MRI Reconstruction via K-Space Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current magnetic resonance imaging (MRI) techniques face challenges in accelerating data acquisition in radial imaging without compromising resolution, contrast, and artifact susceptibility, particularly due to the need for extensive coil calibration and increased noise in reconstructed images.
Innovation Solution
The method involves defining a grid in k-space with equidistant sample points along one-dimensional edges, allowing for separate reconstruction of missing data in internal and external regions using raw data from the same magnetic resonance coil, reducing the need for multiple coil data and simplifying the calculation of GRAPPA weights, thereby accelerating image reconstruction while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If parallel imaging methods are used to accelerate data acquisition in radial imaging, then acquisition time is reduced, but the complexity of coil calibration increases and noise in reconstructed images increases
Solution Approach 1:
The patent segments the k-space into internal and external regions with different grid dimensions. The internal region uses a finer grid for accurate reconstruction, while the external region uses a coarser grid for faster processing. This segmentation allows the method to achieve acceleration without requiring full coil calibration across the entire k-space, thereby reducing calibration complexity while maintaining image quality.
Solution Approach 2:
The patent applies different reconstruction strategies to different regions of k-space. The internal region (containing central k-space points) is reconstructed with higher precision using raw data from the same coil, while the external region is reconstructed with lower precision using combined coil data. This local differentiation reduces the overall computational burden and calibration requirements while preserving image quality in critical regions.
2Loss of time
If parallel imaging methods are used to accelerate data acquisition, then acquisition time is reduced, but noise in reconstructed images increases
Solution Approach 1:
By segmenting k-space into internal and external regions, the patent applies appropriate reconstruction methods to each region. The internal region benefits from high-precision reconstruction using same-coil data, while the external region uses efficient reconstruction with combined coil data. This segmentation prevents noise propagation across the entire image while maintaining acceleration benefits.
Solution Approach 2:
The patent uses an intermediary approach by combining raw data from multiple coils in a weighted manner for the external region reconstruction. This intermediary combination strategy allows the method to achieve acceleration while controlling noise levels, as the weighting factors are optimized to minimize noise amplification while maintaining image quality.
3Manufacturing precision
If conventional GRAPPA methods are used for radial imaging reconstruction, then image reconstruction is achieved, but numerical effort increases significantly
Solution Approach 1:
The patent segments the reconstruction process into two stages: first reconstructing the internal region using raw data from the same coil (low computational cost), then using this intermediate result to reconstruct the external region with combined coil data (moderate computational cost). This segmentation reduces the overall numerical effort compared to conventional GRAPPA while maintaining reconstruction accuracy.
Solution Approach 2:
The patent applies partial reconstruction to the internal region using only the necessary raw data from the same coil, rather than performing full GRAPPA reconstruction across the entire k-space. This partial action approach reduces computational complexity while achieving sufficient accuracy for the critical central region, thereby improving overall productivity.
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 enables exact artifact-free image reconstruction with a significantly reduced numerical effort, comparable to conventional GRAPPA methods, while maintaining the advantages of radial imaging's reduced movement sensitivity and shortening acquisition periods.
Implementation Method 1
radio frequency excitation signals (RF pulses) are emitted via a radio frequency transmission system, which result in the nuclear spins of specific nuclei being resonantly excited (i.e., at the Larmor frequency that exists at the specific location) so as to be deflected by a defined flip angle with respect to the magnetic field lines of the basic magnetic field
Implementation Method 2
a gradient system, superimposes a magnetic field gradient thereon, allowing the magnetic resonance frequency (Larmor frequency) of nuclear spins at a respective location to be determined
Implementation Method 3
In the process of relaxing the excited nuclear spins, radio frequency signals (magnetic resonance signals) are resonantly emitted, which are received by an appropriate receiving antenna (also called magnetic resonance coils)
Data Source
AI summary
In a method and a reconstruction device for reconstructing an image from MR raw data acquired with multiple coils and entered at sample points on a grid in k-space, the sample points are arranged on data entry trajectories, respectively, along one-dimensional edges in an equidistant grid dimension characteristic for the respective edge. If raw data were acquired at all sample points, k-space would be sufficiently scanned, but the raw data are entered only on a portion of the sample points so that sufficient scanning exists only in an internal region of k-space, with undersampling existing in an external region of k-space. Reconstruction of the missing raw data is performed by reconstructing raw data for a specific coil for the non-sampled sample points in the internal region using the raw data acquired with that coil for other sample points in the internal region, without using raw data acquired with other coils, and for the non-sampled sample points in the external region, using the raw data acquired for the internal region, as well as reconstructed raw data, and using raw data acquired with different coils.