Passive shimming of a superconducting magnet using the L1-norm regularized least square algorithm

Area of Science:

• Medical Imaging
• Applied Physics
• Magnetics

Background:

• Uniformity of the static magnetic field (B0) is critical for Magnetic Resonance Imaging (MRI) system performance.
• Passive shimming, using steel shims, is a standard technique to enhance B0 field uniformity.
• Current methods, like linear programming (LP), face challenges in balancing field uniformity and minimizing shim mass due to thermal stability concerns.

Purpose of the Study:

• To develop a new algorithm for passive shimming in MRI systems.
• To better balance the dual constraints of achieving high B0 field uniformity and minimizing the total mass of steel shims.
• To overcome the limitations of the conventional linear programming (LP) method in terms of speed and constraint balancing.

Main Methods:

• Developed a novel algorithm employing the least squares method to minimize magnetic field inhomogeneity.
• Incorporated an L1-norm based constraint to control the total mass of steel used for shimming.
• Tested the algorithm using practical field data from an MRI system.

Main Results:

• The proposed algorithm achieved superior field uniformity compared to the conventional LP approach.
• The new method demonstrated a 43% improvement in field uniformity for the tested case.
• The algorithm maintained similar computational cost and shim material mass while enhancing field quality.

Conclusions:

• The novel algorithm effectively balances the dual constraints of field uniformity and shim mass.
• This approach offers a significant improvement over traditional LP methods for MRI passive shimming.
• The developed technique has the potential to enhance MRI performance and stability.