改进的经验模态分解法分离超声多普勒血流与管壁信号

超声多普勒血流信号常包含管壁信号的干扰,准确分离二者对提高血流检测的精度具有重要作用。本文提出两种改进的经验模态分解(EMD)方法,先将含管壁信号的超声多普勒信号分解成多层本征模态函数(IMF),然后根据血流信号与管壁信号的不同特性,对既含管壁信号又含血流信号的IMF分量进行分离处理,最后将各层IMF分量中的管壁成分叠加得到管壁信号的估计,而血流信号可通过原信号减去估计的管壁信号而得到。将本方法用于计算机仿真信号和人体实测的超声多普勒信号,并与高通滤波器法、空间选择性降噪法和原EMD法进行比较,结果表明:本文提出的两种方法能在较大的管壁搏动速度范围内准确地分离血流信号和管壁信号,其平均相对误差比高通滤波器的结果降低了约52%和57%。可见,本文提出的两种方法有望用于血流信号与管壁信号的准确分离。 

Separating Doppler ultrasound blood flow and vessel wall signals by improved empirical mode decomposition methods

Doppler ultrasound blood flow signals usually contain vessel wall components.Their accurate separation is of great significance in the blood flow detection.In this paper,two improved Empirical Mode Decomposition (EMD) methods are proposed.Firstly,Doppler ultrasound signals with vessel wall components are decomposed into multi-level Intrinsic Mode Functions (IMFs).Then those IMFs containing both blood flow and vessel wall signals are separated according to their different properties.Finally,wall components of each level IMF are added as the estimation of the vessel wall signal and the blood flow signal is obtained by subtracting the estimated vessel wall signal from the original signal.Experiments on both computer simulated and real human carotid Doppler ultrasound signals are carried out to compare these two methods with the high pass filter,the spatially selective noise filtration algorithm and the original EMD method.It is shown that methods proposed in this paper can separate blood flow and vessel wall signals more accurately within a large range of wall velocities and their mean absolute error are about 52%and 57%lower than that of the high-pass filtering.So methods proposed in this paper may be effectively used to separate Doppler ultrasound blood flow and vessel wall signals.