Multi-Organ Segmentation via Level Set Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing discriminative segmentation approaches for volumetric medical images face challenges in accurately segmenting multiple organs due to non-homogeneous distribution of control points, leading to varying segmentation accuracy and difficulties in detecting and removing overlaps or gaps between organ boundaries.
Innovation Solution
A method combining learning-based segmentation with level set optimization, where point-to-point correspondences are preserved, and novel energy terms are used to impose region-specific constraints for non-overlap, coincidence, and shape similarity, allowing for refined segmentation and accurate boundary detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If point-cloud based shape representation is used in discriminative segmentation, then segmentation speed and automatic detection capability are improved, but segmentation accuracy becomes non-uniform due to non-homogeneous distribution of control points
Solution Approach 1:
The patent transforms the shape representation from point-cloud based to level set based, changing the fundamental parameter of how boundaries are represented. Level sets provide homogenous resolution across the entire boundary while maintaining computational efficiency, resolving the contradiction between speed and uniformity.
Solution Approach 2:
The patent uses learning-based segmentation to generate initial point-cloud boundaries, then copies this information into a level set framework. This allows the benefits of both approaches to be combined: the speed of learning-based methods with the uniformity of level set representation.
2Extent of automation
If point-cloud based shape representation is used, then fast automatic landmark detection is achieved, but difficulty increases in detecting and removing overlaps or gaps between organ boundaries
Solution Approach 1:
The patent copies boundary information from point-cloud representation into level set functions, enabling automated processing while improving the detectability of overlaps and gaps through the continuous mathematical representation of level sets.
Solution Approach 2:
The patent replaces the discrete point-cloud mechanical representation with a continuous level set mathematical representation, making it easier to detect and resolve boundary issues through gradient-based optimization and energy minimization.
3Measurement precision
If control points are non-homogeneously distributed to capture local details, then local boundary accuracy is improved, but overall shape representation becomes inconsistent
Solution Approach 1:
The patent changes the representation parameter from variable-density point clouds to uniform-level level sets, achieving consistent global representation while maintaining local accuracy through the mathematical properties of level set functions and their ability to represent boundaries at any resolution.
Data Source
AI summary
A method and system for automatic multi-organ segmentation in a 3D image, such as a 3D computed tomography (CT) volume using learning-base segmentation and level set optimization is disclosed. A plurality of meshes are segmented in a 3D medical image, each mesh corresponding to one of a plurality of organs. A level set in initialized by converting each of the plurality of meshes to a respective signed distance map. The level set optimized by refining the signed distance map corresponding to each one of the plurality of organs to minimize an energy function.