Compressed Sensing for Diffusion Spectrum Imaging Scan Time
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Diffusion Spectrum Imaging (DSI) techniques require extended scanning times due to their high dimensionality, limiting their effectiveness in vivo.
Innovation Solution
The method involves acquiring MR signals at undersampled q-space locations and using compressed sensing techniques to synthesize data, allowing for the generation of evenly distributed q-space encodings, which reduces acquisition time while maintaining resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional DSI techniques are used to generate diffusion information, then acceptable information for diffusion properties is provided, but the scanning time becomes excessively long
Solution Approach 1:
The patent applies partial action by acquiring MR signals at only a subset of q-space locations rather than all possible locations. Specifically, the method acquires signals at undersampled q-space points and uses compressed sensing to reconstruct the remaining data, thereby reducing scan time while maintaining acceptable diffusion information quality.
Solution Approach 2:
The patent introduces compressed sensing as an intermediary technique between the undersampled q-space data and the final diffusion information extraction. This intermediary method enables reconstruction of complete q-space data from incomplete measurements, resolving the contradiction between reduced scan time and maintained information quality.
2Adaptability or versatility
If high dimensionality DSI is performed to characterize water diffusion in 3D q-space, then comprehensive diffusion properties are obtained, but the scan time increases substantially
Solution Approach 1:
The patent performs partial action by sampling only a portion of the required q-space locations. Instead of acquiring complete 3D q-space data, the method acquires signals at strategically selected undersampled points and uses compressed sensing to infer the remaining data, maintaining comprehensive diffusion characterization while reducing scan time.
Solution Approach 2:
The patent changes the sampling parameters by transitioning from uniform Nyquist-rate sampling to non-uniform undersampled sampling. This parameter change allows reduction of the number of required measurements while maintaining the ability to characterize diffusion properties through compressed sensing reconstruction.
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 shortens scan time while maintaining image quality and resolution, optionally increasing resolution within the same scan time as conventional methods.
Implementation Method 1
acquiring the MR signal at a plurality of q-space locations
Implementation Method 2
synthesizing the MR signal for the entirety of for q-space encodings using a compressed sensing technique
Implementation Method 3
The MR signal in q-space is generally related to the water displacement probability density function at a fixed echo time by the Fourier transform
Data Source
AI summary
A method of generating a magnetic resonance (MR) image of a tissue includes acquiring MR signals at undersampled q-space encoding locations for a plurality of q-space locations that is less than an entirety of the q-space locations sampled at the Nyquist rate, wherein the acquired signal at the q-space locations represents the three-dimensional displacement distribution of the spins in the imaging voxel, synthesizing the MR signal for the entirety of q-space encodings using a compressed sensing technique for a portion of q-space locations at which MR data was not acquired, combining the acquired MR signals at q-space encodings and the synthesized MR signals at q-space encodings to generate a set of MR signals at q-space encodings that are evenly distributed in q-space, using the set of MR signals at q-space encodings to generate a function that represents a displacement probability distribution function of the set of spins in the imaging voxel, and generating an image of the tissue based on at least a portion of the generated function. A system and computer readable medium are also described herein.