计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (19): 140-144.

• 数据库、信号与信息处理 • 上一篇    下一篇

基于CUDA的超声脉冲多普勒成像

尹华国1,何兴无1,2,周洪林1   

  1. 1.成都农业科技职业学院 电子信息分院,成都 611130
    2.四川大学 计算机学院,成都 610065
  • 出版日期:2012-07-01 发布日期:2012-06-27

Ultrasound pulsed wave doppler based on CUDA

YIN Huaguo1, HE Xingwu1,2, ZHOU Honglin1   

  1. 1.School of Electronics and Information, Chengdu Vocational College of Agricultural Science and Technology, Chengdu 611130, China
    2.School of Computer Science, Sichuan University, Chengdu 610065, China
  • Online:2012-07-01 Published:2012-06-27

摘要: 医学超声脉冲多普勒成像模式是在临床超声成像系统中获得人体血管中血流分布情况的一种重要的检测工具,与传统的B超,彩超成像模式不同,超声脉冲多普勒成像模式不仅可以通过频谱图显示表示流过取样容积的血流速度变化和测定某一位置的血流,而且相比较于连续波式多普勒模式它可以消除多普勒信号的混叠效应提高检测的空间分辨率。但是脉冲多普勒系统在处理时涉及大量的复杂运算,例如FFT(快速傅里叶变换)和卷积运算等,使其难于应用到临床实时系统中。为此研究并提出了一种基于统一计算设备架构(CUDA)平台的超声脉冲多普勒成像系统的并行处理算法。该算法包括了壁滤波、频谱估计、移频处理和频谱显示后处理等处理步骤的并行实现。数据实验结果表明,基于CUDA的超声脉冲多普勒成像处理结果与基于CPU的实现相比,不仅可以得到相同质量的频谱图,而且可以取得较大的加速效果,满足实时系统需求;数据测试显示,对于65 535×20的信号数据能够达到1秒处理2 770条谱线的计算性能,速度提高了约140倍。

关键词: 高性能并行计算, 超声脉冲多普勒成像, 图形处理器, 图像并行处理算法

Abstract: Medical ultrasound Pulse Wave(PW) spectrum Doppler imaging system is a valuable tool for clinical diagnosis for flow velocity distribution in vessels. It can not only show the velocity variability of blood flow in the sampling volume and estimate the position of the specific blood flow through the displayed spectrum while the traditional imaging mode can’t like B-mode and color mode, but also improve the spatial resolution for these detections for it can remove the alias of Doppler signal which is much better than the  Continuous Wave(CW) Doppler. Thus, there are significant meanings in clinical application. However, because of the massive computation involved in the current PW Doppler spectrum ultrasound imaging system, it will be hard and costly to implement with the traditional processing platform, such as FFT(Fast Fourier Transformation), convolution and so on. A new algorithm of PW Doppler spectrum ultrasound based on CUDA(Compute Unified Device Architecture) parallel processing platform is presented. This method  mainly includes the following parallel procedures such as wall filter,  spectrum averaging, frequency shift, post-processing and so on. Test results from the Doppler I/Q data, not only show the output of Graphics Processing Unit(GPU) is definitely the same as the one of CPU, but also demonstrate the obvious speedup using GPU, that is, it can process 2 770 spectrums per second for the data size(65 535×20 sample points) which is 140 times faster than the CPU implementation.?

Key words: high performance parallel processing, ultrasound pulsed wave Doppler, Graphics Processing Unit(GPU), parallel algorithm for image processing