基于压缩感知的超声成像快速实现方法研究

发布时间：2018-05-02 23:29

本文选题：平面波超声成像 + 压缩感知　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：压缩感知理论是近年来信号处理领域中兴起的研究热点,该理论在降低信号采样速率,提高信号重建质量上有着广阔的应用前景,并在超声成像方向展开了有益的探索。超声成像受其工作模式和理论的制约,一直以来在成像质量上低于其他医学成像方式,在成像实时性方面受超声声速的限制也难以达到较快的成像速度。为提高超声成像的质量和实时性,可以改进超声成像的理论,同时引入压缩感知理论。然而,将压缩感知理论引入超声成像使得在重建信号时需建立较大规模的观测矩阵和较复杂的迭代优化过程,这不利于信号的实时恢复。为了改进超声成像应用压缩感知理论采用的重建算法,把并行计算思想引入压缩感知重建的过程中,将有助于提高信号的恢复速度,实现准实时超声成像。本文重点研究了压缩感知框架下超声成像的快速实现问题,在保证超声成像重建质量的前提下,首先改进了超声成像的理论,建立平面波声散射超声回波模型,在此基础上引入压缩感知理论,建立压缩感知重建数学模型,并改进了压缩感知重建的FISTA(Fast Iterative Shrinkage-Thresholding Algorithm)算法,降低计算复杂度,减少迭代次数,此后,将并行计算的思想引入压缩感知重建过程,针对不同的数据规模,分别在Matlab、MPI、CUDA平台中验证并行计算的加速效果,在CUDA平台中获得最高的加速比。仿真数据表明,本文设计的方案不仅获得了较高质量的超声图像,在超声成像的实时性方面也有较大的提高。
[Abstract]:The theory of compression sensing is a hot topic in the field of signal processing in recent years. It has a broad application prospect in reducing the sampling rate and improving the quality of signal reconstruction. It also has a beneficial exploration in the direction of ultrasonic imaging. The ultrasonic imaging is restricted by its working mode and theory, it is always lower than other medical imaging methods in imaging quality, and it is difficult to achieve faster imaging speed because of the limitation of ultrasonic sound velocity in imaging real-time. In order to improve the quality and real time of ultrasonic imaging, we can improve the theory of ultrasonic imaging and introduce the theory of compression perception. However, the introduction of compression sensing theory into ultrasonic imaging makes it necessary to establish a large scale observation matrix and complex iterative optimization process for signal reconstruction, which is not conducive to real-time signal recovery. In order to improve the reconstruction algorithm used in ultrasonic imaging using compression sensing theory, the parallel computing idea will be introduced into the process of compression perception reconstruction, which will help to improve the speed of signal recovery and achieve quasi-real-time ultrasonic imaging. This paper focuses on the fast realization of ultrasonic imaging under the frame of compression sensing. On the premise of guaranteeing the reconstruction quality of ultrasonic imaging, the theory of ultrasonic imaging is improved, and the ultrasonic echo model of plane wave acoustic scattering is established. On this basis, the compression perception theory is introduced, the mathematical model of compression perception reconstruction is established, and the FISTA(Fast Iterative Shrinkage-Thresholding algorithm of compression perception reconstruction is improved to reduce the computational complexity and the number of iterations. The idea of parallel computing is introduced into the process of compressed perceptual reconstruction. According to different data scales, the acceleration effect of parallel computing is verified in Matlab / MPICUDA platform, and the highest speedup is obtained in CUDA platform. The simulation results show that the proposed scheme not only achieves high quality ultrasonic images, but also improves the real-time performance of ultrasonic imaging.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;TB559

【参考文献】