A Novel and Low Cost Acoustic based Probe for Local Pulse Wave

Velocity Estimation

Experimental Characterization and in-Vivo Feasibility

H. C. Pereira

1,2

, J. Maldonado

, T. Pereira

, M. Contente

V. Almeida

, T. Pereira

, J. B. Simões

2,1

J. Cardoso

and C. Correia

Instrumentation Centre (CI-GEI), Physics Department, University of Coimbra, Coimbra, Portugal

ISA- Intelligent Sensing Anywhere, Coimbra, Portugal

Instituto de Investigação e Formação Cardiovascular, Aveleira, Penacova, Portugal

Escola Superior de Tecnologia da Saúde de Coimbra, S.Martinho do Bispo, Coimbra, Portugal

Keywords: Local Pulse Wave Velocity, Double Headed Probe, Microphones, Test Bench Systems, in-Vivo

Measurements.

Abstract: The use of local pulse wave velocity (PWV) as an independent risk factor for cardiovascular events and as a

marker of atherosclerosis has been gained clinical relevance over the years. A novel acoustic double headed

probe for non-invasive measurement of the local PWV is presented in this paper. The PWV is assessed in

one single location and involves the determination of time delay between the signals acquired

simultaneously by two acoustic sensors, placed ≈11 mm apart. Several tests were performed in special

purposes test bench systems in order to characterize the acoustic probe (AP) regarding the existence of

crosstalk between the transducers, repeatability, waveform analysis and also its time resolution. Results

demonstrate the effectiveness of the AP in acquiring repeatedly the same waveform, with the possibility to

measure higher PWV (14 m/s), with a relative error less than 5%, when using two uncoupled APs. In-vivo

acquisitions were also carried out with the AP in the carotid artery of 17 healthy volunteers with the

intention of local PWV and other hemodynamic parameters estimation, such as left ventricular ejection time

(LVET). For the total of subjects’ sample, the obtained mean carotid PWV was 2.96±1.08 m/s and the

LVET mean value was 288.59±21.42 ms.

1 INTRODUCTION

Over the last years, great emphasis has been placed

on the role of arterial wall stiffening in the

development of cardiovascular diseases and events.

Aortic stiffness which results from the progressive

degeneration of the wall’s elastic fibres is generally

associated with ageing and some pathophysiological

conditions, such as hypertension, end-stage renal

disease and diabetes (Laurent et al., 2001, Shoji et

al., 2001). Currently, arterial stiffness is included in

in the guidelines of the European Society of

Cardiology and European Society of Hypertension

as an independent predictor of cardiovascular

mortality and morbidity and as a new parameter of

target organ damage that must be considered in

cardiovascular risk stratification (Mancia et al.,

2007). The most simple, non-invasive and robust

method to assess arterial stiffness is pulse wave

velocity (PWV), i.e., the velocity at which the

pressure wave, generated by ventricular contraction

propagates along an artery (Pannier, 2002). Carotid-

Femoral PWV is considered the gold standard

measurement of arterial stiffness, being supported by

several clinical studies that highlight the relevant

contribution of PWV to the diagnosis, prognosis and

follow-up of the general population/patient

(Maldonado et al., 2011, Meaume et al., 2001). The

most prominent commercial devices require

low/moderate technical expertise; however they

present several drawbacks in PWV assessment. The

practical solution used by these systems relies on the

acquisition of pulse waves at the carotid and femoral

arteries, to determine the time delay measured

between pressure upstroke at each site. The distance

Pereira H., Maldonado J., Pereira T., Contente M., Almeida V., Pereira T., Simões J., Cardoso J. and Correia C..

A Novel and Low Cost Acoustic based Probe for Local Pulse Wave Velocity Estimation - Experimental Characterization and in-Vivo Feasibility.

DOI: 10.5220/0004251900780088

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2013), pages 78-88

ISBN: 978-989-8565-34-1

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

between the two acquisition locations is usually

measured using a tape and then PWV is determined

using the linear ratio between distance and time

delay (Boutouyrie et al., 2009). Nevertheless, this

method not only negligence opposite wave

propagation but also presents errors in estimating the

distance between the recording locations (for

example, the curvature of the arteries cannot be

taken into account) (Segers et al., 2009). On the

other hand, it introduces a rough estimate of local

properties of the artery, since it integrates the

different segments of arterial stiffness (carotid,

aorta, iliac, femoral), being unable to differentiate

between muscular and elastic segments. The

possibility to assess to the local hemodynamics is in

fact very useful, particularly in the carotid artery due

to its predisposition to atherosclerotic plaques

formation and its significance in the development of

cerebrovascular diseases (Laurent et al., 2006).

Variables such as the local PWV, arterial

distensibility, cross-sectional compliance or Young’s

modulus assess the local intrinsic properties of the

arterial wall itself, being more related the

biomechanical properties of the artery (Gamble et

al., 1994). Although local PWV has already been

used as an independent risk factor for cardiovascular

events such as coronary disease and stroke and as a

marker for cardiovascular disease including

atherosclerosis (Gaszner et al., 2012, Laurent et al.,

2003), at the present moment there is no gold

standard to the assessment of local PWV. In the past

few years, several methods have been investigated

with the intention of local PWV assessment. The

generalized methods require simultaneous

measurements of pressure, velocity or diameter at

the same site (Rabben et al. 2004, Borlotti et al.,

2012), while the most recent studies show that this

hemodynamic parameter can also be obtained using

the time delay calculated between the piezoelectric

elements of an ultrasound probe, placed at a fixed

distance (Hermeling et al., 2007). In all the methods,

the assessment to the local variables require

burdensome or specialized imaging technologies

(ultrasound and echo tracking techniques), limiting

the use to clinical practice.

The present work intends to present, characterize

and validate an efficient and low-cost tool based on

a non-invasive device that is placed over the carotid

artery and can be easily handled in diagnostic trials

by an operator. Based on a previous work, where a

double headed piezoelectric probe was developed

and characterized in laboratory (Pereira et al., 2010

and 2011), it was developed a simpler and novel

system for local PWV estimation and other

parameters extraction, such as left ventricular

ejection time (LVET). The device is based on a

double configuration of two acoustic sensors that are

placed at a fixed distance, d, allowing simultaneous

acquisition of two (sound) pulse waves. The

measurement of time delay between the waves, t,

allows local PWV to be determined, simply, as:

smPWV



)/(

(1)

2 MATERIALS

2.1 The Double Headed Acoustic Probe

The developed probe, presented in figure 1, consists

of two acoustic transducers (Pro-Signal, ABM-712-

RC, microphone-solder pad) that are placed

approximately 11mm apart and fixed at the top of a

plastic box (Multicomp, 77 mm x 49 mm x 26.6

mm). The transducers, based on 9.7 mm diameter

electret condenser microphones with an operating

frequency of 100Hz to 10 kHz and noise cancelling

directivity, are placed at the minimum separating

distance, while avoiding mechanical contact. These

elements form an ergonomic configuration that

allows a safe and effective way of collecting the

pulse wave on the carotid artery for both the patient

and the operator.

Figure 1: Representation of the double headed probe. a-

Microphone 1; b- Microphone 2; d1-distance between

transducers centres: 11mm; d2- sensors height: 2mm.

The probe does not include any type of signal

conditioning circuit, so the acoustic signals are

acquired directly by a Personal Computer (PC)

Sound Card. The AP uses parallel audio cable to

convey the information obtained directly from the

transducers, to the microphone input of the PC

ANovelandLowCostAcousticbasedProbeforLocalPulseWaveVelocityEstimation-ExperimentalCharacterization

andin-VivoFeasibility

Figure 2: Schematic representation of the overall system used in in-vivo measurements. A- Acoustic Probe; B- External

USB Sound Card; C-PC; D- Cardiocheck DB; E- Cardiocheck GUI; F-Data Pre-processing; G- Hemodynamic Parameters

Extraction.

Sound Card. In circumstances in which the PC

Sound Card does not have stereo input, the probe

connects first to an external Sound Card (7.1

Sweex® USB Sound Card, 16-bit, 48 kHz

Maximum Sampling Rate, 90 dB Signal to Noise

Ratio) that then delivers the collected signals to the

PC, via USB. The data acquisitions are displayed in

real time, through a dedicated Matlab® Based

Graphical User Interface (Cardiocheck GUI) and

automatically stored in a non-commercial Microsoft

Access® based database (Cardiocheck DB). The

data is subsequently processed using different

algorithms (detailed in section 3.3) that aim the

extraction of several hemodynamic parameters,

namely the PWV and the LVET (figure 2).

2.2 Test Setups

For characterizing the AP, as well as the various

parameters extraction algorithms, it was developed

two special purpose sets of test bench systems. The

test setup I was designed to evaluate the ability of

the probe in reproducing repeatedly different types

of waveforms but also to evaluate the existence of

crosstalk between both transducers. The setup uses a

700 µm stroke actuator, ACT, driven by a high

voltage linear amplifier, HVA (Physik Intrumente

GmbH P-287 and E-508, respectively) to generate a

pressure wave that is fed to the acoustic probe by

means of a “mushroom” shaped PVC interface

(figure 3). This PVC interface (10 mm diameter),

coupled to the ACT, is in mechanical contact with

the transducer, without affecting the output voltage

since the sensors does not respond to DC excitation.

With this mechanical adapter it is possible to

transmit the ACT’s motion associated to the pressure

wave, in such a way that the longitudinal forces are

responsible for the transducer electric response. The

waveforms are programmed and downloaded into an

arbitrary waveform generator, AWG, Agilent

33220A that delivers the signal that is generated by

the ACT and also the synchronism that triggers the

data acquisition system, DAS (National

Instruments®, USB6210). Although the AP is a

prototype suitable for clinical tests, designed to be

combined with a PC Sound Card, it was necessary to

use a different DAS in test bench experiments, in

order to acquire additional reference signals.

Figure 3: Representation of the mechanical structure of the

test setup I. 1-ACT. 2- PVC interface. 3- AP.

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

Figure 4: Schematic representation of the test setup II.

The test bench system II, schematically

presented in figure 4, was developed aiming the

assessment of the temporal resolution of the AP.

This test bench emulates the main arterial

pressure wave propagation features of the

cardiovascular system, presenting an upgrade in

relation to the experimental setup developed

previously by Pereira et al. (Pereira et al., 2009). The

main difference is based on the use of a natural

rubber latex tube, considered to be a reliable

material to simulate the compliance of a human

artery and providing a higher distensibility than the

silicon tube originally used. As in the test bench I,

the pressure waveform is firstly generated using the

AWG and then delivered to the ACT/HVA, that

through a piston (“mushroom” shaped PVC, 15 mm

diameter) - rubber membrane mechanism launches

the wave into a 200 cm long latex tube (Primeline

Industries, 7.9 mm inner diameter, 0.8 mm wall

thickness), filled with water. The tube’s sealing is

made using a T-shaped scheme, guaranteeing

geometric homogeneity. The wave is captured by the

AP placed along the tube and by two pressure

sensors PS1 and PS2 (Honeywell, 40PC015G1A),

placed transversely and longitudinally to the tube.

These sensors are used as the main reference for

time delay/pulse wave velocity assessment but also

monitor the DC pressure level of the tube, imposed

by a piston P and a mass m at one of the tube’s

extremities. The range of DC pressure levels in the

tube varies, approximately, from 30 mmHg to 400

mmHg, including (and exceeding) the pressure

variations registered in a healthy and non-healthy

human system. Although the variation in the DC

pressure level interferes with the wave propagation

velocity, the AP was tested at a constant DC

pressure (≈ 66 mmHg), since it was not crucial for

the present work having several wave propagation

velocities.

To record simultaneously the different sensors

response it was used the aforementioned DAS,

triggered by the AWG.

3 METHODS

3.1 Experimental Characterization

The experimental characterization of the AP

consisted in the evaluation of its performance

regarding three main aspects: repeatability in

assessing pressure waveform, occurrence of

crosstalk phenomenon and estimation of time

resolution. Several pressure waveforms were

programmed and used as inputs in these studies,

including Gaussian-like and Cardiac-like pulses.

The last ones, synthesized using a weighted

combination of exponentially shaped sub-pulses

(Almeida et al., 2009), reproduce different states of

arterial wall elasticity: type A and type B,

correspond respectively to cases of pronounced and

slight arterial stiffness (non-healthy subjects) and

type C, commonly seen in healthy individuals,

characterize elastic arteries (Murgo et al., 1980).

In all the experiments, the data acquisition was

performed through a dedicated acquisition module

of National Instruments (NI© USB6210) and data

ANovelandLowCostAcousticbasedProbeforLocalPulseWaveVelocityEstimation-ExperimentalCharacterization

andin-VivoFeasibility

logging was accomplished by NI LabView™ 2010

SignalExpress. All the signals were sampled at 5

kHz and stored for offline analysis using Matlab®.

Data processing was undertaken in Matlab® 2009a

and statistical analysis was performed using

Microsoft Excel 2010.

3.1.1 Waveform Analysis/Repeatability

The first part of this study aimed to examine the

probe’s response for different types of waveforms

generated by the Agilent 33220A and exerted by the

ACT. To obtain the best response of the transducer it

was selected for each input signal, the best

amplitude (3.5V) and frequency (1Hz). All the

sensors signals were submitted to a 300 Hz low pass

filter and to a band cut filter of 49Hz-51Hz, in order

to avoid, respectively, the presence of the resonant

frequency of the actuator ( ~ 380Hz ± 20%) and the

50Hz power line interference. It was also performed

an integration of the transducer signals using the

Matlab® function cumsum to compare them with

the original input signals.

In the second part, it was intended to measure

the same waveform repeatedly and under the same

conditions by the AP. For this study, each sensor

was excited with fifty independent impulses (with

the same amplitude and width (Gaussian, 1 s width,

1Hz frequency). With those signals, it was

determined the average signal which was used as

reference to determine the root mean square error

(RMSE), for each one. The RMSE was then

computed to each signal.

3.1.2 Crosstalk Analysis

Since the two transducers composing the AP were

incorporated in the same case with a very small

separating distance, it was important to analyse

whether some kind of interaction between them

existed. The first part of this study was done

simultaneously with the repeatability test, where one

of the acoustic transducers was being actuated

(microphone 1) and the other one was left free

(microphone 2), that is to say without any contact

with the PVC adapter/ACT (figure 3). The responses

of both transducers for fifty independent impulses

(Gaussian, 1 s width, 1 Hz frequency) were recorded

and the average signals were estimated. This

procedure was then applied to the other sensing

element, such that the actuated transducer was

microphone 2 and the free transducer was

microphone 1.

The actuated transducers generated a typical

differential signal with a good signal-to-noise ratio,

while the free transducers generated a much lower

amplitude signal, with a profile substantially

opposite to that obtained for the actuated transducers

(see results section 4.1). Due to the characteristics of

the signal obtained for the free transducers, it was

necessary to perform an additional experiment to

determine whether this transmission might interfere

with one of the most important aspects of the probe:

its time delay assessment. Thus, the second test

consisted in the direct and simultaneous actuation of

both transducers, with the purpose of time delay

assessment. Both sensors were excited with three

independent impulses (Gaussian, 1 s width, 1Hz

frequency) and for each acquisition it was

determined the time delay between both transducers,

using different algorithms yet to be described on

subsection 3.3. In this particular experiment, the

signals were sampled at 12.5 kHz, the same

sampling frequency used in in-vivo tests.

3.1.3 Time Resolution Evaluation

One of the main goals of the AP characterization

was the evaluation of its ability of precisely

assessing the time delay between two distinct points,

separated from a very small distance.

In order to evaluate its time resolution, it was

used two different APs (AP1 and AP2) that were

placed on the tube of the test bench II, with the help

of two external clamps (figure 4). One of the probes

was kept fixed at the 50 cm position, while the other

one was moving from 100 cm position to 54 cm

position by 2 cm intervals. For each position, a

Gaussian waveform (100 ms width, 10 Hz

frequency) was delivered to the system, and then

time delay and PWV were estimated between the

first microphones of both probes and also between

the pressure sensors (PS1 and PS2), attached at the

extremities of the tube.

The relative errors between the reference PWV

and the PWV obtained with the uncoupled

transducers for each separating distance (x) were

calculated, using the algorithms of section 3.3. The

test was repeated for more two times, for a constant

DC pressure of ≈ 66 mmHg.

3.2 In Vivo Measurements

3.2.1 Participants and Study Protocol

Seventeen young volunteers aged 22.12 ±1.96 years

were recruited and gave written informed consent

prior to recording. Each participant was properly

weighed and measured and after 5m of rest of supine

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

position, a blood pressure measurement was

obtained from his right brachial artery, using an

automatic clinically validated sphygmomanometer

(MAM Colson BP 3AA1-2 ®; Colson, Paris).

Next, a straight arterial segment of the right common

carotid artery was identified by a skilled operator

and a record of approximately 20s-30s was obtained

with the probe longitudinally aligned to the artery.

Data acquisition was performed with the dedicated

real-time software Cardiocheck GUI and

automatically stored in the Cardiocheck DB. Age,

sex, weight, height, waist, systolic blood pressure

(SBP) and diastolic blood pressure (DBP) were also

stored in the same database.

All the signals were acquired at a sample rate of

12.5 kHz and were processed offline in Matlab

2009a®, aiming the extraction of carotid pulse wave

velocity and other hemodynamic parameters.

3.3 Signal Processing

In the first part of this work (experimental

characterization), a set of dedicated algorithms have

been developed aiming the estimation of time delay

in two main situations: between the signals of the

AP transducers and between the signals of pressure

sensors (test setup II). After a common pre-

processing, based on a low-pass filter with a cut-off

frequency of 100 Hz to reduce high frequency noise,

four different methods were used for time delay

estimation: a) maximum of cross-correlation

function, b) maximum and c) minimum amplitude

identification and d) zero-crossing detection. The

cross correlation method uses the xcorr function of

Matlab’s Signal Processing Toolbox to determine

the peak of cross-correlogram that allows delay

estimation by subtracting the peak time position

from the pulse length. The other methods ensure an

accurate detection of some fiducial points of the

signal, such as the maximum, the minimum and the

zero. As so, the methods of maximum and minimum

amplitude identification use a 6th polynomial fit in

the maximum and the minimum region of the

signals, while the zero-crossing method applies a

linear fit to the region where the signal crosses the

zero. For all the methods, time delay is estimated

between the maxima, minima and zero points

detected in each set of signals.

In the last part of this work, the AP was used to

assess PWV and other hemodynamic parameters in

human carotid arteries. Since the acquisitions were

constituted by several cardiac cycles, it was

necessary to apply a dedicated segmentation routine,

based on a minima detection approach to divide the

data stream into single periods. Before applying the

segmentation algorithm, the signals were filtered

with the aforementioned 100 Hz low pass-filter and

then heart rate was determined. For each cardiac

cycle, the maximum of cross-correlation was used

for carotid PWV estimation and an average value

was obtained. Besides PWV, it was also possible to

determine hemodynamic parameters, such as:

LVET, defined as the period of time from the start of

the pulse (aortic valve open) to the dicrotic notch

(closure of the aortic valve) and diastole phase (DP),

defined as the period of time from the dicrotic notch

to the end of the pulse. These parameters were

extracted based on the conviction that the onsets of

the first and second carotid sounds (S1 and S2)

coincide respectively with the onset and with the

dicrotic notch of the carotid pulse waveform

(Hasegawa et al., 1991). The onsets of carotid

sounds S1 and S2 were identified as the maxima of

the second time derivative of the acoustic signal.

LVET and DP were calculated for each cardiac

cycle. Data were expressed as mean ± SD.

4 RESULTS AND DISCUSSION

4.1 Experimental Characterization

The first part of probe’s experimental

characterization consisted in the evaluation of the

AP output to different waveforms and its

repeatability. The response obtained by the AP for

each one of the waveforms is presented in figure 5.

The AP profiles are similar to those expected by a

differentiator circuit; however it is not possible to

precisely recover the original pressure waveform.

When the acoustic signals are integrated there are

noticeable similarities with the input signals,

however the RMSE between both signals is quite

high (approximately 13% for each case). This

performance was predictable, since the sensitivity of

the acoustic sensors must be reduced for low

frequencies that are below the microphone’s 3dB

bandwidth (100 Hz-10 kHz). Since low frequencies

are determinant for the precise reconstruction of

arterial pressure waveform, the use of these acoustic

sensors limits the possibility of the AP for waveform

estimation purposes. Nevertheless, this fact does not

disqualify the use of this probe for its main purpose:

PWV estimation, once the method does not depend

on the waveform accuracy.

The results regarding the repeatability test are

shown in table 1 and figure 6.

ANovelandLowCostAcousticbasedProbeforLocalPulseWaveVelocityEstimation-ExperimentalCharacterization

andin-VivoFeasibility

(a) (b)

(c)

(d)

Figure 5: Acoustic sensor responses to different excitation pressure waveforms. (a) Gaussian-like Pulse. (b) Type A

Cardiac-like Pulse. (c) Type B Cardiac-like Pulse. (d) Type C Cardiac-like Pulse. AP- Acoustic Sensor Signal. ACT- Input

Signal. INT- Integrated Sensor Signal.

Table 1: Statistics of the measurements obtained in the

repeatability test.

Transducer Mic 1 Mic 2

Nº Acquisitions 50 50

Mean (%) 1.1781 0.6198

Std. Deviation (%) 0.0345 0.0298

Maximum (%) 1.2849 0.7379

Minimum (%) 1.1174 0.5890

Figure 6: Graphic representation of the RMSE distribution

between the reference signal and the microphone 2 output,

obtained in the repeatability test.

Although the microphone 2 (0.6198±0.0298)

exhibits a better performance than the microphone 1

(1.1781±0.0345), the RMSE variance values

obtained for both probes are identically low,

evidencing the reliability of the system.

The second part of the AP’s characterization

intended to study the presence or absence of

crosstalk effect in both transducers. The first results

achieved in this study are illustrated in figure 7. The

actuated sensors present a good signal-to-noise ratio

and a typical profile when compared with the one

obtained previously in the waveform analysis test

(figure (5 (a)).

The free transducers also present a slight profile

but of much lower amplitude. Although the results

suggest the existence of crosstalk effect, this

phenomenon was seen as a mass inertial effect

(transducer resistance to conserve its idle state),

since the profile of the free transducer had an

inversed shape relatively to the actuated one. This

assumption could not be proven in the present work

but it will be aim of futures studies. Nevertheless,

and since the main purpose of this probe is the

assessment of local PWV, it was employed a

different approach, in order to determine if this

“transmission” might interfere with the AP’s time

delay. For this purpose, both transducers were

simultaneously actuated with three independent

0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

Time

(

)

Normalized Amplitude (V)

ACT

INT

0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

Time

(

)

Normalized Amplitude (V)

ACT

INT

0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

Time(s)

Normalized Amplitude (V)

ACT

INT

0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

Time

(

)

Normalized Amplitude (V)

ACT

INT

0 10 20 30 40 50

0.1

Acquisition (N)

RMSE (dB)

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

Gaussian waves and time delay was calculated using

different algorithms. The results regarding this

experiment are presented in table 2 and figure 8.

Table 2: Time delay values obtained for each algorithm

when both transducers are simultaneously actuated with

three independent Gaussian impulses.

Time Delay Estimation Method

xcor

max min zc

8e-5

0.0134 0.0136 0.0037

8e-5 0.0109 0.0142 0.0038

8e-5 0.0103 0.0140 0.0035

Figure 7: Crosstalk phenomenon study: average response

of both AP’s transducers to fifty independent pulses. The

actuated transducer is microphone 2 and the free

transducer is microphone 1.

Figure 8: Crosstalk phenomenon study: typical response of

both AP’S transducers to a Gaussian pulse, simultaneously

delivered to them.

The time delay obtained for each one of the

algorithms is very different and actually surprising.

It was not expected to obtain such a variable and

elevated time delay values for maximum, minimum

and zero crossing algorithms. In contrast, the cross-

correlation algorithm presented a great performance,

where time delay always matched the minimum

detectable time, limited by the system, i.e.:, the

sampling time (1/12500Hz). In order to understand

the achieved results, the AP’s response was also

analysed (figure 8). It is visible that the profiles

obtained for each one of the transducers are

identical; however, they present important

differences in terms of amplitude and peaks

correspondence. It was expected that the maxima

and the minima of both signals were in agreement,

but actually that didn’t happen. These slight profiles

difference can be justified with the experiment level

of difficulty. It is extremely important that the

simultaneous actuation of both transducers is made

rigorously under the same conditions; otherwise the

waveforms of each transducer can be affected. This

also suggests that time delay algorithms that depend

only on a fiducial point are more susceptible to

error, especially if the waveforms don’t have exactly

the same profile. Finally and in what concerns to

crosstalk effect, it can be concluded that the

existence of a possible transmission between sensors

does not affect the time delay, when the cross-

correlation algorithm is used. In order to prove the

effectiveness of the other algorithms, it will be

necessary to proceed to additional experiments.

The last test concerning AP’s experimental

characterization intended to evaluate its time

resolution. In this test, it was determined the PWV

for two uncoupled AP’s in successively smaller

separation distances and the PWV reference

obtained using the pressure sensors PS1 and PS2.

The PWV results obtained for each algorithm and

the relative errors between the reference PWV and

the PWV obtained with the uncoupled transducers,

for each separating distance and method are

presented, respectively, in figures 9 and 10.

Figure 9: Time Resolution study: PWV values of

uncoupled acoustic sensors and pressure sensors, yielded

by the four algorithms. Each point is an average of three

trials.

The statistics of the measurements are

synthesized in table 3.

0 0.2 0.4 0.6 0.8 1

-0.1

-0.05

0.05

0.1

Time

(

)

Amplitude (V)

Actuated

Free

0 0.2 0.4 0.6 0.8 1

-0.02

-0.01

0.01

0.02

Time(s)

Amplitude (V)

Mic a

Mic b

ANovelandLowCostAcousticbasedProbeforLocalPulseWaveVelocityEstimation-ExperimentalCharacterization

andin-VivoFeasibility

Table 3: Statistics of the measurements obtained in the

time resolution test.

Algorithm

PWV(m/s)

Relative

Error

Mean (%)

Pressure

Sensors

(reference)

Acoustic

Sensors

Maximum

13.856±0.037 13.742±0.372 2.083

Xcorr

13.716±0.037 13.308 ±0.524 4.246

13.702±0.034 13.592 ±0.560 2.863

Min

13.540±0.039 13.176±0.547 3.550

Figure 10: Time Resolution study: relative errors for each

distance and method.

The algorithms with the best and worst general

performance are the maximum and the cross-

correlation with an average error less than 3% and

5%, respectively. However, and for the minimum

distance achieved (2 cm) the magnitude of the errors

is less than 1%, when considering cross-correlation

and zero-crossing algorithms.

The results obtained for AP time resolution for

each algorithm, exhibit a very good performance

suggesting that the AP have enough accuracy to be

considered an interesting stand-alone instrument for

local PWV assessment.

4.2 In-vivo Measurements

Following the preliminary tests of the probe in the

test benches, it was performed a set of measurements

in human carotid arteries, in order to test the AP in

in-vivo conditions (figure 11). The characteristics of

the patients, as also the results of the parameters

assessed by the AP, (heart rate, local PWV, LVET

and DP) are given in table 4.

In order to assess pulse wave velocity, it was

only used the cross-correlation algorithm, since it

has presented the best performance both on crosstalk

and on time resolution studies. The range of

obtained values for carotid PWV are slightly lower

than the values obtained by other reference studies

that also assess the carotid PWV (≈ 4 m/s) ( Borlotti

et al., 2012 and Hermeling et. al., 2006). However,

the number of analysed subjects not only is small as

also include very young people (22.12±1.96 years),

which can justify a lower PWV mean (≈3 m/s), due

to the high elasticity of young and healthier arteries.

Although the obtained PWV variance is high (

N=17, ≈ 1 m/s), it is concordant with the PWV

variance obtained in a recent study for a significant

number of healthier subjects (N=1774, ≈ 1.64 m/s ) (

Borlotti et al., 2012).

Figure 11: Preliminary results of the AP acquiring data in

a healthy young subject. CS

- First Carotid Sound. CS

Second Carotid Sound.

Table 4: Main characteristics of the volunteers and AP

parameters assessment.

Variable Mean ± SD

Age, years 22.12 ±1.96

N(Male/Female) 17(6/11)

Height, cm 166.82±11.73

Weight, Kg 66.18±18.87

BMI, Kg/m

23.50±4.94

Waist, cm 75.35±14.85

Brachial SBP, mmHg 114.35±12.62

Brachial DBP, mmHg 70.94±7.11

Heart Rate, bpm 67±12.09

Local PWV(m/s) 2.96±1.08

LVET (ms) 288.59±21.42

DP (ms) 611.07±148.84

Nevertheless and in order to address more

accurate results, it will be necessary to assess to a

higher number of subjects, not only with a broader

range of ages but also with pathologies, such as

hypertension or atherosclerosis, where is expected to

observe an increase of local PWV. The use of a

reference method is also indispensable to validate

the developed algorithms for AP hemodynamic

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

parameters extraction. Currently, the probe

presented a good performance in acquiring signals

with a good bandwidth and signal- to-noise ratio in

human carotid arteries, allowing the application of

various algorithms that extract clinically relevant

information.

Regarding the LVET values, we believe that is

actually possible to determine this parameter as the

time delay between the main onsets of each carotid

sounds, since the estimated values are generally

close to the expected for healthy subjects (Willems

et. al., 1970). This parameter will be the subject of a

further study, to evaluate the robustness of the

algorithm.

5 CONCLUSIONS

A novel and low-cost doubled headed probe

specifically designed to assess local PWV has been

developed and characterized in dedicated test setups.

The probe demonstrated good performance on

the dedicated test setups and results showed that its

signals are repeatable and crosstalk effect do not

interfere with its time resolution when the cross-

correlation algorithm for time delay estimation is

used.

It is also possible to conclude favourably towards

the effectiveness of the AP in the measurement of

local PWV. The maximum amplitude and the cross

correlation algorithms exhibited the capability of

measuring higher PWV (≈ 14m/s) with an error less

than 10%, for the several separating distances (50

cm to 2 cm).

The natural follow-up of this work will be the

continuation of the assessment of local PWV and

other hemodynamic parameters in a significant

numbers of patients (normal and with different

pathologies), under medical control. The obtained

AP values must also be compared with the values

obtained with standard commercial systems.

Although studies to validate the clinical use of

AP are still required, this device seems to be a valid

alternative system, to local PWV stand-alone

devices.

ACKNOWLEDGEMENTS

The authors acknowledge Fundação para a Ciência e

Tecnologia for funding SFRH/BDE/15669/2007 and

PTDC/SAU-BEB/100650/2008, project developed

under the initiative of QREN, funding by

UE/FEDER, through COMPETE - Programa

Operacional Factores de Competitividade. The

authors also thank to the company ISA-Intelligent

Sensing Anywhere and the medical collaboration of

Instituto de Investigação e Formação

Cardiovascular, Clínica da Aveleira and Escola

Superior de Tecnologia da Saúde de Coimbra.

REFERENCES

Almeida, V., Pereira, T., Borges, E., Figueiras, E.,

Cardoso, J., Correia, C., Pereira, H. C., Malaquias, J.

L. and Simões, J. B. 2009. Synthesized cardiac

waveform in the evaluation of augmentation index

algorithms. In Proceedings of BIOSIGNALS 2010,

Valencia, Spain.

Borlotti, A., Khir, A. W., Rietzschel, E. R., De Buyzere,

M. L., Vermeersch, S. & Segers, P. 2012. Noninvasive

determination of local pulse wave velocity and wave

intensity: changes with age and gender in the carotid

and femoral arteries of healthy human. J Appl Physiol,

113, 727-35.

Boutouyrie, P., Brie, M., Collin, C., Vermeesch, S. and

Pannier, B. 2009. Assessment of pulse wave velocity.

Artery Research, 3, 3-8.

Gamble, G., Zorn, J., Sanders, G., Macmahon, S. &

Sharpe, N. 1994. Estimation of arterial stiffness,

compliance, and distensibility from M-mode

ultrasound measurements of the common carotid

artery. Stroke, 25, 11-6.

Gaszner, B., Lenkey, Z., Illyes, M., Sarszegi, Z., Horvath,

I. G., Magyari, B., Molnar, F., Konyi, A. & Cziraki, A.

2012. Comparison of aortic and carotid arterial

stiffness parameters in patients with verified coronary

artery disease. Clin Cardiol, 35, 26-31.

Hasegawa, M., Rodbard, D. & Kinoshita, Y. 1991. Timing

of the carotid arterial sounds in normal adult men:

measurement of left ventricular ejection, pre-ejection

period and pulse transmission time. Cardiology, 78,

138-49.

Hermeling, E., Reesink, K. D., Reneman, R. S. & Hoeks,

A. P. 2007. Measurement of local pulse wave velocity:

effects of signal processing on precision. Ultrasound

Med Biol, 33, 774-81.

Laurent, S., Boutouyrie, P., Asmar, R., Gautier, I., Laloux,

B., Guize, L., Ducimetiere, P. & Benetos, A. 2001.

Aortic stiffness is an independent predictor of all-

cause and cardiovascular mortality in hypertensive

patients. Hypertension, 37, 1236-41.

Laurent, S., Cockcroft, J., Van Bortel, L., Boutouyrie, P.,

Giannattasio, C., Hayoz, D., Pannier, B.,

Vlachopoulos, C., Wilkinson, I., Struijker-Boudier, H.

& European Network for Non-Invasive Investigation

of Large, A. 2006. Expert consensus document on

arterial stiffness: methodological issues and clinical

ANovelandLowCostAcousticbasedProbeforLocalPulseWaveVelocityEstimation-ExperimentalCharacterization

andin-VivoFeasibility

applications. Eur Heart J, 27, 2588-605.

Laurent, S., Katsahian, S., Fassot, C., Tropeano, A.,

Gautier, I., Laloux, B. and Boutouyrie, P. 2003. Aortic

stiffness is an independent predictor of fatal stroke in

essential hypertension. Stroke, 34, 1203-1206.

Maldonado, J., Pereira, T., Polonia, J., Silva, J. A., Morais,

J., Marques, M. & Participants in The, E. P. 2011.

Arterial stiffness predicts cardiovascular outcome in a

low-to-moderate cardiovascular risk population: the

EDIVA (Estudo de DIstensibilidade VAscular)

project. J Hypertens, 29, 669-75.

Mancia, G., De Backer, G., Dominiczak, et al.,

Management of Arterial Hypertension of the European

Society Of, H. & European Society Of, C. 2007. 2007

Guidelines for the Management of Arterial

Hypertension: The Task Force for the Management of

Arterial Hypertension of the European Society of

Hypertension (ESH) and of the European Society of

Cardiology (ESC). J Hypertens, 25, 1105-87.

Meaume, S., Rudnichi, A., Lynch, A., Bussy, C., Sebban,

C., Benetos, A. & Safar, M. E. 2001. Aortic pulse

wave velocity as a marker of cardiovascular disease in

subjects over 70 years old. J Hypertens, 19, 871-7.

Murgo, J. P., Westerhof, N., Giolma, J. P. & Altobelli, S.

A. 1980. Aortic input impedance in normal man:

relationship to pressure wave forms. Circulation, 62,

105-16.

Pannier, B. M., Avolio, A. P., Hoeks, A., Mancia, G. &

Takazawa, K. 2002. Methods and devices for

measuring arterial compliance in humans. Am J

Hypertens, 15, 743-53.

Pereira, H. C., Cardoso, J. M., Almeida, V. G., Pereira, T.,

Borges, E., Figueiras, E., Ferreira, L. R., Simões, J.,

Correia, C. 2009. Programmable testbench for

hemodynamic studies. IFMBE Proceedings 25/IV

1460ff.

Pereira, H. C., Pereira, T., Almeida, V., Borges, E.,

Figueiras, E., Simoes, J. B., Malaquias, J. L., Cardoso,

J. M. & Correia, C. M. 2010. Characterization of a

double probe for local pulse wave velocity assessment.

Physiol Meas, 31, 1449-65.

Pereira, H. C., Basílio, J. B., Malaquias, J., Pereira, T.,

Almeida, V., Borges, E., Figueiras, E., Cardoso, J. and

Correia C. 2011. Double Headed Probe for Local

Pulse Wave Velocity Estimation – A new device for

hemodynamic parameters assessment. In Proceedings

of BIOSIGNALS 2011, Rome, Italy.

Rabben, S. I., Stergiopulos, N., Hellevik, L. R., Smiseth,

O. A., Slordahl, S., Urheim, S. & Angelsen, B. 2004.

An ultrasound-based method for determining pulse

wave velocity in superficial arteries. J Biomech, 37,

1615-22.

Segers, P., Kips, J., Trachet, B., Swillens, A., Vermeersch,

S., Mahieu, D., Rietzchel, E., Buyzere, M. and Bortel,

L. 2009. Limitations and pitfalls of non-invasive

measurement of arterial pressure wave reflections and

pulse wave velocity. Artery Research,

3, 79-88.

Shoji, T., Emoto, M., Shinohara, K., Kakiya, R.,

Tsujimoto, Y., Kishimoto, H., Ishimura, E., Tabata, T.

& Nishizawa, Y. 2001. Diabetes mellitus, aortic

stiffness, and cardiovascular mortality in end-stage

renal disease. J Am Soc Nephrol, 12, 2117-24.

Willems, J. L., Roelandt, J., De Geest, H., Kesteloot, H. &

Joossens, J. V. 1970. The left ventricular ejection time

in elderly subjects. Circulation, 42, 37-42.

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices