GPU accelerated FDTD solver and its application in B1-shimming

Introduction FDTD is widely used for EM analyses in MRI applications, due to its simplicity and efficiency in wave modelling and its ability to handle field-tissue interactions. However, contemporary MRI problems, such as high field-tissue interaction modelling requires high performance computing. Conventional CPU-based FDTD calculations offer compromised computing performance without expensive super-computer support. This study extends our recent works on CPU-base FDTD simulations into a Graphics Processing Unit (GPU)-based parallel-computing framework, producing substantially boosted computing efficiency at only PC-level cost. The new computational strategy enables intensive computing feasible for solving forward-inverse EM problems in modern MRI, as illustrated in the high-field B1shimming investigation presented herein. Moreover, the new rotating RF excitation technique proposed here can compensate for B1 inhomogeneities while simultaneously controlling SAR, and as such, may have a number of applications in high-field MRI.