Lung Image Segmentation Using Gradient and Cavity Sensitivity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for 3D lung segmentation and lung vessel exclusion from MRI images are inadequate due to low contrast boundaries between lung parenchyma and chest walls, and are ineffective in processing cavity structures like vessels and nodules.
Innovation Solution
An image processing apparatus and method that utilizes sensitivity to gradient and cavity structure for accurate lung segmentation, employing a feature sensitivity indicator to distinguish and exclude chest walls while retaining cavity structures, using a combination of gradient and cavity structure sensitivity indicators in the level set evolution process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional segmentation methods (threshold method, binarization segmentation method) are used for lung segmentation, then the processing is simple and fast, but the accuracy is poor due to low contrast boundaries between lung parenchyma and chest wall
Solution Approach 1:
The patent segments the segmentation process into multiple stages: preliminary segmentation using simple methods, followed by refinement using level set methods with gradient and cavity structure sensitivity indicators. This allows the system to achieve high accuracy while keeping the overall process manageable through structured division of tasks.
Solution Approach 2:
The patent applies preliminary segmentation using threshold methods or binarization segmentation methods before the refinement stage. This preliminary action provides an initial segmentation result that can be quickly obtained, and then the level set method refines this result to achieve high accuracy, thereby reducing the overall complexity by avoiding the need for complex methods throughout the entire process.
2Measurement precision
If level set method with gradient sensitivity is used for segmentation, then the boundary distinction between lung parenchyma and chest wall is improved, but the cavity structures (vessels and nodule) are not correctly included
Solution Approach 1:
The patent merges two types of sensitivity indicators: gradient sensitivity indicator for boundary distinction and cavity structure sensitivity indicator for cavity structure detection. By combining these two indicators in the level set evolution equation, the system simultaneously achieves accurate boundary distinction and correct cavity structure inclusion, resolving the contradiction between the two requirements.
3Ease of manufacture
If chest wall tissue is included in the extracted lung image, then the segmentation is simpler, but the diagnosis accuracy is reduced due to false positive areas
Solution Approach 1:
The patent uses feedback mechanisms in the level set evolution process where the sensitivity indicators continuously guide the segmentation boundary. The gradient sensitivity indicator provides feedback on boundary regions, and the cavity structure sensitivity indicator provides feedback on cavity structures, allowing the segmentation to automatically adjust and exclude chest wall tissue while maintaining simplicity through automated refinement.
Data Source
AI summary
An image processing apparatus according to an embodiment includes processing circuitry. The processing circuitry recognizes at least one of chest wall, vessels, nodules, and tumors by increasing sensitivity to a gradient and a sensitivity to a cavity structure in three-dimensional data including a lung area. The processing circuitry generates lung image data corresponding to the lung area by performing a data removing process for removing data derived from the chest wall and a data holding process for holding data derived from at least one of the vessels, the nodules, and the tumors on a basis of a result obtained by the recognizing. The processing circuitry outputs the lung image data.