超声多普勒管壁搏动和血流信号的分离研究

The Separation in Rreal-time of the Doppler Blood Flow and Moving Wall Signals Using EMD Algorithm

本研究提出利用经验模式分解(EMD)算法分解混叠有管壁成分的超声多普勒血流信号来实现管壁搏动和血流信号的分离。该方法首先将混叠有管壁搏动的超声多普勒血流信号分解为少量有限的分量,即内模函数(IMFs),然后根据管壁搏动信号与血流信号的功率比变化曲线,用比值法自动确定并去除低频管壁博动成分。在仿真实验中用提出的方法处理模拟的多普勒信号,对于靠近管腔内壁的血流信号其在频域功率谱上的相对误差为50%,在时域幅度的相对误差为45%,与高通滤波器方法的相对误差95%相比,准确性得到提高。基于个人计算机用C语言编程实现提出的算法,对实际采集的人体颈动脉多普勒信号可实现实时分离处理。结果表明:基于经验模式分解的滤波方法能有效客观地滤除管壁搏动信号,更准确地保留低频血流信号成分。

In this paper, the empirical mode decomposition （EMD） was proposed to decompose mixed Doppler signals for separating the Doppler blood flow and the wall signals. The EMD was firstly to be used to decompose a mixed Doppler signal into a finite and usually small number of individual components named intrinsic mode functions （IMFs）. Then a strategy based on the curve of the wall-to-blood signal energy ratio （WBSR） was developed to automatically identify and remove low frequency components which contributed to the wall pulsation. In the simulation study, this method was applied to process the simulated Doppler signals. Compared with the estimated low blood flow signal in frequency and time domain with relative error of 95% based on the tradition high-pass filter, the new approach obtained improved accuracy with the errors of 50% and 45 %, respectively. This method was implemented based on a personal computer using C language, and could process the Doppler ultrasound signals recorded from carotid arteries in real-time. From the simulation and clinical results, it could be concluded that the proposed method improved the performance for wall components removal from the mixed signals effectively and objectively.