Implicit GROG Kernel Reconstruction for MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional MRI reconstruction from non-Cartesian data sampling is complex and time-consuming, especially when dealing with field imperfections, leading to noise amplification and reconstruction artifacts.
Innovation Solution
The method employs an ensemble of linear estimators to transform non-Cartesian data into Cartesian data, using multiple neighboring non-Cartesian k-space coordinates to estimate a single Cartesian k-space coordinate, and utilizes a neural network to implicitly represent the collection of linear estimators for efficient interpolation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If non-Cartesian sampling is used to improve acquisition efficiency, then scan time is reduced, but reconstruction complexity increases and becomes time-consuming
Solution Approach 1:
The patent introduces Cartesian k-space coordinates as an intermediary representation between non-Cartesian sampling and final image reconstruction. By mapping non-Cartesian samples to the nearest Cartesian grid points and using Cartesian-based reconstruction algorithms, the method maintains the acquisition efficiency of non-Cartesian sampling while achieving the computational simplicity of Cartesian reconstruction.
2Productivity
If non-Cartesian sampling is used to improve acquisition efficiency, then scan time is reduced, but noise amplification and reconstruction artifacts occur
Solution Approach 1:
The patent employs feedback mechanisms through iterative reconstruction algorithms that adjust the weighting and selection of non-Cartesian samples based on their contribution to the final image. By evaluating reconstruction quality metrics and adjusting sampling weights accordingly, the method reduces noise amplification and artifacts while maintaining fast acquisition.
Solution Approach 2:
The patent dynamically adjusts reconstruction parameters such as interpolation kernels, weighting factors, and regularization terms based on the specific non-Cartesian trajectory and sampling density. This adaptive parameter adjustment optimizes reconstruction quality for different sampling patterns while maintaining computational efficiency.
3Device complexity
If single input sample linear estimators are used in GROG to simplify reconstruction, then computational load is reduced, but estimation errors increase leading to noise amplification
Solution Approach 1:
The patent merges multiple neighboring non-Cartesian k-space samples to estimate a single Cartesian k-space coordinate, rather than using single-input estimators. By combining information from multiple samples through weighted averaging or iterative optimization, the method reduces estimation errors and noise amplification while maintaining computational feasibility.
Solution Approach 2:
The patent transitions from one-dimensional linear estimators to multi-dimensional estimation by considering samples from multiple spatial dimensions and orientations. This dimensional expansion allows the use of ensemble estimators that leverage correlations across different directions, improving accuracy without proportionally increasing computational load.
4Reliability
If ensemble of linear estimators with multiple neighboring coordinates is used to improve estimation accuracy, then noise amplification is reduced, but computational complexity increases
Solution Approach 1:
The patent segments the estimation process into multiple stages: first identifying neighboring non-Cartesian samples, then applying weighted estimation, and finally interpolating to Cartesian grid points. This segmentation allows the use of efficient algorithms at each stage, reducing overall computational complexity while maintaining the benefits of ensemble estimation.
Solution Approach 2:
The patent uses a limited number of neighboring samples (partial action) rather than all available data, selecting only those within a defined spatial window or distance threshold. This partial sampling approach maintains reconstruction quality while significantly reducing computational complexity compared to using all neighboring points.
Data Source
AI summary
A method for magnetic resonance imaging includes a) performing by an MRI scanner an MRI scan to acquire non-Cartesian k-space MRI acquisition data; b) estimating by the MRI scanner Cartesian k-space data from the non-Cartesian k-space MRI acquisition data, wherein the estimating comprises estimating each Cartesian k-space coordinate in the Cartesian k-space data from multiple neighboring non-Cartesian k-space coordinates using an ensemble of GRAPPA kernels, where each of the GRAPPA kernels is obtained from a non-linear model trained on calibration data from an MRI calibration scan; and c) reconstructing by the MRI scanner an MRI image from the estimated Cartesian k-space data.