MR Image Reconstruction Using Hybrid Sampling Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MR imaging techniques face challenges in acquiring MR images with desired spatial resolution and signal-to-noise ratio in a short acquisition time while minimizing artifacts caused by undersampling.
Innovation Solution
A method that combines parallel imaging and iterative reconstruction by using a coherent sampling pattern in the central raw data region and an incoherent sampling pattern in the outer region, incorporating coil characteristics into the reconstruction process to effectively deconvolute larger structures and suppress artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If coherent sampling pattern is used in the central raw data region, then the acquisition time is reduced and parallel imaging advantages are utilized, but undersampling artifacts may occur
Solution Approach 1:
The patent applies different sampling patterns to different regions of the raw data space: coherent sampling is used in the central region where low spatial frequency information resides, while incoherent sampling is used in the outer regions containing high spatial frequency information. This local differentiation allows the system to benefit from the speed of coherent sampling in the central region while using incoherent sampling in the periphery to suppress undersampling artifacts, thereby resolving the contradiction between acquisition time and artifact suppression.
2Object-affected harmful factors
If incoherent sampling pattern is used in the outer raw data region, then artifacts are suppressed, but the ability to deconvolute larger structures is reduced
Solution Approach 1:
The patent differentiates the sampling approach by spatial region: the central raw data region is sampled coherently to preserve the low spatial frequency information necessary for deconvolving large structures, while the outer regions are sampled incoherently to suppress radio-frequency artifacts. This regional differentiation resolves the contradiction by assigning each region the sampling pattern most suited to its information content.
3Measurement precision
If complete sampling of raw data space is performed, then image quality and spatial resolution are maximized, but measurement time increases significantly
Solution Approach 1:
The patent changes the sampling pattern parameter from fully coherent (parallel imaging) to a hybrid approach combining coherent and incoherent sampling. By applying incoherent sampling to the outer regions of k-space while maintaining coherent sampling in the center, the system achieves a balance that preserves sufficient spatial resolution information while reducing the overall measurement time compared to complete sampling.
Data Source
AI summary
In a method to generate magnetic resonance (MR) images of an examination subject, MR signals are detected simultaneously with multiple coils, each coil having its own coil characteristic. In the detection of the MR signals, raw data space is incompletely filled with MR signals; with raw data space being undersampled in a central raw data region with a coherent acquisition pattern that is composed of a spatially repeating set of raw data points; and raw data space outside of the central raw data region is sampled with an incoherent acquisition pattern. The MR image is reconstructed from the detected MR signals, step-by-step in an iterative reconstruction procedure using a reconstruction matrix A, starting from an initial estimate; wherein the reconstruction matrix has continuing information about the coil characteristics with which the MR signals were detected.