CT Material Density Image Decomposition With Noise and Artifact Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CT imaging methods face challenges in achieving high signal-to-noise ratios and low noise in material density images due to degraded signal-to-noise ratios from direct matrix inversion or least square decomposition, and iterative decompositions are slow and prone to overfitting.
Innovation Solution
An image processing system utilizing a convolution neural network with a penalty function for iterative noise reduction and artifact removal, enhancing signal-to-noise ratios through iterative decomposing and noise reduction processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct matrix inversion or least square decomposition is used to decompose CT images, then material decomposition can be achieved, but the signal-to-noise ratio of the material density image is seriously degraded
Solution Approach 1:
The patent introduces a penalty function as an intermediary between the decomposition model and the optimization process. This penalty function incorporates image quality constraints (such as total variation or sparsity) to guide the decomposition while maintaining signal-to-noise ratio, effectively mediating between the mathematical decomposition requirement and the physical image quality requirement.
Solution Approach 2:
The patent transforms the decomposition problem from direct matrix inversion to a parameter optimization problem by introducing regularization parameters in the penalty function. By adjusting these parameters, the system can control the trade-off between decomposition accuracy and noise suppression, thereby improving signal-to-noise ratio while maintaining image quality.
2Measurement precision
If iterative decomposition methods are used to improve signal-to-noise ratio, then image quality can be enhanced, but the processing speed becomes slow and overfitting occurs
Solution Approach 1:
The patent applies partial iteration by combining a limited number of iterative decomposition steps with direct decomposition methods. Instead of performing extensive iterative processing that causes overfitting and slows down computation, the system performs a controlled number of iterations just enough to improve signal-to-noise ratio without sacrificing processing speed or inducing overfitting.
Solution Approach 2:
The patent implements feedback control in the iterative decomposition process by monitoring convergence criteria and stopping conditions. The system adjusts the number of iterations based on feedback from intermediate results, preventing overfitting while ensuring sufficient noise reduction, thereby balancing image quality improvement with processing efficiency.
3Measurement precision
If iterative decomposition methods are used to reduce noise, then signal-to-noise ratio can be improved, but the processing time increases significantly
Solution Approach 1:
The patent performs preliminary preprocessing of the CT images before decomposition, including noise filtering and data normalization. This preliminary action reduces the initial noise level and improves the conditioning of the decomposition problem, allowing subsequent iterative processing to converge faster and achieve the desired signal-to-noise ratio with fewer iterations, thereby reducing overall processing time.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The embodiments of the present disclosure provide methods and systems for image processing. The method may include: obtaining one or more initial material density images; and inputting the one or more initial material density images into a trained image processing model to obtain one or more target material density images; wherein the trained image processing model is configured to simultaneously perform decomposition processing and at least one of noise reduction processing or artifact removal processing on the one or more initial material density images.