Vessel Segmentation in CT Angiography Using Boundary Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic segmentation algorithms often misidentify vessels as bone in CT angiography data, leading to occlusion of vessel regions, especially in areas with little intensity variation between vessels and bone, resulting in incomplete visualization of vascular structures.
Innovation Solution
An image data processing system that includes a boundary identification unit to select points near occluded vessel regions, perform further segmentation using level-set processes, and generate a revised representation of vessels by filling the bone domain and applying morphological filters to distinguish and restore missing vessel regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automatic bone segmentation algorithms are used to identify and remove bone from visualization, then bone regions can be effectively masked out, but vessels with intensity similar to bone are incorrectly identified as bone and occluded
Solution Approach 1:
The patent divides the segmentation process into multiple stages: initial automatic bone segmentation, followed by identification of potential vessel regions within bone masks, and final classification using trained classifiers. This multi-stage segmentation approach allows the system to handle the complexity of distinguishing vessels from bone by breaking down the problem into manageable steps, thereby maintaining automation while improving reliability.
Solution Approach 2:
The system employs feedback mechanisms where the initial segmentation results are evaluated, potential vessel regions are identified within bone masks, and classifiers are trained using these findings. The results feed back into refining the segmentation, creating an iterative process that continuously improves vessel identification accuracy while maintaining automatic operation.
2Productivity
If automatic segmentation algorithms are used to distinguish vessels from bone, then processing speed is maintained, but misidentification occurs in areas with little intensity variation
Solution Approach 1:
The patent changes multiple parameters simultaneously: it adjusts intensity thresholds to identify potential vessel regions within bone masks, modifies spatial constraints to define search regions, and transforms the classification problem by using trained classifiers that consider multiple features beyond simple intensity. These parameter changes enable the system to maintain processing speed while significantly improving distinction accuracy in challenging areas.
Solution Approach 2:
The system creates a composite approach by combining multiple methods: automatic intensity-based segmentation, morphological operations to define search regions, and machine learning classifiers. This composite methodology integrates the speed of automatic algorithms with the precision of trained models, achieving both productivity and measurement precision.
3Reliability
If manual adjustment is enabled for users to identify vessel regions, then segmentation accuracy improves, but processing time increases significantly
Solution Approach 1:
The system performs self-correction by automatically identifying potential vessel regions within bone masks and applying trained classifiers to distinguish vessels from bone. This self-service capability reduces or eliminates the need for manual user intervention, maintaining high segmentation accuracy while minimizing time loss. The algorithm autonomously handles the challenging cases that would otherwise require manual adjustment.
Data Source
AI summary
An image data processing apparatus comprising a image data processing unit for obtaining segmented image data segmented using a segmentation process and including a representation of a vessel, wherein a region of the vessel is missing from the representation, a boundary identification unit for identifying at least one point at or near a boundary of the missing region, wherein the image data processing unit is configured to perform a further segmentation process to identify the missing region of the vessel, using the location of the at least one identified point at or near the boundary of the missing region, and the image data processing unit is further configured to generate a revised representation of the vessel including the missing region of the vessel.