Constrained Multi-Shape Evolution for Overlapping Cytoplasm Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for segmenting overlapping cytoplasm in medical images face challenges due to insufficient intensity information, reliance on finite shape hypotheses, and lack of consideration for shape relationships between cells, leading to inaccurate and inconsistent segmentation results.
Innovation Solution
A method and system that establish an infinite shape hypothesis set, perform constrained multi-shape evolution, and learn the importance of shape instances to accurately segment overlapping cytoplasm by combining local and overall prior shapes with intensity information, ensuring the final shape is constrained within the hypothesis set.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional segmentation methods (threshold, watershed, image segmentation) are used, then the segmentation process is simple, but the segmentation accuracy is poor because intensity information in overlapping areas is confusing and misleading
Solution Approach 1:
The patent introduces shape prior information as an intermediary element to mediate between the confusing intensity information and the segmentation result. By incorporating shape constraints (elliptical or star-shaped priors) into the segmentation process, the method guides the segmentation to produce anatomically plausible results even when intensity information is ambiguous or misleading in overlapping regions
Solution Approach 2:
The patent changes the parameters used for segmentation by adding shape-based constraints to the traditional intensity-based approaches. The energy function is modified to include shape prior terms, transforming the segmentation from purely intensity-driven to a combined intensity-and-shape-driven process, thereby improving accuracy in overlapping regions
2Device complexity
If finite shape hypothesis methods are used to model prior shapes, then the segmentation process is constrained, but the segmentation accuracy is insufficient because specific shape hypotheses cannot well restore the occluded boundary part of the cytoplasm
Solution Approach 1:
The patent employs dynamic shape evolution through level set methods, allowing the shape to adapt and evolve during the segmentation process rather than being fixed to pre-defined hypotheses. The shape prior is dynamically adjusted to fit the actual cytoplasm boundaries while maintaining the constraint of plausible shapes (elliptical or star-shaped), enabling accurate restoration of occluded boundaries
Solution Approach 2:
The patent transforms the static finite shape hypotheses into dynamic shape evolutions by incorporating level set equations. This allows the shape parameters to change continuously during segmentation, adapting to the actual data while maintaining the prior shape constraints, thus resolving the contradiction between constraint and accuracy
3Productivity
If local prior shape methods are used to evolve cytoplasm shape, then the segmentation is computationally efficient, but the segmentation results are inconsistent with clump evidence because the shape relationship between all cytoplasm and clumps is not considered
Solution Approach 1:
The patent merges local prior shape information with global clump-level evidence by formulating a joint energy function that incorporates both cytoplasm-specific shape priors and overall clump boundary constraints. This combination ensures that individual cytoplasm segmentations are consistent with the global clump structure while maintaining computational efficiency through the level set framework
4Ease of operation
If intensity information is relied upon for segmentation, then the segmentation process is straightforward, but the segmentation results are incredible when intensity evidence contradicts the local prior shape
Solution Approach 1:
The patent introduces shape prior information as an intermediary that mediates between intensity evidence and segmentation results. When intensity information is ambiguous or contradictory, the shape prior acts as a guiding constraint to produce anatomically plausible results, preventing incredible segmentation outcomes while maintaining the simplicity of intensity-based approaches where applicable
Data Source
AI summary
The present invention relates to a method for segmenting overlapping cytoplasm in a medical image, including: establishing a cytoplasm shape hypothesis set (201); and selecting a shape hypothesis for each cytoplasm from the established cytoplasm shape hypothesis set to perform constrained multi-shape evolution (202), thereby segmenting overlapping cytoplasm in the medical image, wherein the constrained multi-shape evolution (202) includes: segmenting a clump area composed of a plurality of overlapping cytoplasm to provide clump evidence (301); performing shape alignment to assess quality of the selected shape hypotheses (302); and performing shape evolution to determine a better shape hypothesis for each cytoplasm (3). The present invention further relates to a system for segmenting overlapping cytoplasm in a medical image. The method and system of the present invention can perform shape prior-based overlapping cytoplasm segmentation more accurately and more effectively.