high performance computing on graphics processing units: hgpu.org

To foster innovative uses of micro-ultrasound in biomedicine, it is beneficial to develop flexible research-purpose systems that allow researchers to easily reconfigure its system-level operations such as transmit firing sequence and receive processing. In this paper, we present the development of a programmable micro-ultrasound research platform that is capable of realizing various micro-imaging algorithms. The research platform comprises a linear-array-based scanning front-end and a PC-based data processing back-end, which employs a graphical processing unit (GPU) as the processor core. The front-end operations can be configured from the PC via the parallel port and the two blocks are synchronized by an external clock. Acquired data from the front-end is first digitized and relayed to the PC through an data acquisition card (200 MHz, 14-bit). They are then transferred to the GPU (GTX 275) in which the image formation is carried out via multi-thread processing. Results are displayed on-screen in real-time and can be saved to the PC’s hard disk for offline analysis. Through a module-based programming approach, this platform can facilitate realization of custom-designed imaging algorithms developed by researchers. In this work, B-mode imaging and adaptive color flow imaging have been implemented as demonstrations of the research platform’s programmability. The performance results show that real-time processing frame rates can be achieved for both imaging modes.