Optimization of a boundary element approach to electromagnet design with application to a host of current problems in Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) has proven to be a valuable methodological approach in both basic research and clinical practice. However, significant hardware advances are still needed in order to further improve and extend the applications of the technique. The present dissertation predominantly addresses gradient and shim coil design (sub-systems of the MR system). A design study to investigate gradient performance over a set of surface geometries ranging in curvature from planar to a full cylinder using the boundary element (BE) method is presented. The results of this study serve as a guide for future planar and pseudo-planar gradient systems for a range of applications. Additions to the BE method of coil design are developed, including the direct control of the magnetic field uniformity produced by the final electromagnet and the minimum separation between adjacent wires in the final design. A method to simulate induced eddy currents on thin conducting surfaces is presented. The method is used to predict the time-dependent decay of eddy currents induced on a cylindrical copper bore within a 7 T MR system and the induced heating on small conducting structures; both predictions are compared against experiment. Next, the method is extended to predict localized power deposition and the spatial distribution of force due to the Lorentz interaction of the eddy current distribution with the main magnetic field. New methods for the design of actively shielded electromagnets are presented and compared with existing techniques for the case of a whole-body transverse gradient coil. The methods are judged using a variety of shielding performance parameters. A novel approach to eliminate the interactions between the MR gradient system and external, non-MR specific, active devices is presented and its feasibility is discussed. A completely new approach to shimming is presented utilizing a network of current pathways that can be adaptively changed on a subject-by-subject basis and dynamically controlled. The potential benefits of the approach are demonstrated using computer simulations and a prototype coil is constructed and tested as a proof-of-principle