FPGA Acceleration of 3D FDTD for Multi- Antennas Microwave Imaging Using HLS

Abstract

Microwave Imaging (MI) for biomedical applications has attracted attention due to its harmless radiation compared to X-ray or MRI. One of the commonly used computing methods in MI is Finite Difference Time Domain (FDTD), which is executed several times in iterative loops, hence resulting in a high execution time. Although several hardware accelerators for FDTD have been recently introduced, they are not specifically designed for MI applications. In particular, only simple absorbing boundary conditions have been investigated, and the impact of dispersive materials on FDTD has not been considered. In this paper, we propose a multi-FPGA accelerator for 3D FDTD that is integrated in an MI algorithm, with Convolutional Perfectly Matched Layer (CPML) boundary conditions and an exact model for dispersive materials. By using High Level Synthesis (HLS), we obtain an optimized hardware accelerator that uses an efficient blocking method to reduce the data transfer time between external and local memories. We propose two alternative architectures that trade off performance and resource usage. In addition, our code, being developed at a high level, can also be run on GPUs whenever necessary. The results show that our multi-FPGA accelerator is superior to three similar GPU-based designs in terms of execution time and power consumption.

Multistatic radar-based imaging in layered and dispersive media for biomedical applications

“…For example, the computation time is of paramount importance in the case of stroke detection as the patient seeks urgent help. Various algorithms are used to improve computational efficiency, such as hardware resources, including graphical processing units, field-programmable gate arrays [26] and software frameworks such as Spark and Hadoop [27]. For applications like AD monitoring, the computation time becomes less important than accuracy.…”

Accelerated Medical Microwave Tomography via Heterogeneous Computing

“…The computation in the CPML regions is likewise incorporated into the kernels updating the H and E fields. Other implementations either expand the CPML region to whole domain and assign zero values at the non-boundary regions [32], [33] or update the CPML regions into multiple independent kernels [33], [34]. The first approach increases unnecessarily the memory requirements.…”

“…Moreover, we opted to merge the update of the polarization current with the update of the E field, similar to [34]. This can avoid storing and reloading intermediate values from memory, which are required for approximating the derivative in (4) if the update is done in separate kernels.…”