Computers in Biology and Medicine 116 (2020) 103540

Contents lists available at ScienceDirect

Computers in Biology and Medicine

journal homepage: http://www.elsevier.com/locate/compbiomed

Implementation and validation of real-time algorithms for atrial fibrillation

detection on a wearable ECG device
Italo Agustin Marsili a, Luca Biasiolli a, Michela Mas�e b, c, *, Alberto Adami a,
Alberto Oliver Andrighetti a, Flavia Ravelli c, Giandomenico Nollo b, d
a
Medicaltech Srl, Rovereto, Italy
b
IRCS-HTA, Healthcare Research and Innovation Program, Fondazione Bruno Kessler, Trento, Italy
c
Department of Physics, University of Trento, Trento, Italy
d
BIOtech Labs, Department of Industrial Engineering, University of Trento, Trento, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Background: Due to the growing epidemic of atrial fibrillation (AF), new strategies for AF screening, diagnosis,
Wearable devices and monitoring are required. Wearable devices with on-board AF detection algorithms may improve early
Mobile health diagnosis and therapy outcomes. In this work, we implemented optimized algorithms for AF detection on a
Smart health
wearable ECG monitoring device and assessed their performance.
Entropy
Embedded algorithms
Methods: The signal processing framework was composed of two main modules: 1) a QRS detector based on a
Cardiac arrhythmias finite state machine, and 2) an AF detector based on the Shannon entropy of the symbolic word series obtained
Cardiac rhythm monitoring from the instantaneous heart rate. The AF detector was optimized off-line by tuning its parameters to reduce the
computational burden while preserving detection accuracy. On-board performance was assessed in terms of
detection accuracy, memory usage, and computation time.
Results: The on-board implementation of the QRS detector produced an overall accuracy of 99% on the MIT-BIH
Arrhythmia Database, with memory usage ¼ 672 bytes, and computation time �90 μs. The on-board imple­
mentation of the optimized AF algorithm gave an overall accuracy of 98.1% (versus 98.3% of the original
version) on the MIT-BIH AF Database, with increased sensitivity (99.2% versus 98.5%) and decreased specificity
(97.3% versus 98.2%), memory usage ¼ 4648 bytes, and computation time � 75 μs (consistent with real-time
detection).
Conclusions: This study demonstrated the feasibility of real-time AF detection on a wearable ECG device. It
constitutes a promising step towards the development of novel ECG monitoring systems to tackle the growing AF
epidemic.

1. Introduction of AF events, which are often paroxysmal and asymptomatic. These

Atrial fibrillation (AF) is the most common arrhythmia in clinical
practice, affecting 1–2% of the general population [1]. The prevalence of
AF increases with age, reaching 18% in the population older than 85
years [2]. Due to population ageing, by 2030 AF is expected to affect
14–17 million people in Europe and 12.1 million people in the US [3,4].
AF is characterized by the absence of proper atrial activation and
contraction and by an irregular ventricular response [5,6], which are
associated with worsened hemodynamics and an increased risk of
ischemic stroke [7]. AF diagnosis can be hindered by the episodic nature
features make AF one of the most important public health issues and a
significant cause of healthcare expenditure in western countries.
New strategies for AF screening may improve stroke prevention and
patient stratification for different treatment options, ultimately leading
to better patient outcomes [8]. Mobile health technologies may enable
earlier AF detection and assessment through an extended physiological
monitoring by mobile and wearable devices [9–14]. The use of smart­
phones and smartwatches for AF detection has recently received atten­
tion as a mean for low-cost mass-screening [9–13]. While trials are
ongoing to evaluate the accuracy of these devices [9,11], AF detection

* Corresponding author. IRCS-HTA, Healthcare Research and Innovation Program, Fondazione Bruno Kessler, Department of Physics, University of Trento, 38123,
Trento, Italy.
E-mail address: michela.mase@unitn.it (M. Mas�e).

https://doi.org/10.1016/j.compbiomed.2019.103540
Received 31 July 2019; Received in revised form 11 November 2019; Accepted 11 November 2019
Available online 12 November 2019
0010-4825/© 2019 Elsevier Ltd. All rights reserved.
I.A. Marsili et al.
from standard electrocardiographic (ECG) signals remains the gold
standard for diagnosis. Therefore, new technologies for continuous
recording of ECG signals during daily activity and for extended periods
of time are under development. In particular, tele-Holter or event
recorder systems with on-board algorithms for AF detection may in­
crease the likelihood of detecting events by allowing longer recording
times in real-life scenarios. However, on-board AF detectors require not
only high performance, but also minimization of energy consumption,
which is mostly driven by data storage and computational complexity.
Previous studies have proposed to detect AF by analyzing ECG and P-
wave morphology [15–17], the ventricular interval (RR) series [10,
18–27], or a combination of the two [28–30]. For mobile applications,
RR-based algorithms are generally preferred to morphology-based al­
gorithms, due to the lower amount of incoming data and the lower noise
sensitivity of QRS-complex detection with respect to P-wave analysis.
RR-based algorithms can detect AF by quantifying the complexity of RR
series using different metrics, such as coefficient of variation and density
histograms [18], heart rate variability parameters from Poincar� e plots
[31], pattern similarity [20], sample entropy [10,25] or coefficient of
sample entropy [22,24,25], Shannon entropy [19,21], or normalized
fuzzy entropy [27].
Despite the large number of studies on AF detection algorithms, the
efforts to implement them on hardware devices are sparse [10,32]. In
this study we addressed the implementation of an algorithmic frame­
work for real-time AF detection on a prototype ECG monitoring device.
The framework included a finite-state machine QRS detector [33] and a
Shannon entropy-based AF detector [19], which had been previously
selected for their low computational burden and high accuracy [34]. We
aimed to: 1) optimize the off-line version of the AF detector to further
reduce the computational burden; and 2) measure the real-time per­
formance of the on-board implementation in terms of detection accu­
racy, memory usage, and computation time.
2. Methods
2.1. Hardware
The algorithms were designed to work on a prototype monitoring
system developed by Medicaltech srl (Rovereto, Italy), which allowed
continuous acquisition and recording of ECG signals and other
Fig. 1. Hardware and software components of the AF detection system. (A). Prototype board of the wearable device with main processing units highlighted and
labelled. (B). Block diagram of the algorithmic framework, composed of QRS detector (orange) and AF detector (cyan).
2
Computers in Biology and Medicine 116 (2020) 103540
physiological parameters in mobility, for hospital, ambulatory, and
domestic monitoring. The device was designed to work in different
configurations, such as 12-lead ECG, Holter, and tele-Holter/event
recorder. In real-time monitoring configuration, the device recorded
and analysed 3 ECG channels, at 250 Hz sampling rate, 16 bits, and 1
μV/bit resolution. A high-pass digital filter (0.5 Hz) removed baseline
wandering from the raw ECG signal. Notch (50–60 Hz) and low-pass
(35 Hz) filters removed line and muscular activity noise before dis­
playing the signal. Acquired signals could be transmitted via Wi-Fi to a
telemedicine server workstation for data storage and further analysis.
The device included an integrated CPU with a 32-bit ARM Cortex M3
processor (NXP LPC1788), a 512 kB internal FLASH memory bank and a
64 kB internal RAM memory bank, a low-power 24-bit analog front-end,
a RedPine BT/WiFi module, a Telit 2G/3G cellular module, two RAM
memory external banks of 2 MB each, and a μSD slot (Fig. 1A). The
firmware was loaded into the FLASH memory to execute low-level
functions for peripheral management and higher-level functions, such
as signal processing and analysis. After subtracting the memory required
by other operations, we estimated that the maximum amount of memory
available to the QRS and AF detection algorithms was <20 kB.
2.2. Framework for AF detection
The framework for AF detection consisted of two main modules
(Fig. 1B). The QRS detector (first module) was based on the algorithm
proposed by Guti� errez-Rivas et al. [33], which analysed a single-lead
ECG signal and detected the sequence of R-waves, yielding the heart
rate (hr) series as output. The AF detector (second module) was based on
the algorithm by Zhou et al. [19], which converted the hr series into a
symbolic word series, then calculated the Shannon entropy of the series,
and finally classified heart beats as ‘AF’ and ‘non-AF’. The two algo­
rithms were selected for their intuitive implementation, real-time
capability, and low computational burden [34].
2.2.1. QRS detector
Algorithm. The QRS detector included a pre-processing step, which
reduced baseline wandering (discrete derivative with Nd sample inter­
val) and high-frequency artefacts (moving average filtering of window
length NQRS), and emphasized R-peaks (square operation). After that,
QRS detection was performed by an adaptive threshold, based on a
I.A. Marsili et al.
finite-state machine with three states:
- State 1 (Peak Detector): the algorithm searched for the maximum
peak in a fixed search window of 260 ms (i.e., minimum RR þ QRS
complex duration); the detected maximum was classified as R-peak;
the threshold th was updated as the mean amplitude of all the
detected R-peaks.
- State 2 (Wait): the algorithm waited for a time period equal to the
minimum RR (200 ms) minus the time between the last detected R-
peak and the end of State 1.
- State 3 (Threshold): the threshold th was set at the last updated value
and was progressively reduced at each sample according to the
function:
PTh
th ½n� ¼ th ½n 1� � e fs (1)
where PTh was a parameter and fs was the sampling frequency; this state
ended when the signal became higher than the threshold; the system
returned to State 1 and a new R-peak was searched.
As indicated in ½33�; the parameters were adapted to the sampling
frequency (fs ¼ 250 ​ Hz), and set to: NQRS ¼ 6; Nd ¼ 5; PTh ¼ 6:07:
The output of the QRS detector was the R-peak time series (in sec­
onds), which was then converted into the hr series (in bpm). Given its
low computational burden, in the present work the algorithm was
implemented without modifications with respect to the original version.
Off-line testing. The QRS detection algorithm was implemented ‘off-
line’ in MATLAB (version R2019a, MathWorks, Inc., Natick, Massa­
chusetts, USA) to provide the reference baseline for detection perfor­
mance, which was assessed in terms of accuracy, sensitivity, and positive
predictive value (PPV). The algorithm was tested on the American Heart
Association Database (AHADB), the MIT-BIH Noise Stress Test Database
(NSTDB [35]), and the MIT-BIH Arrhythmia Database (MITDB [36]), as
required by the International Electrotechnical Commission (IEC
60601-2-47 [37]). The analysis started 5 min after the beginning of each
ECG recording, following the guidelines of the American National
Standards (ANSI/AAMI EC38:1998 [38] and ANSI/AAMI EC57:1998
[39]).
The AHADB is composed of 80 ECG recordings (½ hour each),
sampled at 250 Hz with 12-bit resolution over a 10 mV range [40]. After
removal of the first 5 min of each recording, the total number of QRS
complexes annotated by the cardiologists (Ground Truth) was 181,564.
The MITDB is a class 1 core database [41], which has been exten­
sively used as a benchmark in the literature, including [33]. The MITDB
contains 48 ECG recordings (½ hour each) obtained from 47 subjects,
sampled at 360 Hz (here re-sampled at 250 Hz) with 11-bit resolution
over a 10 mV range. The total number of QRS complexes annotated by
the cardiologists was 91,285.
The NSTDB [35] is a class 1 core database, which includes 12 ECG
recordings (½ hour each, same characteristics of the MITDB) from 2
subjects and 3 noise recordings (baseline wander, muscle artefacts, and
electrode motion ‘em’ artefacts) [42]. The ECG recordings with different
signal-to-noise ratios (SNR from 24 to 6 dB) were obtained by adding
the ‘em’ noise to 2 ‘clean’ ECG recordings from the MITDB. The total
number of QRS complexes annotated by the cardiologists was 21,462.
The results of the QRS detector were validated against Ground Truth
annotations using the ‘bxb’ function of the WaveForm DataBase (WFDB)
Software Package ([43], version 10.6.0 for Linux downloaded from
Ref. [44]).
On-board testing. For the on-board implementation, the source code
of the QRS detection algorithm was re-written in C using μVision (ARM,
version 5.23), integrated in the firmware, compiled with Armcc (version
5.06), and loaded into the internal FLASH memory of the prototype ECG
device. The algorithm parameters were set as in the ‘off-line’ imple­
mentation for comparison.
MITDB recordings were re-sampled at 250 Hz, loaded into the μSD
memory card of the device, and then loaded one by one into the external
3
Computers in Biology and Medicine 116 (2020) 103540
RAM to be used as input of the on-board algorithm. Raw signals were not
pre-processed by the prototype filters, but only by the QRS module fil­
ters. The results of the QRS detector were validated using ‘bxb’.
The amount of on-board resources used by the QRS detector was
assessed in terms of memory usage, number of operations, and compu­
tation time for each of the three machine states. The memory usage
indicated the size of the static structure used by the QRS detector and
was measured from the change of heap size using RealTerm software
(CRUN, version 2.0.0.70) via serial connection. The number of times
each operation (shift, compare, sum or subtraction, division or multi­
plication) appeared in the source code was summarized as minimum and
maximum sets, depending on conditional statements being satisfied or
not.
The range of computation times was estimated by probing the
voltage of a motherboard pin that was activated at the beginning and
deactivated at the end of each state, and measuring the interval between
activation and deactivation with a digital oscilloscope (Tektronix model
MSO2024B).
2.2.2. AF detector
Algorithm. The AF detector proposed by Zhou et al. [19] consisted of
four main processing steps: 1. conversion of the hr series into a symbolic
sequence (sy); 2. combination of the symbolic series into a word series
(wv); 3. quantitative characterization of the word distribution by
Shannon Entropy; 4. thresholding of Shannon Entropy values for binary
classification of each beat.
The conversion stage encoded the hr series into few symbols, sy,
according to a predefined base of dimension b, formed by the sequence
of integers [0, …, b-1]. The conversion was performed by the mapping
function:
8
< b 1 if hr½n� � hrmax
sy½n� ¼ hr½n� (2)
: other cases
d
where ⌊ ⌋ was the floor operator, hrmax was a cut-off value, and d was a
quantization factor equal to hrmax/(b-1).
Symbols were combined into words, wv, of m symbols, according to:
wv½n� ¼ sy½n� þ ðsy½n 1�Þ � ðbÞ… þ ðsy½n m�Þ � ðbÞm 1
​ (3)
Word values were in the range [0, wvmax], with maximum given by:
X
m
wvmax ¼ ðb 1Þ � ðbÞi 1
(4)
i¼1
The information stored in a sequence of N words, wv [n-Nþ1],… wv
[n], was quantified by a coarse version of Shannon Entropy, H:
k Xk
H½n� ¼ pi � log2 ðpi Þ (5)
N � log2 ðNÞ i¼1
where 1 � k � N was the number of characteristic elements in the
sequence, and 0 � pi � 1 was the probability associated to the ith char­
acteristic element. This was estimated as pi ¼ Ni/N, where Ni was the
number of repetitions of the element in the sequence. The calculation of
H was repeated beat-to-beat, lowering the computational cost by a
recursive implementation [19].
Finally, H values were compared with an Optimal Threshold (OT),
and cardiac beats were classified as ‘AF’ if H � OT, and as ‘non-AF’
otherwise.
Optimization. The AF detector was optimized to minimize resource
consumption for on-board implementation by testing different config­
urations of the encoding base (b), word length (m), and window length
for H computation (N). In the original implementation by Zhou et al.
[19], indicated as ‘Zhou x64 3f’, the parameters were set to b ¼ 64,
m ¼ 3, and N ¼ 127 (‘f’ ¼ full window size), which required a memory
buffer size of 256 kB, i.e. much larger than what was available on the
I.A. Marsili et al.
device. In this study, we tested four modified versions of the AF detec­
tion algorithm:
- ‘Zhou x64 2f’, with parameters b ¼ 64, m ¼ 2, and N ¼ 127, corre­
sponding to a buffer size of 4 kB.
- ‘Zhou x64 2h’, with b ¼ 64, m ¼ 2, and N ¼ 63 (‘h’ stands for half
window), corresponding to a buffer size of 4 kB.
- ‘Zhou x32 2f’, with b ¼ 32, m ¼ 2, and N ¼ 127, corresponding to a
buffer size of 1 kB.
- ‘Zhou x32 2h’, with b ¼ 32, m ¼ 2, and N ¼ 63, corresponding to a
buffer size of 1 kB.
For ‘x64’ configurations hrmax was set to 315 bpm and d to 5, while
for ‘x32’ ones hrmax was set to 310 bpm and d to 10. It is worth noting
that the decrease of m from 3 to 2 led to a decrease in buffer size from
256 kB to 4 kB for ‘x64’ configurations and further to 1 kB for ‘x32’ ones.
Off-line training. The AF detection algorithm proposed by Zhou
et al. [19] was implemented off-line in MATLAB using the original set of
parameters (‘Zhou x64 3f’) and the four modified versions. AF detection
performance was evaluated by Receiver Operating Characteristic (ROC)
analysis on training data to find the OTs for each version. RR series from
the Long Term AF Database (LTAFDB) [45] were used as training
dataset. LTAFDB is a class 3 core database, which includes 84 long-term
(24–25 h each) ECG recordings of subjects with paroxysmal or sustained
AF. Each record contains two simultaneously recorded ECG signals
sampled at 128 Hz with 12-bit resolution over a 20 mV range. The
LTAFDB contains 8,995,973 cardiac beats in total, of which 58.2% are
classified by experts as AF.
The AF detection algorithm performed a binary classification
(AF ¼ true or false). Sensitivity and specificity were calculated from the
number of true positives (TP), true negatives (TN), false positives (FP),
and false negatives (FN). To obtain the ROC curves, sensitivity and
specificity of the binary classification were calculated while changing
the detection threshold from 0 to 1. ROC analysis used empirical prior
probabilities derived from the class frequencies, and the OT for each
version was found at the intersection between the ROC curves and a
straight line with slope S, calculated as:
costðFPÞ N
S¼ � (6)
costðFNÞ P
where P ¼ TP þ FN was the total number of positives (‘AF’ beats) and N
¼ TN þ FP the total number of negatives (‘non-AF’ beats), while cost(FP)
and cost(FN) were the misclassification costs of FP and FN, respectively.
OTs were found assuming equal misclassification costs, i.e. cost
(FP) ¼ cost(FN) ¼ 0.5. Additional thresholds TA and TB were calculated
using misclassification costs that favoured either sensitivity or speci­
ficity (i.e. changed the slope of the line intersecting the ROC curves).
Specifically, threshold TA, favouring sensitivity, was obtained by setting
cost(FP) ¼ 0.4 and cost(FN) ¼ 0.6; while threshold TB, favouring speci­
ficity, by setting cost(FP) ¼ 0.6 and cost(FN) ¼ 0.4.
Off-line testing. All the different versions of the AF detector were
validated against Ground Truth annotations, evaluating sensitivity,
specificity, PPV, and overall accuracy at their respective OTs (obtained
from the off-line training phase). RR series from the MIT-BIH Atrial
Fibrillation Database (AFDB [46]) were used as test dataset. The AFDB is
a class 3 core database that is used as benchmark in the literature,
including [19]. It is made of 25 ECG recordings (~10 h each) of subjects
with AF (mostly paroxysmal), sampled at 250 Hz with 12-bit resolution
over a range of 20 mV. Some records (00735, 03665, 04936 and 05091)
were excluded due to missing data or incorrect annotations, as in Zhou
et al. [19]. The resulting database included 1,039,748 cardiac beats in
total, of which 45.2% were classified by experts as AF.
Finally, the performance of the complete analysis framework, i.e.
combined QRS and AF detectors, was assessed on ECG recordings from
the AFDB. The resulting AF classification for each RR interval was
4
Computers in Biology and Medicine 116 (2020) 103540
validated against Ground Truth annotations that were re-indexed to the
QRS complexes detected by the proposed method.
On-board testing. The source code of the AF detection algorithm was
re-written in C using μVision, integrated in the firmware, compiled with
Armcc, and loaded into the internal FLASH memory of the prototype
ECG device. For the on-board implementation, we selected the set of
parameters of the best performing algorithm among the modified
versions.
The test dataset included the same RR series from the AFDB that were
used for off-line testing. All series were loaded into the device μSD
memory card, and then one by one into the external RAM. Results were
validated against Ground Truth annotations. On-board resource con­
sumption for encoding and complexity assessment in the AF detector
were measured, as described for the QRS detector.
3. Results
3.1. QRS detection
3.1.1. Off-line testing
The performance of the off-line implementation of the QRS detector
on the AHADB, NSTDB, and MITDB are reported in Table 1. Excellent
sensitivity and PPV (>99%) were obtained on the AHADB and MITDB,
whereas PPV and detection error worsened significantly on the NSTDB,
mostly due the extreme cases with SNR ¼ 0 and 6 dB (Table 2).
However, at SNR ¼ 24, 18, and 12 dB, optimal performance was ach­
ieved, indicating robustness to noise at realistic SNR levels.
3.1.2. On-board testing
The performance of the on-board implementation of the QRS de­
tector was assessed on MITDB signals loaded into the μSD card of the
device. The results were practically identical to the off-line imple­
mentation (Table 1). The on-board memory usage of the QRS detector
was 672 bytes, with an indicative range of computation times between 2
and 10 μs for State 1, 10–70 μs for State 2, and 1–10 μs for State 3
(Table 3 and Fig. 2). The number of operations for each state depended
on the ECG signal satisfying certain conditions. The minimum and
maximum sets of operations are detailed in Table 3.
3.2. AF detection
3.2.1. Off-line training
The off-line performance of different versions of the AF detector on
the LTAFDB (training dataset) are shown as ROC curves in Fig. 3A. The
highest score for the Area Under the Curve (AUC) was achieved by the
original ‘Zhou �64 3f’ algorithm (98.4%), while the best performance
among the modified versions was achieved by ‘Zhou �64 2h’ with
AUC ¼ 97.9%. Due to different encoding lengths and window sizes, H
values computed by the modified versions tended to be lower, which
resulted in lower OTs with respect to the original algorithm.
At OT ¼ 0.416, ‘Zhou �64 2 h’ yielded sensitivity ¼ 95.7% and
specificity ¼ 93.0%, compared to sensitivity ¼ 96.6% and speci­
ficity ¼ 93.7% of ‘Zhou �64 3f’. Additional detection thresholds for
‘Zhou �64 2 h’ were calculated: setting TA ¼ 0.366 increased sensitivity
to 97.0% (specificity ¼ 90.7%), while setting the threshold to TB ¼ 0.453
increased specificity to 94.3% (sensitivity ¼ 94.6%).
3.2.2. Off-line testing
The off-line performance of different versions of the AF detection
algorithm was measured on the AFDB (test dataset) and the resulting
ROC curves are shown in Fig. 3B. The behaviour of the ROC curves for
the test dataset was similar to that on the training data (Fig. 3A), albeit
with superior overall performance. The highest AUC was achieved by
the original ‘Zhou �64 3f’ algorithm (99.7%), followed by ‘Zhou �64
2h’ (99.5%).
The OTs found on the training dataset were used for off-line testing
I.A. Marsili et al. Computers in Biology and Medicine 116 (2020) 103540

Table 1
Performance of the QRS detection algorithm on the AHADB, NSTDB, and MITDB.
Database Implem Total QRS TP FN FP Sensitivity PPV Error

AHADB Off-line 181564 180426 1138 695 99.4% 99.6% 1.0%

NSTDB Off-line 21462 21186 276 3753 98.7% 85.0% 18.8%
MITDB Off-line 91285 90566 719 204 99.2% 99.8% 1.0%
MITDB On-board 91285 90567 718 211 99.2% 99.8% 1.0%

FN, false negatives; FP, false positives; Implem, implementation; PPV, positive predictive value; TP, true positives.

The results obtained using thresholds TA and TB on the AFDB are

Table 2
reported in Table 4. Using TA ¼ 0.366 increased sensitivity (99.6%), but
Performance of the QRS detector on the NSTDB, detailed at different noise
reduced specificity to a larger extent (95.9%). Differently, TB ¼ 0.453
levels.
produced a more balanced performance with lower sensitivity (98.6%)
SNR [dB] Total QRS TP FN FP Sensitivity PPV Error and higher specificity (97.9%) than OT ¼ 0.416.
24, 18, 12 3577 3577 0 1 100% 100% 0% The performance of the whole ECG analysis framework (QRS þ AF
6 3577 3562 15 164 99.6% 95.6% 5.0% detectors) on the AFDB is shown in Table 4. All results were slightly
0 3577 3521 56 1121 98.4% 75.9% 32.9%
lower than for the AF detector alone, which could be mostly attributed
6 3577 3372 205 2465 94.3% 57.8% 74.6%
to the different number of QRS complexes detected (1,054,238 RR in­
FN, false negatives; FP, false positives; PPV, positive predictive value; TP, true tervals found by our QRS detector versus 1,039,076 in the QRS anno­
positives. tations). It is worth noting that reference QRS annotations in the AFDB
were detected automatically and not manually corrected.

3.2.3. On-board testing

Table 3 The detection performance of the on-board implementation of ‘Zhou
Resource consumption for the on-board implementation of the detection �64 2 h’ on the AFDB was identical to that of the off-line version
algorithms. (Table 4). The on-board memory usage of the AF detector was 4648
Algorithm Memory Number of Operations Computation Bytes, with an indicative range of computation times between 1 and 5 μs
Usage Maximum Time for the encoding step and 60–70 μs for the complexity assessment step
[Bytes] Minimum Range [μs] (Table 3 and Fig. 2).
QRS detector 672
State 1 2 compare; 3 compare; 2–10 4. Discussion
2 sum/sub 7 sum/sub
State 2 2 compare; 2 compare; 10–70
1 div/mult 4 sum/sub; In this study, we optimized, implemented, and validated an algo­
4 div/mult rithmic framework for real-time AF detection on a prototype ECG
State 3 1 compare; 2 compare; 1–10 monitoring device. ECG recordings from gold-standard public databases
1 sum/sub 2 sum/sub;
were loaded directly into the device to assess detection accuracy,
1 div/mult
AF detector 4648 memory usage, and computation time of the on-board algorithms. We
Encoding 2 compare 1 int div; 1–5 demonstrated the feasibility of real-time AF detection on a wearable
2 compare; device by achieving excellent performance with limited memory usage
1 shift; and computation time.
1 sum
Complexity 4 compare; 4 compare; 60–70
Assessment 4 sum/sub; 6 sum/sub;
4.1. Algorithm optimization for hardware implementation
3 div/mult 3 div/mult

Div, division; int, integer; mult, multiplication; sub, subtraction.
on the AFDB. The thresholds are marked on the ROC curves in Fig. 3B to
show that they were close to optimal also on the test dataset, which
indicated their generalizability to unseen data. The performance metrics
for all the algorithms at their respective OTs are reported in Table 4. All
the modified algorithms showed higher sensitivity than the original one.
However, the ranking of algorithms in terms of accuracy remained un­
altered with respect to the training phase, since the increase in sensi­
tivity was balanced by a decrease in specificity. ‘Zhou �64 2h’ was
confirmed the best performer among the modified algorithms, with ac­
curacy of 98.1% compared to 98.3% in ‘Zhou �64 3f’, an increased
sensitivity (þ0.7%), but decreased specificity ( 0.9%) and PPV
( 1.0%).
Overall, ‘Zhou �64 2 h’ improved the number of cardiac beats
correctly classified as AF with respect to ‘Zhou �64 3f’, i.e. increased TP
and reduced FN. This behaviour can be appreciated in the representative
example shown in Fig. 4. However, in some cases the modified algorithm
increased FP, as exemplified in Fig. 5. These results can be explained by
the faster response of ‘Zhou �64 2 h’ to brief AF events and heart rate
changes, mostly due to the shorter word length, half-size window, and
lower OT.
The ECG analysis framework was based on published algorithms for
QRS and AF detection, which were required to work in real-time within
the limitations imposed by the hardware. The QRS detector was spe­
cifically designed by the authors for real-time applications with low
power consumption and performed like the Pan & Tompkins’ algorithm
[47], but with only half the computational cost [33]. Since the resource
consumption of the QRS detector was well-within the limits of our
hardware, we implemented the original algorithm without further
optimization, and tested it on ECG recordings from the AHADB and
MITDB, and on recordings with increasing levels of noise from the
NSTDB. QRS detection performance on the MITDB was only slightly
lower than that reported by the original study [33], likely due to the fact
that we did not run QRS detection on the first 5 min of each ECG
recording. Results on the NSTDB compared favorably with those ob­
tained by top-ranking algorithms from the Physionet Challenge 2014
[48]. In particular, on data with SNR ¼ 6 dB the performance was very
close to the top scores, while on data with SNR ¼ 0 dB both sensitivity
and PPV were the highest. For the on-board implementation, we ob­
tained sensitivity and PPV larger than 99% on the MITDB, memory
usage <1 kB and total computational time <0.1 ms per beat, which
demonstrated the suitability of the QRS detection algorithm for robust
and reliable real-time applications.

5
I.A. Marsili et al. Computers in Biology and Medicine 116 (2020) 103540

Fig. 2. Computational burden of the AF detection framework. Typical computation times for each of the three machine states (1, 2, or 3) of the QRS detector, and for
the encoding (A) and complexity assessment (B) steps of the AF detector. The output voltage was measured directly from the device using an oscilloscope.

Fig. 3. Receiver Operating Characteristic (ROC) curve analysis on training (A) and test (B) datasets for different versions of the AF detection algorithm: original
‘Zhou �64 3f’ algorithm vs. modified ‘Zhou �64 2h’, ‘Zhou �64 2f’, ‘Zhou �32 2h’, and ‘Zhou �32 2f’ versions. Legends report the Area Under the Curve (AUC) and
the Optimal Threshold (OT) for each algorithm. The thresholds applied on the test dataset and reported in panel (B) were the OTs resulting from the off-line training
phase (A). Plots are magnified at different scales.

Table 4
Performance of the original and modified versions of the AF detection algorithm on the AFDB (test dataset).
Algorithm Threshold Implementation Total RR TP FN FP Sensitivity Specificity PPV Accuracy

Zhou �64 3f OT ¼ 0.630 Off-line 1038383 462655 7039 10463 98.5% 98.2% 97.8% 98.3%
Zhou �64 2h OT ¼ 0.416 Off-line 1039076 466227 3731 15600 99.2% 97.3% 96.8% 98.1%
Zhou �64 2f OT ¼ 0.308 Off-line 1038383 466213 3481 17911 99.3% 96.9% 96.3% 97.9%
Zhou �32 2h OT ¼ 0.178 Off-line 1039076 464976 4982 31453 98.9% 94.5% 93.7% 96.5%
Zhou �32 2f OT ¼ 0.104 Off-line 1038383 467080 2614 41470 99.4% 92.7% 91.9% 95.8%
Zhou �64 2h OT ¼ 0.416 On-board 1039076 466227 3731 15600 99.2% 97.3% 96.8% 98.1%
Zhou �64 2h OT ¼ 0.416 QRS þ AF off-line 1054238 469910 4123 19786 99.1% 96.6% 96.0% 97.7%
Zhou �64 2h TA ¼ 0.366 Off-line 1039076 467966 1992 23226 99.6% 95.9% 95.3% 97.6%
Zhou �64 2h TB ¼ 0.453 Off-line 1039076 463386 6572 11852 98.6% 97.9% 97.5% 98.2%

FN, false negatives; FP, false positives; PPV, positive predictive value; RR, ventricular intervals; TP, true positives.

The algorithm by Zhou et al. [19] was selected for its low compu­
tational burden compared to similarly performing algorithms [20,34].
The encoding of the hr series into a symbolic sequence allowed a concise
description of the series, whose complexity could be easily quantified in
terms of Shannon entropy [19]. The encoding step required a small set of
arithmetic operations without any convolution/filtering operation, and
a recursive calculation algorithm reduced the burden of Shannon
entropy beat-to-beat computation [19]. The present work aimed at
further reducing the computational cost by optimizing the main pa­
rameters for symbolic sequence generation and Shannon Entropy
computation. Decreasing the word length from 3 to 2 symbols provided
major improvements (shrinking memory consumption by 64 times)
while not significantly affecting detection accuracy (only 0.2–0.4%
lower on the AFDB). On the other hand, changing the encoding base

6
I.A. Marsili et al. Computers in Biology and Medicine 116 (2020) 103540

Fig. 4. Performance of different AF detectors on a segment of recording ‘04048’ from the AFDB. (A) Heart rate signal with Ground Truth annotation of AF events
(segments highlighted in green). (B) Shannon Entropy estimated by ‘Zhou �64 3f’ (original algorithm) with detected AF events in orange. (C) Shannon Entropy
estimated by ‘Zhou �64 2h’ (modified algorithm) with detected AF events in orange. ‘Zhou �64 2h’ improved the number of cardiac beats correctly classified as AF
(i.e. true positives) compared to ‘Zhou �64 3f’. Optimal thresholds for both algorithms were calculated on the training dataset (LTAFDB).

Fig. 5. Performance of different AF detectors on a segment of recording ‘05261’ from the AFDB. (A) Heart rate signal without AF events. (B) Shannon Entropy
estimated by ‘Zhou �64 3f’ (original algorithm) with detected AF events in orange. (C) Shannon Entropy estimated by ‘Zhou �64 2h’ (modified algorithm) with
detected AF events in orange. ‘Zhou �64 2h’ increased the number of cardiac beats misclassified as AF (i.e. false positives) compared to ‘Zhou �64 3f’. Optimal
thresholds for both algorithms were calculated on the training dataset (LTAFDB).

from 64 to 32 elements led to a further 4-fold memory reduction, but
caused a larger deterioration of the detection accuracy (1.8–2.5%
decrease). Finally, different window lengths for Shannon entropy
computation yielded very similar performance. ’Zhou �64 2h’ offered
the best compromise between detection accuracy and computational
burden, with on-board accuracy of 98.1%, memory usage <5 kB, and
computation time �75 μs per beat. The performance was well within the
limits imposed by the hardware configuration and consistent with
real-time detection. It is worth noting that, compared to the original
version, the modified algorithm demonstrated higher sensitivity
(þ0.7%) in detecting AF beats, due to a faster response to short AF ep­
isodes. On the other hand, lower specificity ( 0.9%) was observed, most
likely due to the lower detection threshold and the coarser description of
hr dynamics in the two-dimensional (versus three-dimensional) word

7
I.A. Marsili et al. Computers in Biology and Medicine 116 (2020) 103540

state space. The proposed algorithm compared favorably with the
best-performing RR-based algorithms in the literature (Table 5),
demonstrating the highest sensitivity for all the considered detection
thresholds (OT, TA, TB). In terms of specificity, it displayed the third-best
performance after Petrenas et al. [20] ( 0.4%) and Huang et al. [49]
( 0.2%), when adopting TB. However, we should bear in mind that
minor differences in performance among algorithms may be due to
differences in the validation procedure on the AFDB (e.g. removal of
recordings, exclusion of the initial 5 min of each recording, etc).
Our results demonstrated the feasibility of implementing real-time
QRS and AF detection algorithms on a prototype device. Studies on
hardware implementations are sparse [10,32], and direct comparison
with other hardware systems may be difficult due to differences in the
implementation, detection settings, and signals. Andersson et al. [32]
designed and fabricated an ASIC (Application-Specific Integrated Cir­
cuit) real-time AF detector performing ultra-low voltage operations for
implantable loop recorders. The detector, based on the combination of
three RR metrics, had a sensitivity of 94.9% and specificity of 95.8%
(90.6% and 97.6% for the memory-saving version) on the AFDB
(Table 5). Lee et al. [10] optimized three RR-based detection algorithms
for implementation on iPhone4S technology. The highest performance
was obtained by the sample entropy detector, which achieved an ac­
curacy of 96.1% on the combined AFDB þ NSRDB datasets, and a
computation time of ~25 ms for the analysis of a 64-beat segment.
4.2. Clinical use of the technology
The feasibility of performing on-board AF detection with high ac­
curacy, as demonstrated in this study, may have implications for the
development of mobile health systems for AF monitoring [9]. Algo­
rithms were implemented on a prototype wearable ECG monitoring
device, characterized by a small size, limited cost, and the ability to
work in different modalities, e.g. tele-Holter and event-recorder.
Wearability and small economic cost of the device might encourage its
adoption for long-term monitoring of cardiac rhythm in large population
studies. In particular, since the device acquires standard ECG signals, it
could be used for AF diagnosis and burden assessment following ther­
apeutic interventions.
Since working in different modalities may require some performance
trade-offs, we identified two detection thresholds that favoured either
sensitivity or specificity. As a first example, when working in event-
recorder modality, the highest possible sensitivity should be pursued,
so that AF events are not missed. In such a scenario, the use of TA would
increase sensitivity to 99.6%, but at the cost of specificity being
decreased to 95.9%. As a future development, in order to reduce the
false alarms associated with lower specificity, one could devise a two-
step detection process, in which “potential AF” episodes detected by
the device are then automatically analysed on a telemedicine worksta­
tion. Remote AF detection could benefit from the use of
computationally-expensive algorithms, e.g. artificial intelligence and
machine learning approaches [30,51,52], since energy and memory
consumption would no longer be a concern. As a second example, when
working in Holter modality for diagnostic purposes, a better balance
between sensitivity and specificity could be obtained by setting the
detection threshold to TB.
Finally, it is worth noting that our prototype device may not be
suitable for mass-screening. However, the reduced computational cost
and memory usage of the optimized algorithms, combined with the
progress in smartphone technology, would potentially support the
implementation of the AF detector on smartphones for screening ap­
plications. With this aim, further studies should assess detection accu­
racy as a function of type, quality, and length of the signals provided by
the wearable technology. For instance, in a recent study on AF detection
by smartphone technology, only 66% of the traces acquired by the
“Kardia Band” were found to be analyzable [8].
4.3. Study limitations
The main limitation of this study was that the algorithms were not
tested on ECG signals acquired directly from AF patients by the proto­
type device. Although validation on public databases allowed compar­
isons with other methods, future studies should use ECG signals acquired
by the device under real-world conditions. This would allow us to
investigate the actual level and type of noise and artefacts corrupting the
signals and their effect on AF detection, as well as to devise suitable
filtering strategies to maximize performance.
5. Conclusion
This study demonstrated the feasibility of real-time AF detection on
wearable devices. A low-complexity algorithmic framework was opti­
mized and implemented on a prototype ECG monitoring device. The on-
board implementation achieved excellent real-time performance (98%
accuracy) with limited memory usage (less than 6 kB) and computation

Table 5
Performance of the optimized AF detector (‘Zhou �64 2h’) in comparison with previously proposed RR-based detectors.
AF Detectors Performed Analysis Database Sensitivity Specificity

Proposed detector (’Zhou �64 2h’) Shannon entropy of symbolic series AFDB1 99.2% (at OT) 97.3% (at OT)
99.6% (at TA) 95.9% (at TA)
98.6% (at TB) 97.9% (at TB)
Liu et al., 2018 [27], Normalized fuzzy entropy AFDB2 98.46% 89.85%
Andersson et al., 2015x [32], Turning point ratio þ Root mean square AFDB 94.9% 95.8%
of ΔRR þ Shannon entropy
Petrenas et al., 2014 [20], Ectopic beat filtering þ Bigeminal AFDB 97.1% 98.3%
suppression þ RR irregularity þ Signal fusion AFDB3 97.1% 98.1%
AFDB4 98.0% 98.2%
Lee et al., 2013x [10], Root mean square of ΔRR AFDB4 þ NSRDB 90.49% 94.17%
Shannon entropy 74.15% 96.81%
Sample entropy 97.26% 95.91
Lee et al., 2013 [23], Time varying coherence function þ Shannon entropy AFDB4 98.22% 97.68%
Huang et al., 2011 [49], ΔRR distribution difference curve AFDB 96.1% 98.1%
Lake and Moorman 2011 [22], Coefficient of sample entropy AFDB 91% 94%
Lian et al., 2011 [50], RR-ΔRR map AFDB 95.8% 96.4%
Dash et al., 2009 [21], Turning points ratio þ Root mean square AFDB4 94.4% 95.1%
of ΔRR þ Shannon entropy
Tateno and Glass 2001 [18], Kolmogorov-Smirnov test on ΔRR AFDB 94.4% 97.2%

AFDB , recordings no.00735, no.03665, no.04936, and no.05091 excluded; AFDB , recordings no.05091 and no.0.7859 manually corrected; AFDB3, recordings
1 2

no.00735 and no.03665 excluded; AFDB4, recordings no.04936 and no.05091 excluded; x, device implementation.

8
I.A. Marsili et al. Computers in Biology and Medicine 116 (2020) 103540

time (less than 0.2 ms per beat). These results are a promising step to­ [15] S. Ladavich, B. Ghoraani, Developing an atrial activity-based algorithm for
detection of atrial fibrillation, in: Conf. Proc. Annu. Int. Conf. IEEE Eng. Med. Biol.
wards the development of novel mobile health monitoring systems for
Soc. IEEE Eng. Med. Biol. Soc. Annu. Conf 2014, 2014, pp. 54–57, https://doi.org/
the management of the growing AF epidemic. 10.1109/EMBC.2014.6943527.
[16] M. García, J. R� odenas, R. Alcaraz, J.J. Rieta, Application of the relative wavelet
energy to heart rate independent detection of atrial fibrillation, Comput. Methods
Funding Progr. Biomed. 131 (2016) 157–168, https://doi.org/10.1016/j.
cmpb.2016.04.009.
This study was partially supported by Provincia Autonoma di Trento, [17] H. Pürerfellner, E. Pokushalov, S. Sarkar, J. Koehler, R. Zhou, L. Urban,
G. Hindricks, P-wave evidence as a method for improving algorithm to detect atrial
Industrial Project Ri.Car.Do. LP N� 6, 1999 (2017–2019).
fibrillation in insertable cardiac monitors, Heart Rhythm 11 (2014) 1575–1583,
https://doi.org/10.1016/j.hrthm.2014.06.006.
[18] K. Tateno, L. Glass, Automatic detection of atrial fibrillation using the coefficient of
Declaration of competing interest variation and density histograms of RR and deltaRR intervals, Med. Biol. Eng.
Comput. 39 (2001) 664–671.
[19] X. Zhou, H. Ding, W. Wu, Y. Zhang, A real-time atrial fibrillation detection
Marsili IA, Biasiolli L, Adami A, and Andrighetti AO are with Medi­
algorithm based on the instantaneous state of heart rate, PLoS One 10 (2015),
caltech Srl, which developed the prototype device the algorithms were e0136544, https://doi.org/10.1371/journal.pone.0136544.
implemented on. [20] A. Petrenas,
_ V. Marozas, L. S€ornmo, Low-complexity detection of atrial fibrillation
in continuous long-term monitoring, Comput. Biol. Med. 65 (2015) 184–191,
Nollo G served as scientific advisor for Medicaltech srl.
https://doi.org/10.1016/j.compbiomed.2015.01.019.
[21] S. Dash, K.H. Chon, S. Lu, E.A. Raeder, Automatic real time detection of atrial
References fibrillation, Ann. Biomed. Eng. 37 (2009) 1701–1709, https://doi.org/10.1007/
s10439-009-9740-z.
[22] D.E. Lake, J.R. Moorman, Accurate estimation of entropy in very short
[1] I. Savelieva, J. Camm, Update on atrial fibrillation: part I, Clin. Cardiol. 31 (2008)
physiological time series: the problem of atrial fibrillation detection in implanted
55–62, https://doi.org/10.1002/clc.20138.
ventricular devices, Am. J. Physiol. Heart Circ. Physiol. 300 (2011) H319–H325.
[2] Y.C. Lau, G.Y.H. Lip, New advances in the treatment of atrial fibrillation: focus on
[23] J. Lee, Y. Nam, D.D. McManus, K.H. Chon, Time-varying coherence function for
stroke prevention, Expert Opin. Pharma 15 (2014) 2193–2204, https://doi.org/
atrial fibrillation detection, IEEE Trans. Biomed. Eng. 60 (2013) 2783–2793,
10.1517/14656566.2014.948418.
https://doi.org/10.1109/TBME.2013.2264721.
[3] S. Colilla, A. Crow, W. Petkun, D.E. Singer, T. Simon, X. Liu, Estimates of current
[24] M. Carrara, L. Carozzi, T.J. Moss, M. de Pasquale, S. Cerutti, M. Ferrario, D.E. Lake,
and future incidence and prevalence of atrial fibrillation in the U.S. adult
J.R. Moorman, Heart rate dynamics distinguish among atrial fibrillation, normal
population, Am. J. Cardiol. 112 (2013) 1142–1147, https://doi.org/10.1016/j.
sinus rhythm and sinus rhythm with frequent ectopy, Physiol. Meas. 36 (2015)
amjcard.2013.05.063.
1873–1888.
[4] M. Zoni-Berisso, F. Lercati, T. Carazza, S. Domenicucci, Epidemiology of atrial
[25] M. Mas�e, M. Disertori, M. Marini, F. Ravelli, Characterization of rate and regularity
fibrillation: European perspective, Clin. Epidemiol. 6 (2014) 213–220.
of ventricular response during atrial tachyarrhythmias. Insight on atrial and nodal
[5] F. Ravelli, M. Mas� e, A. Cristoforetti, M. Marini, M. Disertori, The logical operator
determinants, Physiol. Meas. 38 (2017) 800–818, https://doi.org/10.1088/1361-
map identifies novel candidate markers for critical sites in patients with atrial
6579/aa6388.
fibrillation, Prog. Biophys. Mol. Biol. 115 (2–3) (2014) 186–197, https://doi.org/
[26] J.H. Park, H.N. Pak, S.K. Kim, J.K. Jang, J.I. Choi, H.E. Lim, C. Hwang, Y.H. Kim,
10.1016/j.pbiomolbio.2014.07.006.
Electrophysiologic characteristics of complex fractionated atrial electrograms in
[6] M. Mas� e, M. Marini, M. Disertori, F. Ravelli, Dynamics of AV coupling during
patients with atrial fibrillation, J. Cardiovasc. Electrophysiol. 20 (2009) 266–272.
human atrial fibrillation: role of atrial rate, Am. J. Physiol. Heart Circ, Physiology
[27] C. Liu, J. Oster, E. Reinertsen, Q. Li, L. Zhao, S. Nemati, G.D. Clifford, A comparison
309 (2015) H198–H205, https://doi.org/10.1152/ajpheart.00726.2014.
of entropy approaches for AF discrimination, Physiol, Measure 39 (2018), 074002,
[7] A.J. Camm, P. Kirchhof, G.Y.H. Lip, U. Schotten, I. Savelieva, S. Ernst, I.C. Van
https://doi.org/10.1088/1361-6579/aacc48.
Gelder, N. Al-Attar, G. Hindricks, B. Prendergast, H. Heidbuchel, O. Alfieri,
[28] S. Babaeizadeh, R.E. Gregg, E.D. Helfenbein, J.M. Lindauer, S.H. Zhou,
A. Angelini, D. Atar, P. Colonna, R. De Caterina, J. De Sutter, A. Goette, B. Gorenek,
Improvements in atrial fibrillation detection for real-time monitoring,
M. Heldal, S.H. Hohloser, P. Kolh, J.-Y. Le Heuzey, P. Ponikowski, F.H. Rutten, ESC
J. Electrocardiol. 42 (2009) 522–526, https://doi.org/10.1016/j.
committee for practice guidelines, guidelines for the management of atrial
jelectrocard.2009.06.006.
fibrillation: the task force for the management of atrial fibrillation of the European
[29] A. Petrenas,
_ L. S€
ornmo, A. Luko�sevi�cius, V. Marozas, Detection of occult
society of cardiology (ESC), Eur. Eur. Pacing Arrhythm. Card. Electrophysiol. J.
paroxysmal atrial fibrillation, Med. Biol. Eng. Comput. 53 (2015) 287–297,
Work. Groups Card. Pacing Arrhythm. Card. Cell. Electrophysiol. Eur. Soc. Cardiol
https://doi.org/10.1007/s11517-014-1234-y.
12 (2010) 1360–1420, https://doi.org/10.1093/europace/euq350.
[30] R. Colloca, A.E. Johnson, L. Mainardi, G.D. Clifford, A Support Vector Machine
[8] D. Kotecha, G. Breithardt, A.J. Camm, G.Y.H. Lip, U. Schotten, A. Ahlsson,
approach for reliable detection of atrial fibrillation events, Comput. Cardiol. 2013
D. Arnar, D. Atar, A. Auricchio, J. Bax, S. Benussi, C. Blomstrom-Lundqvist,
(2013) 1047–1050.
M. Borggrefe, G. Boriani, A. Brandes, H. Calkins, B. Casadei, M. Castell� a, W. Chua,
[31] J. Park, S. Lee, M. Jeon, Atrial fibrillation detection by heart rate variability in
H. Crijns, D. Dobrev, L. Fabritz, M. Feuring, B. Freedman, A. Gerth, A. Goette,
Poincare plot, Biomed. Eng. Online 8 (2009) 38, https://doi.org/10.1186/1475-
E. Guasch, D. Haase, S. Hatem, K.G. Haeusler, H. Heidbuchel, J. Hendriks,
925X-8-38.
C. Hunter, S. K€ a€ab, S. Kespohl, U. Landmesser, D.A. Lane, T. Lewalter, L. Mont,
[32] O. Andersson, K.H. Chon, L. Sornmo, J.N. Rodrigues, A 290 mV sub-V(T) ASIC for
M. Nabauer, J.C. Nielsen, M. Oeff, J. Oldgren, A. Oto, L. Pison, T. Potpara,
real-time atrial fibrillation detection, IEEE Trans. Biomed. Circuits Syst 9 (2015)
U. Ravens, I. Richard-Lordereau, M. Rienstra, I. Savelieva, R. Schnabel, M.
377–386, https://doi.org/10.1109/TBCAS.2014.2354054.
F. Sinner, P. Sommer, S. Themistoclakis, I.C. Van Gelder, P.E. Vardas, A. Verma,
[33] Raquel Guti�errez-Rivas, Juan Jesús García, William P. Marnane, Alvaro� Hern�andez,
R. Wakili, E. Weber, D. Werring, S. Willems, A. Ziegler, G. Hindricks, P. Kirchhof,
Novel real-time low-complexity QRS complex detector based on adaptive
Integrating new approaches to atrial fibrillation management: the 6th AFNET/
thresholding, IEEE sens. J. 15 (2015) 6036–6043.
EHRA Consensus Conference, Eur. Eur. Pacing Arrhythm. Card. Electrophysiol. J.
[34] I.A. Marsili, M. Mas�e, V. Pisetta, E. Ricciardi, A.O. Andrighetti, F. Ravelli, G. Nollo,
Work. Groups Card. Pacing Arrhythm. Card. Cell. Electrophysiol. Eur. Soc. Cardiol
Optimized algorithms for atrial fibrillation detection by wearable tele-holter
20 (2018) 395–407, https://doi.org/10.1093/europace/eux318.
devices, in: 2016 IEEE Int. Smart Cities Conf. ISC2, 2016, pp. 1–4, https://doi.org/
[9] M.V. McConnell, M.P. Turakhia, R.A. Harrington, A.C. King, E.A. Ashley, Mobile
10.1109/ISC2.2016.7580863.
health advances in physical activity, fitness, and atrial fibrillation: moving hearts,
[35] G.B. Moody, W.E. Muldrow, R.G. Mark, A noise stress test for arrhythmia detectors,
J. Am. Coll. Cardiol. 71 (2018) 2691–2701, https://doi.org/10.1016/j.
ComputCardiol. (1984) 381–384.
jacc.2018.04.030.
[36] G.B. Moody, R.G. Mark, The impact of the MIT-BIH arrhythmia database, IEEE Eng.
[10] J. Lee, B.A. Reyes, D.D. McManus, O. Maitas, O. Mathias, K.H. Chon, Atrial
Med. Biol. Mag. Q. Mag 20 (2001) 45–50.
fibrillation detection using an iPhone 4S, IEEE Trans. Biomed. Eng. 60 (2013)
[37] International Electrotechnical Commission Standards, IEC 60601-2-47:2012-Med­
203–206, https://doi.org/10.1109/TBME.2012.2208112.
ical electrical equipment - Part 2-47: particular requirements for the basic safety
[11] M.P. Turakhia, M. Desai, H. Hedlin, A. Rajmane, N. Talati, T. Ferris, S. Desai,
and essential performance of ambulatory electrocardiographic systems.
D. Nag, M. Patel, P. Kowey, J.S. Rumsfeld, A.M. Russo, M.T. Hills, C.B. Granger, K.
https://webstore.iec.ch/publication/2666. (Accessed 1 June 2019).
W. Mahaffey, M.V. Perez, Rationale and design of a large-scale, app-based study to
[38] American National Standards, ANSI/AAMI EC38:1998-Ambulatory
identify cardiac arrhythmias using a smartwatch: the Apple Heart Study, Am. Heart
electrocardiographs, 2ed. https://webstore.ansi.org/standards/aami/ansiaam
J. 207 (2019) 66–75, https://doi.org/10.1016/j.ahj.2018.09.002.
iec381998. (Accessed 1 June 2019).
[12] J.M. Bumgarner, C.T. Lambert, A.A. Hussein, D.J. Cantillon, B. Baranowski,
[39] American National Standards, ANSI/AAMI EC57:1998/(R)2003-Testing and
K. Wolski, B.D. Lindsay, O.M. Wazni, K.G. Tarakji, Smartwatch algorithm for
reporting performance results of cardiac rhythm and ST segment measurement
automated detection of atrial fibrillation, J. Am. Coll. Cardiol. 71 (2018)
algorithms. https://webstore.ansi.org/standards/aami/ansiaamiec5719982003.
2381–2388, https://doi.org/10.1016/j.jacc.2018.03.003.
(Accessed 1 June 2019).
[13] K.H. Chon, D.D. McManus, Detection of atrial fibrillation using a smartwatch, Nat.
[40] American Heart Association Database. https://physionet.org/content/aha
Rev. Cardiol. 15 (2018) 657–658, https://doi.org/10.1038/s41569-018-0057-1.
db/1.0.0/. (Accessed 1 June 2019).
[14] V.D.A. Corino, R. Laureanti, L. Ferranti, G. Scarpini, F. Lombardi, L.T. Mainardi,
[41] MIT-BIH Arrhythmia Database. https://alpha.physionet.org/content/mit
Detection of atrial fibrillation episodes using a wristband device, Physiol. Meas. 38
db/1.0.0/. (Accessed 1 June 2019).
(2017) 787–799, https://doi.org/10.1088/1361-6579/aa5dd7.

9
I.A. Marsili et al.
[42] MIT-BIH Noise Stress Test Database. https://alpha.physionet.org/content/nst
db/1.0.0/. (Accessed 1 June 2019).
[43] A.L. Goldberger, L.A. Amaral, L. Glass, J.M. Hausdorff, P.C. Ivanov, R.G. Mark, J.
E. Mietus, G.B. Moody, C.K. Peng, H.E. Stanley, PhysioBank, PhysioToolkit, and
PhysioNet: components of a new research resource for complex physiologic signals,
Circulation 101 (2000) E215–E220, https://doi.org/10.1161/01.cir.101.23.e215.
[44] The WaveForm DataBase - WFDB - Software Package. https://archive.physionet.
org/physiotools/wfdb.shtml. (Accessed 1 July 2018).
[45] Long Term Atrial Fibrillation Database. https://alpha.physionet.org/content/ltaf
db/1.0.0/. (Accessed 1 June 2019).
[46] MIT-BIH Atrial Fibrillation Database. https://alpha.physionet.org/content/afdb/
1.0.0/. (Accessed 1 June 2019).
[47] J. Pan, W.J. Tompkins, A real-time QRS detection algorithm, IEEE Trans. Biomed.
Eng. 32 (1985) 230–236, https://doi.org/10.1109/TBME.1985.325532.
10
Computers in Biology and Medicine 116 (2020) 103540
[48] M. Vollmer, Noise resistance of several top-scored heart beat detectors, in: 2017
Comput. Cardiol. CinC, 2017: pp. 1–4. doi:10.22489/CinC.2017.182-178.
[49] C. Huang, S. Ye, H. Chen, D. Li, F. He, Y. Tu, A novel method for detection of the
transition between atrial fibrillation and sinus rhythm, IEEE Trans. Biomed. Eng.
58 (2011) 1113–1119, https://doi.org/10.1109/TBME.2010.2096506.
[50] J. Lian, L. Wang, D. Muessig, A simple method to detect atrial fibrillation using RR
intervals, Am. J. Cardiol. 107 (2011) 1494–1497, https://doi.org/10.1016/j.
amjcard.2011.01.028.
[51] R.J. Martis, M.M.R. Krishnan, C. Chakraborty, S. Pal, D. Sarkar, K.M. Mandana, A.
K. Ray, Automated screening of arrhythmia using wavelet based machine learning
techniques, J. Med. Syst. 36 (2012) 677–688, https://doi.org/10.1007/s10916-
010-9535-7.
[52] M. Athif, P.C. Yasawardene, C. Daluwatte, Detecting atrial fibrillation from short
single lead ECGs using statistical and morphological features, Physiol. Meas. 39
(2018), 064002, https://doi.org/10.1088/1361-6579/aac552.