Passive shimming performance in 3 T MRI systems: Influence of shim parameters under varying magnet field distributions

This study proposes a simple and computationally efficient method to optimize the structural design parameters of passive shimming slots, aiming to improve magnetic field homogeneity in cryogen-free 3 T/200 mm superconducting magnets used across diverse application environments. The proposed method combines Latin Hypercube Sampling (LHS), utilizing over 300 sampled configurations, with a linear programming (LP)-based optimization framework to explore high-dimensional design spaces while adhering to structural constraints. The method was applied to four distinct magnets, each characterized by unique field inhomogeneity patterns resulting from manufacturing and assembly variations. Through harmonic decomposition, system-specific sensitivities were identified and effectively mitigated using customized passive shimming strategies tailored to each magnet. The optimization process achieved substantial improvements in magnetic field homogeneity, with peak-to-peak (PP) values enhanced to 12.16, 10.04, 27.28, and 54.59 parts per million (ppm) for Magnets 1 to 4, respectively. Correspondingly, the root-mean-square error (RMSE) homogeneity improved to 2.28, 1.98, 5.07, and 9.68 ppm. Furthermore, the magnitudes of all harmonic terms were reduced by 1-2 orders of magnitude, with suppression levels exceeding 90%, while minimizing the use of ferromagnetic materials. The practical feasibility of the proposed strategy was validated on-site: Magnet 1 successfully delivered high-quality animal MRI imaging with excellent signal-to-noise ratios (SNRs), and the remaining magnets are currently undergoing final calibration and delivery. This work presents a robust and scalable optimization framework for precise and resource-efficient passive shimming, offering valuable guidance for future magnet design, customization, and deployment in biomedical and industrial applications.