Assessing electromagnetic (EM) interactions between radiofrequency (RF) fields and implanted electrophysiological recording devices is important for magnetic resonance imaging (MRI) safety and compatibility during simultaneous MRI. This is difficult for micrometre-scale neural interfaces, where features spanning tens to hundreds of micrometres must be resolved within centimetre-scale RF coil and subject models. We present a hybrid EM simulation framework combining the Huygens’ Box (HB) method with three-dimensional finite-difference time-domain (FDTD) sub-gridding to model microscopic, ultra-thin implant structures at high resolution while reducing computational. Using intracortical and epicortical graphene solution-gated field-effect transistor (gSGFET) arrays as representative worst-case microscale test structures, RF coil models of a cylindrical phantom, mouse, and rat were simulated with and without the arrays present. Transmit magnetic field, electric field, and specific absorption rate distributions were benchmarked against conventional full-wave multi-port FDTD simulations using voxel-wise percentage differences and Bland-Altman analysis. At matched resolution, the HB approach reduced runtime by 2- to 4.3-fold relative to the MP approach across phantom and mouse models, with the largest gains in the array-included, high-resolution configuration. HB simulations preserved key spatial features, including field symmetry and localised hotspot regions. Differences relative to full-wave simulations were small for the transmit magnetic field and modest for the electric field and SAR. Absolute values remained below safety limits, and the gSGFET arrays produced limited perturbation of RF transmission and energy deposition. The HB approach also enabled higher-resolution mouse and rat simulations that were not feasible with the conventional MP approach. These results show that the hybrid framework provides a computationally feasible approach for evaluating EM interactions between MRI RF systems and micrometre-scale neural interfaces. The method supports high-resolution multi-scale simulation where conventional full-wave modelling is impractical and provides a workflow for RF safety evaluation and design optimisation of graphene-based neural implants.