Variable-Density MRI Sampling for Sparse Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MRI sampling strategies face challenges in achieving asymptotic incoherence in real-world scenarios due to physical and physiological limitations, limiting the effectiveness of compressed sensing techniques for improving image quality and reducing scan time.
Innovation Solution
A method is developed to modify conventional sampling patterns by maximizing the sampled k-space area without increasing scan time, using a base variable-density sampling pattern, such as a spiral phyllotaxis pattern, and applying sparse reconstruction techniques to enhance image quality, particularly for complex anatomical structures and textures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If random sampling schemes are used to achieve incoherence for compressed sensing, then image reconstruction quality improves, but scan time increases due to physical limitations of MRI systems
Solution Approach 1:
The patent modifies sampling pattern parameters (density, distribution, k-space coverage) to achieve asymptotic incoherence while maintaining scan time constraints. By changing the sampling density function and k-space trajectory parameters, the system achieves compressed sensing benefits without requiring fully random sampling that would exceed hardware capabilities.
Solution Approach 2:
The sampling pattern is made adaptive and dynamic rather than static. The system adjusts sampling density and distribution based on image content and reconstruction requirements, allowing optimization of both image quality and scan time through iterative refinement of sampling parameters during the acquisition process.
2Measurement precision
If the number of high-frequency samples in k-space is increased to achieve asymptotic incoherence, then compressed sensing benefits are obtained, but hardware and physiological constraints are violated
Solution Approach 1:
Different regions of k-space are sampled with different densities and patterns. The center region uses one sampling strategy while high-frequency regions use another, allowing each region to be optimized for its specific requirements. This local differentiation enables achievement of asymptotic incoherence in critical regions without violating hardware constraints overall.
Solution Approach 2:
Instead of attempting to achieve perfect incoherence across all of k-space, the patent applies partial incoherence strategies to specific regions where it provides the most benefit. By focusing incoherence achievement on high-frequency regions rather than uniformly applying it throughout, the system obtains compressed sensing benefits while remaining within hardware capabilities.
3Loss of time
If conventional sampling patterns are used, then scan time is maintained within constraints, but image quality and resolution are limited
Solution Approach 1:
The system performs preliminary analysis of the image content and identifies regions that would benefit most from enhanced sampling. Based on this preliminary assessment, it pre-plans the sampling strategy to allocate samples preferentially to high-value regions, thereby improving image quality in critical areas without increasing overall scan time.
Solution Approach 2:
The sampling and reconstruction process incorporates feedback mechanisms where reconstruction quality is continuously assessed and used to guide subsequent sampling decisions. This feedback loop allows the system to adaptively allocate sampling resources to regions that most need them, improving overall image quality while maintaining scan time constraints.
Data Source
AI summary
A computer-implemented method of selecting a Magnetic Resonance Imaging (MRI) sampling strategy includes selecting a base variable-density sampling pattern and determining a scan time associated with the base variable-density sampling pattern. A modified variable-density sampling pattern is created by modifying one or more parameters of the base variable-density sampling pattern to maximize a sampled k-space area without increasing the scan time. Next, a scan is performed on an object of interest using the modified variable-density sampling pattern to obtain a sparse MRI dataset. Then a sparse reconstruction process is applied to the sparse MRI dataset to yield an image of the object of interest.