Time-variable blood flow averaged waveforms

The computation of the average of random signals is addressed in this article. These averaging techniques are particularly relevant for Doppler blood flow clinical evaluation and research that usually accesses clinical indicators related to maximum and mean frequency, bandwidth and power variation over time. A novel technique is presented and compared to one previously described in the literature (Kitney and Giddens, "Analysis of blood velocity waveforms by phase shift averaging and autoregressive spectral estimation'', J. Biomech. Eng., 105, 398 - 404, 1983.). While the latter method is an iterative one, which sometimes does not converge, the former is a sequential technique, guaranteeing convergence. Clinical Doppler blood flow signals were averaged using both methods and the resulting waveforms were evaluated against the original signals, emphasising the impact on the relevant spectral parameters. The results obtained showed that, except for the maximum error in the computation of power variation over time, the errors produced are reduced by the proposed new algorithm.