Spinal Canal Center Line Extraction via Dynamic Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting the spinal canal center line from medical images are sensitive to segmentation errors, leading to incorrect reformatted images with unrealistic distortions, which can hinder accurate diagnosis in time-critical trauma settings.
Innovation Solution
A method and apparatus for extracting a spinal canal center line that involves receiving 3D scan data, performing voxel-wise spinal canal segmentation, generating an inverse density array, and using dynamic programming-based optimal path finding to iteratively refine the spinal canal path until it meets a predetermined accuracy threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional segmentation-based center line extraction is used, then the process is simple when segmentation is accurate, but the result becomes highly sensitive to segmentation errors causing unrealistic image distortions
Solution Approach 1:
The patent replaces the mechanical segmentation-based approach with a physics-inspired optimal transport model. Instead of relying on segmentation masks to define the spinal canal path, the method uses density functions and optimal transport theory to compute the center line as the path of least resistance through the image intensity field, making it insensitive to segmentation errors.
Solution Approach 2:
The patent transforms the problem from discrete segmentation-based path extraction to continuous density-based optimal path finding. By changing the parameter representation from binary segmentation masks to continuous density functions and using optimal transport metrics, the method achieves robustness against segmentation inaccuracies while maintaining computational feasibility.
2Measurement precision
If high-resolution sampling is performed throughout the entire 3D array, then the accuracy of the spinal canal path is improved, but the computational time and resources increase significantly
Solution Approach 1:
The patent divides the 3D imaging volume into multiple 2D slices along the spinal canal trajectory. By processing each slice independently at high resolution and then integrating the results through optimal transport, the method achieves high overall accuracy without the computational burden of processing the entire 3D volume at maximum resolution simultaneously.
Solution Approach 2:
The patent applies high-resolution sampling selectively only in regions containing the spinal canal structure, rather than uniformly across the entire 3D volume. The optimal transport framework allows concentrated computational effort where needed (along the spinal path) while using coarser sampling in surrounding regions, achieving accuracy efficiency trade-off.
Data Source
AI summary
Method and apparatus for finding an optimal spinal canal path center line are disclosed. Imaging scans such as a Computer Tomography (CT) scan are used to diagnosis rib and spine fractures. The spine image is usually reformatted in a straightened CPR (curved planar reformat). A dynamic programming-based techniques to find an optimal path from a spinal canal segmentation is disclosed. The disclosed method and apparatus overcomes spinal canal segmentation errors resulting in spinal canal gaps and unrealistic image distortions. The disclosed techniques are not sensitive to segmentation errors even is the presence of high segmentation noise. The disclosed techniques generate a binary three-dimensional image corresponding to the scanned spine image, perform density estimation to create a three-dimensional having the same size as the original binary three-dimensional, perform dynamic programming based optimal spinal canal path finding with a coarse to fine optimal path finding strategy to accelerate the optimal spinal canal path finding.