A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform

Measurement

Dan Wang

, Dean Krusienski

and Zhili Hao

Department of Mechanical and Aerospace Engineering, Old Dominion University, Norfolk, VA, U.S.A.

Electrical and Computer Engineering, Old Dominion University, Norfolk, VA, U.S.A.

Keywords: Wearable Sensors, Microfluidics, Arterial Pulse Waveform, Baseline Drift, Health Monitoring.

Abstract: In light of the need of health monitoring, the paper presents a flexible polyethylene terephthalate (PET)-

based wearable sensor for arterial pulse waveform measurement. The sensor encompasses a

polydimethylsiloxane (PDMS) microstructure embedded with an electrolyte-enabled 5×1 transducer array,

which spans 6mm and has a spatial resolution of 1.5mm. A pulse signal exerts a deflection on the

microstructure and is recorded as a resistance change by a transducer at the site of the pulse. An untrained

individual can easily align the sensor on a targeted artery with a negligible margin and then acquire the

arterial pulse waveform continuously and non-invasively. This sensor is fabricated using microfluidics

technology and thus features low cost for mass production. The sensor is hand-held on an artery and records

its pulse signal for a 10s period, which bears baseline drift, due to the respiration and the motion artifact.

Discrete Meyer Wavelet Transform (DMWT) and Cubic Spline Estimation (CSE) are employed to remove

baseline drift in a pulse signal. The pulse waveform is expressed in terms of the sensor deflection as a

function of time. Carotid arterial pulse waveforms are measured by the sensor on three subjects at rest and

on two subjects post-exercise. Additionally, radial arterial waveforms are measured on one subject at rest.

The measured pulse pattern change of the two subjects between at rest and post-exercise is consistent with

the literature. As the pulse transmits from central (carotid) to peripheral (radial) for one subject, the ratio of

amplitude of main peak to amplitude of dicrotic wave goes up and the up-stroke time becomes shorter. This

is consistent with the related observations in the literature. Thus, the limited amount of data collected here

demonstrates the feasibility of using the sensor as a wearable health monitoring device.

1 INTRODUCTION

Arterial pulse waveforms are intimately associated

with the physiological conditions of the

cardiovascular system and thus provide valuable

information of the diagnosis and treatment of

Cardiovascular disease (CVD) (Lin et al., 2013). As

reported by the World Health Organization, CVDs

are the number one cause of death globally: more

people die annually from CVDs than from any other

cause (Lin et al., 2013). For this reason, various

devices and techniques have been developed for

arterial pulse waveform measurement, or arterial

tonometry. To date, a few devices for arterial

tonometry, including a flexible pulse monitoring

system from Pressure Profile System (PPS) (Hu et

al., 2012) and CASPal system that measures Central

Aortic Systolic Pressure (CASP) from HealthStats

(Saugel et al., 2014), are commercially available and

have been successfully employed for arterial pulse

waveform measurement. However, current

tonometric devices are unsuitable for not only

wearing with relative comfort but also for an

untrained individual to use at home.

Recently, based on microfluidics technology, a

polyethylene terephthalate (PET)-based sensor array

to monitor arterial pressure waveforms is developed

(Digiglio et al., 2014). Although this PET-based

sensor array offers quite a few attractive features for

daily use by an untrained individual, it entails

complex fabrication process, including bonding

three layers together and injecting electrolyte into

each individual sensor in the sensor array.

Meanwhile, its spatial resolution of 5mm×5mm is

well above the typical size of a carotid and radial

artery, and thus the sensor array may risk having no

transducers placed right above the target artery.

To address the above-mentioned two issues

Wang, D., Krusienski, D. and Hao, Z.

A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement.

DOI: 10.5220/0005698000660075

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 1: BIODEVICES, pages 66-75

ISBN: 978-989-758-170-0

preserving the attractive features of the sensor array,

this paper presents a PET-based sensor for arterial

pulse waveform measurement. With a flexible

substrate, this sensor bears the same design of a

Pyrex-based sensor previously developed by our

group (Cheng et al., 2013; Gu et al., 2013). The core

of the sensor is a single polydimethylsiloxane

(PDMS) microstructure embedded with an

electrolyte-enabled resistive transducer array

underneath. The spatial resolution of 1.5mm of the

transducer array allows aligning one of them right at

the site of an artery with a negligible margin. A

simple, low-cost fabrication process is developed for

realizing this PET-based sensor, where a new

bonding process is employed to strengthen the

bonding between the PDMS microstructure and the

PET substrate with indium titanium oxide (ITO)

electrodes. The fabricated sensor is further utilized

to measure carotid arterial pulse waveforms of three

subjects and radial arterial pulse waveforms of one

subject, and the measured results are compared with

the related information in the literature for

demonstrating its feasibility.

2 SENSOR DESIGN AND

FABRICATION

2.1 Sensor Design

Figure 1 depicts the configuration of the PET-based

sensor. The sensor encompasses a rectangular

PDMS microstructure embedded with an electrolyte-

filled microchannel, and a set of ITO electrode pairs

distributed along the microchannel length. The

portion of electrolyte across an electrode pair

functions as a resistive transducer, whose resistance

varies with the bottom deflection of the

microstructure at its location and is routed out by the

electrode pair. Thus, together with the set of

electrode pairs, one body of electrolyte in the

microchannel forms a 5×1 transducer array with a

spatial resolution of 1.5mm.

Distributed deflection acting on the top of the

microstructure translates to the bottom deflection of

the microstructure and thus geometrical changes of

the microchannel, which register as resistance

changes by the transducer array. Table 1 summarizes

the key design parameters of the sensor. The details

of the sensor design can be found in the literature

(Cheng et al., 2013; Gu et al., 2013).

(a)

(b)

Figure 1: Configuration of microfluidic sensor: (a) 3D

view with labelling; (b) side view with key design

parameters being labelled.

Table 1: Key design parameters of the PET-based sensor.

Parameter Value Symbol

Microchannel cross-section

1mm×80µm w

×h

Microchannel length 30mm L

Spatial resolution 1.5mm d

Microstructure thickness 1.2mm h

2.2 Fabrication Process

Figure 2 illustrates a low-cost, two-mask fabrication

process for realizing the PET-based sensor. The

process starts with a commercial ITO/PET sheet (a

0.2mm-thick PET substrate coated with

120nm~160nm-thick ITO layer). To pattern ITO

electrodes on the PET substrate, a 15m-thick dry

film (Alpho NIT 215, NichigoMorton Co., Ltd.) is

laminated onto the PET substrate. Via the first mask,

electrode pattern is transferred to the dry film, which

is followed by wet etching of the ITO layer to form

ITO electrodes. Afterward, the dry film is removed

using ethanol. Via the second mask, a SU8 mold is

created on a Pyrex substrate. Then, a mixture of

curing agent to PDMS elastomer with a weight ratio

of 1:10 is poured over the SU8 mold. After being

cured at room temperature over 24hrs, the

microstructure is peeled off from the SU8 mold and

a hole is punched into each reservoir using a needle.

To strengthen the bonding strength between the

PDMS microstructure and the PET substrate with

Pulse Signal

Microchannel

Reservoi

PET Substrate

PDMS

ITO Electrodes

A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement

patterned ITO electrodes, a chemical gluing strategy

is adopted (Tang and Lee, 2010, Tsuwaki et al.,

2014). First, photoresist is placed onto the

microchannel of the microstructure. Then, the

patterned ITO electrodes and the microstructure are

activated with hydroxyl groups by an oxygen plasma

treatment for 1 minute, which are followed by

immersing the microstructure and ITO electrodes

into 1% (v/v) 3-Glycidyloxypropyltrimethoxysilane

(GOPTS) and 5% (v/v) 3-

Aminopropyltriethoxysilane (APTES) for 20

minutes, respectively. Afterward, the microstructure

is rinsed with acetone, isopropanol and DI water,

and, sequentially, the PET substrate with patterned

ITO electrodes is rinsed with ethanol and DI water.

Finally, the microstructure and the PET substrate are

aligned and bonded under a contact pressure at

100 ℃ for 5 minutes, then at 50 ℃ for 24hrs.

Electrolyte, 1-ethyl-3-methylimidazolium

tricyanomethanide (EMIM TCM), is injected into

the microchannel via a reservoir using a syringe.

Two reservoirs are then sealed with PDMS with a

weight ratio of 1:10. Conductive epoxy is used to

make electrical connection between the contact pads

of the sensor and the associated electronics on PCBs.

Figure 2 shows a couple of pictures of the fabricated

PET-based sensor.

Figure 2: Fabrication process for the PET-based sensor.

(a) Dry film lamination. (b) Patterning of dry film. (c)

Patterned ITO electrodes. (d) Alignment and bonding with

microstructure. (e) 3D view of the sensor.

3 ARTERIAL PULSE WAVEFORM

MEASUREMENT AND SIGNAL

PROCESSING

The measurements of arterial pulse waveform are

conducted on three subjects (a 16yr-old female

teenager, a 28yr-old female adult and a 28yr-old

male adult) in a quiet environment, after resting for

several minutes and no drinking and eating for over

1hr. All the pulse signals are taken from the carotid

artery at the right of the neck, except one is taken

from a radial artery. Additionally, two subjects are

required to do 5min-long strenuous exercise and

their carotid pulse signals are taken immediately

after exercise from the same location as at rest.

3.1 Pulse Waveform Measurement

As shown in Figure 3(c), the PET-based sensor is

placed at the site of the carotid artery and is pressed

against it with two fingers. Note that the hold-down

pressure against an artery is uncontrollable. The

arterial pulse exerts a time-varying deflection on the

top of the PDMS microstructure, which registers as a

resistance change by the transducer at the site of the

artery.

Figure 3: Pictures of the fabricated PET-based sensor. (a)

The transducer array with labelling. (b) The whole sensor.

arterial pulse waveform measurement: two fingers holding

the sensor above the carotid artery.

To monitor resistance changes in the transducer

array, a function generator is used to apply an

Alternating Current (AC) signal (100kHz, peak-peak

amplitude: 220mv) as the input signal for all the

transducers (Cheng et al., 2013). The output of a

transducer is connected to its electronics

implemented on PCB for both amplifying the AC

signal coming out from the transducer and

(b)

(c)

(a)

=1.5mm

Transducer

Array

Transducer

Array

=1mm

=30mm

(a)

(b)

(c)

(d)

(e)

PET

Dry film

Patterned Dry film

PDMS

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

converting the AC signal to a Direct Current (DC)

output, which is recorded by a LabVIEW program.

Note that the same design of electronics is used for

all the transducers, but is implemented on separate

PCBs. The circuit design can be found in our

previous work (Gu et al., 2013). Each pulse is

recorded for a 10s period. The sampling rate is kept

at 500Hz. Later on, 25 data points per second are

utilized for the extracted pulse waveform from an

originally recorded pulse signal.

3.2 Signal Processing for Converting

Resistance Changes to Pulse

Waveforms

As mentioned above, the recorded parameter of a

transducer is a DC voltage output, V

out

, which is

related to the resistance of a transducer by (Cheng et

al., 2013; Gu et al., 2013):

out



(1)

where v

is the peak-to-peak value of the AC signal,

and R

is the feedback resistance of the electronics

used. Therefore, the resistance of a transducer can be

obtained by:

pp F

out



(2)

The bottom deflection at a transducer is

represented by the resistance change:

pp F

out out









(3)

where V

out0

is the DC voltage of a transducer, when

it is free of deflection.

Owing to fabrication variation in transducer

height, h

, (the smallest design parameter), the

original resistance (defined as the resistance of a

transducer when it is free of deflection) may vary

among the transducers. The original resistance of the

transducer is roughly calculated as (Yang et al.,

2015):

eei











(4)

where is the electrical conductivity of EMIM

TCM, d

, w

and h

e-i

are the length, width and height

of the i

transducer, respectively. We further define

the resistance of a transducer after being pressed

against an artery as its initial resistance, R

0-i

’

. As

such, the resistance change is calculated relative to

the initial resistance, instead of the original

resistance:

1- 1

() /2

/2 ( )

esi

eei ei

eie

isi

eei e

RRR

dh h

wRd

dh w











 









 















 



(5)

where z

s-i

is the deflection for the i

transducer, h

e-i

is the initial height of the transducer.

Since the hold-down pressure against an artery is

not controllable, the initial resistances of the sensor

vary among all the measurements. According to

Equation (5), the sensor deflection at the i

transducer can be obtained:

()













(6)

3.3 Signal Processing for Baseline Drift

Removal

The respiration and motion artifact (i.e., motion of

the sensor and the body during measurement) can

introduce baseline drift to the recorded pulse

waveform. The Discrete Meyer Wavelet

Transformation (DMWT) and Cubic Spline

Estimation (CSE) have been implemented for

removing baseline drift from the recorded data (Xu

et al., 2007). DMWT is well known for representing

localized variations in a signal simultaneously in the

time and frequency domains. CSE is used to detect

the amplitude envelope of the signal.

Because the baseline drift introduced by

respiration and the body and sensor’s motion has

nonlinear and quasi-periodic contents, linear

interpolation estimation has been proven ineffective

(Xu et al., 2007). In contrast, a high-degree

polynomial is smooth, but it may cause the Runge

phenomenon, which increases the error of the signal

(Xu et al., 2007). Thus, CSE is widely used and is

adopted in this work to remove the baseline drift

when the Energy Ratio (ER) of the recorded data

reaches a threshold.



A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement

Figure 4: Flow chart of signal processing for pulse waveforms (Xu et al., 2007).

In DMWT, a function is defined in frequency

domain:

1/2 /2

1/ 2 / 2

(2 ) sin 1

(2 ) cos 1

(w)

evw















  









  









, ,



























(7)

where v(x) is an auxiliary function expressed as (Xu

et al., 2007):

423

() (35 84 70 20 ), [0,1]vx x x x x x 

(8)

Figure 4 depicts the flow chart of signal

processing for removing the baseline drift from the

original pulse signal, while Figure 5 illustrates the

corresponding intermediate results of the signal

processing. First, we utilize CSE to obtain the

baseline envelope of the recorded resistance signal.

Figure 5(a) shows both the recorded resistance

signal, converted from the DC output voltage signal

using Equation (2), and the approximated baseline of

the signal via CSE as a function of time.

During the measurement post-exercise, there are

three to five respiration cycles for a 10s period. The

motion artifacts are the low frequency components.

The main frequency of baseline drift is less than 1Hz.

The cutoff frequency of the fourth-level scale

function is 1.56Hz. And the frequency content of the

pulse waveform is less than 40Hz. Thus, (A1−A4)

and A4 are used to approximate the pulse signal and

its baseline drift, respectively. Thus, we apply

DMWT to the resistance signal to obtain its first-

level approximation, A1, and its fourth-level

approximation, A4, as shown in Figure 5(b). The

two approximations are utilized to compute the ER

of the recorded pulse signal as below:

(a)

(b)

(c)

(d)

(e)

Figure 5: The corresponding intermediate results of the

signal processing using DMWT and CSE. (a) Resistance

with estimated baseline. (b) The first-level and fourth-

level approximation content of pulse decomposition. (c)

The resistance change with estimated baseline. (d) The

resistance change. (e) Sensor deflection at the 4

transducer.

Wav ele t

Decomposition

Point

Detection

Cubic

Spline

Estimation

Computing

the ER

Wavelet

Filter

ER > 10dB

R and

Deflection

∆

Original

Pulse

Signal

Filtered

Pulse

Signal

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

14 (14)

20log

4(4)

AAmeanAA

AmeanA

 





(9)

where || || represents the order-two norm, mean|| A1-

A4|| represents the average of A1-A4.

ER is used to quantify the extent of the baseline

drift. It is found that a pulse signal exhibits little

base drift when its ER is higher than 10dB.

Therefore, 10dB is selected as the ER threshold for

removing the baseline drift from the recorded

resistance signal. If ER is higher than 10 dB, the

baseline drift is removed by subtracting the

approximated baseline in Figure 5(a) from the

recorded resistance signal. If ER is lower than 10dB,

the baseline drift is removed by subtracting A4 in

Figure 5(b) from the recorded resistance signal.

Then, we obtain the resistance change of the

recorded resistance signal, as shown in Figure 5(c).

However, the pulse waveform still contains baseline

drift. CSE is used again to obtain the baseline drift

of the resistance change, as shown in Figure 5(c).

Afterward, the resistance change in Figure 5(c) is

subtracted by this new baseline drift to obtain the

resistance change in Figure 5(d). Finally, based on

Equation (6), the resistance change in Figure 5(d) is

converted to the sensor deflection, as shown in

Figure 5(e). The sensor deflection captures the pulse

waveform without being distorted by the baseline

drift. All the pulse signals measured on a subject at

rest have an ER value higher than 10dB, while the

pulse signals measured on a subject post-exercise

have an ER value lower than 10dB.

4 MEASURED RESULTS

Our goal for the sensor in this work is to aquire the

pulse waveform of a subject so as to evaluate some

physiological parameters related to the

cardiovascular system. Certainly, with an

oscillometric device (Digiglio et al., 2014), the

measured results can be further processed to obtain

the absolute values for blood pressure. To

demonstrate the feasiblity of the sensor for pulse

waveform measurement, carotid arterial pulse

waveforms are measured by the sensor on three

subjects at rest and on two subjects post-exercise.

Additionally, radial arterial waveforms are measured

on one subject at rest. Note that since we recently

started to explore the sensor for pulse waveform

measurement, the data collected here is not

comprehensive. Nevertheless, the measured pulse

waveforms are compared with the related

information in the literature for the feasibility of

using the sensor for pulse waveform monitoring.

4.1 Robustness to Motion Artifact

Motion artifact from the respiration and the

handshaking during measurement are unavoidable.

Sometimes, it is needed to assess arterial pulse

waveform difference between before-exercise and

post-exercise. As such, the sensor needs to be

immune to motion artifact. Figure 6 shows the

measured resistances of the five transducers

measured on the 16-yr female post-exercise. The 4

and 5

transducers capture not only a clear patterned

pulse signal but also the heavily-breathing pattern,

indicating that the carotid artery is between the two

transducers. The breathing pattern introduces

extremely large baseline drift to the recorded pulse

signal. However, as will be seen in the next

subsection, the sensor is capable of obtaining the

undistorted pulse waveform under such severe

motion artifact.

The rest three transducers obtain random signals

with a respiration pattern, indicating that these

transducers are away from the carotid artery. Note

that this measurement indicates that the transducers

do not interfere with each other. Thus, the sensor

provides a negligible alignment margin for a

untrained individual to use.

4.2 Carotid Pulse Waveforms at-Rest

and Post-Exercise

The same sensor is used to conduct all the pulse

measurements on the carotid artery. Thus, variation

in pulse waveform among the three subjects arises

mostly from their cardiovascular system conditions,

in the sense that hold-down pressure may affect the

pulse waveform slightly, according to Equation (6).

Quantification of the effect of hold-down pressure

on the measured pulse waveform needs to be further

studied in the future. The measurement on the

carotid artery of each subject was repeated several

times. For each measurement, pulse signal typically

shows up in two or three transducers. For

consistence, we simply choose the pulse signal from

the 4

transducer for comparison. Thus, the sensor

deflection at this transducer is used to present a

measured pulse waveform.

Figure 7(a) compares the pulse waveforms

measured on the three subjects at rest and Figure 7(b)

compares the pulse waveforms measured on onsets

of the first pulses of the subjects are set at the same

A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement

(a)

(b)

(c)

(d)

(e)

Figure 6: Arterial waveform patterns of five transducers

from the 16yr-old female teenager post-exercise. (a)-(e)

are recorded resistance of each transducer.

time instant. While the hold-down pressure against a

carotid artery is larger at-rest than post-exercise, the

sensor deflection is larger at rest than post-exercise,

as can be seen in Figure 7. This indicates that a

strong pulse from post-exercise does not directly

translate to a large sensor deflection, without a large

hold-down pressure.

(a)

(b)

Figure 7: Carotid pulse waveform in terms of the sensor

deflection at the 4

transducer at-rest (a) and post-exercise

(b).

Figure 8(a) shows the aorta pressure waveform

in the comparison between two adults: a 68-year old

individual and his 37-year-old son (O'Rourke and

Hashimoto, 2007). The pulse pattern varies with

ages. Since carotid artery is central and thus is

representative of aorta pressure. Evidently, the pulse

waveforms of the 28yr-old male and 16yr-old female

(both physically active) at rest exhibit a quite

vertical up-stroke and are similar to the one of the

young man. In constrast, containing an inclined up-

stroke, the pulse waveform of the 28yr-old female

(not physically active) at rest is closer to the one of

the old man.

The 28yr-old female has the lowest pulse rate but

the highest pulse strength. This high pulse strength

might be due to the artery being close to the skin.

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

(a)

(b)

Figure 8: Aortic (a) and radial (b) pressure wave in 36-

year-old man (young) and his 68-year-old father (old)

(O'Rourke and Hashimoto, 2007).

Figure 9: Carotid artery distension wave and blood

pressure recordings before, during and after dynamic

exercise. Carotid artery distension wave recordings and

corresponding blood pressure values in one representative

subject during the control period, at peak-exercise, and at

5 and 55 min post-exercise. Time scale = 1s. (Studinger et

al., 2003, with permission).

The 16yr-old female has a faster pulse rate than the

28yr-old male both at rest and post-exercise. The

dicrotic notch in the pulse waveforms of the two

subjects post-exercise is much lower than its

counterpart at rest. Figure 9 shows a similar

difference in pulse waveform between at rest

(control) and post-exercise. Thus, our measurement

(Figure 7) is consistent with that in the literature.

4.3 Radial Arterial Pulse Waveforms

Another PET-based sensor is used to measure the

radial pulse of the 28yr-old female, with the sensor

being aligned in parallel with and perpendicular to

the radial artery. Figure 10 illustrates the measured

pulse waveforms under the two alignments. In the

parallel alignment, three transducers, 1, 2 and 3, are

at the site of the artery, while only transducer 3 is at

the site of the artery in the perpendicular alignment.

(a)

(b)

Figure 10: The radial artery pulse measurement for 28yr-

old female adult. (a) The sensor aligned parallel with the

radial artery. (b) The sensor aligned perpendicular to the

radial artery.

The pulse waveform varies between the two

alignments, which is believed to result from the

sensor being more loosely in contact with the artery

in the perpendicular alignment than in the parallel

Young

Old

130

Time

(

)

Radial

mmHg

Young

Time

(

)

Old

AORTA

130

mmHg

A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement

alignment. Thus, the pulse waveform in Figure 10(a)

is believed to be correct and is compared with the

one in Figure 8(b), showing a little bit away from the

one of the young man.

4.4 Arterial Tonometric Parameters

Figure 11 illustrates a typical cycle of a measured

pulse signal, which includes main arterial tonometric

parameters often used to characterize the waveform

(Velik, 2015). The physical meaning of the

tonometric parameters in the figure is given in Table

Figure 11: Parameters of pulse wave signal (Velik, 2015).

According to Figure 11, the arterial tonometric

parameters of the subjects are calculated and

summarized into Table 2. The ratio of h1/h5 is

calculated, instead of the absolute values of h1 and

h5, due to the lack of a device for calibration. The

radial pulse measurement under the parallel

alignment is included in the table. As mentioned

previously, the ratio of h1/h5 post-exercise is higher

than at rest, which is observed on the 16yr-old

female and the 28yr-old male. It has been found that

as the pulse transmits from central (carotid) to

peripheral (radial), the ratio of h1/h5 goes up

(McEniery et al., 2014) and the up-stroke time (t1)

becomes shorter (Hurst, 1982). The measured results

on the 28yr-old female are in good agreement with

these findings. Note that two different sensors of the

same design are used to measure the carotid and

radial arterial pulse waveforms of the 28yr-old

female. The consistency of the results on the carotid

artery and the radial artery (the ratio of h1/h5 going

up and the up-stroke time being shorter from carotid

to radial) simply prove the robustness of the sensor.

We envision that this flexible sensor can be

attached on a bracelet and can then be worn on the

wrist loosely, together with integrated circuit for

signal routing and wireless transmission. Whenever

an individual needs to measure the radial pulse

waveform, he can align the sensor at the site of the

artery and press the sensor against the artery with his

fingers. To measure the carotid artery, the bracelet

can be taken off the wrist and put the sensor on the

neck. Owing to the small size and flexibility of the

sensor, it is expected to be comfortable to wear a

bracelet with the sensor.

Table 2: Measured arterial tonometric characteristics from the subjects.

Variables Description

Carotid artery

Radial

artery

Female Teenager Male Adult Female Adult

Female

Adult

(parallel)

At-rest

Post-

exercise

At-rest

Post-

exercise

At-rest At-rest

h1/h5

Amplitude of main peak /

Amplitude of dicrotic wave

1.551 3.202 1.415 3.251 1.876 2.296

t1(s)

Duration between onset and

peak point

0.24 0.12 0.24 0.12 0.24 0.1

t2(s)

Duration between onset and

incisura

0.4 0.24 0.32 0.28 0.36 0.3

t3(s)

Duration between onset and

dicrotic

0.32 0.16 0.56 0.16 0.52 0.5

T(s) Time in one pulse cycle 0.72 0.4 0.88 0.44 0.88 0.8

Amplitude

Time

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

5 CONCLUSION

This paper presents a PET-based wearable sensor for

arterial pulse waveform measurement for untrained

individuals to conduct the arterial pulse waveform

measurement. The sensor contains a PDMS

microstructure embedded with a 51 resistive

transducer array, spanning 6mm and with a spatial

resolution of 1.5mm. Built on PET substrate, the

sensor is fabricated using a low-cost, two-mask

fabrication process. To demonstrate its feasibility for

arterial pulse waveform measurement, one sensor is

used to measure carotid arterial pulse waveform of

three subjects at rest and two subjects post-exercise,

while another sensor of the same design is used to

measure radial arterial pulse waveform of one

subject at rest. The respiration and motion artifact

introduces baseline drift to originally recorded pulse

signal. A combination of DMWT and CSE is

utilized to effectively remove the baseline drift in a

pulse signal. The robustness of the sensor to baseline

drift is demonstrated by the pulse signals measured

on a subject post-exercise. After its baseline drift

being removed, an arterial pulse waveform is

expressed in terms of the sensor deflection as a

function of time. All the measured pulse waveforms

of carotid and radial arteries of the three subjects at

rest and post-exercise are consistent with their

counterparts in the literature, thus demonstrating the

feasibility of using the sensor as a wearable health

monitoring device.

REFERENCES

Cheng, P., Gu, W., Shen, J., Ghosh, A., Beskok, A. and

Hao, Z., 2013. Performance study of a PDMS-based

microfluidic device for the detection of continuous

distributed static and dynamic loads. Journal of

Micromechanics and Microengineering, 23(8), p.

085007.

Digiglio, P., Li, R., Wang, W. and Pan, T., 2014.

Microflotronic arterial tonometry for continuous

wearable non-invasive hemodynamic monitoring.

Annals of biomedical engineering, 42(11), pp.2278-

2288.

Gu, W., Cheng, P., Ghosh, A., Liao, Y., Liao, B., Beskok,

A. and Hao, Z., 2013. Detection of distributed static

and dynamic loads with electrolyte-enabled distributed

transducers in a polymer-based microfluidic device.

Journal of Micromechanics and Microengineering,

23(3), p.035015.

Hu, C. S., Chung, Y. F., Yeh, C. C., and Luo, C. H., 2011.

Temporal and spatial properties of arterial pulsation

measurement using pressure sensor array. Evidence-

Based Complementary and Alternative Medicine, 2012.

Hurst, J. W. and Logue, R. B., 1982. The heart: arteries

and veins, McGraw -Hill, pp.170-179.

Lin, W. H., Zhang, H. and Zhang, Y. T., 2013.

Investigation on cardiovascular risk prediction using

physiological parameters. Computational and

mathematical methods in medicine, 2013.

McEniery, C. M., Cockcroft, J. R., Roman, M. J.,

Franklin, S. S. and Wilkinson, I. B., 2014. Central

blood pressure: current evidence and clinical

importance. European heart journal, 35(26), pp.1719-

1725.

O'Rourke, M. F. and Hashimoto, J., 2007. Mechanical

factors in arterial aging: a clinical perspective. Journal

of the American College of Cardiology, 50(1), pp.1-13.

Saugel, B., Fassio, F., Hapfelmeier, A., Meidert, A. S.,

Schmid, R. M., and Huber, W., 2012. The T-Line TL-

200 system for continuous non-invasive blood pressure

measurement in medical intensive care unit patients.

Intensive care medicine. 38(9), pp.1471-1477.

Studinger, P., Lenard, Z., Kovats, Z., Kocsis, L. and

Kollai, M., 2003. Static and dynamic changes in

carotid artery diameter in humans during and after

strenuous exercise. The Journal of physiology, 550(2),

pp.575-583.

Tang, L. and Lee, N. Y., 2010. A facile route for

irreversible bonding of plastic-PDMS hybrid

microdevices at room temperature. Lab on a Chip,

10(10), pp.1274-1280.

Tsuwaki, M., Kasahara, T., Edura, T., Matsunami, S.,

Oshima, J., Shoji, S., Adachi, C. and Mizuno, J., 2014.

Fabrication and characterization of large-area flexible

microfluidic organic light-emitting diode with liquid

organic semiconductor. Sensors and Actuators A:

Physical, 216, pp.231-236.

Velik, R., 2015. An objective review of the technological

developments for radial pulse diagnosis in Traditional

Chinese Medicine. European Journal of Integrative

Medicine, 7(4), pp.321-331.

Xu, L., Zhang, D., Wang, K., Li, N. and Wang, X., 2007.

Baseline wander correction in pulse waveforms using

wavelet-based cascaded adaptive filter. Computers in

Biology and Medicine, 37(5), pp. 716-731.

Yang, Y., Shen, J. and Hao, Z., 2015. A two-demensional

(2D) distributed deflection sensor for tissue palpation

with correction mechanism for its performance

variation. under review.

A Flexible PET-based Wearable Sensor for Arterial Pulse Waveform Measurement