Gradient coil design for a 0.23 T NICU MRI system using an improved two-step Target-Field Method with enhanced...

This study presents an optimized gradient coil design for a miniature 0.23 T MRI system, aimed at improving absolute and relative magnetic field linearity while accommodating various gradient thicknesses. The design uses a two-step optimization approach: the first step uses Tikhonov regularization to solve a linear problem, providing a stable solution, and the second step refines the solution through nonlinear constrained optimization to further enhance field linearity. An explicit objective function for the inductance matrix of biplanar gradient coils is simplified to enhance computational efficiency. Validation through MATLAB and COMSOL finite element analysis showed excellent performance. Imaging experiments were conducted on small animals (cats and dogs, whose sizes are similar to neonates) while awaiting ethical approval for human neonatal studies. Results demonstrated that all gradient coils achieved absolute and relative linearity errors below 5%. Cubic phantom scans showed slight displacement at the edges, but the structured phantom MRI lines align precisely with the physical markers, indicating negligible geometric distortion. The shield design maintained Z-leakage fields below 5 Gauss, with eddy current compensation achieving a 90% reduction (residual X/Y-gradient < 0.05%, Z-gradient < 0.20%). T1 and T2-weighted images depicted clear brain structures, while FLAIR and STIR sequences effectively highlighted tissue changes. The proposed gradient coil design method significantly improves absolute and relative linearity while accommodating various gradient thicknesses, demonstrating strong resistance to interference and broad applicability. The comprehensive design-to-manufacturing process ensures optimal parameter selection, resulting in high-quality imaging across multiple MRI sequences. This design demonstrates strong potential for precise in-vivo brain imaging in further NICU applications.