Model-Based Iterative Reconstruction for MRI EPI Distortion Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Echo planar imaging (EPI) in MRI is sensitive to hardware non-idealities like B0 field inhomogeneity, eddy-currents, and gradient nonlinearity, leading to geometrically distorted images, which existing post-processing methods only partially correct and degrade resolution.
Innovation Solution
A model-based iterative reconstruction method that accesses k-space data and signal model parameters associated with non-idealities, constructing a signal model to optimize an objective function that prospectively accounts for these issues, using nonlinear regularization techniques like wavelet-based sparsity regularization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image-based post-processing methods are used to correct distortions, then geometric accuracy is improved, but resolution is degraded
Solution Approach 1:
The patent applies preliminary action by incorporating distortion correction models into the image reconstruction process itself, rather than performing corrections after image acquisition. The signal model includes terms for B0 field inhomogeneity, eddy currents, and gradient nonlinearity that are accounted for during reconstruction, preventing distortions from occurring in the first place rather than correcting them afterward.
Solution Approach 2:
The patent merges distortion correction with image reconstruction by integrating multiple correction models (B0 inhomogeneity, eddy currents, gradient nonlinearity) into a unified iterative reconstruction framework. This combination allows simultaneous optimization of both geometric accuracy and resolution through a single processing pipeline rather than sequential operations.
2Manufacturing precision
If hybrid reconstruction models are used to leverage image-based corrections, then some distortion correction is achieved, but comprehensive correction of multiple non-idealities is not fully realized
Solution Approach 1:
The patent implements universality by creating a single reconstruction framework that simultaneously handles multiple types of non-idealities (B0 field inhomogeneity, eddy currents, gradient nonlinearity, ramp-sampling) through a unified signal model. This multi-functional approach allows the system to correct all listed distortions using the same iterative reconstruction process rather than requiring separate correction pipelines for each distortion type.
3Speed
If standard EPI reconstruction is used, then speed is maintained, but geometric distortion is introduced
Solution Approach 1:
The patent applies feedback through iterative reconstruction where the signal model continuously refines the image based on the acquired k-space data and the distortion models. The iterative process adjusts the image estimate until convergence, with each iteration incorporating feedback from the distortion correction models to progressively improve geometric accuracy while maintaining computational efficiency.
Data Source
AI summary
Images are reconstructed from k-space data using a model-based image reconstruction that prospectively and simultaneously accounts for multiple non-idealities in accelerated single-shot-EPI acquisitions. In some implementations, nonlinear regularization (e.g., sparsity regularization) is also incorporated to mitigate noise amplification. The reconstructed images have reduced distortions and noise amplification effects relative to those images that are processed using conventional post-reconstruction techniques to correct for non-idealities.