Subspace K-Space Parallel Imaging for Aliasing-Robust MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing magnetic resonance imaging (MRI) methods face challenges with aliasing issues in parallel imaging, particularly in k-space, leading to artifacts and computational intensity, especially when dealing with unexpected aliasing and multi-channel spaces.
Innovation Solution
The implementation of subspace convolutional kernels for combining information across neighboring k-space locations and multiple channels to generate subspace compressed k-space data, using deep learning techniques to learn and apply these kernels for efficient data reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If k-space versions of parallel imaging (e.g., GRAPPA, ARC, SPIRIT) are used, then robustness to aliasing is improved, but computational intensity increases and challenges for effective regularization arise
Solution Approach 1:
The patent transforms the parallel imaging problem from multi-channel k-space domain to a reduced subspace domain. By identifying and exploiting the inherent low-rank structure of the sensitivity encoding problem, the method projects the high-dimensional multi-channel data onto a lower-dimensional subspace spanned by the dominant singular vectors. This dimensional reduction maintains aliasing robustness while dramatically reducing computational complexity and enabling effective regularization in the compressed subspace.
2Device complexity
If image space versions of parallel imaging (e.g., SENSE, ESPIRIT) are used, then channel combination benefits are achieved, but aliasing issues cause reconstruction failures
Solution Approach 1:
The patent introduces a subspace projection as an intermediary step between raw multi-channel k-space data and final image reconstruction. Instead of directly combining channels in image space (which fails with aliasing) or working fully in k-space (which is computationally intensive), the method projects data onto a calibrated subspace that captures the essential signal structure. This intermediate representation enables both efficient channel combination and robust handling of aliased data through subsequent reconstruction algorithms.
3Reliability
If ESPIRIT is used to derive multiple sensitivity maps, then handling of additional aliasing is improved, but computational complexity and artifacts increase
Solution Approach 1:
The patent extracts only the most significant subspace components (dominant singular vectors) from the sensitivity encoding data, rather than computing multiple full sensitivity maps as in ESPIRIT. By taking out and retaining only the essential low-rank structure that captures the majority of signal energy, the method achieves effective aliasing handling while avoiding the computational burden and phase singularity artifacts associated with deriving and processing multiple complete sensitivity maps.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances image quality by being robust to aliasing, allows higher acceleration rates, reduces reconstruction time, and minimizes memory overhead, resulting in faster and more accurate MRI scans.
Implementation Method 1
the resulting set of received nuclear magnetic resonance (NMR) signals are digitized and processed
Implementation Method 2
magnetic field gradients (Gx, Gy, and Gz) are employed. Typically, the region to be imaged is scanned by a sequence of measurement cycles in which these gradient fields vary
Data Source
AI summary
A computer-implemented method includes obtaining, via a processing system including one or more processors, multi-channel k-space data of a subject acquired with a magnetic resonance imaging scanner. The computer-implemented method also includes utilizing, via the processing system, subspace convolutional kernels on the multi-channel k-space data to combine information from local neighboring k-space locations and across multiple channels to generate subspace compressed k-space data having fewer channels than a number of channels utilized to acquire the multi-channel k-space data.