An effective beamforming algorithm for a GPU-based ultrasound imaging system

Abstract

In this paper, four beamforming algorithms (i.e., interpolation and phase rotation with pre-and post-filtering, IBF-PRE, IBF-POST, PRBF-PRE and PRBF-POST, respectively) implemented on a high-performance graphicsprocessing unit (GPU) were presented. Each beamforming method was divided into two kernels consisting of various beamforming and mid-processing blocks and efficiently implemented on a NVIDIA's Computer Unified Device Architecture (CUDA) platform (GeForce GTX560 Ti, NVIDIA, Santa Clara, CA, USA). To evaluate the performance of each method, pre-beamformed radio-frequency (RF) data were captured by a commercial ultrasound machine equipped with a research package (G40, Siemens Healthcare, Mountain View, CA, USA) from a tissue mimicking phantom. The execution time for each beamforming algorithm was measured by using a time stamp produced by a CUDA timer. The IBF-PRE outperforms over other methods (i.e., IBF-POST, PRBF-PRE, PRBF-POST), in terms of execution time, i.e., 7.89 ms vs. 16.19 ms, 21.89 ms, and 10.62 ms, respectively. This result indicates that the IBF-PRE method is suitable for the fully software based ultrasound imaging system.

Efficient Parallel-Beamforming Based on Shared FIFO for Ultra-Compact Ultrasound Imaging Systems

“…The consideration of software implementation would be an interesting point to further investigate. Previously, there was a comprehensive comparison between various architectures for sBF [38], which indicates that the post-fractional filteringbased dynamic receive beamforming method, the basic architecture used in the pBF-sFIFO method, yields the inferior execution time due to more memory access. However, when it comes to the realization of pBF, the memory access could be significantly reduced by the proposed pBF-sFIFO method since the single block data sets fetched in the shared memory can be effectively accessed by multiple parallel-beamforming paths.…”