MRI Nyquist Ghost Reduction via 2D Convolution Kernels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Magnetic resonance imaging (MRI) using echo planar imaging (EPI) often suffers from Nyquist ghost artifacts due to phase inconsistencies between odd and even echoes, which traditional one-dimensional phase corrections fail to fully address, leaving residual artifacts.
Innovation Solution
A method involving a computing device that determines convolution kernels from measured data sets to generate synthetic and combined data sets, allowing for two-dimensional phase correction and iterative processing to reduce or eliminate Nyquist ghost artifacts in MRI images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-dimensional phase correction is applied to correct phase errors along the readout direction, then phase inconsistencies are partially corrected, but residual Nyquist ghost artifacts remain
Solution Approach 1:
The patent extends the phase correction from one-dimensional (readout direction only) to two-dimensional by incorporating both readout and phase encoding directions. The 2D convolution kernels are applied in k-space to simultaneously correct phase errors in both spatial dimensions, thereby eliminating residual Nyquist ghost artifacts that persist after 1D correction.
2Productivity
If traditional one-dimensional phase correction techniques are used, then the correction process is simple and fast, but Nyquist ghost artifacts cannot be fully removed
Solution Approach 1:
The patent introduces two-dimensional convolution kernels that operate in both readout and phase encoding directions within k-space. This 2D approach maintains computational efficiency while achieving complete removal of Nyquist ghost artifacts, overcoming the limitation of 1D techniques that leave residual artifacts.
3Measurement precision
If two-dimensional phase correction is implemented to remove residual artifacts, then image quality improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary 1D phase correction along the readout direction before applying the 2D convolution kernels. This preliminary action simplifies the subsequent 2D correction by pre-aligning the phase errors, thereby reducing the overall computational complexity while maintaining the artifact removal effectiveness.
Solution Approach 2:
The patent divides the phase correction process into two segments: first applying 1D correction along the readout direction, then applying 2D convolution kernels to correct remaining errors in both readout and phase encoding directions. This segmentation makes the complex 2D correction more computationally manageable.
4Measurement precision
If iterative processing is performed to update combined data sets, then phase correction accuracy is enhanced, but processing time increases
Solution Approach 1:
The patent implements an iterative feedback mechanism where combined data sets are generated by merging synthetic and measured data, then fed back into the convolution kernel application process. This feedback loop progressively refines the phase correction accuracy by continuously updating the combined data sets with corrected information from previous iterations.
Data Source
AI summary
A method and system for reducing Nyquist ghost artifact is provide. The method may include: obtaining a plurality of measured data sets; determining, based on the plurality of measured data sets, in a data space, a plurality of convolution kernels, each convolution kernel relating to all of the plurality of measured data sets; generating, based on the plurality of convolution kernels and the plurality of measured data sets, in the data space, a plurality of synthetic data sets; generating, based on the plurality of synthetic data sets and the plurality of measured data sets, in the data space, a plurality of combined data sets, each combined data set relating to one of the plurality of synthetic data sets and a corresponding measured data set of the plurality of measured data sets; and reconstructing, based on the plurality of combined data sets, an image.