Accelerating Non-Cartesian MRI Reconstruction Convergence Using k-Space Preconditioning.

Area of Science:

• Medical Imaging
• Computational Science
• Signal Processing

Background:

• Iterative Magnetic Resonance Imaging (MRI) reconstructions from non-uniformly sampled k-space data are computationally intensive.
• Current acceleration methods often compromise reconstruction accuracy or increase computational cost per iteration.

Purpose of the Study:

• To develop a novel k-space preconditioning formulation for accelerating MRI reconstruction convergence.
• To overcome the limitations of existing methods regarding accuracy and computational efficiency.

Main Methods:

• A dual formulation of the reconstruction problem is utilized to enable k-space preconditioning.
• Density-compensation-like operations are employed for preconditioning in k-space.
• The primal-dual hybrid gradient method is used, avoiding inner loops for efficiency.

Main Results:

• The proposed preconditioning method accelerates convergence without sacrificing reconstruction accuracy.
• The method demonstrates competitive convergence rates compared to existing algorithms.
• Experimental results show practical convergence within approximately ten iterations.

Conclusions:

• The novel k-space preconditioning formulation offers a significant advancement in accelerating MRI reconstructions.
• This approach provides an accurate and computationally efficient solution for handling non-uniformly sampled k-space data.
• The method holds promise for faster and more reliable MRI scans.