Iterative CT Reconstruction with Angle-Dependent Blur Kernel
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Iterative reconstruction techniques in X-ray computed tomography face a trade-off between noise suppression and spatial resolution improvement, with existing methods either compromising on noise or spatial clarity.
Innovation Solution
The implementation of an angle-dependent spatially variant low pass filter in the iterative reconstruction process, which is angularly sensitive and applied during the reconstruction of images in X-ray CT systems, enhances spatial resolution while maintaining low noise levels by simulating system optics blur and using a line search strategy for total variation minimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If sharp convolution kernels with high-frequency boost are applied in FBP to improve spatial resolution, then spatial resolution is improved, but noise levels increase
Solution Approach 1:
The patent applies different filtering strategies to different frequency components and spatial regions. The iterative reconstruction algorithm applies sharp deconvolution kernels selectively to enhance high-frequency components while applying noise suppression in low-frequency regions, achieving local optimization of both resolution and noise control
Solution Approach 2:
The patent dynamically adjusts reconstruction parameters including kernel sharpness, regularization strength, and iteration count based on the specific imaging scenario. By changing these parameters adaptively, the system optimizes the balance between spatial resolution enhancement and noise suppression for different clinical applications
2Object-affected harmful factors
If iterative reconstruction with noise compensation means is used, then noise levels are reduced, but spatial resolution improvement is limited
Solution Approach 1:
The patent segments the reconstruction process into distinct stages: initial reconstruction with noise robustness, followed by deconvolution for resolution enhancement, and final refinement. This segmentation allows each stage to optimize for its specific goal without compromising the other
Solution Approach 2:
The patent employs dynamic adjustment of deconvolution kernel strength and iteration parameters during the reconstruction process. The algorithm adaptively increases resolution enhancement in later iterations when noise has been sufficiently suppressed, and adjusts parameters based on local image characteristics to maximize resolution improvement
3Object-affected harmful factors
If enlarged voxel footprint is used in forward model, then noise suppression is improved, but spatial resolution deteriorates
Solution Approach 1:
The patent performs preliminary deconvolution operations during the iterative reconstruction process to compensate for the blurring effect of enlarged voxels. By applying deconvolution kernels that reverse the point spread function effects beforehand, the system recovers fine spatial details that would otherwise be lost due to voxel enlargement
Solution Approach 2:
The patent introduces deconvolution kernels as an intermediary mechanism between the forward model with enlarged voxels and the final image reconstruction. These kernels act as a mathematical filter that reverses the blurring effect, allowing the system to benefit from noise suppression through voxel averaging while recovering spatial resolution through deconvolution
Data Source
AI summary
Spatial resolution is substantially improved by simulating a system blur kernel including an angle variable in the forward projection during a predetermined iterative reconstruction technique. The iterative reconstruction acts as a deconvolution, which overcomes certain restrictions of system optics. In general, resolution is substantially improved with cone beam and helical data without a large increase in noise.