A Novel Multi-Modal Sensing System Prototype for Cardiovascular and

Cardiopulmonary Monitoring

Yusuf Ziya Hayirlioglu and Beren Semiz

Department of Electrical and Electronics Engineering, Koc University, Istanbul, Turkey

Keywords:

Wearable Devices, Health Monitoring, Seismocardiogram, Photoplethysmogram, Electrocardiogram.

Abstract:

Cardiopulmonary disease treatments can highly beneﬁt from remote monitoring systems, allowing for early

diagnosis and enabling personalized treatment programs. In this paper, the feasibility and performance of

such a system is demonstrated. Continuous and simultaneous monitoring of electrocardiogram (ECG), seis-

mocardiogram (SCG), photoplethysmogram (PPG), and body temperature signals from a total of six sensors

is achieved by a microcontroller-based setup, which consists of a ﬁxed main body mounted on mid-sternum

and a mobile daughter body mounted on the wrist. The data is stored in an SD card and transmitted by a

Bluetooth to PC in real-time, allowing easy data access. The proposed system’s performance is examined in

comparison to the heart rate (HR), heart rate variability (HRV), and respiration rate metrics derived from the

BIOPAC system’s ECG and respiration data. Low margins of error in all test cases show that the system works

at high performance.

1 INTRODUCTION

According to the 2020 report of the World Health

Organization, cardiopulmonary diseases are among

the leading causes of death, constituting a signiﬁ-

cant fraction of the total deaths worldwide (∼30%)

(WHO, 2020). In addition, Hospital Readmissions

Reduction Program (2010), which was passed un-

der the Affordable Care Act, reported heart failure,

acute myocardial infarction and pneumonia as the top

three health problems that should be focused on to re-

duce re-hospitalization rates (Zuckerman et al., 2016).

Considering these facts, remote and continuous mon-

itoring systems could potentially allow early diag-

nosis and intervention, enable the generation of per-

sonalized treatment programs, and provide proactive

and preventive treatment protocols for individuals at

risk of cardiovascular and cardiopulmonary diseases

(Sana et al., 2020).

The main methods used in the clinic for diagno-

sis and follow-up can be listed as assessing (i) the

vital signs (blood pressure, heart rhythm, respira-

tory rhythm, body temperature), (ii) chest and lung

sounds, (iii) heart sounds, and (iv) vascular health

(Gupta and Shea, 2021). Hence, there is a com-

pelling need for novel sensing modalities and anal-

https://orcid.org/0000-0002-7544-5974

ysis pipelines to digitize these examination steps and

to achieve continuous health monitoring regardless of

time and environmental stressors.

Since the electrical, mechanical, acoustic and op-

tical signals recorded from the human body have

physiological origins, they can provide clinically use-

ful information about the underlying anatomical and

physiological conditions. Among the aforementioned

signals, the most commonly used ones are the elec-

trocardiogram (ECG), seismocardiogram (SCG) and

photoplethysmogram (PPG) waveforms. While the

ECG measures the electrical activity of the heart,

the SCG assesses the mechanical activity originat-

ing from the contraction of the heart and ejection of

blood from the ventricles (Inan et al., 2014). Stud-

ies have shown that the SCG signal can be lever-

aged in estimating systolic time intervals (Shandhi

et al., 2019), assessment of myocardial contraction

(Tavakolian et al., 2012), heart failure classiﬁcation

(Inan et al., 2018), studying respiration rate and

phases (Pandia et al., 2012), and derivation of hemo-

dynamic parameters such as stroke volume (Semiz

et al., 2020). On the other hand, the PPG is gener-

ated as a result of the variance in the amount of light

absorbed in the arteries due to the changes in arte-

rial blood volume during the cardiac cycle. In the lit-

erature, it has been shown that analysis of the PPG

signal can provide important information regarding

126

Hayirlioglu, Y. and Semiz, B.

A Novel Multi-Modal Sensing System Prototype for Cardiovascular and Cardiopulmonary Monitoring.

DOI: 10.5220/0011663700003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 126-134

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)

Figure 1: (a) The main body of the proposed wearable device was designed in such a way that ECG, SCG, PPG and body

temperature measurements could be taken continuously. (b-e) The removable daughter body was planned to be designed to

allow additional SCG and PPG measurements to assist in the analysis.

blood oxygen level, blood pressure and vascular re-

sistance assessment (cheol Jeong et al., 2018).

Although there are several wearable system pro-

totypes utilizing ECG, SCG and PPG signals sepa-

rately in the literature (Fattah et al., 2017; Li et al.,

2017; Di Rienzo et al., 2011; Hernandez et al., 2014;

Da He et al., 2011), they have generally been lim-

ited to focusing on heart rhythm assessment. Thus,

none of them allows comprehensive cardiovascular

and cardiopulmonary monitoring as the aforemen-

tioned examination steps could not be digitized alto-

gether. On the other hand, systems allowing multi-

ple measurements only enable the recording of sig-

nals from the location where the device is attached,

thus cannot provide any mobility function (Etemadi

et al., 2015). Since such a system is also expected to

evaluate lung health, respiratory and blood pressure

parameters, having an adaptable and mobile system is

crucial. In addition to all these, it is necessary to de-

velop a system that allows real-time data transfer to

enable timely intervention.

In this paper, for the ﬁrst time to the best of our

knowledge, an adaptable wearable patch prototype

is designed to enable (i) comprehensive cardiovascu-

lar and cardiopulmonary monitoring through simul-

taneous acquisition of various physiological signals

(ECG, SCG, PPG and body temperature). The pro-

posed design includes a main body to be mounted

on the mid-sternum and a removable daughter body

to allow additional SCG and PPG measurements to

assist in comprehensive cardiopulmonary and cardio-

vascular assessment (Fig. 1). The proposed pro-

totype (ii) supports both real-time data transfer and

SDcard recording, is (iii) developable and adaptable.

Once improved and converted into printed-circuit-

board (PCB) form, the system will be (iv) mobile,

thus will be convenient for the patients and healthcare

professionals for use in various environments, both

inside and outside clinic.

2 METHODS

2.1 Hardware

As previously mentioned, the long-term aim is to

develop a PCB-based multi-modal wearable system,

which would consist of the following two parts:

• MAIN BODY to be mounted on the chest, which

houses the sensors for measuring the body tem-

perature, ECG, proximal PPG and SCG signals,

the microcontroller to which all sensors will be

connected, the battery of the system, and the Blue-

tooth and microSD card modules (Fig. 1(a)).

• DAUGHTER BODY to be mounted on distal lo-

cations, which houses the sensors that can be at-

tached or detached from the main body and allows

recording of distal SCG and PPG signals (Fig.

1(b-e)).

Based on this design, in this paper, a development-kit-

based system prototype is presented (Fig. 2(a)). First,

the sensors in the market and literature were investi-

gated and the most suitable ones were determined.

2.1.1 Seismocardiogram (SCG) and

Photoplethysmogram (PPG)

In the literature, analysis of the SCG signal has gener-

ally been performed at <100 Hz frequency values and

it has been shown that having a sample rate around

500 Hz is sufﬁcient (Semiz et al., 2020). Although

analog accelerometers have higher bandwidth com-

pared to digital accelerometers, the main aim was to

select an accelerometer having low noise, high reso-

lution, and sensitivity to allow measurement of peak-

to-peak amplitude values within the limits of <10 mg.

Hence, it was decided to use the ADXL355 (Analog

Devices, Norwood, MA, USA) accelerometer with

25 µg/

√

Hz noise ﬂoor and 0.003 mV/bit resolution,

A Novel Multi-Modal Sensing System Prototype for Cardiovascular and Cardiopulmonary Monitoring

127

Figure 2: (a) General view of the system, (b) Temperature, MAX30102 and ADXL355 sensors to be mounted on the chest

area, (c) DS3231 RTC module, (d) HW-125 microSD card module, (e) MAX30102 and ADXL355 sensors to be mounted on

the wrist area, (f) AD8232 analog front-end IC, (g) HC-05 Bluetooth module.

which has been proven successful in previous studies

(Ganti et al., 2020). In accordance with the design,

two ADXL355 accelerometers were included in the

system, one for each of the main (chest) and daughter

bodies (Fig. 2(b) and (e)).

For the PPG signal, ﬁrst the systems available in

the market were examined. Smart watches on the

market mostly rely on the use of green light emitting

diode (LED) as green LED has relatively higher sig-

nal quality and is less susceptible to motion artifacts

compared to red and infrared lights. However, green

LED-based systems could only allow the measure-

ment of the microvascular network on the skin sur-

face due to green light’s relatively lower wavelength.

On the contrary, red and infrared lights have longer

wavelengths, thus can penetrate into the depths where

the arteries are located in the tissue (Lee et al., 2013;

Maeda et al., 2011). Hence, in the design of the PPG

system, the use of red and infrared LEDs was pre-

ferred. To this end, a digital MAX30102 (Maxim

Integrated, Sunnyvale, CA, USA) sensor with a res-

olution of 0.05 mV/bit and a sampling rate of 3200

samples/second has been deemed appropriate. Two

MAX30102 sensors were included in the system,

one for the main (chest) and one for the daugh-

ter body (Fig. 2(b) and (e)). It should be noted

that MAX30102 only has one I2C address, therefore,

an I2C multiplexer, TCA9548A (Texas Instruments,

Dallas, TX, USA), was used to prevent any commu-

nication conﬂict when utilizing two of them.

While the ADXL355 and MAX30102 sensors

both support I2C communication, the maximum

clock frequency ADXL355 supports is 3.4 MHz, and

the maximum clock frequency MAX30102 supports

is 400 kHz. Therefore, the maximum clock frequency

of the I2C protocol would be 400 kHz when the

same signal bus is used. 400 kHz was deemed to

be sufﬁcient considering the requirements of the pro-

totype. Additionally, the ADXL355 is a 3 V sensor

with no on-board regulators and the current prototype

runs with 5 V. To prevent any damage to the sen-

sor, we used logic level enhancement mode ﬁeld ef-

fect transistors, BSS138 (ON Semiconductor Corpo-

ration, Phoenix, Arizona, USA), in the I2C bus. Both

ADXL355 and MAX30102 packages include built-in

FIFOs, allowing temporary storage of data.

2.1.2 Electrocardiogram (ECG) and Body

Temperature

To record the ECG signal, AD8232 (Analog Devices,

Norwood, MA, USA) analog front-end integrated cir-

cuit (IC) with a noise ﬂoor of 100 nV/

√

Hz, gain of

100, and gain-bandwidth product of 100 kHz was se-

lected (Fig. 2(f)). Three gel electrodes were mounted

in the Einthoven’s triangle arrangement and the sig-

nals obtained through these electrodes were condi-

tioned by AD8232. The resulting signal was then

sampled at 10 bit resolution and 500 Hz by AT-

MEGA2560’s analog-to-digital converter (ADC).

For body temperature measurement, LM35 (Texas

Instruments, Dallas, TX, USA) analog IC with

±0.5

◦

C accuracy, 10 mV/

◦

C linear scale factor, -55

◦

- 150

◦

C measurement range and 600 nV/

√

Hz noise

ﬂoor was chosen and integrated into the part to be

mounted on the chest (Fig. 2(b)). The signal was

sampled by the ATMEGA2560’s ADC at 10-bit reso-

lution and 1 Hz.

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

128

2.1.3 Microcontroller

The features considered in the selection of the mi-

crocontroller were as follows: having enough pro-

cessing power to collect data at the desired sampling

rates, working with as low power as possible, hav-

ing a high storage area and EEPROM, and having a

sufﬁcient number of analog and digital pins. With

256 KByte memory, 8 KByte EEPROM, 16 analog

pins, and 86 programmable input/output lines, AT-

MEGA2560 was deemed sufﬁcient for the prototype.

Powered at 5 V in active mode with an 11.0592 MHz

crystal oscillator, it has a current consumption of 14

mA, and a current consumption of 0.1 uA in power-

off mode. Having four 16-bit timers and six hardware

interrupts also had a positive effect on the microcon-

troller choice.

2.1.4 Data Storage and Transfer

To achieve real-time data transfer, Bluetooth func-

tionality was included in the prototype. In addition

to Bluetooth connection, data is simultaneously saved

on the microSD card to act as a back-up. For the SD

card a high write speed is desired for faster opera-

tion. The SD card of choice was a 16GB SanDisk

Ultra with an 80 MB/s write speed. The SD card was

integrated into the system through HW-125 microSD

card module (Fig. 2(d)) and ﬁle naming was provided

by the date/time stamps taken from the DS3231 RTC

module (Fig. 2(c)).

For Bluetooth connection, HC-05, which has a

maximum baud rate of 1328400 and can work as both

slave and master, was selected (Fig. 2(g)). In the

system, 24 KBytes of data was produced per second,

while the Bluetooth could reach a data transmission

rate of 132 KBytes per second. While searching for a

device to be connected, the query time interval, query

time, paging time interval and paging time were set as

640 ms, 0.625 ms, 640 ms, and 0.625 ms, respectively

to ensure that the Bluetooth consumes less power than

it does with its default parameter values. When con-

nected to a device, the Bluetooth was put into sniff

mode, so that it consumes less power while waiting

for a command during communication. The maxi-

mum time, minimum time, test time and waiting time

for the sniff mode were selected as 125 ms, 6.25 ms,

1.25 ms, and 5 s, respectively.

It should be noted that when the data is being

stored and transferred, all data points are processed

simultaneously due to data structure to be used. Al-

though not all sensors are sampled at 500 Hz, all data

is stored and transferred at the highest sampling rate.

Therefore signals with sampling rates smaller than

500 Hz will have repeating samples in them.

2.2 Firmware

First, the time required to read data from the sen-

sors, save it on the SD card, and send it via Blue-

tooth while the system reaches the desired sampling

rates was investigated. I2C communication was es-

tablished at the clock frequency of 400 kHz, due to

MAX30102’s speed constraint. With this communi-

cation speed, samples could be read from both in-

frared and red LEDs of the MAX30102 at approxi-

mately 0.9 ms, and the X, Y and Z axes of ADXL355

at 0.6 ms. And for the analog data, it took around 0.15

ms to read one sample. Overall, two MAX30102s,

two ADXL355s, one ECG and one temperature sen-

sor were included in the system. The ADXL355 and

MAX30102 collect data at 500 Hz and 200 Hz sam-

ple rates respectively. The data MAX30102 collects is

subjected to an on-chip sample averaging operation,

leading to an effective sampling rate of 50 Hz. The

ECG is sampled at 500 Hz and the temperature sen-

sor is sampled at 1 Hz.

Based on these speciﬁcations, reading data from

the accelerometers (0.6 ms each) and the ECG sen-

sor (0.15 ms) were taking 1.35 ms of a 2 ms cycle.

When various other operations were added into the

system (e.g., writing data to SD card, sending data

over Bluetooth, reading data from MAX30102), it

was observed that data could not be timely obtained

from the sensors which required to be sampled at 500

Hz, resulting in data skipping when the FIFOs over-

ﬂow. To overcome this problem, an algorithm was

developed. In this algorithm, the time spent by AT-

MEGA2560 in the statements was calculated. The

primary timer-controlled statement, which executes

most frequently, needs to be completed in less than

2 ms due to 500 Hz sampling rate requirement. If the

microcontroller goes over 2 ms in the statement, the

extra time spent is stored in a variable and is incre-

mented with each occurrence of such an event. In the

case when the extra time it calculates goes over 2 ms,

the microcontroller is forced to read from FIFOs of

the sensors and the virtual FIFO created for the ECG.

With this algorithm, it was ensured that the microcon-

troller will collect all available data from the sensors

without missing any samples.

The reasoning behind using a virtual FIFO for the

ECG is because the ECG samples were not being col-

lected at constant intervals due to variable statement

execution times, causing both data skipping and data

shifting. By using the FIFOs of the MAX30102 and

ADXL355, the data could be collected at the desired

frequencies without any data shift and with the help of

the aforementioned algorithm, data skipping problem

was also solved. However, the analog ECG sensor,

A Novel Multi-Modal Sensing System Prototype for Cardiovascular and Cardiopulmonary Monitoring

129

which lacks any kind of storage element, was experi-

encing data skipping and small data shifts. As a so-

lution, a virtual FIFO and timer interrupts were used.

Whenever data needs to be read from the ECG sensor,

the ATMEGA2560 is interrupted by the timer and a

data point is sampled from the ECG sensor. This en-

sures that the ECG is sampled at constant intervals.

The sample is then stored in the virtual buffer cre-

ated for the ECG until it is written to the SD card and

transmitted by the Bluetooth.

Four different modes (Start Recording, Pause

Recording, Stop Recording and Low Power ) have

been added to the system via buttons using hardware

interrupts to control recording capability and power

consumption. In Start Recording mode, the system

starts reading data by creating a new ﬁle on the SD-

card and writes the data it reads to this ﬁle. The

name of the ﬁle is determined by the timestamp cap-

tured by the DS3231 RTC module. Pause Recording

mode temporarily stops recording while recording is

in progress, but does not close the ﬁle. To continue

recording, Pause Recording button should be pressed

again. In Stop Recording mode, the created ﬁle is

closed and recording is stopped. Data written to the

SDcard is simultaneously sent to Bluetooth via the se-

rial port and to PC via Bluetooth. To test the Blue-

tooth connection with a PC, PuTTY was used. The

data that PuTTY writes to its console is saved in a

log ﬁle. The last mode, Low Power mode turns off

the sensors and communication channels busses on

the system, and puts the microcontroller in low power

mode, minimizing power consumption.

2.3 Data Collection Protocol

This study was conducted under a protocol approved

by the Koc University Institutional Review Board and

all subjects provided written consent. The system was

tested on two subjects (one male and one female) with

no history of cardiovascular or cardiorespiratory dis-

eases. The main body of the system (Fig. 2(b)) in-

cluding a tri-axial accelerometer, PPG module and

temperature sensor was mounted on the mid-sternum

of the subject using hypoallergenic transparent medi-

cal tape. Three ECG cables were attached to the chest

through gel electrodes in accordance with Einthoven’s

triangle. Additionally, the daughter body (Fig. 2(e))

including a tri-axial accelerometer and PPG module

was placed on the wrist with the same medical tape.

It should be noted that the accurate direction of the

accelerometers and skin-contact of PPG sensors were

ensured before starting data collection.

Along with the signals collected through the pro-

totype, reference respiration and ECG signals were

recorded simultaneously with the BIOPAC system

(BIOPAC Systems, Inc. Goleta, CA, USA). Refer-

ence respiration signal was measured using respira-

tion effort transducer and reference ECG signal was

acquired through three gel electrodes. The signals

were transferred to the BIOPAC system using wire-

less Bionomadix RSPEC-R module (BIOPAC Sys-

tems, Inc. Goleta, CA, USA). All reference signals

were sampled at 500 Hz.

As the main goal was to validate the system, the

protocol did not include any physiological modula-

tion. Instead, the subjects stood motionless vertically

for ﬁve minutes. The signals were post-processed us-

ing MATLAB (MathWorks, Natick, MA, USA).

2.4 Data Analysis

In this section, physiological parameters derived from

the reference signals and the ones acquired with the

prototype were compared to assess the performance

of the prototype. First, the signals were ﬁltered us-

ing digital ﬁnite impulse response (FIR) band-pass

ﬁlters in accordance with the bandwidths reported in

the literature (Carek et al., 2017; Pandia et al., 2012;

Shandhi et al., 2019). For ECG and PPG, frequency

ranges were determined as 0.5 - 40 Hz and 1 - 16

Hz, respectively. For the SCG signals, two different

FIR ﬁlters were implemented. The one having 0 - 1

Hz range, which corresponds to the chest movements

originating from exhalation and inhalation, was ap-

plied to derive the respiratory information. On the

other hand, the one having 1- 40 Hz bandwidth was

used to represent the vibration and acoustic infor-

mation originating from the cardiac output and heart

sounds.

Figure 3: PPG signals before and after Gaussian window-

ing.

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

130

Figure 4: 5-second-long, pre-processed segments of the signals acquired with the system.

2.4.1 Derivation of Heart Rate and Heart Rate

Variability

First, performance of the prototype in estimating heart

rate (HR) and heart rate variability (HRV) values

through the acquired ECG and PPG signals was as-

sessed. It has been observed that the PPG signals

taken from the chest in particular contain a high level

of noise, therefore it was more difﬁcult to determine

the peak and dicrotic points compared to the PPG

taken from the wrist. This was an expected result,

as it was more difﬁcult to reach the microvascular re-

gion in the chest area compared to the wrist. To mit-

igate the negative impact of signal quality, Gaussian

smoothing (window length: 100, width factor: 2.75)

was applied to smooth out these noise-based oscilla-

tions observed in the signals. 5-second-long segments

from the PPG signals before and after Gaussian win-

dowing were presented in Fig. 3. HR and HRV were

then calculated using the following steps:

1. First, reference HR and HRV values were cal-

culated using the reference ECG signal recorded

with the BIOPAC system. To this end, R-points on

the ECG were determined and an RR vector was

created by calculating intervals between succes-

sive R-points. HR in beats per minute (bpm) and

HRV in milliseconds were calculated on this vec-

tor using Eq. 1 and Eq. 2, respectively (std: stan-

dard deviation, F

: sampling rate). The standard

deviation of inter-beat-interval (SDNN) is mea-

sured in ms. The short-term recording for SDNN

is 5 minutes (Shaffer and Ginsberg, 2017).

2. After computing the reference values, HR and

HRV were derived from the ECG signal acquired

through the prototype using the same formulas.

3. In the last step, PPG signals acquired through the

prototype from the wrist and chest were used to

derive the HR and HRV values. By calculating

the time interval between consecutive peaks on

the Gaussian-windowed red and infrared PPG sig-

nals, a PP vector similar to the RR vector was gen-

erated. HR and HRV values were calculated using

the same formulas, but inserting the PP vector in-

stead of the RR vector.

4. To calculate the percent error (%error), Eq. 3

A Novel Multi-Modal Sensing System Prototype for Cardiovascular and Cardiopulmonary Monitoring

131

Table 1: Heart rate and heart rate variability calculations (bpm: beats per minute, ms: milliseconds).

First Subject

Heart Rate (bpm) Error (%) Heart Rate Variability (ms) Error (%)

BIOPAC ECG (Reference) 65.2 - 15.2 -

Wearable ECG 65.9 1.07 15.9 4.61

Wearable PPG Infrared (Wrist) 65.6 0.61 15.3 0.66

Wearable PPG Red (Wrist) 65.6 0.61 15.3 0.66

Wearable PPG Infrared (Chest) 66.3 1.68 15.5 1.97

Wearable PPG Red (Chest) 67.2 3.07 15.7 3.29

Second Subject

BIOPAC ECG (Reference) 88.8 - 15.0 -

Wearable ECG 90.2 1.58 17.2 14.6

Wearable PPG Infrared (Wrist) 90.2 1.58 17.2 14.6

Wearable PPG Red (Wrist) 89.8 1.12 17.2 14.6

Wearable PPG Infrared (Chest) 88.9 0.11 17.2 14.6

Wearable PPG Red (Chest) 91.9 3.49 17.4 16.0

was used where actual stands for the reference

HR and HRV values and calculated stands for the

ones computed using the ECG, PPG-red and PPG-

infrared signals acquired from the prototype.

HR =

60 ∗F

mean(RR))

(1)

HRV =

std(RR))

∗1000 (2)

%error =

actual −calculated

actual

∗100 (3)

2.4.2 Derivation of Respiration Rate

SCG signal, which was ﬁltered in 0 - 1 Hz range, was

used to extract respiratory information. The analy-

sis was speciﬁcally performed on the dorso-ventral

axis in accordance with the literature (Pandia et al.,

2012). Subsequently, Gaussian windowing was ap-

plied as previously done in PPG case to smooth out

the oscillations observed in the signal. Using the

peaks on these signals, reference and SCG-derived

respiration rates (number of exhalation-inhalations in

one minute) were calculated.

3 RESULTS AND DISCUSSION

3.1 System Performance

5-second-long, pre-processed segments of the signals

acquired with the system from one of the subjects

were presented in Fig. 4. As can be seen from the

plots, the sensors could successfully collect the phys-

iological signals at the desired sampling rates. In ad-

dition, the representative SCG signals written to SD

Figure 5: Comparison of representative SCG signals written

to SD card and transferred to computer via Bluetooth. There

was a correlation of 100% and no data loss was observed

during Bluetooth transmission.

card and transferred to computer via Bluetooth are

presented in Fig. 5. It should be noted that the sig-

nal received via Bluetooth has been shifted down in

the y-axis to facilitate visualization. As seen, there is

a 100% correlation between the signals and there is

no data loss during transmission via Bluetooth.

3.2 Comparison with Reference

Measurements

3.2.1 Heart Rate and Heart Rate Variability

First, the performance of the prototype in estimating

HR and HRV values through the acquired ECG and

PPG signals was assessed. As detailed in Section

2.4.1, HR and HRV were calculated using the time

intervals between the consecutive peak locations on

ECG and PPG signals. On the other hand, the ref-

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

132

Table 2: Breathing Rate Calculations ( breaths per minute).

First

Subject

Error

(%)

Second

Subject

Error

(%)

BIOPAC Belt

(Reference)

24.5 - 21.1 -

Wearable SCG-Z

(Chest, 0-1 Hz)

24.8 1.22 20.7 1.89

erence HR and HRV were calculated using the peak

locations on the ECG acquired with the BIOPAC sys-

tem. The resulting HR and HRV values for both sub-

jects are presented in Table 1. The low margin of error

between the reference and prototype-derived values

shows the high performance of the designed system.

Speciﬁcally, there is a maximum of 2-beat difference

between the reference HR values and the ones derived

from ECG and PPG signals acquired from chest and

wrist. Additionally, although the PPG signal quali-

ties obtained from the chest and wrist were differ-

ent (chest being more susceptible to artifacts), this

negative effect could be eliminated with ﬁltering and

Gaussian smoothing. Indeed, the HR values obtained

from the chest and wrist resulted in negligible errors,

both being similar to the reference HR measurements.

3.2.2 Respiration Rate

20-second-long segments from the reference and

SCG-derived respiration signals from one of the sub-

jects were presented in Fig. 6. As seen in the ﬁgure,

the respiratory signal generated from the SCG signal

moves in parallel with the reference respiratory sig-

nal. In addition, the number of breaths per minute

calculated on the reference and SCG signals for both

subjects is presented in Table 2. For the ﬁrst subject,

reference and SCG-based breathing rates were cal-

culated as 24.5 and 24.8 breaths per minute, respec-

tively. For the second subject, these values were 21.1

and 20.7, respectively. Obtaining almost the same

number of respiration cycles with the reference wave-

form in both subjects shows that the sensors in the

system and the signal improvement methods work at

high performance.

4 CONCLUSIONS

In this work, an adaptable wearable patch prototype

is designed to enable comprehensive cardiovascular

and cardiopulmonary monitoring through simultane-

ous acquisition of ECG, PPG, SCG, and body tem-

perature signals. The design includes a main body to

be mounted on the mid-sternum to collect proximal

ECG, SCG, PPG and body temperature signals, and

a removable daughter body to allow distal SCG and

Figure 6: Comparison of the reference and SCG-derived

respiratory signals.

PPG measurements to assist in comprehensive health

assessment. Respiration rate, HR and HRV values

from two subjects obtained from the prototype were

compared with the reference ones acquired with the

BIOPAC system. Obtaining a low margin of error in

all cases showed that the prototype and the signal im-

provement methods work at high performance. In ad-

dition, the signals transferred with the Bluetooth sys-

tem achieved 100% correlation with the signals saved

on the SDcard. Despite receiving data from so many

sensors, having no data loss was indeed a promising

result for the future versions of the current prototype.

Future work will focus on validating the system

with a larger dataset including subjects having vary-

ing physiological and demographic features. Addi-

tionally, since the calculation of oxygen saturation

and blood pressure values from the PPG signal pri-

marily requires a calibration step, studies in this area

will be carried out after collecting data from a large

number of participants. In the long run, the prototype

will be converted into the PCB form having actual

main and daughter bodies to allow cable-free mea-

surements. Such a system will potentially be con-

venient for the patients and healthcare professionals

to achieve continuous health monitoring regardless of

time and environmental stressors.

ACKNOWLEDGEMENTS

This work was supported by the Scientiﬁc and Tech-

nological Research Council of Turkey (TUBITAK)

under grant number 121E133.

A Novel Multi-Modal Sensing System Prototype for Cardiovascular and Cardiopulmonary Monitoring

133

REFERENCES

Carek, A. M., Conant, J., Joshi, A., Kang, H., and Inan,

O. T. (2017). Seismowatch: wearable cufﬂess blood

pressure monitoring using pulse transit time. Proceed-

ings of the ACM on interactive, mobile, wearable and

ubiquitous technologies, 1(3):1–16.

cheol Jeong, I., Bychkov, D., and Searson, P. C. (2018).

Wearable devices for precision medicine and health

state monitoring. IEEE Transactions on Biomedical

Engineering, 66(5):1242–1258.

Da He, D., Winokur, E. S., and Sodini, C. G. (2011). A con-

tinuous, wearable, and wireless heart monitor using

head ballistocardiogram (bcg) and head electrocardio-

gram (ecg). In 2011 Annual International Conference

of the IEEE engineering in medicine and biology so-

ciety, pages 4729–4732. IEEE.

Di Rienzo, M., Meriggi, P., Rizzo, F., Vaini, E., Faini,

A., Merati, G., Parati, G., and Castiglioni, P. (2011).

A wearable system for the seismocardiogram assess-

ment in daily life conditions. In 2011 Annual In-

ternational Conference of the IEEE Engineering in

Medicine and Biology Society, pages 4263–4266.

IEEE.

Etemadi, M., Inan, O. T., Heller, J. A., Hersek, S., Klein, L.,

and Roy, S. (2015). A wearable patch to enable long-

term monitoring of environmental, activity and hemo-

dynamics variables. IEEE transactions on biomedical

circuits and systems, 10(2):280–288.

Fattah, S. A., Rahman, M. M., Mustakin, N., Islam, M. T.,

Khan, A. I., and Shahnaz, C. (2017). Wrist-card: Ppg

sensor based wrist wearable unit for low cost person-

alized cardio healthcare system. In 2017 IEEE Global

Humanitarian Technology Conference (GHTC), pages

1–7. IEEE.

Ganti, V. G., Carek, A. M., Nevius, B. N., Heller, J. A.,

Etemadi, M., and Inan, O. T. (2020). Wearable cuff-

less blood pressure estimation at home via pulse tran-

sit time. IEEE Journal of Biomedical and Health In-

formatics, 25(6):1926–1937.

Gupta, J. and Shea, M. (2021). Cardiovascu-

lar examination. Merck Manual. https:

//www.merckmanuals.com/en-pr/professional/

cardiovascular-disorders/approach-to-the-cardiac-

patient/cardiovascular-examination(visited:2022-09).

Hernandez, J., Li, Y., Rehg, J. M., and Picard, R. W.

(2014). Bioglass: Physiological parameter estimation

using a head-mounted wearable device. In 2014 4th

International Conference on Wireless Mobile Com-

munication and Healthcare-Transforming Healthcare

Through Innovations in Mobile and Wireless Tech-

nologies (MOBIHEALTH), pages 55–58. IEEE.

Inan, O. T., Baran Pouyan, M., Javaid, A. Q., Dowling, S.,

Etemadi, M., Dorier, A., Heller, J. A., Bicen, A. O.,

Roy, S., De Marco, T., et al. (2018). Novel wearable

seismocardiography and machine learning algorithms

can assess clinical status of heart failure patients. Cir-

culation: Heart Failure, 11(1):e004313.

Inan, O. T., Migeotte, P.-F., Park, K.-S., Etemadi, M.,

Tavakolian, K., Casanella, R., Zanetti, J., Tank, J.,

Funtova, I., Prisk, G. K., et al. (2014). Ballistocardio-

graphy and seismocardiography: A review of recent

advances. IEEE journal of biomedical and health in-

formatics, 19(4):1414–1427.

Lee, J., Matsumura, K., Yamakoshi, K.-i., Rolfe, P., Tanaka,

S., and Yamakoshi, T. (2013). Comparison between

red, green and blue light reﬂection photoplethysmog-

raphy for heart rate monitoring during motion. In 2013

35th annual international conference of the IEEE en-

gineering in medicine and biology society (EMBC),

pages 1724–1727. IEEE.

Li, S.-H., Lin, B.-S., Wang, C.-A., Yang, C.-T., and

Lin, B.-S. (2017). Design of wearable and wire-

less multi-parameter monitoring system for evaluat-

ing cardiopulmonary function. Medical Engineering

& Physics, 47:144–150.

Maeda, Y., Sekine, M., and Tamura, T. (2011). Relation-

ship between measurement site and motion artifacts

in wearable reﬂected photoplethysmography. Journal

of medical systems, 35(5):969–976.

Pandia, K., Inan, O. T., Kovacs, G. T., and Giovangrandi, L.

(2012). Extracting respiratory information from seis-

mocardiogram signals acquired on the chest using a

miniature accelerometer. Physiological measurement,

33(10):1643.

Sana, F., Isselbacher, E. M., Singh, J. P., Heist, E. K.,

Pathik, B., and Armoundas, A. A. (2020). Wearable

devices for ambulatory cardiac monitoring: Jacc state-

of-the-art review. Journal of the American College of

Cardiology, 75(13):1582–1592.

Semiz, B., Carek, A. M., Johnson, J. C., Ahmad, S.,

Heller, J. A., Vicente, F. G., Caron, S., Hogue, C. W.,

Etemadi, M., and Inan, O. T. (2020). Non-invasive

wearable patch utilizing seismocardiography for peri-

operative use in surgical patients. IEEE Journal

of Biomedical and Health Informatics, 25(5):1572–

1582.

Shaffer, F. and Ginsberg, J. P. (2017). An overview of heart

rate variability metrics and norms. Frontiers in Public

Health, 5.

Shandhi, M. M. H., Semiz, B., Hersek, S., Goller, N.,

Ayazi, F., and Inan, O. T. (2019). Performance

analysis of gyroscope and accelerometer sensors for

seismocardiography-based wearable pre-ejection pe-

riod estimation. IEEE journal of biomedical and

health informatics, 23(6):2365–2374.

Tavakolian, K., Portacio, G., Tamddondoust, N. R., Jahns,

G., Ngai, B., Dumont, G. A., and Blaber, A. P. (2012).

Myocardial contractility: A seismocardiography ap-

proach. In 2012 Annual International Conference of

the IEEE Engineering in Medicine and Biology Soci-

ety, pages 3801–3804. IEEE.

WHO (2020). The top 10 causes of death. Geneva: World

Health Organization. https://www.who.int/news-

room/fact-sheets/detail/the-top-10-causes-of-death

(visited: 2022-09).

Zuckerman, R. B., Sheingold, S. H., Orav, E. J., Ruhter,

J., and Epstein, A. M. (2016). Readmissions, observa-

tion, and the hospital readmissions reduction program.

New England Journal of Medicine, 374(16):1543–

1551.

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

134