Lung Nodule Segmentation via Level Set and Shape Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automated segmentation of lung nodules in CT scans face challenges due to low contrast, noise, and varying shapes, particularly for nodules attached to the pleural surface or vessels, as they often assume spherical or ellipsoidal shapes, which can lead to inaccurate segmentation.
Innovation Solution
A method using a combination of level set segmentation and a linear combination of training shapes to iteratively evolve a segmenting surface, with a sparse representation that adapts to shape variability, allowing for accurate segmentation of nodules without restricting them to predefined structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If intensity-based segmentation algorithms are used, then segmentation speed is improved, but segmentation accuracy deteriorates due to low contrast and noise
Solution Approach 1:
The method segments the nodule boundary by dividing it into multiple discrete points sampled along the contour, then fits a continuous curve through these points. This segmentation approach allows the algorithm to work with discrete intensity samples while reconstructing the continuous boundary, improving accuracy without sacrificing speed
Solution Approach 2:
A curve fitting function serves as an intermediary between the discrete intensity-based sample points and the continuous nodule boundary. This intermediary model allows the algorithm to infer the true boundary position from noisy intensity data, resolving the contradiction between fast intensity-based processing and accurate boundary detection
2Measurement precision
If shape refinement is applied to improve accuracy, then segmentation precision is improved, but the method becomes restricted to specific nodule types only
Solution Approach 1:
The shape refinement method uses a dynamic curve fitting approach where the boundary is represented as a flexible curve that can adapt to different nodule shapes. The curve is controlled by adjustable parameters including polynomial order and smoothing weight, allowing it to accommodate various nodule geometries without being restricted to specific types
Solution Approach 2:
The method changes parameters such as polynomial order and smoothing weight to adapt the curve fitting behavior to different nodule characteristics. By adjusting these parameters, the same basic algorithm can accurately segment various nodule types including solid, part-solid, and ground glass nodules, maintaining both accuracy and versatility
3Measurement precision
If preprocessing and postprocessing techniques are used to exclude vessels and pleural surface, then segmentation accuracy is improved, but processing complexity increases
Solution Approach 1:
The method performs preliminary sampling of intensity values along the nodule boundary before curve fitting. By pre-sampling discrete points along the contour, the algorithm prepares clean boundary data that reduces the need for complex postprocessing to remove vessels and pleural surface artifacts
Solution Approach 2:
The curve fitting process extracts the true nodule boundary from the sampled points by fitting a smooth curve that naturally excludes vessels and pleural surface irregularities. This extraction approach separates the true boundary signal from noise and artifacts without requiring complex preprocessing or postprocessing steps
Data Source
AI summary
A method and system of automated segmentation of an anatomical object through learned examples include: receiving, by a processing device, an image of the anatomical object; determining a sparse representation of a shape of the anatomical object by iteratively evolving a segmenting surface as a combination of a level set segmentation and a linear combination of training shapes; and outputting, to an output device, the sparse representation of the shape of the anatomical object as the segmentation of the anatomical object.