A comparison of optimization algorithms for localized in vivo B₀ shimming

Area of Science:

• Magnetic Resonance Imaging
• Biophysics

Background:

• Accurate B0 field homogeneity is crucial for high-resolution in vivo magnetic resonance spectroscopy (MRS).
• Localized B0 shimming aims to optimize magnetic field uniformity within a specific region of interest.
• Existing optimization algorithms for B0 shimming can be sensitive to initial conditions and ill-conditioned problems.

Purpose of the Study:

• To compare the performance of various B0 shimming optimization algorithms.
• To introduce and evaluate a novel constrained regularized algorithm, ConsTru, for localized in vivo B0 shimming.

Main Methods:

• Ten optimization algorithms, including least-squares solvers and the ConsTru method, were implemented.
• Algorithms were tested on 66 in vivo datasets across 7T and 9.4T scanners.
• The optimal algorithm was selected for single-voxel spectroscopy in the human frontal cortex at 9.4T.

Main Results:

• Shimming algorithms dependent on starting points or lacking regularization showed unstable solutions.
• The ConsTru algorithm demonstrated superior robustness, speed, and efficiency compared to other tested methods.
• High-quality and reproducible spectra were acquired in the frontal cortex using ConsTru at 9.4T.

Conclusions:

• Dedicated linear least-squares solvers are recommended over generic nonlinear solvers for localized in vivo B0 shimming.
• The constrained regularized method (ConsTru) is identified as a fast and highly robust linear solver for B0 shimming.
• ConsTru enables reliable and efficient B0 shimming for advanced in vivo magnetic resonance applications.