Adaptive Hit-or-Miss Region Growing for Vessel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical imaging techniques face challenges in segmenting complex vessel structures in 3D images due to their intricate shapes and varying diameters, leading to incomplete segmentation and sensitivity to noise.
Innovation Solution
An adaptive Hit-or-Miss Region Growing method that iteratively adjusts template sizes and uses metric-sized structuring elements, combined with low-pass filtering and seed point optimization, to robustly segment vessels with varying diameters and improve noise resistance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If fixed-radius structuring elements are used in HMT for region-growing segmentation, then the method is simple and fast, but it can only segment vessels with diameters matching the fixed template size, leading to incomplete segmentation
Solution Approach 1:
The patent applies dynamics by making the structuring element radius adaptive rather than fixed. The radius is dynamically adjusted based on the local vessel diameter detected during segmentation, allowing the algorithm to accommodate vessels of varying sizes while maintaining computational efficiency. This resolves the contradiction between speed and adaptability.
Solution Approach 2:
The patent changes the parameter of structuring element radius from a fixed value to a variable parameter that adapts to local vessel characteristics. By adjusting the radius parameter dynamically during the segmentation process, the method achieves both speed and versatility in segmenting vessels with different diameters.
2Productivity
If region-growing segmentation is performed with fixed structuring elements, then the algorithm is computationally efficient, but it is sensitive to noise and may incorrectly segment structures with similar intensities
Solution Approach 1:
The patent applies preliminary action by performing noise filtering before the region-growing segmentation process. A morphological opening operation is applied to remove noise and spurious structures prior to segmentation, thereby improving reliability without significantly impacting computational efficiency.
Solution Approach 2:
The patent implements feedback by using the detected vessel diameter at each seed point to adjust the structuring element radius for subsequent segmentation steps. This feedback mechanism allows the algorithm to adapt to local characteristics and maintain reliability while preserving computational efficiency.
3Extent of automation
If automatic seed point detection is implemented using HMT, then automation is improved, but it requires strong a-priori morphological knowledge and only works for restricted places
Solution Approach 1:
The patent applies universality by creating a seed point detection method that works for various vessel types and locations without requiring strong a-priori morphological knowledge. The method uses adaptive structuring elements that can detect vessels across different diameters and configurations, making it universally applicable to different anatomical structures.
Solution Approach 2:
The patent changes the approach to seed detection by using adaptive radius adjustment instead of fixed morphological operations. This parameter change allows the algorithm to detect seed points in diverse locations and for different vessel types without requiring specialized a-priori knowledge for each case.
4Adaptability or versatility
If the structuring element radius is increased to match larger vessels, then larger vessels can be segmented, but smaller vessels with different diameters cannot be properly segmented
Solution Approach 1:
The patent applies local quality by adjusting the structuring element radius according to the local vessel diameter at each seed point. Instead of using a uniform radius throughout the image, the method adapts the radius locally to match the specific vessel size being segmented, thereby achieving both versatility and precision.
Solution Approach 2:
The patent makes the structuring element radius dynamic and adaptive to local conditions. The radius is adjusted during the segmentation process based on the detected vessel diameter, allowing the system to accommodate vessels of various sizes while maintaining segmentation accuracy for each specific case.
Data Source
Figure 1~2
Figure 3(a)~4(d)
Figure 5(a)~6
AI summary
The present invention concerns a method for the segmentation of vessel trees in 3D medical images, comprising steps of: - defining an initial seed corresponding to a first identified point, said seed being a point of the 3D image corresponding to a vessel, - performing region-growing segmentation by iteratively (i) identifying points in the neighborhood of previously identified points which belongs to vessels by using an identification criterion based on the application of a grey-level Hit-or-Miss Transform (HMT) with a set of structuring elements of a template size, and (ii) adding the identified points to the segmented vessel tree. The region-growing segmentation is repeated iteratively for different template sizes. The present invention concerns also a computer program implementing the method.