Pulmonary Artery Segmentation via Bronchial Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting pulmonary arteries from MDCT data face challenges in distinguishing between arteries and veins, particularly due to leakage into pulmonary veins and inaccuracies in image quality, leading to unreliable vessel tree topology and misclassification.
Innovation Solution
A method that aligns bronchial and pulmonary vessel structures through spatial transformation, allowing for improved detection and separation of arterial segments by matching adjacent bronchial and vessel segments, and using this alignment to enhance the segmentation process, particularly by applying spatial transformations to adaptively align the bronchial and pulmonary artery trees.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If seed-point based region expansion methods are used to extract pulmonary arteries, then the extraction process can be automated, but the method suffers from leakage into pulmonary veins leading to misclassification
Solution Approach 1:
The patent segments the pulmonary vessel tree by dividing it into arterial and venous components based on their spatial relationships with the bronchial tree. Each vessel segment is classified by evaluating its positional and orientational adjacency to bronchial segments, creating distinct arterial and venous segments rather than treating the vessel tree as a unified structure.
Solution Approach 2:
The bronchial tree serves as an intermediary structure to mediate the classification of pulmonary vessels. By using the bronchial tree's known anatomy and spatial relationships as a reference framework, the method indirectly classifies vessels as arterial or venous based on their adjacency patterns to bronchial segments, rather than attempting direct arterial-vein differentiation.
2Ease of manufacture
If classification depends on local image properties, then the method can process vessels independently, but it fails when bronchial segments are missed due to locally insufficient image quality
Solution Approach 1:
The patent merges the information from multiple bronchial segments and their adjacent vessel segments to form a comprehensive classification decision. Rather than classifying each vessel segment in isolation, the method integrates spatial relationship data across neighboring segments, allowing the classification to benefit from contextual information that compensates for local image quality deficiencies.
Solution Approach 2:
The bronchial tree structure serves multiple functions: it provides anatomical reference, defines spatial zones for vessel classification, and offers redundancy for classification decisions. The same bronchial tree framework is used across different regions and imaging conditions, providing a universal reference system that maintains reliability even when local image quality varies.
3Extent of automation
If pulmonary arteries and veins are distinguished using orientation relations, then the classification can be performed automatically, but the method sometimes fails to correctly classify vessels
Solution Approach 1:
The patent extends the classification approach from considering only orientational relationships to incorporating positional relationships as well. By evaluating both the position and orientation of vessel segments relative to bronchial segments in three-dimensional space, the method creates a more comprehensive spatial adjacency assessment that improves classification accuracy beyond what orientation alone can provide.
Data Source
AI summary
A method of identifying at least part of a pulmonary artery tree (402) comprises receiving (102) a bronchial tree structure (500) and receiving (104) a pulmonary vessel structure (400). A pair of a first bronchial segment (602) and a first vessel segment (604) is identified (106), wherein the first bronchial segment and the first vessel segment are adjacent with respect to position and orientation. The first vessel segment is identified (108) as arterial segment of the pulmonary artery tree. A spatial transformation is applied (110) such that the first bronchial segment and the first vessel segment substantially coincide (602′). Respective further vessel segments (606, 608) are identified (112) adjacent to bronchial segments (610, 612), wherein the bronchial segments are comprised in the bronchial tree.