Modular & Scalable Ultrasound Platform with GPU Processing
Department of Ultrasound, Institute of Fundamental Technological Research PAS, Warsaw, Poland
Institute of Fundamental Technological Research PAS, 2012
@article{lewandowski2012modular,
title={Modular & Scalable Ultrasound Platform with GPU Processing},
author={Lewandowski, M. and Walczak, M. and Witek, B. and Kulesza, P. and Sielewicz, K.},
year={2012}
}
The objective of our project is to develop a complete ultrasound platform with real-time GPU processing. The platform is designed to be modular and scalable both in number of ultrasound channels (64-256), as well as in communication bandwidth and processing power. By standardizing on the PCIe switch fabric, we are planning to integrate all the ultrasound modules and processing resources (GPU) in a single rack enclosure. Using PCIe direct peer-to-peer communication for transferring the data from the ultrasound acquisition modules to the GPUs, we maximize the system bandwidth and minimize CPU usage. The first developed module of our platform is RX64 – a 64-channel ultrasound acquisition PCIe card. The RX64 contains a high-end FPGA Altera Stratix IV 70 GX interfaced to: two 32-channels mixed-signal front-end ultrasound modules and two 64-bit 8GB DDR3 SO-DIMM memories for data buffering. We also develop GPU kernels for SAFT based ultrasound imaging, as well as GPU Framework for building complete signal processing pipeline.
October 25, 2012 by hgpu
