Spectral CT Projection Data De-noising via Covariance Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spectral/multi-energy computed tomography (CT) systems face challenges in reducing correlated noise in projection domain decomposition, leading to image artifacts and reduced clinical value due to noise amplification, which affects the contrast-to-noise ratio and spatial resolution in reconstructed images.
Innovation Solution
A method is introduced to filter correlated noise from spectral/multi-energy projection data using a statistical model based on covariance matrices and correlation coefficients, specifically targeting the noise correlation between basis material line integrals, allowing for de-noising solely in the projection domain without affecting object structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If projection domain decomposition is performed to determine material properties, then material decomposition accuracy is improved, but noise is strongly magnified in the basis material line integrals
Solution Approach 1:
The patent exploits the negative correlation structure of the magnified noise between different basis material line integrals and converts this harmful effect into a beneficial filtering mechanism. By identifying and utilizing the predictable negative correlation pattern, the system filters out the correlated noise components while preserving the true material decomposition signals, thereby resolving the contradiction between achieving accurate material decomposition and avoiding noise amplification
2Ease of operation
If correlated noise is not filtered in projection domain, then processing simplicity is maintained, but streak artifacts appear in reconstructed images
Solution Approach 1:
The patent introduces an intermediary noise filtering step in the projection domain that processes the basis material line integrals before reconstruction. This intermediary filtering operation, which utilizes the negative correlation information, acts as a mediator between the simple decomposition process and the final image reconstruction, eliminating streak artifacts while maintaining overall processing efficiency
3Measurement precision
If noise is magnified in basis material line integrals, then contrast-to-noise ratio is reduced, but material decomposition is still performed
Solution Approach 1:
The patent addresses the contrast-to-noise ratio degradation by moving the filtering operation to the projection domain rather than the image domain. This dimensional shift allows the system to exploit the correlation structure in the projection data space, where the negative correlation between basis material line integrals can be effectively utilized to suppress noise before reconstruction, thereby preserving contrast-to-noise ratio while maintaining material decomposition capability
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 effectively reduces patterned and non-patterned image noise, improves the contrast-to-noise ratio, and suppresses streak artifacts in reconstructed images, enhancing the quality of basis material images and spectral CT applications without compromising spatial resolution.
Implementation Method 1
A CT scanner includes an x-ray tube that emits radiation that traverses an examination region and an object therein. A detector array located opposite the examination region across from the x-ray tube detects radiation that traverses the examination region and the object therein
Implementation Method 2
multiple projection data sets are acquired, which represent the attenuation properties of the scanned object for different X-ray spectra
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application describes an approach to filter, solely in the projection domain, correlated noise from (or de-noise) spectral/multi-energy projection data. As described herein, this can be achieved based at least on variances of the basis material line integrals and a covariance there between, based on multiple correlation coefficients and hyper-planes that describe thenoise correlation between different basis material line integrals, and/or otherwise.