Voxel-Dependent Iterative Tomographic Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing tomographic image reconstruction algorithms, such as FBP, ART, and OSML, face challenges in achieving a balance between signal-to-noise ratio and computational efficiency, with OSML offering better SNR but requiring more time than ART and FBP.
Innovation Solution
A data processing unit and method that implement a voxel-dependent update in the iterative reconstruction of attenuation coefficients, using equations that incorporate a voxel-dependent factor and weighting functions to improve signal-to-noise ratio and speed, specifically by distributing error updates based on each voxel's contribution to forward projections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OSML algorithm is used for iterative reconstruction, then signal-to-noise ratio is improved, but reconstruction time increases significantly
Solution Approach 1:
The patent divides the set of all projections into multiple subsets (ordered subsets) and processes them in sequential iterations. This segmentation allows the algorithm to converge faster by making progress on multiple projection groups simultaneously across different iterations, rather than processing all projections sequentially as in standard ML methods.
Solution Approach 2:
The patent employs periodic cycling through ordered subsets of projections in a systematic sequence. Each iteration processes a specific subset, and the cycle repeats through all subsets multiple times. This periodic action structure enables faster convergence compared to processing all projections uniformly, achieving better SNR with reduced reconstruction time.
2Loss of time
If FBP or ART algorithm is used for reconstruction, then reconstruction time is reduced, but signal-to-noise ratio deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms where the reconstruction algorithm continuously compares forward projections from the current image estimate against actual measured projections, calculates residuals, and uses these errors to update the image estimate. This iterative feedback process maintains computational efficiency while improving SNR by systematically reducing reconstruction errors.
Solution Approach 2:
The patent modifies reconstruction parameters including the objective function (using likelihood-based formulations), relaxation parameters, and subset weighting factors. These parameter changes enable the algorithm to achieve better convergence properties and higher SNR while maintaining faster reconstruction speeds compared to traditional methods.
3Measurement precision
If voxel-dependent update factor is introduced in iterative reconstruction, then signal-to-noise ratio and convergence speed are improved, but computational complexity increases
Solution Approach 1:
The patent introduces a voxel-dependent update factor that assigns different weights to different voxels based on their local characteristics and contribution to forward projections. This local quality approach ensures that voxels contributing more to projection errors receive larger updates, improving convergence speed and SNR while the computational overhead remains manageable through efficient implementation.
Data Source
AI summary
The invention relates to a device and a method for the iterative reconstruction of the attenuation coefficients μj in a tomographic image of an object (1) from projection measurements mi. In the update equation for μjn during the n-th iteration the backprojected error (mi−mi−(μjn)) is weighted by a voxel dependent factor Formula (I). Such a voxel dependent update may particularly be included in the algorithms of ART or ML.