Evaluation of array signal processing methods for ultrasound-based arterial pulse wave velocity measurements on in vitro and in vivo data

Abstract

In this paper we evaluate an improved version of a signal processing approach recently proposed by the authors for estimating arterial pulse wave velocity (PWV) in a uniform arterial segment based on ultrasound measurements. The modifications are described, along with the experimental set-up for their evaluation. In general, results of physical experiments and simulations indicate that the new modifications provide a robust approach to PWV estimation. Non-invasive, in vivo assessment of the local elastic properties of an artery is a desirable capability, both in experimental physiology and in medicine. The arterial pulse wave velocity (PWV) in a short uniform segment of an artery is indicative of the local elastic state of the segment, but pulse wave reflections are always present throughout the arterial system and are a well-recognized source of error in PWV measurement (1). Elimination of estimation bias due to reflections is required if the PWV is to be used to compute parameters such as the compliance of the artery. Ultrasound is a suitable modality for the local PWV estimation problem, as it is for many problems of noninvasive, in vivo biomechanical assessment. Recently the authors (2) showed that pulse wave velocity estimation from several ultrasound measurements taken along a uniform arterial segment is mathematically equivalent to the broadband directional of arrival problem found in radar and sonar. The presence of a reflection wave traveling at the PWV, but in the opposite direction from the forward wave makes the PWV problem similar to the symmetric multipath case in the direction-of-arrival problem (3). Consequently, there is a great deal of work in signal processing literature dealing with the direction-of-arrival estimation problem in this form that can be applied the arterial PWV estimation using ultrasound by way of the problem formulation given in (2). In (4), a method for the PWV problem based on a least squares approach was presented and results of application to in vivo were presented for data taken using a 10 MHz linear probe positioned to acquire 8 beams spaced by 5.6 mm along the flow axis of the left common carotid artery (CCA). The technique of (4) produced estimates in a realistic range, but with large estimation variance on the order of 30-40%. This paper seeks to provide a more detailed evaluation of the general method, using new signal processing approaches for reduction of estimation bias. This evaluation is done by simulation, test tank experiments and further data from the CCAs of healthy volunteers.