Noise removal from doppler ultrasound blood velocity signals using wavelets

One common form of analysis of the Doppler ultrasound signal for blood velocity estimation is the study of the maximum or the mean frequency envelopes of the sonograms after spectral estimation. These 'blood velocity' signals are usually corrupted by noise due to statistical instability associated to the spectrum analysis. The long-term trend evolution of these envelopes, which is related to mean volumetric blood flow, is also masked by noise. We show that Wavelet analysis can be effectively used for the removal of noise in the frequency envelopes extracted from the sonogram, and that longterm trends can be detected efficiently. Newly emerging techniques of time-frequency or, more precisely in this case, time-scale analysis can provide new insights into the nature of certain biological signals. This paper describes results using wavelet transforms for the analysis of the Doppler ultrasound signals for the assessment of blood flow velocities in arteries. Four different wavelet families are tested and their results compared to select the most appropriate ones for the tasks of noise removal and trend analysis of Doppler blood velocity signals. This study shows that, after an appropriate choice of wavelet and a particular level of decomposition, the systolic peaks can be highly resolved, exemplifying the ability of wavelets to successfully filter Doppler ultrasound frequency envelopes. The evolution of trends in Doppler ultrasound frequency has been clearly detected by the Daubechies wavelet for both the carotid and femoral signals. Another contribution of wavelets to the analysis of Doppler ultrasound frequency envelopes, which was clearly shown by applying continuous wavelet transform, is the detection of similar patterns within both the frequency envelopes of the signals from carotid and femoral arteries.