Medical Image Segmentation Preserving Planar Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image data processing methods, particularly non-rigid registration, often fail to preserve planar boundaries correctly, leading to unreliable segmentation of brain regions like M1-M6, which undermines clinician confidence and accuracy in scoring systems such as ASPECTS.
Innovation Solution
A medical image data processing apparatus and method that refines segmentation by constraining it based on anatomically defined planes, ensuring straight or flat boundaries are preserved during non-rigid registration, using techniques like least squares fitting and decomposition of registration transforms to maintain expected geometric properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If non-rigid registration is used to segment brain images, then the registration flexibility and adaptability are improved, but the planar boundary preservation deteriorates
Solution Approach 1:
The registration transform is decomposed into multiple components: a rigid transform, an affine transform, and a non-rigid transform. This segmentation allows each component to handle specific aspects of the registration, with the affine component preserving planar boundaries while the non-rigid component provides flexibility for anatomical adaptation.
Solution Approach 2:
Different regions of the image undergo different types of transformation. Planar regions (such as the boundaries between M1-M3 and M4-M6) are transformed using the affine component to preserve their planarity, while non-planar regions are allowed to undergo full non-rigid transformation for better anatomical matching.
2Measurement precision
If non-rigid registration is applied, then the anatomical matching accuracy is improved, but the geometric property preservation deteriorates
Solution Approach 1:
The transform is segmented into components with different geometric preservation properties. The affine component preserves parallelism and ratios of lengths, while the non-rigid component allows local deformations for anatomical matching. By applying them in sequence, both geometric preservation and anatomical accuracy are achieved.
Solution Approach 2:
The rigid and affine transformations are applied first to establish the overall geometric framework and preserve planar boundaries. Then the non-rigid transformation is applied as a refinement step to improve anatomical matching without disrupting the previously established geometric properties.
3Adaptability or versatility
If planar boundaries are not preserved in registration, then the registration adaptability is improved, but the clinician confidence deteriorates
Solution Approach 1:
By segmenting the transform into components, the system can selectively preserve planar boundaries in regions where clinicians expect them (such as M1-M3 and M4-M6 separations) while allowing adaptability in other regions. This selective preservation maintains clinician confidence without sacrificing overall registration adaptability.
Data Source
AI summary
A medical image data processing apparatus comprising processing circuitry configured to:receive medical image data;segment a body part included in the medical image data into multiple regions;refine or constrain the segmentation based on at least one plane to obtain a segmentation that includes at least one boundary or other feature having a desired property.