Automated Aorta Detection in CTA Volumes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current bone removal algorithms in medical imaging, such as computed tomography angiography (CTA), fail to perfectly detect vessel structures due to residual fragments, which obstruct diagnosis and require manual cleanup.
Innovation Solution
A computer-implemented method using iterative thresholding and watershed-based segmentation, combined with cluster analysis and classification, to differentiate between bone and vascular tissues in 2D and 3D images, effectively removing bone structures and detecting the aorta or largest vessels, employing a decision tree classifier and post-processing techniques to refine results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If bone removal algorithms are applied to segment and classify bone and vessel regions, then bone structures can be removed from the view, but residual bone fragments remain that obstruct vessel detection
Solution Approach 1:
The patent applies multi-threshold segmentation to divide the image into multiple regions based on density values. Instead of using a single threshold that leaves residual fragments, the method uses multiple thresholds (first threshold for bone, second threshold for vessels) to create distinct segmented regions, ensuring complete bone removal while preserving vessel structures.
Solution Approach 2:
The patent inverts the conventional approach by first segmenting bone regions and then using the complement of this segmentation to identify vessel regions. Instead of directly detecting vessels and subtracting bone, the method segments bone at multiple thresholds and defines vessels as the regions not classified as bone, thereby eliminating residual bone fragments from vessel detection.
2Manufacturing precision
If multiple thresholding operations are performed to improve segmentation accuracy, then vessel detection precision improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary clustering analysis on the first thresholded bone regions before applying the second threshold. By pre-processing and organizing bone regions into clusters with representative values, the method reduces the computational burden of subsequent operations and enables efficient multi-threshold processing without exponentially increasing complexity.
Solution Approach 2:
The patent dynamically adjusts threshold values based on clustered bone region characteristics. Instead of using fixed thresholds, the method determines thresholds from the actual data distribution through clustering, adapting the segmentation parameters to the specific image content. This reduces the need for excessive threshold iterations while maintaining high segmentation accuracy.
3Illumination intensity
If contrast agent is administered to increase radiodensity of vessels, then vessel visibility improves, but bone and vessel density values overlap making classification difficult
Solution Approach 1:
The patent uses multi-level segmentation with different threshold values to separate bone and vessel regions despite overlapping density values. The first threshold isolates high-density bone, while the second threshold captures vessels including contrast-enhanced regions. By creating multiple segmented layers, the method resolves the overlap problem that would prevent clear classification with a single threshold.
Solution Approach 2:
The patent employs dynamic threshold selection based on clustered bone region characteristics. Instead of using static thresholds, the method adapts threshold values according to the actual density distribution in each image, allowing flexible separation of bone and vessel regions even when their density values overlap due to contrast enhancement.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for detecting the main body vessels (e.g. the aorta) in a medical volume by refining the result of a bone removal algorithm.