Dynamic MR Reconstruction With Time-Compressed Non-Cartesian K-Space
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning methods for reconstructing dynamic non-Cartesian MRI sequences face challenges due to high computational and memory requirements, particularly in processes involving NuFFT, and lack of data-consistency steps, which affect reconstruction quality.
Innovation Solution
A time compression network is employed to manage complexity by reducing the time dimensionality of k-space data using a neural network, followed by an unrolled iterative reconstruction with data-consistency steps, and orthogonalization procedures like Gram-Schmidt orthonormalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deep learning methods are used to reconstruct dynamic non-Cartesian MRI sequences, then reconstruction quality can be improved, but computational and memory requirements increase significantly due to NuFFT operations
Solution Approach 1:
The patent segments the reconstruction process into distinct modules: a deep learning network for initial reconstruction, a data consistency check module, and an iterative refinement module. This segmentation allows each component to be optimized independently, reducing overall computational complexity while maintaining reconstruction quality.
Solution Approach 2:
The patent performs preliminary reconstruction using a deep learning network before applying data consistency checks and iterative refinement. This preliminary action provides a good initial estimate that reduces the number of iterations needed in subsequent steps, thereby reducing total computational requirements.
2Manufacturing precision
If data-consistency steps are included in the reconstruction process, then reconstruction quality improves, but computational complexity and processing time increase
Solution Approach 1:
The patent implements a limited number of data consistency check iterations rather than exhaustive processing. This partial action provides sufficient reconstruction quality improvement while avoiding excessive processing time, achieving an optimal balance between quality and efficiency.
3Device complexity
If the time dimensionality of k-space data is reduced, then computational footprint decreases, but information loss may occur
Solution Approach 1:
The patent transforms the time-dimensional compression problem into a spatial transformation problem by applying orthogonalization operations. This dimensionality change allows compression without direct information loss, as the orthogonal transformation preserves all information in a different representation.
Solution Approach 2:
The patent introduces orthogonalization procedures as an intermediary step between data compression and reconstruction. This intermediary ensures that compression is performed in a mathematically rigorous way that preserves information, acting as a mediator between the conflicting requirements of compression and information retention.
Data Source
AI summary
Systems and methods for reconstruction for a medical imaging system. Non-Cartesian k-space data is acquired using a dynamic MR sequence. A time compression network compresses the non-Cartesian data. The compressed data is used for reconstruction of an image. The time compression network is configured to reduce the (time and memory) complexity of the reconstruction process.