超声稀疏矩阵波束合成方法及其GPU并行实现
首发时间:2016-01-11
摘要:传统的逐列聚焦成像限制了超声成像的帧频。线阵探头上的平面波成像以及相控阵探头上的球面波成像只需要一次发射和接收即可获得一幅超声图像,因此可以实现超声超高速成像。超声超高速成像的实时实现主要受制于传统的波束合成器无法达到其对计算能力的要求。采用稀疏矩阵波束合成方法,将普通的逐点计算波束合成方法转换为更加简单高效的矩阵相乘,并采用图形处理器(GPU)并行计算加速该算法。仿体实验结果表明,稀疏矩阵波束合成方法可以获得与传统波束合成器一样的结果;在特定的成像参数下,基于中央处理器(CPU)计算的稀疏矩阵波束合成速度超过50 帧/秒,经过GPU加速后合成速度为1000 帧/秒左右。
关键词: 超高速成像 稀疏矩阵波束合成 图形处理器; 并行计算
For information in English, please click here
Ultrasound Sparse Matrix Beamforming Based on GPU Parallel Computing
Abstract:The conventional line-by-line focused imaging restricts the frame rate of ultrasound imaging. In plane wave imaging on a linear array probe and diverging wave imaging on a phased array probe, only one transmit and receipt is required to form an ultrasound image, and thus ultrafast ultrasound imaging can be achieved. However, conventional beamformers cannot meet the demand of ultrafast ultrasound imaging for massive computation. In this paper, a sparse matrix beamforming method is developed to speed up the beamforming by converting conventional pixel-by-pixel operation to matrix multiplication. Graghic processing unit (GPU) based parallel computation is also employed to further accelerate the beamforming. Results of phantom experiments demonstrate that the sparse matrix beamformer can obtain the same results when compared with the conventional beamformer. Under specific imaging parameters, the frame rate is over 50 frames/second (fps) when using the sparse matrix beamformer on a central processing unit (CPU), while it is approximately 1000 fps when using the sparse matrix beamformer on a GPU.
Keywords: ultrafast ultrasound imaging sparse matrix beamforming GPU parallel computing
论文图表:
引用
No.4674972109082914****
同行评议
共计0人参与
勘误表
超声稀疏矩阵波束合成方法及其GPU并行实现
评论
全部评论0/1000