Iterative Image Reconstruction for Low-Dose CT Noise Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image reconstruction methods in medical imaging, such as CT, suffer from reduced dose or insufficient data leading to degraded image quality, with regularization techniques like Total Variation and GGMRF introducing noise and artifacts, compromising diagnostic accuracy.
Innovation Solution
A method involving iterative image reconstruction using a combination of preliminary and further optimization operations via machine learning models to reduce noise and artifacts, enhancing image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If iterative reconstruction with regularization terms (Total Variation, GGMRF) is used to suppress noises, then noise suppression is improved, but image quality deteriorates due to massive artifacts and cartoon sense
Solution Approach 1:
The patent changes the parameter of regularization strength dynamically during iterative reconstruction. It uses a loss function with a regularization term where the regularization parameter is adjusted based on iteration number and image quality metrics, transitioning from strong regularization early in reconstruction to weaker regularization later, thereby suppressing noise without creating excessive artifacts
Solution Approach 2:
The patent implements feedback by evaluating image quality metrics (noise level, artifact level, structural similarity) during iterative reconstruction and using these evaluations to adjust the loss function weights and reconstruction parameters. This closed-loop control allows the system to adaptively balance noise suppression with artifact prevention based on real-time image quality assessment
2Loss of energy
If reduced dose or insufficient data is used in CT imaging, then radiation dose is reduced, but image quality deteriorates significantly
Solution Approach 1:
The patent replaces traditional mechanical/physical reconstruction methods with machine learning-based iterative reconstruction. It uses a neural network trained on high-quality images to learn optimal reconstruction mappings, allowing high-quality image reconstruction from low-dose or insufficient data without the usual quality degradation
Solution Approach 2:
The patent changes the reconstruction approach from conventional filtered back projection to iterative reconstruction with learned prior knowledge. By incorporating deep learning models that capture anatomical priors and adjusting reconstruction parameters based on data quality, it enables high-quality reconstruction from reduced dose data
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A systems and methods for image reconstruction. The system may obtain an initial image to be processed. The system may generate a reconstructed image by performing a plurality of iteration steps on the initial image. Each of the plurality of iteration steps may include a first optimization operation and at least one second optimization operation. The first optimization operation and the at least one second optimization operation may be executed sequentially. The first optimization operation may include receiving an image to be processed in the iteration step and determining an updated image by preliminarily optimizing the image to be processed. The at least one second optimization operation may include determining an optimized image by reducing interference information of the updated image and designating the optimized image as a next image to be processed in a next iteration step. The interference information may include noise information and/or artifact information.