Spatially Varying Regularization for Uniform Noise in 3D CT Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CT image reconstruction methods using regularization and statistical models often result in non-uniform signal-to-noise ratio (SNR) and resolution across images, leading to undesirable image properties such as over-smoothing in certain regions and loss of resolution in areas with high data fidelity.
Innovation Solution
A CT image processing method that computes a spatially varying regularization parameter based on the square root of back-projected inverse statistical variances, normalized relative to a region of interest (ROI), to balance the data term and regularization term during iterative reconstruction, ensuring uniform noise distribution without compromising resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a uniform regularization parameter is used in iterative reconstruction, then the computational process is simple and fast, but the image quality becomes non-uniform with over-smoothing in some regions and loss of resolution in others
Solution Approach 1:
The patent applies local quality by computing spatially varying regularization parameters based on local noise behavior and data fidelity. The regularization parameter is adjusted for each image location according to the local statistical properties of the projection data, ensuring that regions with high data fidelity maintain high resolution while regions with noisy data receive appropriate smoothing. This resolves the contradiction by making the regularization adaptive to local image characteristics rather than applying a uniform parameter globally.
Solution Approach 2:
The patent implements dynamics by making the regularization parameter dynamic and adaptive during the iterative reconstruction process. The parameter is computed as a function of the current image estimate and projection data statistics, allowing it to evolve throughout the iterations. This dynamic adjustment enables the system to maintain optimal balance between noise suppression and resolution preservation at different stages of reconstruction, resolving the contradiction between simplicity and image quality uniformity.
2Reliability
If strong regularization is applied to reduce noise, then the signal-to-noise ratio improves, but the resolution and fine details are lost due to over-smoothing
Solution Approach 1:
The patent applies local quality by computing spatially varying regularization parameters based on local noise behavior and data fidelity. The regularization parameter is adjusted for each image location according to the local statistical properties of the projection data, ensuring that regions with high data fidelity maintain high resolution while regions with noisy data receive appropriate smoothing. This resolves the contradiction by making the regularization adaptive to local image characteristics rather than applying a uniform parameter globally.
3Manufacturing precision
If weak regularization is applied to preserve resolution, then the image detail and sharpness are maintained, but the noise level increases reducing diagnostic quality
Solution Approach 1:
The patent applies local quality by computing spatially varying regularization parameters based on local noise behavior and data fidelity. The regularization parameter is adjusted for each image location according to the local statistical properties of the projection data, ensuring that regions with high data fidelity maintain high resolution while regions with noisy data receive appropriate smoothing. This resolves the contradiction by making the regularization adaptive to local image characteristics rather than applying a uniform parameter globally.
4Manufacturing precision
If spatially varying regularization parameters are computed to achieve uniform noise distribution, then image quality is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-computing the statistical variances of the projection data and using these to determine the spatially varying regularization parameters before the main iterative reconstruction. The variance maps are calculated from the projection data statistics and used to guide the regularization parameter selection throughout the reconstruction process. This preliminary computation of noise characteristics enables the subsequent reconstruction to proceed efficiently with pre-determined local regularization strengths, reducing the overall computational burden while maintaining uniform noise distribution.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and related apparatus (116) for iterative reconstruction of a volume. A regularization parameter (β) of an iterative update function is spatially adapted and normalized in respect to a region of interest ROI in the volume. The method allows achieving essentially uniform noise distribution across the reconstructed volume.