Abstract:
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.