基于EEMD有效成分优选的超声多普勒血流测速法

Doppler ultrasound blood flow velocimetry based on the optimization of effective EEMD components

为克服短时傅里叶变换(Short-Time Fourier Transform,STFT)在应对非平稳血流信号分析的局限性,提出基于集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD),并根据归一化波动指数(Normalized Fluctuation Index,NFI)选择有效血流成分的新方法EEMD_N.首先,对血流多普勒信号进行EEMD分解,得到本征模态函数(Intrinsic Mode Function,IMF)组,计算IMF的NFI;其次,使用傅里叶函数拟合不同NFI阈值的血流速度测量误差,以最小误差确定NFI最优阈值;然后,根据NFI最优阈值在IMF组中选择有效血流成分;最后,使用多普勒频移公式计算血流速度.仿真结果表明,与传统STFT法相比,EEMD_N法估计的血流速度剖面归一化均方根误差减小9.04%.人体颈动脉临床试验结果进一步验证了EEMD_N法的有效性.EEMD_N法能够有效改善血流速度剖面的测量精度,尤其是靠近血管壁的低速血流,有望为心血管疾病提供更准确的诊断信息.

A new approach based on the optimization of effective blood components using the Ensemble Empirical Mode Decomposition(EEMD)and normalized fluctuation index(EEMD_N)is proposed to overcome the limitations of Short-Time Fourier Transform(STFT)in processing non-stationary blood flow signals.Firstly,blood flow Doppler signals are decomposed by the EEMD method to get a group of Intrinsic Mode Functions(IMFs).Next,the Normalized Fluctuation Index(NFIs)of all IMFs are calculated,the Normalized Root Mean Square Errors(NRMSEs)of blood flow velocities,which are measured by IMFs with different NFI thresholds,are fitted by the Fourier function to determine the optimal threshold of NFI.Then,using the optimal threshold,the effective components in blood flow signals are chosen.Finally,blood flow velocities are computed by the Doppler frequency shift formula.Simulations show that the NRMSEs of blood flow velocity profiles estimated by the EEMD_N method were reduced by 9.04%in comparison with these by the traditional STFT method.Results in clinical experiments from a human carotid artery have further verified the effectiveness of the EEMD_N method.In summary,the EEMD_N method could effectively improve the measurement accuracy of blood flow velocity profiles,especially low velocities close to vessel walls,which is potential to provide more accurate diagnostic information for cardiovascular diseases.