Lung Lobe Segmentation via Intermediate Region Feature Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for segmenting lung lobes in medical imaging, such as EP 1 447 772 A1, rely on the absence of vessels which does not accurately coincide with lobe borders, leading to diminished segmentation quality due to variable vessel visibility and quality.
Innovation Solution
Generating a feature image that enhances intermediate regions between lung lobes, using a distance transformation to improve feature detection and adapting a model object to both the object and feature images, with the feature image providing information on intermediate region positions for enhanced segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vessel absence is used for segmenting lung lobes, then segmentation can be performed, but segmentation quality is diminished because vessel absence does not coincide with lobe borders
Solution Approach 1:
The patent segments the lung image into multiple components: object image, feature image showing intermediate regions, and combined segmentation result. The feature image specifically segments and highlights intermediate regions between lung lobes that were previously invisible, allowing precise boundary detection by combining both images.
Solution Approach 2:
The patent transforms the problem from 2D image analysis to 3D spatial understanding by introducing a feature image dimension that specifically encodes intermediate region information. This additional dimensional representation allows the segmentation algorithm to perceive boundaries that are not visible in the original image plane.
2Extent of automation
If vessel-based segmentation is used, then automated segmentation is achieved, but reliability is reduced due to variable vessel visibility and quality
Solution Approach 1:
The patent introduces an intermediary feature image that mediates between the original object image and the final segmentation result. This feature image acts as a reliable intermediary representation that consistently highlights intermediate regions regardless of vessel visibility, making the automated segmentation more reliable across different cases.
Solution Approach 2:
The patent changes the parameter being analyzed from vessel presence/absence to intermediate region detection. By transforming the segmentation criterion from vascular anatomy to spatial boundary features, the method achieves consistent results independent of vessel quality or visibility conditions.
3Device complexity
If only object image is used for segmentation, then processing is simpler, but intermediate regions between sub-objects cannot be accurately identified
Solution Approach 1:
The patent divides the processing into distinct segments: generating object image, generating feature image showing intermediate regions, and combining them for final segmentation. This segmented approach adds only the necessary complexity for intermediate region detection while keeping other processing steps simple and efficient.
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
The present invention relates to an apparatus (1) for segmenting an object comprising sub-objects shown in an object image. The apparatus comprises a feature image generation unit (2) for generating a feature image showing features related to intermediate regions between the sub-objects and a segmentation unit (3) for segmenting the sub-objects by using the object image and the feature image. Preferentially, the feature image generation unit (2) is adapted for generating a feature image from the object image. In a further embodiment, the feature image generation unit (2) comprises a feature enhancing unit for enhancing features related to intermediate regions between the sub-objects in the object image.