Solution Approach 1:
Transformations are determined in advance through elastic fusion before the atlas element determination process. This preliminary action allows subsequent superimposition of multiple patient images to proceed efficiently using pre-computed transformation data, reducing the overall processing time while maintaining high atlas accuracy.
Solution Approach 2:
Multiple patient images are transformed and superimposed to create an averaged atlas element representation. This copying and averaging process improves atlas accuracy by incorporating variability across multiple subjects while the use of pre-computed transformations keeps processing time manageable.
3Ease of manufacture
If region is assigned homogeneous grey value to obtain transformations, then transformation determination is simplified, but information about anatomical structure is lost
Solution Approach 1:
The image is segmented into a first region with homogeneous grey values for transformation determination and a second region with preserved original values for anatomical structure analysis. This segmentation ensures that information loss in the first region does not affect the second region, resolving the contradiction between simplification and information preservation.
Solution Approach 2:
The homogeneous grey value region acts as an intermediary that facilitates transformation determination without directly providing anatomical information. The transformations obtained from this intermediary are then applied to the original image regions to extract anatomical structures, thus mediating between simplification and information preservation.
Data Source
AI summary
Disclosed is a computer-implemented method of determining an atlas element. The method encompasses acquiring model image data that describes at least one fixed element (an anatomical element such as an anatomical body part, for example, a rib). Patient image data is acquired that describes an improvable element (an anatomical element such as an anatomical body part, for example heart). Region data is acquired, for example by assigning a homogeneous grey value to a region of the model image. The patient image is matched to the model image, wherein no matching is performed within the region. A transformation for mapping the improvable element into the region is determined based on matching. Several patient images are then mapped into the region and superimposed. An atlas element is determined based on the superimposed images. The method may be repeated using the determined anatomical atlas element as a constraint to detect further atlas elements.