Spectral Image Decomposition Accuracy via Residual Region Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spectral image data decomposition in spectral x-ray imaging is often inaccurate due to suboptimal processes and image acquisition inaccuracies, particularly affecting the distribution of contrast agents, which hampers clinical applications.
Innovation Solution
An apparatus and method that identify residual regions of contrast agents in virtual non-contrast images using expected structural characteristics, allowing for recalculating weights of basis functions to improve decomposition accuracy, focusing corrections on these regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spectral image data decomposition is performed using conventional basis functions, then image data can be processed and spectral images can be generated, but the decomposition accuracy is insufficient leading to residual contrast agent artifacts
Solution Approach 1:
The patent applies parameter changes by transforming the decomposition problem from conventional basis functions to a dictionary learning framework with learned basis functions and sparse coefficients. The decomposition model changes from fixed mathematical basis functions to adaptive learned dictionaries that capture the true sparsity patterns of anatomical structures, thereby improving decomposition accuracy and reducing contrast agent residuals
Solution Approach 2:
The patent uses copying by creating virtual non-contrast images from the decomposed spectral data. These virtual images serve as copies that represent what the anatomy would look like without contrast agents, allowing for the identification and removal of contrast agent artifacts while preserving anatomical structures
2Reliability
If conventional decomposition methods are used, then processing speed is maintained, but clinical utility is reduced due to inaccurate contrast agent distribution
Solution Approach 1:
The patent applies preliminary action by pre-learning dictionaries from training data before actual decomposition. The dictionary learning phase is performed in advance on representative anatomical data, so that during clinical processing, the pre-learned dictionaries can be directly applied for fast decomposition without requiring real-time iterative learning, thus maintaining processing speed while improving accuracy
Solution Approach 2:
The patent changes the decomposition parameters from conventional fixed basis functions to adaptive sparse representations with learned dictionaries. This parameter transformation enables more accurate separation of contrast agents from anatomical structures, significantly improving clinical utility for tasks like virtual non-contrast imaging and contrast agent quantification
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
The approach significantly enhances the accuracy of decomposed spectral image data, reducing contrast agent residuals and improving the clinical utility of spectral image data.
Implementation Method 1
functions that describe the dependence of the Compton scattering effect and/or the photoelectric effect on the energy of the radiation
Implementation Method 2
functions that describe the dependence of the Compton scattering effect and/or the photoelectric effect on the energy of the radiation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention refers to an apparatus for determining decomposed spectral image data with an improved accuracy. The apparatus (110) comprises a spectral image data providing unit (111) for providing spectral image data, a spectral image data decomposition unit (112) for calculating a basis decomposition for the spectral image data, a virtual non- contrast image generation unit for generating a virtual non-contrast image based on the decomposed spectral image data, and a contrast agent residual identification unit (113) for identifying a residual region comprising a contrast agent residual in the virtual non-contrast image based on an expected structural characteristic of the contrast agent residual, wherein the spectral image data decomposition unit (114) is configured to calculate a new basis decomposition in the residual region. Utilizing structural characteristics of the contrast agent residual allows for a very accurate determination of contrast agent residuals and an improvement of the decomposition accuracy in this area.