CT Data Set Registration via Elastic Deformation Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In medical imaging, particularly in computer-aided diagnosis, registering multiple scans of a patient over time is challenging due to deformations and displacements caused by body motion and twists, making it difficult to accurately match tumors across different data sets.
Innovation Solution
Automatically detecting data points corresponding to physical features like vertebras, correlating them across data sets, and calculating a piecewise linear similarity transformation to account for body motions and twists, using techniques such as circular template fitting and contrast score computation to establish a registration basis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multiple scans are taken to observe tumor growth over time, then diagnostic information is improved, but registration accuracy deteriorates due to body motion and twists
Solution Approach 1:
The patent applies elasticity theory to transform the rigid registration parameters into flexible transformation parameters. By modeling the body as an elastic medium that deforms under motion and twist, the system can accommodate changes in body position while maintaining accurate tumor location correspondence across multiple scans. This resolves the contradiction by allowing diagnostic information from multiple scans to be integrated while compensating for motion-induced registration errors through elastic deformation models.
2Adaptability or versatility
If elasticity is introduced to address body motions and twists, then registration adaptability is improved, but transformation complexity increases
Solution Approach 1:
The patent segments the transformation process into distinct components: rigid body transformation, elastic deformation, and tumor-specific localization. By dividing the complex elastic transformation into manageable segments, the system achieves high adaptability to various body motions while keeping the computational complexity tractable through modular processing steps.
3Ease of operation
If automatic detection of data points is implemented, then ease of operation is improved, but detection precision requirements increase
Solution Approach 1:
The patent implements automatic detection algorithms that enable the system to self-correct and self-optimize detection precision through iterative refinement. The system automatically identifies corresponding data points across scans, refines the detection through elasticity-based transformation, and improves accuracy without requiring manual intervention, thus achieving both ease of operation and high detection precision.
Data Source
AI summary
Disclosed is a method and apparatus for registering data points in data sets representing scan data. Data points corresponding to a physical feature represented in the scan data are automatically detected. The detected data points in one data set are correlated with detected data points in another data set. A group of similarity transformations between the correlated detected data points is then calculated. The group of similarity transformations is then combined. In one advantageous embodiment, the physical feature is vertebras.