CT Scatter Estimation Using Adaptive Kernel Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current scatter estimation models in computerized tomography (CT) imaging systems result in measurement errors and image degradation due to scattered radiation, leading to uncertainties of approximately +/-50 Hounsfield Units, especially in challenging situations like pelvis scans, which existing solutions have not adequately addressed.
Innovation Solution
A method that generates original projections, reference scatter data, and estimated scatter data using kernels with adjustable parameters, where optimal kernel parameters are determined by minimizing the difference between reference and estimated scatter data, and then interpolated for improved scatter estimation across all projection angles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If kernel methods are used to estimate scatter radiation, then scatter estimation can be performed, but measurement errors and image degradation persist with uncertainties of approximately +/-50 Hounsfield Units
Solution Approach 1:
The patent adjusts kernel parameters (such as sigma values and scaling factors) to optimize scatter estimation accuracy. By modifying these parameters based on object characteristics and projection data, the system reduces measurement errors and uncertainties in Hounsfield Units while maintaining computational efficiency.
Solution Approach 2:
The system dynamically adapts kernel parameters during the scanning process based on real-time detection of object characteristics, density distributions, and projection data variations. This dynamic adjustment allows the scatter estimation to respond to changing conditions, improving both measurement precision and image reliability.
2Productivity
If existing scatter estimation models are applied, then computational speed is maintained, but accuracy deteriorates in challenging situations such as pelvis scans
Solution Approach 1:
The patent applies different kernel parameters and estimation strategies to different regions of the scanned object based on local characteristics such as density, size, and composition. For example, pelvis regions receive specialized parameter adjustments compared to other anatomical areas, improving accuracy without compromising overall computational efficiency.
Solution Approach 2:
The scanning process is divided into segments where object characteristics are analyzed and kernel parameters are adjusted for each segment or region of interest. This segmentation allows tailored scatter estimation for challenging areas like the pelvis while maintaining efficient processing for other regions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances scatter estimation accuracy, reducing measurement errors and image degradation, particularly at challenging angles like 45 and 135 degrees, by applying optimized and interpolated kernel parameters to improve reconstructed image data quality.
Implementation Method 1
generating a first set of estimated scatter data associated with the target object by applying one or more kernels with first values for one or more kernel parameters to a first subset of projections
Implementation Method 2
interpolating the adjusted first values for remaining projections out of the set of original projections to generate second values for the one or more kernel parameters
Data Source
AI summary
Techniques described herein generally relate to estimating scatter. In one embodiment, one example method for estimating scatter associated with a target object may include generating a set of original projections associated with the target object, generating a set of reference scatter data associated with the target object at one or more selected projection angles, generating a first set of estimated scatter data associated with the target object also at the one or more selected projection angles, adjusting first values for one or more kernel parameters of one or more kernels that reduce a difference between the set of reference scatter data and the first set of estimated scatter data, interpolating the adjusted first values for remaining projections out of the set of original projections to generate second values for the one or more kernel parameters, and generating a second set of estimated scatter data associated with the target object.