Tomography Image Reconstruction Using Segmented FBP and Iterative Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image reconstruction techniques, such as filtered back-projection (FBP) and iterative reconstruction (IR), face limitations in achieving accurate and efficient computation of dense images, particularly with large numbers of voxels, due to inefficiencies in projection usage, negative density estimates, and computational burdens.
Innovation Solution
A system and method that computes differences between actual and predicted projection data, uses impulse responses for voxel updates, and reconstructs images efficiently by inverting functions and applying Kalman/Wiener filter gains, allowing for high-fidelity image reconstruction with reduced computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If filtered back-projection (FBP) is used for image reconstruction, then computation speed is improved, but quantitative accuracy deteriorates and negative density estimates occur
Solution Approach 1:
The reconstruction process is segmented into two distinct stages: (1) FBP is used to rapidly generate an initial image estimate, and (2) a limited-number iterative refinement process corrects quantitative inaccuracies. This segmentation allows the system to benefit from both the speed of FBP and the accuracy of iterative methods without suffering from their respective drawbacks.
Solution Approach 2:
The FBP reconstruction is performed as a preliminary action before iterative refinement. By pre-computing the initial image estimate using FBP, the subsequent iterative process only needs to correct errors rather than compute from scratch, significantly reducing the number of iterations needed and maintaining computational efficiency while improving quantitative accuracy.
2Measurement precision
If conventional iterative reconstruction techniques are used, then quantitative accuracy is improved, but computational burden increases significantly
Solution Approach 1:
Instead of performing full iterative reconstruction, the invention applies partial iteration by limiting the refinement process to a small number of iterations (typically 3-10) after FBP. This partial action is sufficient to correct the most significant quantitative errors while avoiding the excessive computational burden of complete iterative convergence.
Solution Approach 2:
The iterative refinement process uses feedback from the difference between measured and calculated projections to correct the initial FBP image. This feedback mechanism allows the system to systematically reduce quantitative errors by adjusting the image estimate based on projection data mismatches, achieving high accuracy with minimal iterations.
3Measurement precision
If matrix-based techniques are used for small problems, then reconstruction accuracy is improved, but computational complexity becomes unmanageable for typical tomographic settings
Solution Approach 1:
The FBP algorithm serves as an intermediary that transforms the ill-posed inverse problem into a well-conditioned initial estimate. This intermediary step converts the complex matrix inversion problem into a simpler convolution operation, providing a starting point that is computationally manageable and sufficiently accurate for subsequent refinement.
Solution Approach 2:
The invention replaces the mechanically intensive matrix-based iterative inversion with a hybrid approach combining FBP (based on fast Fourier transformation and convolution) and limited iterative refinement. This substitution reduces computational complexity from O(N^5) to a manageable level while preserving reconstruction accuracy through the feedback-driven correction step.
Data Source
AI summary
A system comprising: one or more transmitters to transmit an excitation energy into an object under observation; one or more detectors to generate projection space data encoding an energy received by the one or more detectors in response to the transmitted excitation energy into the object under observation; a controller to control the one or more transmitters to transmit the excitation energy and the one or more detectors to generate the projection space data; and an image reconstructor to receive the projection space data and to process the projection space data by: computing a first quantity characterizing a difference between the projection space data and predicted projection data; computing a second quantity corresponding to at least one impulse response, each impulse response corresponding to a reference voxel of unit-intensity; computing an update value using the first quantity and an inverted function of the second quantity; and reconstructing an object space image representing the object under observation using the update value.