Adaptive Morphological Segmentation for Lung Nodule Volume Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting objects in medical imaging, particularly for lung cancer diagnosis, face challenges in accurately segmenting objects of varying sizes, densities, and morphologies connected to surrounding structures, and are influenced by scan and reconstruction parameters, leading to reproducibility issues and computational inefficiencies.
Innovation Solution
A method involving initial segmentation using region growing, followed by erosion and dilation operations with variable thresholds, and a convex hull operation to improve segmentation quality, along with a volume determination approach that accounts for partial volume effects using weighting factors to enhance accuracy and reproducibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional two-dimensional diameter measurement methods are used, then the assessment process is simpler and faster, but the measurement precision is insufficient for irregularly shaped nodules
Solution Approach 1:
The patent replaces manual mechanical measurement processes with automated computer-based image processing algorithms. The system automatically segments nodules from CT images, calculates three-dimensional volumes, and tracks growth over time, eliminating the need for manual slice selection and diameter measurement while providing superior volumetric accuracy for irregularly shaped lesions.
2Reliability
If known segmentation methods are used, then the segmentation process is faster, but the reliability is poor for high-density objects connected to other structures
Solution Approach 1:
The patent employs a multi-stage segmentation approach that divides the complex task of separating connected structures into distinct phases: initial threshold-based segmentation to identify candidate regions, followed by region growing to expand into relevant areas, and finally morphological operations (erosion and dilation) to separate connected components. This staged segmentation improves reliability for high-density objects connected to vasculature or chest wall while maintaining computational efficiency.
Solution Approach 2:
The patent uses dynamic morphological operations where erosion and dilation thresholds are adjusted iteratively to achieve optimal separation of connected structures. The region growing process dynamically expands segmented areas based on intensity similarity criteria, allowing the segmentation to adapt to varying local characteristics of connected structures rather than applying fixed parameters throughout.
3Adaptability or versatility
If fixed threshold segmentation is used, then the device complexity is lower, but the adaptability to objects of different sizes and densities is reduced
Solution Approach 1:
The patent implements adaptive parameter adjustment where segmentation thresholds are not fixed but are modified based on local image characteristics. The region growing process uses intensity similarity criteria that adapt to local density variations, and morphological operations employ thresholds that can be adjusted according to the specific properties of the objects being segmented. This enables the system to handle objects of different sizes, densities, and morphologies effectively.
Data Source
AI summary
The invention relates to a method of segmenting an object in a data set, wherein the object is initially segmented resulting in a first set (N0) of voxels. An erosion operation is performed on the first set (N0) of voxels resulting in an eroded set (N−) of voxels. A dilation operation is performed on the eroded set (N−) of voxels resulting in a dilated set (N+) of voxels. The erosion operation depends on a variable erosion threshold (Θ−), and the dilation operation depends on a variable dilation threshold (Θ+).