Android-based ECG Monitoring System for Atrial Fibrillation

Detection using a BITalino

ECG Sensor

Gabriel Saatkamp Lazaretti

1,2 a

, João Paulo Teixeira

, Eduardo Vinicius Kuhn

and Pedro Henrique Borghi

Research Centre in Digitalization and Intelligent Robotics (CEDRI), Instituto Politécnico de Bragança (IPB), Portugal

Federal University of Technology - Paraná (UTFPR), Toledo, Brazil

Faculty of Engineering of University of Porto (FEUP), Porto, Portugal

Keywords: Android, Atrial Fibrillation, BITalino, ECG, Smartphone.

Abstract: Cardiac arrhythmias are disorders that affect the rate and/or rhythm of the heartbeats. The diagnosis of most

arrhythmias is made through the analysis of the electrocardiogram (ECG), which consists of a graphical

representation of the electrical activity of the heart. Atrial fibrillation (AF) is the most present type of

arrhythmia in the world population. In this context, this work deals with the implementation of a system for

automatic analysis of ECG signals aiming to identify AF episodes. The system consists of a signal acquisition

step performed by an ECG sensor connected to an acquisition platform. The acquired signal is transmitted via

bluetooth to a smartphone with Android

™

operating system. The signal processing is carried out through an

application developed using the IDE Android

™

Studio. When assessed over signals from the MIT-BIH Atrial

Fibrillation database, the R-wave peak detection algorithm showed mean values of sensitivity and positive

predictivity of 98.99% and 95.95%, respectively. The classification model used is based on a long short-term

memory (LSTM) neural network and had an average accuracy of 94.94% for identifying AF episodes.

1 INTRODUCTION

Cardiovascular diseases (CVDs - cardiovascular

diseases) are one of the main causes of death around

the world. According to data from the world health

organization, it is estimated that about 17.9 million

people died from some type of CVD in 2019, which

represents 32% of the deaths in the world with 85%

of them being from stroke and myocardial infarction

(commonly called heart attack) (World Health

Organization, 2021). In this context, arrhythmias are

disorders that affect the frequency and/or rhythm of

heartbeats (Antzelevitch & Burashnikov, 2011).

Among the various types of arrhythmias, atrial

fibrillation (AF) is the most common. In the European

Union, it is estimated that the number of AF cases in

the adult population over 55 years of age reached 8.8

million in 2010 (Krijthe et al., 2013). Worldwide, this

number reached the 33.5 million mark in the same

https://orcid.org/0000-0003-2993-5398

https://orcid.org/0000-0002-6679-5702

https://orcid.org/0000-0003-0881-4888

https://orcid.org/0000-0003-2918-6630

year (Chugh et al., 2014). This predominant

arrhythmia, if not treated, increases the chances of

developing an eventual cardiac arrest (Wang et al.,

2003), dementia (Ott et al., 1997), as well as strokes

(Dulli et al., 2003; Jørgensen et al., 1996; H.-J. Lin et

al., 1996; Marini et al., 2005; Wolf et al., 1991, 1998).

So, since AF is associated with an increased risk of

mortality, it deserves medical attention (Miyasaka et

al., 2007; Stewart et al., 2002).

Developed by Willem Einthoven in 1902

(AlGhatrif & Lindsay, 2012), the electrocardiogram

(ECG), which is a graphical representation of the

heart’s electrical activity, made it possible to observe

variations in the frequency and rhythm of the

heartbeats; thus, the ECG provides a way to visualize

how the electrical system of the heart behaves. In this

context, the 12-lead system is the most widely used

for diagnosing cardiac arrhythmias (Mittal et al.,

2011). Figure 1 shows a typical waveform of a

Lazaretti, G., Teixeira, J., Kuhn, E. and Borghi, P.

Android-based ECG Monitoring System for Atrial Fibrillation Detection using a BITalino

 ECG Sensor.

DOI: 10.5220/0010905400003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 1: BIODEVICES, pages 177-184

ISBN: 978-989-758-552-4; ISSN: 2184-4305

177

complete cardiac cycle (one beat) in an ECG signal,

consisting of a sequence of waves and complexes

called P wave, QRS complex, T wave and U wave.

The AF occurs due to the generation of multiple

electrical impulses in different regions of the atria

(ectopic foci) which cause irregular muscle cell

stimulation, resulting in ineffective or even non-

existent atrial contraction (fibrillatory behavior). The

main symptoms that may indicate the presence of AF

in the patient are the feeling of weakness, chest pain,

shortness of breath, and palpitations. However, in

some cases, AF can occur without any evident

symptoms (asymptomatically). The main features

observed in an ECG signal for diagnosing AF are:

• Irregular heart rate, i.e., unexpected changes in

the R-R interval (termed RRi) over time; and

• Absence of a clearly distinguishable P wave

which is then replaced by fibrillatory waves (F

waves).

Figure 1: Points and intervals of interest in the electrical

representation of a heartbeat.

2 PROBLEM FORMULATION

The diagnosis of CVDs, as AF, is in general

performed through the analysis of the ECG signal by

a qualified health professional. Nevertheless, due to

the advances of medical technology, computer

systems for automatic ECG analysis have been able

to assist professionals in the diagnoses of pathologies.

The use of computer-aided diagnosis systems

(CADx) becomes especially interesting in the

analysis of ECG recordings with long hours of

duration (Hagiwara et al., 2018). [Such systems

consist in general of four main sequential steps,

namely: 1) Pre-processing; 2) Segmentation;

3) Feature extraction; and 4) Classification.] Note

that, with the development of the Holter monitor in

the 1940s, methods for ambulatory and external

monitoring of the ECG (AECG -Ambulatory External

Electrocardiogram), as described in (Mittal et al.,

2011), have emerged as an important tool to perform

AF diagnosis for the 12-lead ECG.

Considering the processing capacity, portability,

and affordable price of smartphones, researchers are

proposing their use as processing devices in

automatic ECG signal analysis systems for the

diagnosis of cardiac arrhythmias. Specifically for the

detection of AF events, the “ECG Check” device,

developed by Cardiac Designs

, and the “Kardia

Mobile”, developed by AliveCor

, are examples of

professional devices which use smartphones to detect

the presence of AF events based on a single lead of

the ECG signal. The effectiveness of these devices is

discussed in (Aljuaid et al., 2020; Chan & Choy,

2017; Evans et al., 2017; Garabelli et al., 2017; T.

Hickey et al., 2017).

In this context, the present research work has the

following objectives:

• To acquire a single lead ECG signal using

3-electrode configuration;

• To transmit the signal from the acquisition

device to the processing device using bluetooth

technology;

• To assess the received signal and implement, if

necessary, a pre-processing step to remove

artifacts from the acquired signal;

• To implement a segmentation algorithm for

detecting QRS complex and R-peaks in ECG

signals;

• To extract features from the ECG signal which

will serve as input to the machine learning

model used in the classification step; and

• To implement a classification step, based on a

Long Short-Term Memory (LSTM) neural

network capable of classifying segments of

ECG signal according to their rhythm (normal

rhythm, AF, or another rhythm).

3 ECG SIGNAL ACQUISITION

AND TRANSMISSION

The ECG acquisition step is concerned with the

proper placement of electrodes and the design of a

signal conditioning system used to amplify, filter, and

digitalize the signal. Particularly, the acquisition of

the ECG signal was performed here by a BITalino

ECG sensor (using a 3-electrode configuration)

connected to the A1 analog port of the BITalino

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

178

(r)evolution Core platform, which works with a

rechargeable battery and has bluetooth technology for

transmitting the acquired signals [for details, see

(BITalino, 2021a)].

Considering the nomenclature presented in (Drew

et al., 2004), the electrodes were positioned as

follows: Positive electrode in RA; Negative electrode

on LL; Reference electrode in LA. The acquired ECG

signal is transmitted to the smartphone via bluetooth,

since both the BITalino

(r)evolution Core platform

and the smartphone support it. Furthermore,

bluetooth is a consolidated communication protocol

for transmitting information wirelessly over a short

distance requiring low-power consumption.

Moreover, the use of bluetooth protocol in an

Android

™

environment is made easy through some

APIs (Android, 2021). Still, the manufacturer of the

BITalino platform, PLUX - Wireless Biosignals S.A,

provides an API that makes it possible to establish

and manage the bluetooth communication [for details,

see (BITalino, 2021)]. Finally, the smartphone

receives a digital ECG signal with resolution of 10

bits (see specifications of the A1 analog port),

sampled at 100 Hz, and with a limited frequency

range between 0.5 Hz and 40 Hz (which is, to some

extent, enough to deliver a high-quality ECG signal).

In turn, the processing of the ECG signal on the

smartphone is done through 20-second windows, i.e.,

a new window containing the last 20-seconds samples

of the ECG signal is ready to be processed every 20

seconds. It is important to mention that the acquisition

of the signal is done asynchronously in relation to the

processing; so, while a 20-second window is being

processed, another one is being acquired. The first

processing step, carried out over the 20-seconds

window of the acquired signal, involves the use of the

ECG sensor transfer function (BITalino, 2021). After

applying the sensor transfer function, the signal

samples are transformed to values whose amplitude is

within the range of -1.5 mV to 1.5 mV. The signal is

then resampled to 250 Hz (using the “resample”

function of the MATLAB

software) to meet the

input conditions of the segmentation step used.

It is worth mentioning that the BITalino device

can acquire the ECG signal with good quality and low

noise level when the measure is done with the user at

rest. On the other hand, the signal obtained while the

user was walking contains several artifacts that end

up mischaracterizing the ECG. For this reason, the

acquisition of the signals considered in this study was

conducted with the user at rest and it was ensured that

there was good contact between the electrodes and the

skin surface at each measurement. Thereby, it was not

necessary to implement any other signal pre-

processing step in the system.

4 R-PEAK DETECTION

ALGORITHM

Figure 2 depicts a block diagram of the implemented

R-wave peak detection algorithm, which is based on

(D. Benitez et al., 2001; Kaur et al., 2019;

Manikandan & Soman, 2012). Although the same

algorithm was used and validated in the work of

(Borghi et al., 2021), some modifications were made

in the original algorithm aiming to process 20-second

windows and to port the code to the C language

through the MATAB Coder

™

application. Basically,

in this algorithm, the ECG signal is first filtered by a

480-order FIR-type bandpass filter, with a passband

between 5Hz and 15Hz, designed using the self-

convolution method of a 60th-order Hamming

window (Kaur et al., 2019). Then, the first-order

derivative of the filtered signal

()

n is calculated

through

'( ) ( 1) ( ).d n fn fn=+− (1)

According to (Manikandan & Soman, 2012), signal

differentiation acts as a high-pass filter to reduce

interference from P and T waves. Next,

'( )dn is

normalized as

'( )

()

max [ '( ) ]

= (2)

where

N denotes the number of samples

corresponding to a 20-second window of the ECG

signal. The nonlinear Shannon energy (SE)

transformation is applied to

(),dn i.e.,

() ()log[ ()].

En d n d n=− (3)

This transformation aims to rectify and to highlight

the region of the QRS complex, facilitating the

detection of R peaks. The resulting Shannon energy

signal is smoothed by a 38-sample window moving

average filter (approximately, 152.7 ms for a

sampling frequency of 250 Hz). The Hilbert

transform is applied to the smoothed signal

(),

resulting in

().hn In turn, ()hn is applied to a

rectangular moving average filter with a duration of

625 samples, yielding

'( ).hn Then, we compute

() () '()zn hn h n=− (4)

Android-based ECG Monitoring System for Atrial Fibrillation Detection using a BITalino

 ECG Sensor

179

aiming to smooth out the signal and identify the lower

amplitude R-peaks. At the end of this process, a

smoothed odd symmetry signal

()zn is obtained,

where its zero-crossing point corresponds to the peak

of ( ),sn which is a strong candidate for the R-peak

position. Finally, using these candidate points as a

reference, an R-peak position detection is done by

searching for the local maximum near each candidate

point. The algorithm was validated considering the

MIT-BIH Atrial Fibrillation database (Moody &

Mark, 1992), which contains 23 recordings of ECG

signals of 10 hours each taken in patients diagnosed

with AF. Each of these recordings has two ECG

signal leads, sampled at 250 Hz, with 12-bit

resolution and with a bandwidth between 0.1 Hz and

40 Hz. Only the first lead of each recording was used

in the validation of the implemented segmentation

algorithm. The same database was used in the work

of (Borghi et al., 2020, Borghi et al., 2021) for

validating the R-peak detection algorithm, as well as

for validating the classification model.

The performance of the algorithm is shown in

Table 1 in terms of the sensitivity metric, defined as

()

TP FN

(5)

and positive predictivity metric, given by

()

TP FP

(6)

where

(True Positive),

(False Positive), and

FN (False Negative) indicate, respectively, the

number of labels that represent a heartbeat, the

number of labels that does not represent a heartbeat,

and the number of beats that were not marked (Luz et

al., 2016). As in (Borghi et al., 2020), it was

considered that the labeling made by the algorithm

regarding the position of the R-peak is correct if

within a context region of 150 ms centered on this

mark (75 ms before and 75 ms after) there is an

original annotation in the database. The algorithm

presents good detection results, with mean values of

Se and P+ equal to 98.99% and 95.95%, respectively.

In (Borghi et al., 2020), the segmentation

algorithm was validated using the measure of

accuracy, calculated for each subject as

100%

ACC

= (7)

where

N represents the total number of beats

marked in the database annotations. The average

accuracy found in (Borghi et al., 2020) is 98.95%,

while the average accuracy found here is 98.99%.

These results are consistent and confirm the

effectiveness of the approach used to port the code

made in MATLAB to C language.

When the algorithm is applied to an ECG signal

acquired by the BITalino platform, the results were

validated through visual inspection. Figure 3 shows the

detection results for a 5-second signal acquired with

the user at rest. Notice that the algorithm can properly

detect the position of R-peaks in ECG signals.

Figure 2: Block diagram of the R-peak detection algorithm

adapted from (Manikandan & Soman, 2012).

Figure 3: R-peaks detected by the segmentation algorithm

with the subject at rest.

5 FEATURE EXTRACTION AND

CLASSIFICATION MODEL

Considering the position of the R-peaks properly

detected, we proceed now to the extraction of features

Bandpass

Filtering

(BPF)

First-Order Forward

Differencing

(FOFD)

Amplitude

Normalization

(AN)

ECG

Linear Digital Filtering

f (n)

d’ (n)

d (n)

s (n)

Peak-Finding Logic

Hilbert

Transformation

Moving Average

Filtering

Zero Crossing Point

Detection

z (n)

Identification of the

real R-Peak

R-Peak Position

Moving Average

Filtering

Shannon Energy

Computation

Smooth SE Envelope Extraction

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

180

Table 1: Performance of the segmentation algorithm over the signals from the MIT-BIH Atrial Fibrillation database.

that will serve as input to the classification model.

Among these features, the interval between two

consecutive R peaks is calculated from

() ( 1)

RRi=

250

nn−−

detec detec

(8)

with vector

()n

detec

R containing the ECG signal

samples referring to the moment of occurrence of the

R-peaks. In a vector

'( ),nseg 60 consecutive RRi

values are stored, thus forming a segment. Each

segment

'( )nseg is then normalized with respect to

its maximum absolute value, i.e.,

()

() .

max [ ( ) ]

seg'

seg

seg'

(9)

So,

()nse

serves as input for the classification

model whose output can be one of the following

categories, namely: normal rhythm, AF rhythm or

other rhythm (see Figure 4). The machine learning

model selected to be used in this work was trained by

(Borghi et al., 2020) to perform heart rhythm

classification and AF event detection, based on a

bidirectional LSTM neural network. Such a model,

which contains 50 nodes in the hidden layer and needs

only the knowledge of RRi, showed an accuracy of

94.94% when validated over the MIT-BIH-Atrial

Fibrillation database in (Borghi et al., 2020). As

discussed in (Borghi et al., 2020), the RRi is the most

significant characteristic to differentiate segments

that present a normal behavior from those that exhibit

AF events. Note that, since the chosen model uses

only 60 RRi values as input features, its

implementation is not computationally expensive,

ratifying its use on a smartphone.

Figure 4: Illustration of the input and possible outputs of the

classification model.

6 ANDROID APPLICATION

When launching the app, the user is asked to activate

the smartphone's bluetooth function. Once the

bluetooth is activated, it is needed to turn on the

BITalino platform and press the “Connect” button on

the smartphone screen. If the BITalino platform is

within the range of bluetooth communication, the

platform name and Media Access Control (MAC)

address will be displayed on the application screen

and, when selecting it, the connection will be

established (see Figure 5). Once the smartphone is

connected to the BITalino platform, a graphical

interface is presented to the user, in which it is

possible to visualize the ECG signal, the connection

status between the smartphone and the BITalino

platform, the name, and MAC address of the

platform. Still, three buttons are visible, one named

“Disconnect” that allows user to disconnect the

platform and smartphone, another named “Find R

Peaks”, and a third one named “Screen for AF”.

When the connection status between the platform and

the smartphone is indicated on the screen as

“CONNECTED”, the user can click on either of the

two processing buttons. Note that the electrodes need

to be properly placed on the body of the user.

SUBJECT Se(%) P+(%) SUBJECT Se(%) P+(%) SUBJECT Se(%) P+(%)

98,62 95,24

97.79 96,25

99,54 96,75

99,69 97,1

99,58 95,98

98,49 96,11

99,71 92,03

98,49 96,12

99,99 97,49

98,42 95,02

99.93 98,09

99,67 96,64

99,78 96,55

99,83 96,88

99,81 96,49

99,87 97,35

92,64 90,24

99,43 97,21

99,75 96,37

99,6 96,48

98,97 94,49

99,95 95,84

98,99

Mean P + (% )

95,95

Mean S e (% )

Classification

Model

(LSTM)

{seg[1]; seg[2]; seg[3]; … seg[60]}

Normal Rhythm

AF Rhythm

Other Rhythm

Android-based ECG Monitoring System for Atrial Fibrillation Detection using a BITalino

 ECG Sensor

181

6.1 Segmentation Functionality

By clicking the “Find R-peaks” button, the

application starts the ECG signal segmentation

function. This function analyzes 60 seconds (3

windows of 20 seconds) of the ECG signal and

detects R-peaks. Meanwhile, it is possible to visualize

in real-time the signal being acquired. The

visualization of the ECG signal on the smartphone

screen is not restricted to the segmentation

functionality. At the end of the measurement, the

application notifies the user about how many beats

were detected and stores the signal and the

annotations regarding the position of the founded R-

peaks in a text file saved in the memory of the

smartphone. This functionality only performs the

segmentation of the ECG signal and does not apply

any classification regarding the rhythm of the

acquired signal.

Figure 5: Graphical interface of the application during the

ECG signal acquisition.

6.2 Quick Check Functionality

By clicking on the “Screen for AF” button, two

options to search for AF episodes are presented to the

user. In the first one, called “Quick check”, the

application will collect one segment with 60 values of

RRi and classify it according to its rhythm. Every 20

seconds, the R-wave peak detection algorithm is

executed and the detected peaks are counted. The

acquisition is interrupted when the system detects the

presence of 61 R-peaks. (The number of 20-second

windows needed depends on the heart rate of the

user.) From these 61 R-peaks, 60 RRi are calculated

to yield a segment which serves as input for the

classification model. The system then classifies the

segment into one of the possible categories (see

Figure 4). At the end of the analysis, the user is

notified with a message that depends on the

classification results. As in the segmentation

functionality, the analyzed ECG signal, the position

of the R-peaks, and the classification results are saved

in a text file in the memory of the smartphone.

6.3 Continuous Monitoring

Functionality

The continuous monitoring function, named

“Monitor”, starts the acquisition of the ECG signal

and segments each new 20-second window available.

Upon detecting the presence of 61 R-peaks, the

application generates a 60 RRi segment, and

classifies it into the 3 possible rhythms (as described

in Figure 4). In this functionality, the acquisition is

not interrupted after the classification of the first

segment, i.e., the system continues acquiring signal

until the user clicks on the “Stop Monitoring” button.

Therefore, this type of monitoring allows collecting

the ECG signal for an indefinite period and

classifying multiple segments, as long as the

bluetooth communication between the platform and

the smartphone is maintained. The construction of

segments is made considering an overlap of about 50

RRi such that each RRi is analyzed more than once in

different segments, reducing the probability of

misclassification. At the end of continuous

monitoring, i.e., when the user clicks on “Stop

Monitoring”, a notification will be presented to the

user regarding the duration of the signal and the

classification results. The acquired signal, the

position of the detected R-peaks, and the results of the

rhythm classification are saved in a text file stored in

the memory of the smartphone. Lastly, it is important

to remark that a MATLAB function (called

“app2mat”) used to read the text file stored in the

memory of the smartphone and a video demonstrating

the use of the developed application are available in

(Lazaretti, G.S. et al., 2021).

7 CONCLUSIONS

A system capable of performing the automatic

analysis of an ECG signal and detecting the presence

of AF episodes was implemented. The application

was developed using the Android

™

Studio IDE and

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

182

the processing steps were developed and validated

using MATLAB

. The processing functions and

routines were ported to C language through the

MATLAB Coder

™

and integrated into the application

written in Java programming language through the

JNI interface. Note that the implementation

methodology adopted here makes it possible to

implement changes to existing functions and/or to

include new processing functions, which can be done

first using MATLAB

. So, this methodology saves

implementation time and allows testing new

processing techniques.

Regarding the acquisition of the signals, it was

verified that the ECG sensor and BITalino platform,

when used with the subject at rest, can deliver to the

smartphone, through the BITalino API, an ECG

signal with good quality. On the other hand, when the

subject is in motion, artifacts were observed in the

acquired signal. Therefore, a pre-processing step

should be implemented aiming the use of this

application in the ambulatory monitoring of patients.

As presented in (Borghi et al., 2020), one of the

most important characteristics for identifying AF

events is the RRi. Despite this aspect, Borghi noted

that the inclusion of information about the P and T

waves of the ECG signal as input characteristics for

the classification model increases the classifier

accuracy (going from 94.94% to 98.17 %). So, the

accuracy of the proposed system could be improved

by replacing the classification model used by another

one that operates also with input characteristics

extracted from P, T and U waves.

In future research works, the application could be

redesigned such that it can operate even when

minimized. Furthermore, adaptations could be made

to include the application in a telemedicine system. In

this sense, the application could then send data and

results to a server, becoming thus available to a health

professional. Still, greater processing capacity can be

exploited and new analysis of the signal could be

carried out using sophisticated algorithms.

New processing functions can be implemented in

the application, since an interesting workflow was

developed that allows adding any function written

using MATLAB. So, the developed application

allows quick implementation and validation of other

ECG signal processing algorithms. Furthermore, it is

possible to evaluate the performance of these

algorithms dealing with real ECG signals acquired by

the BITalino sensor and platform. These

characteristics, along with the integration between the

MATLAB

and the Android

™

Studio platform

through the MATLAB Coder

™

and the JNI interface,

are, therefore, the main contributions of this work.

ACKNOWLEDGEMENTS

This work has supported by Fundação para a Ciência

e Tecnologia within the Project Scope:

UIDB/05757/2020, and by the European Regional

Development Fund (ERDF) through the Operational

Programme for Competitiveness and

Internationalization (COMPETE 2020), under

Portugal 2020 in the framework of the NanoID

(NORTE-01-0247-FEDER-046985) Project.

REFERENCES

AlGhatrif, M., & Lindsay, J. (2012). A brief review: history

to understand fundamentals of electrocardiography.

Journal of Community Hospital Internal Medicine

Perspectives, 2(1). https://doi.org/10.3402/jchimp.v2

i1.14383

Aljuaid, M., Marashly, Q., AlDanaf, J., Tawhari, I.,

Barakat, M., Barakat, R., Zobell, B., Cho, W., Chelu,

M. G., & Marrouche, N. F. (2020). Smartphone ECG

monitoring system helps 54 lower emergency room and

clinic visits in post–atrial fibrillation ablation patients.

Clinical Medicine Insights: Cardiology, 14.

https://doi.org/10.1177/1179546820901508

Android. (2021, May 19). Bluetooth overview.

https://developer.android.com/guide/topics/connectivit

y/bluetooth

Antzelevitch, C., & Burashnikov, A. (2011). Overview of

basic mechanisms of cardiac arrhythmia. Cardiac

Electrophysiology Clinics, 3(1), 23–45. https://doi.org/

10.1016/J.CCEP.2010.10.012

Benitez, D., Gaydecki, P. A., Zaidi, A., & Fitzpatrick, A. P.

(2001). The use of the Hilbert transform in ECG signal

analysis. Computers in Biology and Medicine, 31(5).

https://doi.org/10.1016/S0010-4825(01)00009-9

BITalino. (2021a). Documentation. Available at:

https://bitalino.com/documentation

BITalino. (2021b). Custom software. Available at:

https://bitalino.com/downloads/apis

Borghi, P. H., Teixeira, J. P., & Borges, R. C. (2020).

Classificação de episódios de fibrilação atrial por

análise do ECG com redes neuronais artificiais MLP e

LSTM. Master Thesis in Instituto Politécnico de

Bragança.

Borghi, P. H., Teixeira, J. P., & Borges, R. C. (2021). Atrial

fibrillation classification based on MLP networks by

extracting Jitter and Shimmer parameters. Procedia

Computer Science – Elsevier, 181, pp. 931-939.

https://doi.org/10.1016/j.procs.2021.01.249.

Chan, N., & Choy, C. (2017). Screening for atrial

fibrillation in 13 122 Hong Kong citizens with

smartphone electrocardiogram. Heart, 103(1).

https://doi.org/10.1136/heartjnl-2016- 309993

Chugh, S. S., Havmoeller, R., Narayanan, K., Singh, D.,

Rienstra, M., Benjamin, E. J., Gillum, R. F., Kim, Y.-

H., McAnulty, J. H., Zheng, Z.-J., Forouzanfar, M. H.,

Android-based ECG Monitoring System for Atrial Fibrillation Detection using a BITalino

 ECG Sensor

183

Naghavi, M., Mensah, G. A., Ezzati, M., & Murray, C.

J. L. (2014). Worldwide epidemiology of atrial

fibrillation. Circulation, 129(8). https://doi.org/

10.1161/CIRCULATIONAHA.113.005119

Drew, B. J., Califf, R. M., Funk, M., Kaufman, E. S.,

Krucoff, M. W., Laks, M. M., Macfarlane, P. W.,

Sommargren, C., Swiryn, S., & van Hare, G. F. (2004).

Practice Standards for Electrocardiographic Monitoring

in Hospital Settings. Circulation, 110(17).

https://doi.org/10.1161/01.CIR.0000145144.56673.59

Dulli, D. A., Stanko, H., & Levine, R. L. (2003). Atrial

fibrillation is associated with severe acute ischemic

stroke. Neuroepidemiology, 22(2). https://doi.org/

10.1159/000068743

Evans, G. F., Shirk, A., Muturi, P., & Soliman, E. Z. (2017).

Feasibility of using mobile ECG recording technology

to detect atrial fibrillation in low-resource settings.

Global Heart, 12(4). https://doi.org/10.1016/j.gheart.20

16.12.003

Garabelli, P., Stavrakis, S., & Po, S. (2017). Smartphone-

based arrhythmia monitoring. Current Opinion in

Cardiology, 32(1). https://doi.org/10.1097/HCO.0000

000000000350

Hagiwara, Y., Fujita, H., Oh, S. L., Tan, J. H., Tan, R. S.,

Ciaccio, E. J., & Acharya, U. R. (2018). Computer-

aided diagnosis of atrial fibrillation based on ECG

Signals: A review. Information Sciences, 467.

https://doi.org/10.1016/j.ins.2018.07.063

Jørgensen, H. S., Nakayama, H., Reith, J., Raaschou, H. O.,

& Olsen, T. S. (1996). Acute stroke with atrial

fibrillation: The Copenhagen Stroke Study. Stroke,

27(10). https://doi.org/10.1161/01.STR.27.10.1765

Kaur, A., Agarwal, A., Agarwal, R., & Kumar, S. (2019).

A novel approach to ECG R-Peak detection. Arabian

Journal for Science and Engineering, 44(8).

https://doi.org/10.1007/s13369-018-3557-8

Krijthe, B. P., Kunst, A., Benjamin, E. J., Lip, G. Y. H.,

Franco, O. H., Hofman, A., Witteman, J. C. M.,

Stricker, B. H., & Heeringa, J. (2013). Projections on

the number of individuals with atrial fibrillation in the

European Union, from 2000 to 2060. European Heart

Journal, 34(35). https://doi.org/10.1093/eurheartj/

eht280

Lazaretti, G. S., Kuhn, E. V e Teixeira, J. P. R. (29 October

2021). Illustrative video of the use of the application

and function to read the stored variables [Online].

Available at: http://lapse.td.utfpr.edu.br/downloads/

TCC_Lazaretti_2021.zip

Lin, H.-J., Wolf, P. A., Kelly-Hayes, M., Beiser, A. S.,

Kase, C. S., Benjamin, E. J., & D’A ostino, R. (1996).

Stroke severity in atrial fibrillation. Stroke, 27(10).

https://doi.org/10.1161/01.STR.27.10.1760

Luz, E. J. da S., Schwartz, W. R., Cámara-Chávez, G., &

Menotti, D. (2016). ECG-based heartbeat classification

for arrhythmia detection: A survey. Computer Methods

and Programs in Biomedicine, 127. https://doi.org/

10.1016/j.cmpb.2015.12.008

Manikandan, M. S., & Soman, K. P. (2012). A novel

method for detecting R-peaks in electrocardiogram

(ECG) signal. Biomedical Signal Processing and

Control, 7(2). https://doi.org/10.1016/j.bspc.20

11.03.004

Marini, C., De Santis, F., Sacco, S., Russo, T., Olivieri, L.,

Totaro, R., & Carolei, A. (2005). Contribution of atrial

fibrillation to incidence and outcome of ischemic

stroke. Stroke, 36(6). https://doi.org/10.1161/

01.STR.0000166053.83476.4a

Mittal, S., Movsowitz, C., & Steinberg, J. S. (2011).

Ambulatory external electrocardiographic monitoring.

Journal of the American College of Cardiology, 58(17).

https://doi.org/10.1016/j.jacc.2011.07.026

Miyasaka, Y., Barnes, M. E., Bailey, K. R., Cha, S. S.,

Gersh, B. J., Seward, J. B., & Tsang, T. S. M. (2007).

Mortality trends in patients diagnosed with first atrial

fibrillation. Journal of the American College of

Cardiology, 49(9). https://doi.org/10.1016/j.jacc.20

06.10.062

Moody GB, Mark RG. MIT-BIH Atrial fibrillation

Database 1992. https://doi.org/10.13026/C2MW2D.

Ott, A., Breteler, M. M. B., de Bruyne, M. C., van

Harskamp, F., Grobbee, D. E., & Hofman, A. (1997).

Atrial fibrillation and dementia in a population-based

study. The Rotterdam Study. Stroke, 28(2).

https://doi.org/10.1161/01.STR.28.2.316

Stewart, S., Hart, C. L., Hole, D. J., & McMurray, J. J. V.

(2002). A population-based study of the long-term risks

associated with atrial fibrillation: 20-year follow-up of

the Renfrew/Paisley study. The American Journal of

Medicine, 113(5). https://doi.org/10.1016/S0002-

9343(02)01236-6

T. Hickey, K., B. Biviano, A., Garan, H., Sciacca, R. R.,

Riga, T., Warren, K., Frulla, A. P., Hauser, N. R.,

Wang, D. Y., & Whang, W. (2017). Evaluating the

utility of mHealth ECG heart monitoring for the

detection and management of atrial fibrillation in

clinical practice. Journal of Atrial Fibrillation, 9(5).

https://doi.org/10.4022/jafib.1546

Wang, T. J., Larson, G., Levy, D., Vasan, R. S., Leip, E. P.,

Wolf, P. A., D’Agostino, R. B., Murabito, J. M.,

Kannel, W. B., & Benjamin, E. J. (2003). Temporal

relations of atrial fibrillation and congestive heart

failure and their joint influence on mortality.

Circulation, 107(23). https://doi.org/10.1161/01.CIR.0

000072767.89944.6E

Wolf, P. A., Abbott, R. D., & Kannel, W. B. (1991). Atrial

fibrillation as an independent risk factor for stroke: the

Framingham Study. Stroke, 22(8).

https://doi.org/10.1161/01.STR.22.8.983

Wolf, P. A., Mitchell, J., Baker, C. S., Kannel, W., &

D’Agostino, R. B. (1998). Impact of atrial fibrillation

on mortality, stroke, and medical costs. Archives of

Internal Medicine, 158(3). https://doi.org/10.1001/

archinte.158.3.229

World Health Organization. (2021, June 11).

Cardiovascular diseases (CVDs). https://www.who.int/

news-room/fact-sheets/detail/cardiovascular-diseases-

(cvds)

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

184