Radial Parallel MRI Reconstruction Using Expectation Maximization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing parallel MR imaging methods, such as radial GRAPPA and iterative SENSE, face issues with image degradation due to reconstruction coefficient inaccuracies and noise sensitivity, leading to divergent errors in image reconstruction.
Innovation Solution
A radial parallel MR imaging method that uses an expectation maximization technique instead of conjugate gradient methods, calculating sensitivity information and adjusting signals to minimize noise and improve image quality through projection and coefficient calculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If radial GRAPPA method is used for parallel imaging reconstruction, then image acquisition time is shortened and computation requirement is reduced, but image quality is degraded due to reconstruction coefficient errors
Solution Approach 1:
The patent changes the mathematical parameters and algorithms used in the reconstruction process. Specifically, it employs a modified GRAPPA approach with adjusted coefficient calculation methods and incorporates iterative refinement techniques to correct reconstruction errors, thereby improving image quality while maintaining the speed benefits of parallel imaging
Solution Approach 2:
The patent implements feedback mechanisms by iteratively refining the reconstruction coefficients and image data. The method uses the initially reconstructed image to calculate improved sensitivity profiles and reconstruction coefficients, which are then used to generate a corrected image, continuously feedback-looping to reduce errors and enhance quality
2Manufacturing precision
If iterative SENSE method is used for parallel imaging reconstruction, then image resolution is improved, but image error diverges as reconstruction repetitions increase due to noise sensitivity
Solution Approach 1:
The patent modifies the iterative reconstruction parameters and convergence criteria. It adjusts the sensitivity profile calculation, incorporates regularization terms to penalize noise amplification, and optimizes the iterative steps to stop before divergence occurs, thereby maintaining both high resolution and stability
Solution Approach 2:
The patent converts the harmful effect of noise into a beneficial constraint by using noise-aware reconstruction techniques. It incorporates noise modeling and regularization that actually leverage the presence of noise information to guide the reconstruction toward physically plausible solutions, preventing divergence while maintaining resolution
3Ease of operation
If conjugate gradient technique is used for image correction in SENSE method, then initial image reconstruction is achieved, but noise in acquired data causes reconstruction error to diverge
Solution Approach 1:
The patent replaces the conventional conjugate gradient optimization approach with an alternative reconstruction algorithm that is less sensitive to noise. It uses a combination of analytical solutions and iterative refinement that avoids the problematic conjugate gradient steps, substituting a more robust mathematical approach that maintains stability in the presence of noise
Data Source
AI summary
A parallel imaging (PI) method has been frequently used as a method for shortening an image acquisition time in the MRI field. The PI technique is a method for acquiring data using multi-channel coils, that is, several coils, when acquiring the data in MRI. According to this technique, data, the amount of which is smaller than that when the data is obtained using only one coil, is acquired, and then an image is obtained using coil information. According to an embodiment, a new image reconstruction method is proposed which adopts an expectation maximization (EM) technique that is different from the existing GRAPPA or SENSE technique when an image is obtained using PI data acquired through the radial trajectory.