Low-Power Ultrasound Imaging on Mixed FPGA/GPU Systems

Abstract

Portable and hand-held ultrasound imagers have the potential to revolutionize Point-of-Care medical diagnostics. There is great need for low-cost, portable scanners with extended battery life. In this paper, we focus on hardware-software partitioning in heterogeneous systems where both field-programmable gate array (FPGA) and graphics processing unit (GPU) resources are available. We present the architecture of a prototype test scanner for the evaluation of various hardware-software partitioning strategies. The system is equipped with the Intel Arria 10 FPGA and the Nvidia Tegra X2 mobile GPU. FPGA-based beamformers: Delay-and-Sum and Filtered Multiply-and-Sum, were implemented. These 32-channel beamformer blocks are integrated into a complete dataflow along with the data acquisition, RF filter, quadrature demodulator, and envelope detector. The designed dataflow allows one to allocate processing functions to either hardware (FPGA) or software (GPU) to explore various imaging scenarios and optimize power consumption. A dedicated measurement setup facilitates measuring power consumption of both FPGA and GPU. The developed setup will provide a reliable experimental system power characterization.