Measurement of local pulse wave velocity: effects of signal processing on precision

Abstract

Pulse wave velocity (PWV) provides information about the mechanical properties of the vessel: the stiffer the artery is, the higher the PWV will be. PWV measured over a short arterial segment facilitates direct characterization of local wall properties corrected for prevailing pressure without the necessity of measuring pulse pressure locally. Current methods for local PWV assessment have a poor precision, but it can be improved by applying linear regression to a characteristic time-point in distension waveforms as recorded simultaneously by multiple M-line ultrasounds. We investigated the precision of this method in a phantom scaled according to realistic in vivo conditions. Special attention was paid to the identification of the foot of the wave, using the maximum of the second derivative, the intersecting tangent and the 20% threshold method. Before foot detection, the distension waveforms were subjected to preprocessing with various filters. The precision of the maximum of the second derivative had a coefficient of variation (CV) of 0.45% and 10.45% for an eighth and second order low pass filter, respectively. The intersecting tangent and the threshold method were less sensitive to filtering; the CVs were 0.66% and 0.68% for the high order filter and 2.36% and 1.43% for the low order filter, respectively. We conclude that foot detection by a threshold of 20% or by the tangent method are more suitable to identify the foot of the wave to measure local PWV. Both methods are less sensitive to (phase) noise than the maximum of the second derivative method and exhibit good precision with a CV of less than 1%.