A Novel Microfluidic Pressure Sensor for Traditional Chinese

Medicine (TCM) Pulse Data Collection and Analysis

Zhu Hong

, Lin Liutong

, Er Jui Pin

, James Wong

and Sun Lingling

Healthcare Engineering Centre, School of Engineering, Temasek Polytechnic, 21 Tampines Ave. 1, Singapore

TP-HRG Robotics Innovation Centre, School of Engineering, Temasek Polytechnic, 21 Tampines Ave. 1, Singapore

Keywords: Wearable Electronics, Microfluidic Pressure Sensor, Traditional Chinese Medicine, TCM Pulse Analyzer.

Abstract: Recently, we developed a novel microfluidic pressure sensor which can accurately sense and collect human

wrist arterial pulse signals to be used in a wearable TCM pulse analyzer enabled with artificial intelligence

for self-monitoring of cardiovascular disease. Various micro tactile sensor structures had been explored and

fabricated using in-house microfabrication facilities. The connection between the parameters of sensor and its

output has been investigated and found that the parameters of pressure sensor had great influence on its

performance. An easy-to-use mechanical structure to hold the sensor and pulse signal reading and processing

electronics on both hardware and firmware have also been designed and fabricated.

1 INTRODUCTION

1.1 Significance of TCM Pulse

Diagnosis

TCM has been practiced for more than 2,500 years.

The 11th revision of the World Health Organization’s

(WHO) International Statistical Classification of

Diseases and Related Health Problems (ICD-11),

which was released on 18 Jun 2018, included TCM

for the first time. A disease classification system,

based on TCM was established in the ICD-11. It is of

great practical and historical significance to promote

the integration of TCM and the medical and health

systems in the world, and to lay the foundation for the

world to understand and use TCM.

TCM pulse diagnosis has been used by TCM

physicians to assess patients’ health conditions. As

shown in Figure 1, a TCM physician palpates six

locations, three on each wrist, with the three points

called "Cun" (寸), "Guan" (关) and "Chi" (尺) at three

pulse depths called "Fu" ( 浮 ), "Zhong" ( 中 ) and

"Chen" (沉) and describes pulses in terms of various

characteristics. By comparing the pulses, the health

https://orcid.org/0000-0003-2574-6258

https://orcid.org/0000-0003-2401-3853

https://orcid.org/0000-0002-7071-9462

https://orcid.org/0000-0002-9465-7198

status of individual organs and the whole body can be

determined

(

Meng et al., 2001 and Wang, 2000).

Figure 1: An illustration of the TCM palpation: (a) A TCM

physician using his 3 fingers (index, middle, ring finger) to

collect pulse signal, (b) 3 pulse-taking positions (Cun,

Guan, and Chi) and (c) 3 pulse-taking depths (Fu, Zhong

and Chen).

The basic rationale behind TCM pulse diagnosis is

that pathologic changes in the body are reflected in

the radial pulse. This premise is supported by western

clinical research studies, which have found evidence

for loss of arterial elasticity and alterations in pulse

amplitude, rhythm, and shape in patients with

cardiovascular disease, hypertension, diabetes, etc.

(Malinauskas et al., 2013). In several articles, the

pulse wave is described as the superposition of a

forward traveling wave caused by the ventricular

output of blood and a phase-shifted backward

traveling wave reflected from the peripheral blood

Zhu, H., Lin, L., Er, J., Wong, J. and Sun, L.

A Novel Microﬂuidic Pressure Sensor for Traditional Chinese Medicine (TCM) Pulse Data Collection and Analysis.

DOI: 10.5220/0011679000003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 135-141

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

135

vessels (Malinauskas et al., 2013; Jeon et al., 2011

and Shu, 2007). It was indicated that the forward

traveling component is correlated with information

about the heart itself and the backward traveling wave

contains information about the arterial tree and the

organs it flows through (Malinauskas, et al., 2012).

Therefore, the pulse diagnosis has important clinical

values on diagnosing diseases of various organs and

assessing the patient's status holistically.

1.2 Modern Scientific Approaches on

the Objectification of TCM Pulse

Diagnosis

Despite its importance, TCM pulse diagnosis is

subjective and mysterious as it is based purely on the

physicians’ finger feels. The wrist pulse

characteristics are mainly described qualitatively and

cannot be clearly defined quantitatively. The

fuzziness and ambiguity of pulse concepts make it

difficult to study and master pulse diagnosis. As long

as pulse diagnosis remains in the realm of fingertip-

reading, it will be a difficult skill to master and have

a great deal of subjectivity in interpretation.

Objectification of the pulse diagnosis is highly

desired. Over the last two decades, considerable

research efforts have been put into the attempt to

objectify the pulse diagnosis. Various pulse diagnosis

instruments have been developed to obtain pulse

signals quantitatively (Zhang et al., 2011; Luo et al.,

2012 and Hu et al., 2012), and many analytical

methods have been proposed for exploring the

mechanism of pulse conditions and correlations with

diseases (Jeon et al., 2011; Wei et al., 2009, Wang et

al., 1997; Liao et at., 2012). These modern practices

of pulse diagnosis have revived the TCM and

demonstrated the value of pulse diagnosis is not only

non-invasive diagnosis but also an early prediction of

diseases. Yi-Chia Huang (Huang et al., 2019) has

identified predictive factors of pulse spectrum to

increase the prediction rate of coronary artery disease

(CAD) diagnosis in patients with chest pain or angina

pectoris prior to cardiac catheterization. For

predicting the attack of type 2 diabetes, an algorithm

developed by Yiming Hao (Hao et al., 2019) achieved

an accuracy of 96.35%. Many other studies also

provide important evidence of using pulse

characteristics for various disease diagnoses, such as

lung cancer recognition (Zhang et al., 2018), atopic

eczema diagnosis (Liou et al., 2011), fatty liver

disease diagnosis (Wang et al., 2015), dyspepsia and

the rhinitis detection (Huang et al., 2011).

1.3 The Interest in Developing

Wearable TCM Pulse Analyzer

The pulse analysers currently developed are mainly

desktop instruments (Luo et al., 2012). However, the

desktop pulse analysers are cumbersome and need

complicated pressure adjustment for pulse taking,

which prevents them from being extensively used by

clinicians and patients, especially for continuously

monitoring the pulse signals. Continuous monitoring

of the pulse signal is preferable for many disease

diagnoses and early prediction of risks. With

continuous monitoring, the history of a health

condition related to the diseases can be retrieved for

helping the diagnosis process, and certain important

physiological and pathological characteristics will

not be missed.

In the prospect of the increase in demand,

developing a low-cost, wearable pulse analyzer that

can offer continuous monitoring, immunizes

mechanical and electronic noises, as well as displays

some basic diagnostic results is becoming one of the

most popular research topics (Wang et al., 2016 and

Goyal et al., 2017). However, the wearable pulse

analysers currently available in the market or under

development can only exert non-adjustable pressure,

which is inconsistent with the palpation theory in

TCM. There is no easy-to-use, wearable TCM pulse

analyzer which can palpate at three positions and

three depths.

1.4 Proposed Wearable TCM Pulse

Analyzer

We aim to develop a TCM pulse collector or analyser

to collect the pulse signal at three positions (Cun,

Guan, Chi). A simple mechanical structure will hold

the pulse sensor and to control the exerting pressure

at three levels (Fu, Zhong, Chen) according to TCM

theory has been designed and fabricated. Both

electronics and firmware of the pulse analyser have

been developed for pulse signal reading and

processing. Various filters and algorithms have been

explored to de-noise and remove signal drift. The

signal processing sub-system further controls the

adjustment of the pulse signal collection at three

positions (Cun, Guan, Chi) and three depths (Fu,

Zhong, Chen).

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

136

2 DESIGN AND FAFRICATION

2.1 Sensor Fabrication

The flexible microfluidic pressure sensors have been

designed with different patterns and or geometry

parameters (microfluidic channel width ranging from

200 um to 500 um, channel spacing ranging from 200

um to 500 um). Plastic masks with different patterns

have been printed to fabricate microfluidic sensor

SU8 mould using lithography technique. Figure 2

shows some typical designs of the sensor. Figure 3

shows a SU8 mould for flexible sensor fabrication

using lithography process.

We developed both the bonding process of

Ecoflex and Polydimethylsiloxane (PDMS) to

Polyethylene terephthalate (PET) film with metal

electrodes. The sensors made by bonding PDMS to

PET film have poor sensitivity compared with sensors

made by bonding Ecoflex to PET film, due to the

higher Young's modulus of PDMS than Ecoflex.

Figure 2: Some designs of the flexible microfluidic pressure

sensors.

Figure 3: A 4-inch size SU8 mould for flexible sensor

fabrication.

The flexible microfluidic sensor fabrication

process is composed of following several individual

steps as shown below:

a) SU8 mould fabrication process, including spin

coating of SU8, prebaking of the SU8, ultraviolet

(UV) exposure lithography pattern formation, post-

exposure baking of the SU8 and SU8 pattern

development process.

b) Ecoflex sensor moulding by mixing the

Ecoflex A and B solution, pouring the Ecoflex

solution on the SU8 mould, degassing the Ecoflex

liquid in the vacuum oven, baking at 70

C for 1hour.

c) Demoulding of the Ecoflex from the SU8

mould and separation of the individual Ecoflex sensor

from the whole piece of Ecoflex.

d) Reactive Ion Etching (RIE) O

plasma

treatment on both surfaces of Ecoflex and PET film,

or by hand-held Corona Treaters having air plasma

treatment on both surfaces of Ecoflex and PET film.

The RIE is a standard microfabrication machine, O

plasma treatment by RIE is a more standard method.

e) Bonding of the Ecoflex sensor to the PET film

with metal electrode immediately after the surface

treatment, which is the most critical step and with the

lowest yield in all steps of the fabrication steps.

f) Baking of the bonded sensor at 80

C for 30

mins to further strengthen the bonding strength

between Ecoflex and PET film.

g) Injection of the Eutectic GaIn liquid metal into

the channel between the Ecoflex and PET film and

sealing of the injection hole by Ecoflex liquid; h)

Soldering of the electrically conductive cloth to the

PET film.

2.2 Testing Methods of the Pressure

Sensor

Characterization of the sensor was carried out by

measuring the pressure sensor's electrical resistance

output versus the force applied to the sensor. All the

fabricated microfluidic pressure sensors are tested

using mainly two equipment, force load machine (A)

and a digital multimeter (B) as shown in Figure 4. The

force load machine allows a varied force from 0 N to

5 N which corresponds to the force generated by the

Figure 4: Testing setup for microfluidic pressure sensor.

A Novel Microﬂuidic Pressure Sensor for Traditional Chinese Medicine (TCM) Pulse Data Collection and Analysis

137

human pulse on the wrist to be applied onto the

pressure sensor. The multimeter allows the

measurement of the resistance between the electrode

when a force is applied.

3 RESULTSAND DISCUSSION

3.1 Sensor Parameters and Its Output

Figure 5 shows the structure of the microfluidic

pressure sensor. The working principle of the

microfluidic pressure sensor is based on the resistive

mechanisms where an external force applied to the

device causes a resistivity change in the conducting

path of the microfluid. The conductivity of the

microfluid liquid is determined by the eutectic metal

in the channel. When under pressure, the channel

height will become smaller, the channel length will

become longer, thus the resistance of the liquid metal

inside the channel will become larger under pressure.

An external force deforms this conductive path which

results in an increase in resistance of the sensor.

Figure 5: The structure of the microfluidic pressure sensor.

Figure 6 shows the cross-section diagram of the

microfluidic pressure sensor. The fabrication

parameters of the sensor include channel width,

channel spacing, channel height, and Ecoflex

thickness or height. Below, the impacts of the

different parameters of the sensors on the electrical

resistance between the electrode are discussed.

It is found that the sensors with small dimensions

like 8 mm x 8 mm were very difficult to fabricate, due

to the poor bonding strength between the Ecoflex and

PET film. Sensors with bigger sizes have better yields

in fabrication. It is not recommended to fabricate very

small sensors like 8 mm X 8 mm in size. Currently,

our sensor with good yields is the one with a circular

channel shape with dimension 15 mm X 22 mm.

Figure 7 shows a fabricated microfluidic pressure

sensor, composed of a microchannel filled with liquid

metal between the Ecoflex and PET film.

Figure 8 shows the effects of the Ecoflex

thickness on the output resistance with same channel

width and spacing 500 µm/500 µm. channel height

Figure 6: Cross-Sectional diagram of microfluidic pressure

sensor.

about 60 µm. It can be seen clearly, while the

thickness of the Ecoflex decreased from 1.411 mm to

0.858 mm from top curve to the bottom curve, the

gradient of the force- resistance curve decreased

dramatically, which means the sensitivity of the

sensor decreased while thickness of the sensor

increased. This is reasonable, as a thin Ecoflex tends

to have more deformation when under same pressure,

which results in a higher change of the resistance of

the metal liquid, thus a better response. When the

thickness of the sensor was about 1.4 mm, the

sensor’s resistance almost has no change under

pressure.

Figure 7: Fabricated microfluidic pressure sensor.

Figure 8: Electrical resistance of pressure sensor with same

channel width and spacing 500 µm /500 µm and same

channel height 60 μm but with different sensor thickness.

01234

500/500, T=0.858mm

500/500, T=0.961mm

500/500, T=1.398mm

500/500, T=1.411mm

Resistance, Ohm

Force, N

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

138

Figure 9 shows the electrical resistance response

of the pressure sensor with different channel widths

and spacings but with the same channel height and

around same sensor thickness. It can be seen clearly

that sensor with channel dimension of 200μm width

and 300 μm spacing has the highest gradient of the

input and output curve, and it is the most sensitive one.

Even though the sensor with small channel width

and space has better resistance response. We still did

not fabricate any sensor with channel width or space

less than 200 μm. The main reason is the low yield

when fabricating sensors with channel dimension less

than 200 μm. The smallest sensor dimension we made

in this experiment is 200 μm channel width and

300μm spacing. Since we are looking for a sensor

with better sensitivity, we decided to focus on the

fabrication of sensor with channel width and spacing

of 200 and 300 μm respectively.

Figure 9: Electrical resistance of pressure sensor with

different channel width and spacing ranging from 500/500

µm to 200/300 µm but with same channel height 60 μm and

about same sensor thickness.

We have also carried out the fabrication process

for sensor with different channel height but with same

channel pattern (same channel width and spacing) and

same sensor thickness. The experiment results shows

that sensor with low channel height has better

response under pressure, which is also reasonable.

Currently, the yield of the fabrication process is

good, but still the process is challenging. The most

challenging step in the fabrication process is the

bonding of Ecoflex to the PET film. As we know, in

microfabrication the bonding process is the most

challenging step in all processes since it is greatly

related to the surface condition or surface state of the

two bonding parts. The surface condition is difficult

to control, and hence the bonding process is difficult

to control. The bonding strength of PDMS to PET

film is better than that of Ecoflex to PET film. During

the experiments, two things need to be prioritized.

The first one is to optimize the bonding process, how

to make it stable, and with high yield, while the

second one is to design for manufacturability, i.e.,

how to design the sensor structure or pattern for a

better yield and a sensor with high stability in usage.

Among the different designs of the sensor, based

on the testing results, it is concluded that to fabricate

a microfluidic sensor with high sensitivity, the

channel width, the channel spacing, the channel

height, the sensor height, should be small, which is

limited by the current microfabrication process. In

our case, the microfluidic pressure sensor with

physical design of channel width 200 um, channel

spacing 300 um, channel height about 60 um, and

sensor height 1mm has the best sensitivity.

3.2 Physical Design and Mechanical

Structure of the Wearable TCM

Pulse Reader

In this wearable TCM pulse reader, a microfluidic

force sensor is fixed in a housing. When put on and

properly located on the wrist, the pulsation of radial

artery will be transferred through a mechanism to the

sensor, causing the variation of resistance of the

sensor. The variation of the resistance would be

converted into varying voltage signals, which would

be picked up by a control board connected to it. Signal

would be processed and analysed by an intelligent

system to diagnose or predict certain diseases from it.

Hence, proper mechanism, structure, and housing is

one of the crucial parts of the pulse reader.

Two signal wires connect the sensor to the control

board, with one end connected to the sensor, and the

other end to a snap button. This snap button in turn

connects to another half of the snap button, which is

connected to the control board with another piece of

wire. The signal wires are wrapped in plastic film for

protective and easy handling purposes.

3.3 Pulse Signal Collection, Processing

Electronics and Firmware

A block diagram for the pulse data acquisition, signal

processing, and machine learning is shown in Figure

10. The microcontroller system includes an anti-

aliasing filter before data sampling and some

instrumentation front-end electronics before the

analog-to-digital conversion (ADC). Our sensors

have very low impedance, from a fraction of an ohm

in the unobstructed state to a few tens of ohms under

pressure. The pulses on the wrist only change the

resistance by a few milliohms. A constant excitation

012345

400/400, T=1.372mm

200/300, T=1.372mm

500/500, T=1.398mm

500/500, T=1.411mm

300/300, T=1.509mm

Resistance, Ohm

Force, N

A Novel Microﬂuidic Pressure Sensor for Traditional Chinese Medicine (TCM) Pulse Data Collection and Analysis

139

current to the sensor, a precision reference resistor,

differential inputs to ADC with differential and

common-mode filtering may be necessary. User’s

breathing motion and the air pressure changes will

cause some signal drift (baseline wander). The

removal of such drift, together with other noise if

existing, will be done in the firmware, possibly using

some wavelet decomposition algorithms. The clean

pulse signals will then be used in the following

processes:

a) The measured air pressure value will be

used to control the air-pump and air-release valve to

adjust the pressure on each sensor for pulse taking.

b) Pulse waveform can be displayed on the

device monitor. Additionally, the signals can be

transmitted via Bluetooth Low Energy (BLE) to a

mobile phone and shown on it. The mobile phone can

further connect to the Cloud through Wi-Fi or GSM

for further signal analysis.

c) The 17 pulse features (or potentially just the

key required features for heart disease diagnosis in

the initial prototypes) will be extracted, and features

pertaining to heart diseases will be selected for

diagnosis classification using artificial intelligence

technology.

Figure 10: A block diagram of the pulse signal acquisition,

signal processing, and classification.

3.4 Preliminary Results on Pulse Signal

Acquisition

A preliminary prototype built using an ESP32-PICO-

D4 microcontroller with a 24-bit ADS1220 ADC chip

connected to our in-house fabricated microfluidic

sensor has shown some promising results. Figure 11

shows the pulse signals collected from one subject in

one measurement session, clearly showing different

pulse patterns at different depths/pressures. The air

pressure was not controlled and maintained in this

test, and the signal drift was removed with an

algorithm in the firmware. The signals were not too

noisy at 20 Hz sampling rate. The sensor resistance

varies by 0.3 to 1 milliohm with the arterial pulse.

Figure 11: Pulse signal acquisition and data processing

electronics device and human pulse signal collection.

4 CONCLUSIONS

A novel microfluidic pressure sensor with good

performance which can sense the human pulse signal

has been developed using microfabrication process.

The effects of different sensor design patterns and

different sensor parameters such as Ecoflex thickness,

channel height, channel width and spacing on its

output performance have also been investigated. The

following parameters: Ecoflex thickness less than

1mm, 200 µm channel width, 300 µm channel

spacing and 60 µm channel height is an optimized

sensor parameter which can produce sensor with a

good sensitivity and high yield in fabrication.

A wearable TCM pulse analyzer based on the

microfluidic pressure sensor, which can accurately

sense and collect wrist arterial pulse signals has also

been fabricated. The mechanical structure of the pulse

analyser, the signal processing electronics both on

hardware and software have also been developed,

which can be applied in future diagnosis of diseases

according to TCM theory.

In future work, we will also develop pulse pattern

feature extraction, selection, and classification

models with a premier focus on cardiovascular

disease diagnosis based on TCM theory. We will use

techniques such as time-domain feature extraction,

wavelet decomposition, and ensemble-learning

method combining deep Convolutional Neural

Network (CNN) and Fuzzy Neural Network (FNN) to

enhance the predictive performance.

ACKNOWLEDGEMENTS

We thank Singapore Ministry of Education (MOE)

Translation and Innovation Fund (TIF) to support this

project. We also acknowledge Prof Lim Chwee Teck,

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

140

Dr. Yeo Joo Chuan and their research team from

National University of Singapore (NUS) for all

helpful discussions and valuable suggestions.

REFERENCES

Meng, J., Zhou, Z. (2001). Introduction to Chinese

medicine, Jyin Publishing Company, Taipei.

Wang, L. (2000). Diagnostics of traditional Chinese

medicine, Publishing House of Shanghai University of

Traditional Chinese Medicine, Shanghai.

Malinauskas, K., Palevicius, P., Ragulskis, M.,

Ostasevicius, V., Dauksevicius, R. (2013). Validation

of non-invasive MOEMS-assisted measurement system

based on CCD sensor for radial pulse analysis. Sensors

13 (4); 5368–5380.

Jeon, Y., Kim, J., Lee, H. et al. (2011). A clinical study of

the pulse wave characteristics at the three pulse

diagnosis positions of Chon, Gwan and Cheok.

Evidence-Based Complementary and Alternative

Medicine, vol. 2011, Article ID 904056.

Shu, J. (2007). Developing classification indices for

Chinese pulse diagnosis, Complement. Ther. Med. 15

(3) 190–198.

Malinauskas, K., Ostaševičius, V., Daukševičius, R.,

Jūrėnas V. (2012). Non-invasive micro-opto-electro-

mechanical system adaptation to radial blood flow

pulse and velocity analysis. Mechanics 18 (4) 479–483.

Zhang, J., Wang, R., Lu, S., Gong, J., Zhao, Z., Chen, H.,

Cui, L., Wang, N., Yu, Y. (2011). EasiCPRS: Design

and implementation of a portable Chinese pulse-wave

retrieval system. Proceedings of the 9th ACM

Conference on Embedded Networked Sensor Systems,

149–161.

Luo, C., Chung, Y., Hu, C., et al. (2012). Possibility of

quantifying TCM finger-reading sensations: I. Bi-

sensing pulse diagnosis instrument. Eur J Integr Med;

4: e255–62.

Hu, C., Chung, Y., Yeh, C., Luo, C. (2012) Temporal and

spatial properties of arterial pulsation measurement

using pressure sensor array. Evidence Based

Complement Alternat Med; Vol. 2012, Article ID

745127.

Wei, C., Huang, C., Liao, T. (2009). The exponential decay

characteristic of the spectral distribution of blood

pressure wave in radial artery. Computers in Biology

and Medicine, 39:453–459.

Wang, Y., Chang, C., Chen, J., Hsiu, H., Wang, W. (1997).

Pressure wave propagation in arteries. IEEE Eng Med

Biol Mag, 16:51–56.

Y. Liao, H. Chen, C. Huang, M. Ho et al. (2012). The pulse

spectrum analysis at three stages of pregnancy. J

Alternat Complement Med, 18:382–386.

Huang, Y., Chang, Y., Cheng, S., et al. (2019). Applying

pulse spectrum analysis to facilitate the diagnosis

of coronary artery disease, Evidence-Based

Complementary and Alternative Medicine, vol. 2019,

Article ID 2709486.

Hao, Y., Cheng, F., Pham, M., et al. (2019). A non-invasive,

Economical, and instant-result method to diagnose and

monitor type 2 diabetes using pulse wave: case-control

study. JMIR Mhealth Uhealth, vol. 7, iss. 4, e 11959.

Zhang, Z., Zhang, Y., Yao, L., et al. (2018). A sensor-based

wrist pulse signal processing and lung cancer

recognition. J. Biomed Inform, 79, 107-116.

Liou, J., Huang, C., Chiu, C., et al. (2011). Differences in

pulse spectrum analysis between atopic dermatitis and

nonatopic health children, The Journal of Alternative and

Complementary Medicine, vol 17, no. 4, pp. 325-328.

Wang, N., Yu, Y., Huang, D. (2015). Pulse diagnosis

signals analysis of fatty liver disease and cirrhosis

patients by using machine learning. The Scientific

World Journal, Volume 2015, Article ID 859192.

Huang, C., Chang, H., Kao, S., et al. (2011). Application of

sphygmography to detection of dyspepsia and the

rhinitis. The American Journal of Chinese Medicine,

vol. 39, no. 2, pp. 271–285.

Wang, D., Krusienski, D., Hao, Z. (2016). A Flexible PET-

based wearable sensor for arterial pulse waveform

measurement. Proceedings of the 9th International

Joint Conference on Biomedical Engineering Systems

and Technologies (BIOSTEC 2016), Volume 1:

BIODEVICES, pages 66-75.

Goyal, K., Agarwal, R. (2017). Pulse based sensor design

for wrist pulse signal analysis and health diagnosis.

Biomedical Research; 28 (12): 5187-5195.

A Novel Microﬂuidic Pressure Sensor for Traditional Chinese Medicine (TCM) Pulse Data Collection and Analysis

141