Technical ReportPDF Available

From Seizure Detection to Prediction: A Review of Wearables and Related Devices Applicable to Epilepsy via Peripheral Measurements


Abstract and Figures

Epilepsy is a disorder that affects people of all ages and can cause unpredictable seizures, with severe consequences for patients and their families. Although wearables capable of detecting seizures have started to appear, the prediction still relies on cumbersome and intrusive setups. Nowadays, wearable technologies are growing at an overwhelming pace, with numerous new devices emerging every year. Applications range from fitness tracking to health monitoring and modalities range from simple motion signals to neural activity. In this review, a comprehensive comparison is conducted on some of the clinically validated or peer-reviewed wearable and related devices for physiological sensing, with potential application to epilepsy prediction. Based on previous studies with patients, we focused our review on peripheral measurements prone to be acquired discreetly and non-intrusively. The devices herein mentioned present some type of validation and are currently commercially available for research and personal use. With this analysis, we were able to compare a great variety of specifications and identify points of improvement for all devices, in order to better meet both the needs of researchers in the field of epilepsy and the usability considerations of the end-users.
Content may be subject to copyright.
IT - IST, Torre Norte, Piso 10
Av. Rovisco Pais, 1
Lisboa 1049-001, PORTUGAL
Tel: +351 218 41 84 54
Av.ª das Forças Armadas
Lisboa 1649-026, PORTUGAL
Tel: +351 217 90 30 00
IT - Campus Universitário de Santiago
Aveiro 3810-193, PORTUGAL
Tel: +351 234 37 79 00
IT - Universidade de Coimbra - Polo II
Coimbra 3030-290, PORTUGAL
Tel: +351 239 79 62 36
IT - UBI, Convento Santo António
Covilhã 6201-001, PORTUGAL
Tel: +351 275 31 98 99
IT - UP, Rua Dr. Roberto Frias, s/n
Porto 4200-465, PORTUGAL
Tel: +351 225 08 14 00
IT - PL, Morro do Lena - Alto do Vieiro
Leiria 2411-901, PORTUGAL
Tel: +351 244 82 03 40
From Seizure Detection to
Prediction: A Review of
Wearables and Related
Devices Applicable to
Epilepsy via Peripheral
Mariana Abreu, Ana Fred, Hugo Pl´acido da Silva
IT - Instituto de Telecomunica¸oes
Instituto Superior T´ecnico, 1049-001 Lisboa, Portugal
{mariana.abreu, ana.fred, hugo.placido.silva}
Chen Wang
Future Media Convergence Institute
Xinhuanet, 100031 Beijing, China
Date: 05/09/2019
Version: 1.0
Number: IT-FMCI-20190905
Epilepsy is a disorder that affects people of all ages and can cause un-
predictable seizures, with severe consequences for patients and their families.
Although wearables capable of detecting seizures have started to appear, pre-
diction still relies on cumbersome and intrusive setups. Nowadays, wearable
technologies are growing at an overwhelming pace, with numerous new devices
emerging every year. Applications range from fitness tracking to health moni-
toring, and modalities range from simple motion signals to neural activity. In
this review, a comprehensive comparison is conducted on some of the clini-
cally validated or peer-reviewed wearable and related devices for physiological
sensing, with potential application to epilepsy prediction. Based on previous
studies with patients, we focused our review on peripheral measurements prone
to be acquired discreetly and non-intrusively. The devices herein mentioned
present some type of validation and are currently commercially available for
research and personal use. With this analysis we were able to compare a great
variety of specifications and identify points of improvement for all devices, in
order to better meet both the needs of researchers in the field of epilepsy and
the usability considerations of the end users.
Wearable Technology; Epilepsy; Health Monitoring; Physiological Signals
1 Introduction
Epilepsy is a brain disorder characterized by abnormal hypersynchronous
neuronal activity known as seizures. Seizures are a serious concern for patients
are caretakers since they can lead to violent muscle convulsions, loss of con-
sciousness and even sudden death. Millions of people worldwide suffer from
epilepsy, whereas one third is still unresponsive to anti-epileptic drugs. The
detection, but especially the prediction of seizure events can have a meaning-
ful impact on the lives of epileptic patients and their families, by reducing the
anxiety associated to not knowing when the next seizure will occur and apply
specific treatments when a seizure is about to occur.
While seizure detection consists on triggering during the occurrence of the
seizure, seizure prediction consists on knowing if a seizure will occur in the
near future. Video-electroencephalography (V-EEG) is an established prac-
tical and cost-effective way to measure the neurological activity, reason for
which it has long been the gold standard in epilepsy studies. In recent years,
the growth of EEG databases led to the improvement of seizure prediction
[1]. Assi and colleagues [2] reviewed seizure prediction algorithms with EEG
and concluded the importance of long-term continuous acquisitions and the
advantage of intracranial EEG over scalp EEG. To address this issue, a previ-
ous experiment was conducted using the NeuroVista advisory system with 16
intracranial electrodes [3]. This long-term data has enabling the development
of patient-specific algorithms, which perform better [4, 5]. Despite the benefit
of intracranial EEG for seizure prediction, it involves the implantation of 16
or more electrodes, which might drift to unwanted locations [3], hindering the
data collection and the patient himself. An implantable device seems to be
indicated to close the loop between prediction and action [6]. Still, some issues
should be considered, such as the requirement of periodic replacement due to
battery depletion and the contraindication of MRI [7]. Even though seizure
prediction is evolving towards implantable electrodes and devices, the asso-
ciated risks and invasive procedures leave patients reluctant to accept these
solutions [8]. Furthermore, the long-term use of scalp EEG devices are also
not well accepted by patients [9]. Epilepsy needs a paradigm shift, from EEG
to non-EEG signals, to encounter patient’s preferences.
Although some work has been done in seizure detection [10–13], prediction
using non-intrusive setups is still mostly an open problem as current devices
and setups pose significant usability constraints, leading to poor acceptance
by patients and caregivers [14, 15]. Development of biomedical sensing tech-
nology and its adoption in society has radically changed over the years [16].
Amongst the most innovative outcomes with widespread adoption and numer-
ous applications, are wearable biomedical sensing devices. Wearable is the
common term for devices integrated in garments or designed as body-worn
accessories. In Figure 1, the most common sites for these devices are shown,
Blood Volume Pulse
Muscular Activity
Electrodermal Activity
Cardiac Activity
Brain Activity
Figure 1: Location of peripheral physiological signal sources applicable to
epilepsy prediction across the human body.
annotated with the different information sources they provide in each location,
based on the application of specific physiological sensors across the body.
Wearables with built-in sensors, internal battery and processor, allow con-
tinuous long-term monitoring of movement patterns or even physiological vari-
ables. According to Empatica [17], their device, similar to a smartwatch, can
be decomposed in four main layers: top cover, electronics board, lithium bat-
tery and finally bottom cover. The top cover serves as a protective shield for
the other components; the electronics board contains motion sensors, blue-
tooth antena, memory and CPU; the lithium battery provides long battery
life; finally the bottom cover contains the electrodermal activity electrodes.
This example of a wearable device unfolds the physical constraint associated
with each new layer, hence the trade-off between size and capacity. On the one
hand multiple embedded sensors lead to more information, and larger battery
sizes lead to longer continuous acquisitions; on the other hand, devices should
be small and unobtrusive. Thus, there is a compromise between the device’s
size and its modus operandi.
Wearables open new possibilities in healthcare, by enabling continuous and
objective symptom monitoring in clinical as well as out-of-hospital settings [18,
19]. Due to its nature, epilepsy is a condition that can greatly benefit from the
added value of such devices. Even for EEG-based algorithms, previous reviews
stated the added value of multimodal physiological sensing [7, 20]. Thus it
is of utmost importance the development of multimodal continuous data, to
empower epilepsy prediction research through peripheral measurements.
Multiple recent studies provide a comprehensive analysis of wearable de-
vices for specific purposes, such as autism, fitness, activity monitoring or cog-
nitive function [21–24]. However, many of the devices are still not validated
or reliably tested. In this review, we address devices that were previously vali-
dated or compared to gold standards, or already used in experimental studies.
The main contribution of our study is to summarize and analyze the devices
that are potentially suitable for research in epilepsy prediction.
In this review we focus only on devices that go beyond simple motion
sensing, although most of the devices characterized in this study also measure
motion. The multimodal approach of physiological and motion sensing is
desirable to encompass most types of epileptic seizures. Each seizure type
can entail particular manifestations, and peripheral multimodal approaches
are highly suggested for addressing epilepsy [25, 26].
Furthermore, patients have clearly identified their desire to have devices
capable of monitoring other functions [27]. As such, we focused our analysis
on devices that are capable of monitoring peripheral measurements, with non-
discriminatory body placement. As it will be further explained in Section 2.2
emotional balance also plays a major role in epilepsy, reason for which we also
highlight this component in our survey.
The novelty of wearable technology and epilepsy is self-explanatory from
the analysis of Figure 2. Herein it is visible that most publications cited
throughout this paper are very recent and the increase of papers since 2017
highlights the interest of the research community in these topics.
Figure 2: Number of publications per year cited in this paper.
The remainder of the paper is organized as follows. In Section 2 we pro-
vide a brief explanation on physiological sensing and its importance in health
monitoring. This topic is followed by a review of the state of the art in the
application of peripheral sensing to epilepsy, in order to understand what to
expect from wearable technology. In the following sections, the comparison
between different wearable devices is undertaken. Section 3 focuses on devices
that provide multiple sensors, in Section 4 we describe devices that only record
Electrodermal Activity (EDA), while in Section 5 the devices mentioned can
record Electrocardiographic (ECG) or Photoplethysmographic (PPG) signals.
After the description and characterization of all the devices, Section 6 dis-
cusses the overall specifications, current trends and convergence points, pro-
viding also the final conclusions and thoughts about future work perspectives.
2 Background
2.1 Physiological Sensing
Physiological signals are indicators of one’s homeostasis. From head to toe,
our body is in constant activity, which is measurable through the appropriate
sensors. Most bodily processes are mediated by the brain, and especially
in epilepsy this source plays a major role. However, some neural activity is
also manifested through the Autonomic Nervous System (ANS), which can
be measured peripherally. If we are more anxious or joyful, our body will
express it by means of different heart rates, increased perspiration, different
breathing rates, muscle contractions, and other phenomena. Therefore, from
less intrusive measurements, the neurological well-being and physical state can
still be partially assessed.
One physiological cue directly related with the ANS is the EDA, which is
given by the variations in our skin conductance due to changes in perspiration.
When exposed to arousing situations, our perspiration increases, leading to
distinctive events in the EDA, as illustrated in Figure 3.
Figure 3: Representation of arousal events in EDA signal, during heavy breath-
ing and consciously provoked apneas, which induce a reflex response of the
sympathetic nervous system. The vertical lines divide periods of provoked
apneas (A) and recuperation of breath (R).
Another common alteration in stressful situations is the heart rate increase.
A gold standard for cardiac signal acquisition is the ECG, although it is also
possible to infer the heart rate variability (HRV) through the Blood Volume
Figure 4: Comparison of BVP and ECG signal acquired simultaneously with
the presence of an artifact
Pulse (BVP), which is commonly measured through PPG. The study of HRV
has brought significant insights for different applications, from the discrimina-
tion of non-ictal/ictal periods in focal seizures [28], to emotional assessment
[29]. Despite both PPG and ECG being capable of informing about HRV, and
the PPG’s ease of being measured across the body, as it is shown in Figure
1, ECG is more precise in HRV detection [30] being less sensitive to motion
artifacts (e.g.) as it is visible in Figure 4 .
Furthermore, the cardiac signal’s morphology, seen in ECG, has shown ev-
idence of deeper insights about the subject’s mental health state, not only
for epilepsy seizure detection [31], but also for applications like attention
assessment [32]. The relation of heart-related signals with respiration has
contributed to the extraction of respiratory rate and signal from both PPG
and ECG [33, 34]. While some prefer ECG derived respiration [30], others
say there is no statistical difference between respiration parameters extracted
from ECG and PPG [34]. The importance of respiration is its variability be-
tween stressful and calm situations. Besides its derivation from heart signals,
with chest-mounted sensors, it is also possible to measure the frequency of
inhaling/exhaling, in what is usually called Respiration (Resp) signal. Body
temperature is also a health status indicator, and can be easily measured at
the skin surface.
The importance of each particular sensor is directly related with the pur-
pose of the study and/or acquisition procedure. To achieve a higher degree of
convenience to the users, it is desirable that most sensors can be embedded in
the same wearable device. However, some physiological signals might require
a specific position on the body. In Figure 1, it is shown which signals can be
recorded in each body position, for the most common positions of wearable
devices applicable in epilepsy prediction.
Devices capable of physiological sensing are not necessarily medical de-
vices; if they are not designed with a specific medical intended use, they do
not require any special approval [35]. Still, some validation of their perfor-
mance is required for public, clinical and research use, and reliable results
are essential. With a vast amount of wearable devices currently available, re-
views on these devices are preponderant, to understand the available options
and their limitations. Peake and coworkers [24], recently performed a critical
review providing a thorough comparison of devices within the topics of: hy-
dration, stress, muscle stimulation, cognitive feedback, sleep monitoring and
concussion evaluation. This review comes to the conclusion that despite the
vast amount of wearable devices, few of them were clinically validated or were
submitted to reliability testings. For that reason, most of the devices therein
mentioned will not be addressed in our study. Another recent review explored
several wearable solutions and their suitability to people with autism spectrum
disorders. Taj-Eldin and colleagues [21] performed a comprehensive study on
wearable solutions that could provide some assistance to such patients and
their caregivers.
2.2 Epileptic Seizure Prediction
Epileptic seizures are characterized by abnormal synchronization of brain ac-
tivity, which happen periodically in time. Several types of seizures are char-
acterized, according to the source’s origin, the amount of neurons being acti-
vated, and the physical manifestations of the seizure itself [36]. Seizure detec-
tion aims to identify the seizure’s onset moment, in order to monitor attacks
and alert the caregivers and medical personnel, thereby avoiding sudden unex-
pected death in epilepsy (SUDEP). A significant amount of work can be found
on seizure detection. With just simple motion sensing, such as accelerometers
(Acc), gyroscopes (Gyr) and magnetometers (Mag), results are promising for
generalized tonic-clonic seizures detection [37, 38] and can even be increased
just by adding information of the arm contraction, through Electromyography
(EMG) [39]. One interesting finding by Poh and colleagues [40] is that the
ANS contains meaningful information about the seizure that, when measured
through EDA sensors and combined with Acc data, has improved significantly
state-of-the-art results. Still, these solutions are insensitive to seizures without
motor convulsions, such as temporal lobe epilepsy.
Vandecasteele and coworkers [41] monitored both PPG and ECG using
wearable devices on 11 patients with temporal lobe epilepsy. The study
achieved promising results in temporal seizures detection, by being able to
evaluate accurately the variability of the patient’s heart rate. All these recent
achievements in seizure detection are boosting the use of wearable technology
in epilepsy. In terms of seizure detection, the Empatica Embrace [17] achieved
high accuracy for real-world settings, which allowed its approval by Food and
Drugs Administration (FDA).
In seizure prediction, HRV have revealed a predictive ability [42, 43] as-
sociated to its relation with ANS [44]. Concerning the algorithm, some ap-
proaches are threshold-based, whereas others use machine learning techniques.
Recently, Behbahani [45] reviewed previous works and techniques in epilepsy
detection and prediction through HRV. This study addressed the algorithms,
common features and overall analysis, and they concluded the importance of
a reliable algorithm. In this scope, Support Vector Machines achieved very
promising results with a patient-specific framework [46]. Moreover patient-
specific algorithms are also supported by the evaluation of heart rate differ-
ences across gender [47].
Many interesting studies are arising in peripheral seizure prediction still, no
devices or solutions are feasible on predicting the seizure onset. To enhance
research on this topic, reliable multimodal data is essential and should be
pursued with current wearable technology. One should understand how to
benefit from the already available devices, for data collection process, and
what are the most important factors when considering on developing new
wearable devices.
Following the human-centred design principles, stated by Motti and col-
leagues [48], the development of new solutions for epilepsy should consider the
opinion of patients themselves and their caregivers. Ozanne and coworkers
[49] performed a qualitative study, reaching some considerations for the ap-
plication of wearables to both epilepsy and Parkinson’s disease. This study
highlighted the importance of the end-user opinion in both development and
design processes. The authors concluded that users would prefer less specific
devices, which could provide more than one solution, with aesthetic and unob-
trusive form factors. A deeper study was conducted by Bruno and colleagues
[27] with 87 filled out surveys from patients with epilepsy (52), their caregivers
(13) and health practitioners (22). This study highlighted the interest within
the epilepsy community for the possibilities of wearable technology. In terms
of form factor preference, smartwatch or ring are acceptable, due to their dis-
crete and common shape, whereas head-mounted devices, like those required
for Electroencephalography (EEG), are less appealing.
Wearable devices in epilepsy are reaching the community with unobtrusive
form factors and reliable seizure detection algorithms [12]. Even though these
devices are life-changing, some improvements could take seizure monitoring
to a whole new level and meet the preference of multipurpose devices stated
by patients themselves [49]. If the system is only able to detect seizures, it
is working on immediacy and not prevention. Thus, it would be important
to understand what influences seizure occurrence. The frequency of seizures
is mostly related to epilepsy itself, however it can be influenced by the per-
son’s lifestyle. Stress-precipitated seizures are commonly reported by patients
with epilepsy [50], where general stress reduction techniques are promising,
not only for controlling such events [51], but also for prevention of SUDEP
[52]. Bautista [53] completed a survey with 266 patients with epilepsy, where
patients considered emotional stress to worsen seizures. Moreover, this survey
identified other lifestyle indicators as related to seizure’s frequency, such as
sleep activity and social support. It stated that sleep deprivation is consis-
tently identified as a seizure-trigger amongst patients with epilepsy, and poorly
treated sleep apnea can lead to drug-resistant seizures. These knowledge over
the lifestyle of patients and promotion of healthier habits has shown benefits
for reducing seizures. Besides yoga and meditation for stress reduction, and
promotion of better sleep, physical activity is also achieving promising results
as a complementary therapy for reducing seizures [54].
The commercial solutions presented by wearable devices in epilepsy are
mostly focused just on epilepsy. Since lifestyle and emotional states are also
influencing seizures and playing an important role in their reduction, patients
would benefit from devices with multiple functions. In this study we address
mostly commercial wearable devices which might be applied to epilepsy, with-
out being their only purpose. Thus, most devices addressed are not considered
medical devices, since they are not serving a particular health condition.
3 Multi-modal Sensing Devices
Currently, most of the available devices are fitness oriented, since parame-
ters such as physical activity and heart rate monitoring are easy to track with
simple sensors. However, to address more complex solutions such as epilepsy
prediction (EP), some challenging constraints emerge, such as the placement
and form factor for multiple sensing, associated to a higher cost. We can
divide sensing devices into two different categories: those created to gather
information for investigation purposes or general applications, and those cre-
ated for specific purposes, from fitness, to emotion recognition or even health
monitoring in general. In terms of the devices already on the market, our goal
is to explore the available options for physiological sensing and compare their
main features, in order to understand the value each device can bring to the
field of EP.
The devices herein approached are categorized regarding the type of in-
formation they can provide, with their built-in physiological sensors. The
categories hereby addressed are multimodal sensing devices, devices which
measure EDA (Section4) and lastly, devices which measure heart-related sig-
nals (Section 5). Some devices are developed as more general purpose tools
to record physiological signals, which can then be used for various applica-
tions, including those targeted by our work. These devices usually possess
multiple sensors, to cover a wider range of research purposes. In this scope,
we could include common smartwatches and smartphones, but since they only
record a limited set of physiological signals, they are excluded from this work.
Many commercially available wristbands are able to measure parameters such
as heart rate and skin temperature, however the collected data lacks sufficient
detail to create reliable health applications.
The devices that can record simultaneously multiple sensors are presented
on Table 1. Here seven devices are presented for comparison. The EQ02
developed by Equivital [55] provides accurate ambulatory monitoring. Even
Table 1: Devices with multiple sensors, characterised with respect to their applicability to research, validation status, form
factor or body positioning, battery duration, method to access the data, measured signals and, finally, their applicability to
epilepsy prediction (EP).
Product Research use Validated/ Reliable Site Battery Data Accessibility Sensors Suitability for EP
EQ02 [55] Ambulatory [56], Athletes Monitoring [57] Validation with Holter [58] Vest 48h Bluetooth transmission ECG, Derived RESP, Temp X
E4 [59] Migraines [60], Stress Detection [61] CE clearance [12] Bracelet +48h Cloud Storage EDA, PPG, Temp X
BITalino [62] Biometrics [63], Parkinson’s [64] Validation with BioPac [65] Non Specific 12-24h Bluetooth transmission ECG, EDA, EMG, BVP, RESP X
VisiMobile [66] Ambulatory Monitoring [67] FDAaclearance Wristband +12h WiFi transmission ECG, RESP, SpO2, Temp X
Bioharness3 [68] Sp orts Monitoring HR validation [69] Chest Strap 12-24h BLE transmission ECG, RESP, Temp, ACC X
Hexoskin [70] Remote Monitoring Experimental Validation [71] Smart Shirt 12h Cloud Storage or BLE transmission ECG, RESP, ACC X
Vivosmart4 [72] Stress Assessment Not yet validated Smartwatch 5 days BLE transmission Optical HRV, SpO2, Light, ACC x
aFDA is the United States Food and Drug Administration
though its particular chest band form factor is designed to be comfortable, a
recent study with military soldiers in long-term monitoring showed a negative
preference of chest monitoring, compared to less intrusive sites, such as the
wrist or hip [73].
Empatica developed the E4 [59]. Its shape, similar to a watch, is more
acceptable for long-term acquisitions with common users. The E4 is very com-
mon in health research, since it measures EDA signals unobtrusively and ac-
curately, which is of utmost importance in continuous stress detection [61, 74]
and emotion recognition [75]. Besides, its ability to measure heart rate, con-
tinuously, on the wrist, was studied and compared to state-of-the-art methods
on continuous ECG monitoring and PPG derived heart rate [76, 77]. More-
over, the E4 is accurate enough for use in monitoring of pathologies, such as
migraine attacks [60] or even epileptic seizure detection [12]. Although the E4
allows access to the raw data, the data is stored and managed on the vendor’s
cloud system.
BITalino is an all-in-one board, with the ability of recording a wide range
of physiological signals [62]. Since it is a simple board, it can be placed every-
where on the body, textiles or even external objects. The contact of electrodes
with the skin can be performed through common wire and gel electrodes,
through conductive textiles or any other type. BITalino’s manifoldness en-
larges the variety of applications on which to use this device. BITalino is
currently deployed in biometric solutions [63, 78], rehabilitation [79], Parkin-
son’s early detection [64], continuous vital and sleep monitoring [80, 81]. De-
spite BITalino ’s adequateness for experimental trials, its lack of form factor
and short battery life could hamper real-world testings. On the other hand,
Kutt and coworkers [82] compared the recordings of five state-of-the-art de-
vices, including the Empatica E4, and concluded that BITalino was the most
reliable device for continuous monitoring of EDA and HR, thus for further
investigations in affective computing and telemedicine.
Visi Mobile, by Sotera Wireless, aims to speed up patient’s recovery by
continuous monitoring of the physiological status, providing the doctor with
additional insights on the patient state. Its wristband form factor is tethered
to a chest ECG and armband monitor, providing an accurate recording of the
heart signal. Since we are looking for unobtrusive devices, Visi Mobile may
have limited acceptance by the patients.
The ZephyrT M Bioharness 3 can be slotted onto a chest-band or shirt,
with the purpose of monitoring the athletes during training. Even though it
can measure ECG, its validation was only conducted for heart rate monitoring,
therefore it requires more testing before assessing its overall performance.
With similar purpose, Hexoskin is a smart shirt for remote monitoring.
With embedded sensors on fabric, it measures heart and breathing rates,
physical activity among other parameters. This device was deployed in many
recent investigations [83], bringing wearable tech to a new level of usability.
However, it requires the use of their platform for accessing data, and regular
battery charging for a daily use.
Finally, Vivosmart 4 by Garmin is a small smartwatch that measures heart
rate, SpO2, Acc and ambient light data. Since it is not validated to this date
and the sensors are not very informative to meet our specific solutions, its
applicability is questioned. Still, both form factor and battery duration are
adequate for real-world acquisitions.
The aforementioned devices possess both positive and negative proper-
ties. While some do not have the best form factor or battery duration, others
lack validation or more informative sensors. In terms of their applicability
in epilepsy, all except Vivosmart4 could be adapted. However, in terms of
emotion recognition applications, both EQ02 and VisiMobile are too intru-
sive and do not measure electrodermal activity, which is usually an important
peripheral measurement related with the ANS. Moreover, E4 and Hexoskin
rely on cloud data storage, which can compromise privacy agreements with
the patients and the General Data Protection Regulations (GDPR) currently
enforced in Europe, therefore potentially leading to issues before the Ethical
Review Boards (ERB).
4 Devices Centered on Electrodermal Activity
Devices that measure EDA are usually related to stress management. Some
of these devices are ready to use, providing already a final application, while
others offer access to raw data and programming APIs that make them usable
in new research directions. On Table 2, five devices are compared in terms
of their specifications and applicability to epileptic seizure prediction. Mood-
metric’s smart ring aims to help managing stress. With 4 days of battery life
and a very discrete form factor, it seems ideal for real-life acquisitions of EDA.
Still, it requires more testing and validation, which is already taken care off,
with two studies being published later on this year [84].
Empatica created Embrace for patients with epilepsy, in order to detect
seizures and alert their caregivers. Its robust performance granted the FDA ap-
proval as a medical device. Even though it is specifically designed for epilepsy,
it is not adequate for other applications or improved solutions, since it does
not provide raw data access.
Another device with a smartwatch-like form factor, is GoBe2 by HealBe.
This device works both as a fitness and health tracker and a as research device.
Its form factor and battery lifetime are reasonable, however its performance
requires more reliable testings.
MyFeel is an up-and-coming wristband for emotion sensing. This wrist-
band is still on preliminary testings and further validation is required. If its
performance is proved to be reliable in ambulatory monitoring it could be a
solid solution for the development of mobile applications related with epilepsy
and emotion assessment, since it allows access to raw data and has a 24h long
The last device of Table 2 is Pip, a teardrop shape device, which measures
EDA when the user places his fingers inside. Due to its form factor, associated
with a short battery duration, Pip cannot be used for continuous monitoring,
which is impractical in light of the scope of our work.
In this section, devices that measure skin conductance are compared for
their applicability to epileptic seizure prediction. Since these devices have
less sensors, their battery lifetime if longer than the devices characterized in
Table 1. This property also allows more discrete form factors, which is an-
other advantage for real-world deployment. If a final solution can rely just on
these sensors (e.g. EDA combined with heart rate and skin temperature), the
wearable devices hereby proposed are enough. Nevertheless, further research
is required to confirm the quality of the acquired signals.
5 Devices Centered on Cardiovascular Measurements
Cardiovascular measurements can be quite revealing about the ANS. Al-
though heart rate is the most commonly found feature, additional information
can be extracted from the ECG and PPG signal traces. For example, the am-
plitude of the heart signal is influenced by the breathing amplitude whereas
the heart rate is influenced by physical activity or stress levels [30]. The blood
volume pulse, measured through PPG can be a less intrusive way of inferring
our internal status, since it can be measured non-intrusively, even just at the
fingertip. On Table 3, seven different devices which record these signals are
presented for comparison.
To help managing stress and tracking sleep cycles, the ring created by Oura
is very discrete and it is capable of measuring signals during sleep. Although
this device seems promising, since it is very small and with great battery
lifetime and recording capabilities, data is stored in their cloud, which, as
previously mentioned is an issue of concern.
With an infrared sensor, Fitbit Charge 3 performs both activity tracking
and sleep monitoring, providing a deeper understanding of the user’s physical
state. The concept is similar to Oura and both (Oura and Fitbit Charge)
have been compared to polysonographic exam, to validate their performance
on sleep monitoring [91, 92]. This experimental study confirmed the ability of
both devices to detect different sleep stages, with similar reasonable perfor-
mance. Nevertheless, these studies stated the importance of further validation
for both devices. The advantage of using a device such as Fitbit Charge is an
easier application to the users, since it already has a solid presence on the
market and the user’s trust. Its 5 day battery, smartwatch form factor and
BLE streaming to the mobile, are ideal for real-world applications.
Another real-world wearable is Spire, an adhesive patch that can be at-
tached daily to our clothes, allowing for unobtrusive continuous monitoring.
Even though it is a disposable device, its battery lasts for more than 1 year,
Table 2: Devices based on Electrodermal Activity, with their applicability to research, validation status, form factor or body
positioning, battery duration, how to access data, which signals are measured and finally their suitability to EP.
Product Research use Validated/ Reliable Site Battery Data Accessibility Sensors Suitability for EP
Moodmetric [85] Stress Assessment [56] Experimental comparison [86] Ring 4 days BLE Transmission EDA x
Embrace2 [59] Seizure Detection FDA and CE clearance Smartwatch +48h Onboard Processing EDA, Temp, Acc, Gyr X
GoBe2 [87] Nutrition, Hydration, Emotion Only for calorie intake [88] Smartwatch 24h BLE transmission EDA, Acc, Gyr, Mag X
MyFeel [89] Emotion Preliminary study Wristband +24h BLE Transmission EDA, Temp x
Pip [90] Stress Assessment Not yet validated External +8h Cloud Storage and BLE Transmission EDA x
which reduces its environmental impact. This innovative device is very dis-
crete, being a great option for ambulatory monitoring if further testings con-
firm its recordings reliability.
PulseOn OHR is a smartwatch tracker which provides raw data as output,
for both Acc and PPG. Even though it seems suitable for research, its battery
duration is shorter, when compared to its peers. Qardiocore is a chest-mounted
device, allowing for continuous ECG monitoring. Discrete and battery wise
seems adequate if the prediction process can work based solely on the ECG.
Also an ECG monitoring wearable is AIO Sleeve, a smart sleeve. Since it is a
sleeve and not a chest piece, it can be used during sports practice. However,
its performance was not experimentally tested to this date and its form factor
is not adequate for continuous real-world acquisitions. One device approved
by FDA for continuous monitoring is VitalPatch. As an adhesive patch, is
it placed on the chest for 5 days, monitoring 8 different features. Since it is
a short duration disposable device and data is stored on the cloud, it has a
limited use in the scope of our work.
In Table 3, different devices centered around cardiovascular measurements
are presented, from smartwatches to smart shirts and patches. Wearable tech-
nologies that record PPG signals are less battery intensive and can be mea-
sured easily on the skin surface, while ECG monitoring devices are more chal-
lenging to design. Some of the reviewed devices were not thoroughly tested
while the signals they acquire can be insufficient for epilepsy prediction and
emotion recognition. Nevertheless, some of them provide a good compromise
between form factor and functionality [93, 94].
6 Discussion and Conclusion
In this review, 20 physiological sensing wearable devices were analysed and
evaluated for their applicability in epilepsy research. In order to encompass
multiple types of seizures and the combination with other non-epileptic appli-
cations, like stress or physical monitoring, multimodal devices are preferable.
From all of the aforementioned wearables, only two possess a single sensor and
most have three to four modalities, as it is shown in Figure 5. This number
is explained by the position of the wearable. From Figure 1, a fixed position
allows the measurement of approximately four modalities.
Table 3: Devices with PPG or ECG sensors, with their applicability to research, validation status, form factor or body
positioning, battery duration, how to access data, which signals are measured and finally their suitability for EP.
Product Research use Validated/ Reliable Site Battery Data Accessibility Sensors Suitability in EP
Oura [93] Sleep Management Independent Validation [91] Ring 2-3 days Cloud Storage PPG, Acc, Temp X
Fitbit Charge [95] Activity and Sleep Monitoring Experimental validation [92, 96] Smartwatch 5 days BLE transmission PPG, Acc X
Spire [94] Ambulatory unobtrusive monitoring Peer-reviewed study [97] Clothing Adhesive 1 year Download or API access Resp, PPG, Acc X
PulseOn OHR [98] Atrial Fibrillation [99] Experimentally validated [100] Smartwatch 10h BLE transmission PPG, Acc x
Qardiocore [101] ECG Monitoring [99] CEaand TGAbapproved [102] Chest 24h BLE transmission ECG, Temp X
AIO Sleeve [103] ECG Monitoring Not validated yet Sleeve BLE transmission ECG, Acc x
VitalPatch [104] Continuous Monitoring [67] FDA clearance Adhesive Patch 5 days Cloud Storage ECG, Temp, Acc X
aCE - European Conformity
bTGA - Australia Therapeutic Goods Administration
Figure 5: Comparison between all aforementioned wearable devices, according
to their number of sensors.
In the universe of the wearable devices addressed, we can also analyze
which modalities are more frequent. In Figure 6, Temp and Acc are present
in ten devices. These two sensors are cheap and do not require a specific body
position. EDA and PPG are present in eight devices each, thus we infer the
interest of community in both of them. The reduced number of devices with
Resp sensor, is explained by the ability of measuring respiration from both
ECG and PPG [34]. Both Mag and SpO2do not add great information to
epilepsy prediction compared to other physiological measurements, whereas
EMG is too specific for a generalist approach to epilepsy prediction.
Figure 6: Comparison between all aforementioned sensors, according to the
number of devices with each sensor.
Each device possesses qualities and drawbacks, however no single device
has emerged as a ”de facto” solution for enabling relevant and continuous
data collection with patients in ambulatory environments. The devices from
Table 1, with multiple sensors, are preferred for initial research in these topics,
since multimodal approaches could lead to a better understanding of each
signal’s additional insight. From this group, we can select BITalino and the
Empatica E4 as the most appropriate choices for an experimental development
phase. The advantage of BITalino is the adaptability to any form factor, which
enables experimentation with prototypes in a form factor similar to that of a
potential final solution.
Within off-the-shelve solutions, if EDA is more valuable than ECG and/or
PPG, we can refine the data collection process to work with the devices from
Table 2, or with similar inspired versions. In the devices herein mentioned,
Moodmetric’s Smart Ring is an innovative device since it is shaped like a ring
and its battery lasts for 4 days. Even though it requires more validation, it
appears to be a promising device for unobtrusive EDA monitoring. On the
other hand, if the cardiovascular signals are preferable to EDA, the wearables
of Table 3 may be a better fit for ambulatory data monitoring. From this table
the most innovative devices are Oura, due to its ring shape, and Spire, due to
its long battery life and unobtrusiveness. The ability of both Moodmetric and
Oura to record different physiological signals from the finger, with day-long
batteries, enlightens into future possibilities of very small multimodal devices.
Many other devices, a summary of which is presented in Table 4, are
currently under-development and will be commercialized/validated in a near
future: Rhythm24 [105], Emotibit [106], Moxo [107], Heartguide [108], Kid-
dowear [109], Bitbrain Ring [110], UpMood [111], SKIIN [112] and CART
[113]. These were not mentioned on our analysis for lack of validation or
being still on a prototype/starting phase. Notwithstanding, they also seem
promising forthcoming devices which will add value to research in epilepsy
prediction using peripheral measurements.
In conclusion, wearable technologies are providing many interesting so-
lutions prone to be implemented in epilepsy research, which can provide a
deeper assessment of both physical and mental states. This additional infor-
mation could benefit current EEG-based solutions for seizure prediction or
even sustain the development of innovative solutions based only on peripheral
measurements. Even if prediction is still far from being feasible, the promising
results based on HRV measurements are opening doors for the paradigm shift.
Moreover, seizure prediction systems could act more as forecast appraisers
instead of the competitive approaches led by intracranial EEG predictions.
With this comprehensive analysis of off-the-shelve devices, we were able to
understand the advantages of certain specifications, taking into account the
user’s preference. The evaluation of suitability of the aforementioned devices
in multiple research topics, can be extended to many other devices with similar
In this paper we conducted a comprehensive analysis to several physiolog-
ical sensing devices, giving an overview of applicability and already existing
Table 4: New upcoming devices, with their applicability to research, validation status, form factor or body positioning, battery
duration, how to access data, which signals are measured and finally their suitability for EP. The spaces were left blank when
no information was found on the topic.
Product Research use Validated/ Reliable Site Battery Data Accessibility Sensors Suitability in EP
Rhythm24 [105] Heart Rate Only for Calorie Consumption Smartwatch 24 h BLE transmission EDA, IMP, Acc, Gyr, Mag x
EmotiBit [106] Signal Monitoring Scientific Validation [106] Non Specific Wifi, Bluetooth EDA, PPG, Temp, Acc, Gyr, Mag X
Moxo [107] Empathy Ambulatory EDA [67] Not validated Smartwatch +12 h ECG, RESP, SpO2, Temp X
Heartguide [108] Blood pressure monitor FDA clearance [114] Smartwatch 2 days BLE transmitted Blood Pressure X
Kiddowear [109] Kids Monitoring Not validated Wristband +24 h BLE transmission PPG, Temp, EDA, Acc X
Bitbrain Ring [110] Signal Monitoring Not validated Ring +12 h Bluetooth transmission PPG, EDA, Acc X
Upmood [111] Ambulatory Monitoring Not validated Wristband +12 h Bluetooth transmission PPG X
SKIIN [111] Textile Monitoring Not validated Textile 24 h Bluetooth transmission ECG, Acc, Temp X
CART [113] Atrial Fibrillation Detection Not validated Ring Bluetooth transmission PPG, Acc X
valences, as well as a critical assessment of functionalities, which may be useful
to consider in the development of new systems.
[1] Andreas Schulze-Bonhage, Hinnerk Feldwisch-Drentrup, and Matthias
Ihle. The role of high-quality eeg databases in the improvement and
assessment of seizure prediction methods. Epilepsy & Behavior, 22:S88–
S93, 2011.
[2] Elie Bou Assi, Dang K Nguyen, Sandy Rihana, and Mohamad Sawan.
Towards accurate prediction of epileptic seizures: A review. Biomedical
Signal Processing and Control, 34:144–157, 2017.
[3] Mark J Cook, Terence J O’Brien, Samuel F Berkovic, Michael Murphy,
Andrew Morokoff, Gavin Fabinyi, Wendyl D’Souza, Raju Yerra, John
Archer, Lucas Litewka, et al. Prediction of seizure likelihood with a long-
term, implanted seizure advisory system in patients with drug-resistant
epilepsy: a first-in-man study. The Lancet Neurology, 12(6):563–571,
[4] Ramy Hussein, Mohamed Osama Ahmed, Rabab Ward, Z Jane Wang,
Levin Kuhlmann, and Yi Guo. Human intracranial eeg quantitative
analysis and automatic feature learning for epileptic seizure prediction.
arXiv preprint arXiv:1904.03603, 2019.
[5] Philippa J Karoly, Hoameng Ung, David B Grayden, Levin Kuhlmann,
Kent Leyde, Mark J Cook, and Dean R Freestone. The circadian profile
of epilepsy improves seizure forecasting. Brain, 140(8):2169–2182, 2017.
[6] Sriram Ramgopal, Sigride Thome-Souza, Michele Jackson, Navah Ester
Kadish, Iv´an S´anchez Fern´andez, Jacquelyn Klehm, William Bosl, Claus
Reinsberger, Steven Schachter, and Tobias Loddenkemper. Seizure de-
tection, seizure prediction, and closed-loop warning systems in epilepsy.
Epilepsy & behavior, 37:291–307, 2014.
[7] Maxime O Baud and Vikram R Rao. Gauging seizure risk. Neurology,
91(21):967–973, 2018.
[8] Susan Arthurs, Hitten P Zaveri, Mark G Frei, and Ivan Osorio. Patient
and caregiver perspectives on seizure prediction. Epilepsy & behavior,
19(3):474–477, 2010.
[9] Andreas Schulze-Bonhage, Francisco Sales, Kathrin Wagner, Rute
Teotonio, Astrid Carius, Annette Schelle, and Matthias Ihle. Views of
patients with epilepsy on seizure prediction devices. Epilepsy & behavior,
18(4):388–396, 2010.
[10] Isa Conradsen, andor Beniczky, Karsten Hoppe, Peter Wolf, and
Helge BD Sorensen. Automated algorithm for generalized tonic–clonic
epileptic seizure onset detection based on semg zero-crossing rate. IEEE
Transactions on Biomedical Engineering, 59(2):579–585, 2011.
[11] Uri Kramer, Svetlana Kipervasser, Arie Shlitner, and Ruben Kuzniecky.
A novel portable seizure detection alarm system: preliminary results.
Journal of Clinical Neurophysiology, 28(1):36–38, 2011.
[12] Giulia Regalia, Francesco Onorati, Matteo Lai, Chiara Caborni, and
Rosalind W Picard. Multimodal wrist-worn devices for seizure detection
and advancing research: Focus on the empatica wristbands. Epilepsy
research, 153:79–82, 2019.
[13] Jesper Jeppesen, Anders Fuglsang-Frederiksen, Peter Johansen, Jakob
Christensen, Stephan W¨ustenhagen, Hatice Tankisi, Erisela Qerama,
Alexander Hess, and S´andor Beniczky. O-45 automated seizure detection
for epilepsy patients using wearable ecg-device. Clinical Neurophysiology,
130(7):e36, 2019.
[14] Christian Hoppe, Mieke Feldmann, Barbara Blachut, Rainer Surges,
Christian E Elger, and Christoph Helmstaedter. Novel techniques for
automated seizure registration: patients’ wants and needs. Epilepsy &
Behavior, 52:1–7, 2015.
[15] Jessica Askamp and Michel JAM van Putten. Mobile eeg in epilepsy.
International journal of psychophysiology, 91(1):30–35, 2014.
[16] Hugo Pl´acido da Silva, Ana Fred, and Ra´ul Martins. Biosignals for
everyone. IEEE Pervasive Computing, 13(4):64–71, 2014.
[17] Empatica. Embrace watch — smarter epilepsy management — em-
patica., 2019 (accessed August 6,
[18] Dongni Johansson, Kristina Malmgren, and Margit Alt Murphy. Wear-
able sensors for clinical applications in epilepsy, parkinson’s disease, and
stroke: a mixed-methods systematic review. Journal of neurology, pages
1–13, 2018.
[19] Hugo Pl´acido da Silva, Carlos Carreiras, Andr´e Louren¸co, Ana Fred,
Rui C´esar das Neves, and Rui Ferreira. Off-the-person electrocardio-
graphy: performance assessment and clinical correlation. Health and
Technology, 4(4):309–318, 2015.
[20] Levin Kuhlmann, Klaus Lehnertz, Mark P Richardson, Bj¨orn Schelter,
and Hitten P Zaveri. Seizure prediction—ready for a new era. Nature
Reviews Neurology, 14(10):618–630, 2018.
[21] Mohammed Taj-Eldin, Christian Ryan, Brendan O’Flynn, and Paul
Galvin. A review of wearable solutions for physiological and emotional
monitoring for use by people with autism spectrum disorder and their
caregivers. Sensors, 18(12):4271, 2018.
[22] Kanitthika Kaewkannate and Soochan Kim. A comparison of wearable
fitness devices. BMC public health, 16(1):433, 2016.
[23] Subhas Chandra Mukhopadhyay. Wearable sensors for human activity
monitoring: A review. IEEE sensors journal, 15(3):1321–1330, 2014.
[24] Jonathan M Peake, Graham Kerr, and John P Sullivan. A critical review
of consumer wearables, mobile applications, and equipment for provid-
ing biofeedback, monitoring stress, and sleep in physically active popu-
lations. Frontiers in physiology, 9:743, 2018.
[25] A Ulate-Campos, F Coughlin, M Ga´ınza-Lein, I S´anchez Fern´andez,
PL Pearl, and T Loddenkemper. Automated seizure detection systems
and their effectiveness for each type of seizure. Seizure, 40:88–101, 2016.
[26] Solveig Vieluf, Rima El Atrache, Sarah Hammond, Fatemeh Moham-
madpour Touserkani, Tobias Loddenkemper, and Claus Reinsberger.
Peripheral multimodal monitoring of ans changes related to epilepsy.
Epilepsy & Behavior, 96:69–79, 2019.
[27] Elisa Bruno, Sara Simblett, Alexandra Lang, Andrea Biondi, Clarissa
Odoi, Andreas Schulze-Bonhage, Til Wykes, Mark P Richardson,
RADAR-CNS Consortium, et al. Wearable technology in epilepsy: The
views of patients, caregivers, and healthcare professionals. Epilepsy &
Behavior, 85:141–149, 2018.
[28] Giorgos Giannakakis, Manolis Tsiknakis, and Pelagia Vorgia. Focal
epileptic seizures anticipation based on patterns of heart rate variability
parameters. Computer Methods and Programs in Biomedicine, 178:123–
133, 2019.
[29] Jianping Zhu, Lizhen Ji, and Chengyu Liu. Heart rate variability moni-
toring for emotion and disorders of emotion. Physiological measurement,
40(6), 2019.
[30] Hao-Yu Jan, Mei-Fen Chen, Tieh-Cheng Fu, Wen-Chen Lin, Cheng-Lun
Tsai, and Kang-Ping Lin. Evaluation of coherence between ecg and ppg
derived parameters on heart rate variability and respiration in healthy
volunteers with/without controlled breathing. Journal of Medical and
Biological Engineering, 39:783–795, 2019.
[31] Carolina Varon, Katrien Jansen, Lieven Lagae, and Sabine Van Huffel.
Detection of epileptic seizures by means of morphological changes in the
ecg. In Computing in Cardiology, pages 863–866, 2013.
[32] Carlos Carreiras, Andr´e Louren¸co, Helena Aidos, Hugo Pl´acido da Silva,
and Ana LN Fred. Morphological ECG analysis for attention detection.
In Proc. of the Int’l Joint Conf. on Computational Intelligence (IJCCI),
pages 381–390, 2013.
[33] Vignesh Ravichandran, Balamurali Murugesan, Vaishali Bal-
akarthikeyan, Sharath M Shankaranarayana, Keerthi Ram, Jayaraj
Joseph, Mohanasankar Sivaprakasam, et al. Respnet: A deep learning
model for extraction of respiration from photoplethysmogram. arXiv
preprint arXiv:1902.04236, 2019.
[34] Christina Orphanidou. Derivation of respiration rate from ambulatory
ecg and ppg using ensemble empirical mode decomposition: Comparison
and fusion. Computers in biology and medicine, 81:45–54, 2017.
[35] WeMake. Wemake d2.4 osmd final cclicense.
uploads/2019/06/WeMake OSMD Final CClicense.pdf, 2019 (accessed
August 13, 2019).
[36] Jerome Engel Jr. Seizures and epilepsy, volume 83. Oxford University
Press, 2013.
[37] Sandor Beniczky, Tilman Polster, Troels W Kjaer, and Helle Hjalgrim.
Detection of generalized tonic–clonic seizures by a wireless wrist ac-
celerometer: a prospective, multicenter study. Epilepsia, 54(4):e58–e61,
[38] Aaron Marquez, Michael Dunn, Jaime Ciriaco, and Farid Farahmand.
iseiz: A low-cost real-time seizure detection system utilizing cloud com-
puting. In Proc of the IEEE Global Humanitarian Technology Conference
(GHTC), pages 1–7, 2017.
[39] Mostafa Gheryani, Osman Salem, and Ahmed Mehaoua. An effective
approach for epileptic seizures detection from multi-sensors integrated in
an armband. In Proc. of the IEEE Int’l Conf. on e-Health Networking,
Applications and Services (Healthcom), pages 1–6, 2017.
[40] Ming-Zher Poh, Nicholas C Swenson, and Rosalind W Picard. A wear-
able sensor for unobtrusive, long-term assessment of electrodermal ac-
tivity. IEEE Transactions on Biomedical Engineering, 57(5):1243–1252,
[41] Kaat Vandecasteele, Thomas De Cooman, Ying Gu, Evy Cleeren,
Kasper Claes, Wim Paesschen, Sabine Huffel, and Borb´ala Hunyadi. Au-
tomated epileptic seizure detection based on wearable ECG and PPG in
a hospital environment. Sensors, 17(10):2338, 2017.
[42] Hirotsugu Hashimoto, Koichi Fujiwara, Yoko Suzuki, Miho Miyajima,
Toshitaka Yamakawa, Manabu Kano, Taketoshi Maehara, Katsuya
Ohta, Tetsuo Sasano, Masato Matsuura, et al. Heart rate variability
features for epilepsy seizure prediction. In Proc of the Asia-Pacific Sig-
nal and Information Processing Association Annual Summit and Conf.
(APSIPA), pages 1–4, 2013.
[43] Koichi Fujiwara, Miho Miyajima, Toshitaka Yamakawa, Erika Abe, Yoko
Suzuki, Yuriko Sawada, Manabu Kano, Taketoshi Maehara, Katsuya
Ohta, Taeko Sasai-Sakuma, et al. Epileptic seizure prediction based on
multivariate statistical process control of heart rate variability features.
IEEE Transactions on Biomedical Engineering, 63(6):1321–1332, 2015.
[44] Ebru Kolsal, Ay¸se Serdaro˘glu, Erman C¸ilsal, Serdar Kula, Az-
ime S¸ebnem Soysal, Ay¸seg¨ul Ne¸se C¸ıtak Kurt, and Ebru Arhan. Can
heart rate variability in children with epilepsy be used to predict
seizures? Seizure, 23(5):357–362, 2014.
[45] Soroor Behbahani. A review of significant research on epileptic seizure
detection and prediction using heart rate variability. Turk Kardiyol Dern
Ars, 46(5):414–421, 2018.
[46] Lucia Billeci, Daniela Marino, Laura Insana, Giampaolo Vatti, and Mau-
rizio Varanini. Patient-specific seizure prediction based on heart rate
variability and recurrence quantification analysis. PloS one, 13(9), 2018.
[47] Soroor Behbahani, Nader Jafarnia Dabanloo, Ali Motie Nasrabadi, and
Antonio Dourado. Gender-related differences in heart rate variability
of epileptic patients. American journal of men’s health, 12(1):117–125,
[48] Vivian Genaro Motti and Kelly Caine. Human factors considerations in
the design of wearable devices. In Proc. of the Human Factors and Er-
gonomics Society Annual Meeting, volume 58, pages 1820–1824. SAGE
Publications Sage CA: Los Angeles, CA, 2014.
[49] Anneli Ozanne, D Johansson, Ulla H¨allgren Graneheim, K Malmgren,
F Bergquist, and M Alt Murphy. Wearables in epilepsy and parkin-
son’s disease—a focus group study. Acta Neurologica Scandinavica,
137(2):188–194, 2018.
[50] Michael Privitera, Michael Walters, Ikjae Lee, Emily Polak, Adrienne
Fleck, Donna Schwieterman, and Sheryl R Haut. Characteristics of peo-
ple with self-reported stress-precipitated seizures. Epilepsy & Behavior,
41:74–77, 2014.
[51] Heather R McKee and Michael D Privitera. Stress as a seizure precipi-
tant: identification, associated factors, and treatment options. Seizure,
44:21–26, 2017.
[52] Stephan U Schuele, Peter Widdess-Walsh, Adriana Bermeo, and Hans O
Luders. Sudden unexplained death in epilepsy: the role of the heart.
Cleveland Clinic journal of medicine, 74(1):S121, 2007.
[53] Ramon Edmundo D Bautista. Understanding the self-management skills
of persons with epilepsy. Epilepsy & Behavior, 69:7–11, 2017.
[54] Ricardo Mario Arida, Antonio-Carlos Guimar˜aes de Almeida, Es-
per Abr˜ao Cavalheiro, and Fulvio Alexandre Scorza. Experimental and
clinical findings from physical exercise as complementary therapy for
epilepsy. Epilepsy & behavior, 26(3):273–278, 2013.
[55] Hidalgo. Sense and transmit.
tnr/sense-and-transmit, 2019 (accessed August 6, 2019).
[56] Christina Orphanidou and Ivana Drobnjak. Quality assessment of ambu-
latory ECG using wavelet entropy of the HRV signal. IEEE Biomedical
and Health Informatics, 21(5):1216–1223, 2016.
[57] John Kenny, SarahJane Cullen, and Giles D Warrington. The “ice-mile”:
Case study of 2 swimmers’ selected physiological responses and per-
formance. International journal of sports physiology and performance,
12(5):711–714, 2017.
[58] Abimbola A Akintola, Vera van de Pol, Daniel Bimmel, Arie C Maan,
and Diana van Heemst. Comparative analysis of the equivital EQ02
lifemonitor with holter ambulatory ECG device for continuous measure-
ment of ECG, heart rate, and heart rate variability: A validation study
for precision and accuracy. Frontiers in physiology, 7:391, 2016.
[59] Hidalgo. Real-time physiological signals — e4 eda/gsr sensor. https:
//, 2019 (accessed August 6,
[60] Heli Koskim¨aki, Henna M¨onttinen, Pekka Siirtola, Hanna-Leena Hut-
tunen, Raija Halonen, and Juha R¨oning. Early detection of migraine
attacks based on wearable sensors: experiences of data collection us-
ing empatica E4. In Proc. of the ACM Int’l Joint Conf. on Pervasive
and Ubiquitous Computing and Proc. of the ACM Int’l Symposium on
Wearable Computers, pages 506–511, 2017.
[61] Simon Ollander, Christelle Godin, Aur´elie Campagne, and Sylvie Char-
bonnier. A comparison of wearable and stationary sensors for stress
detection. In Proc. of the IEEE Int’l Conf. on Systems, Man, and Cy-
bernetics (SMC), pages 004362–004366, 2016.
[62] Hugo Pl´acido da Silva, Jos´e Guerreiro, Andr´e Louren¸co, Ana LN Fred,
and Ra´ul Martins. BITalino: A novel hardware framework for physiolog-
ical computing. In Proc. of the Int’l Conf. on Physiological Computing
(PhyCS), pages 246–253, 2014.
[63] Afonso Eduardo, Helena Aidos, and Ana LN Fred. Ecg-based biometrics
using a deep autoencoder for feature learning-an empirical study on
transferability. In ICPRAM, pages 463–470, 2017.
[64] Farika T Putri, Mochammad Ariyanto, Wahyu Caesarendra, Rifky Is-
mail, Kharisma Agung Pambudi, and Elta Diah Pasmanasari. Low
cost parkinson’s disease early detection and classification based on voice
and electromyography signal. In Computational Intelligence for Pattern
Recognition, pages 397–426. Springer, 2018.
[65] Diana Batista, Hugo Pl´acido da Silva, Ana Fred, Carlos Moreira, Mar-
garida Reis, and Hugo Alexandre Ferreira. Benchmarking of the BITal-
ino biomedical toolkit against an established gold standard. Healthcare
Technology Letters, 6(2):32–36, 2019.
[66] SoteraWireless. Why visi - sotera wireless. https://, 2019 (accessed August 6,
[67] Mariska Weenk, Harry van Goor, Bas Frietman, Lucien JLPG Enge-
len, Cornelis JHM van Laarhoven, Jan Smit, Sebastian JH Bredie, and
Tom H van de Belt. Continuous monitoring of vital signs using wearable
devices on the general ward: pilot study. JMIR mHealth and uHealth,
5(7):e91, 2017.
[68] BIOPACSystems. Bioharness — biopac.
bioharness/, 2019 (accessed August 6, 2019).
[69] Daniele Nepi, Agnese Sbrollini, Angela Agostinelli, Elvira Maranesi, Mi-
caela Morettini, Francesco Di Nardo, Sandro Fioretti, Paola Pierleoni,
Luca Pernini, Simone Valenti, et al. Validation of the heart-rate signal
provided by the Zephyr bioharness 3.0. In Proc. of the IEEE Computing
in Cardiology Conf. (CinC), pages 361–364, 2016.
[70] Carre Technologies Inc (Hexoskin). Hexoskin smart shirts - cardiac,
respiratory, sleep and activity metrics.,
2019 (accessed August 6, 2019).
[71] Nour H Cherif, Neila Mezghani, Nathaly Gaudreault, Youssef Ouakrim,
Imane Mouzoune, and Pierre Boulay. Physiological data validation of
the hexoskin smart textile. In Proc. of the Int’l Joint Conf. on Biomed-
ical Engineering Systems and Technologies (BIOSTEC), pages 150–156,
[72] Garmin. vivosmart 4 — activity tracker with pulse ox — garmin., 2019 (accessed August
6, 2019).
[73] Thomas Wyss, Lilian Roos, Nadja Beeler, Bertil Veenstra, Simon Delves,
Mark Buller, and Karl Friedl. The comfort, acceptability and accuracy
of energy expenditure estimation from wearable ambulatory physical
activity monitoring systems in soldiers. Journal of Science and Medicine
in Sport, 20:S133–S134, 2017.
[74] Mert Sevil, Iman Hajizadeh, Sediqeh Samadi, Jianyuan Feng, Caterina
Lazaro, Nicole Frantz, Xia Yu, Rachel Brandt, Zacharie Maloney, and
Ali Cinar. Social and competition stress detection with wristband phys-
iological signals. In Proc. of the IEEE Int’l Conf. on Wearable and
Implantable Body Sensor Networks (BSN), pages 39–42, 2017.
[75] Martin Ragot, Nicolas Martin, Sonia Em, Nico Pallamin, and Jean-Marc
Diverrez. Emotion recognition using physiological signals: laboratory vs.
wearable sensors. In Proc. of the Int’l Conf. on Applied Human Factors
and Ergonomics, pages 15–22. Springer, 2017.
[76] Cameron McCarthy, Nikhilesh Pradhan, Calum Redpath, and Andy
Adler. Validation of the empatica e4 wristband. In Proc of the IEEE
EMBS Int’l Student Conf. (ISC), pages 1–4, 2016.
[77] Julia Pietil¨a, Saeed Mehrang, Johanna Tolonen, Elina Helander, Holly
Jimison, Misha Pavel, and Ilkka Korhonen. Evaluation of the accuracy
and reliability for photoplethysmography based heart rate and beat-to-
beat detection during daily activities. In Proc. of the Int’l Conf. Euro-
pean Medical and Biological Engineering Conf. (EMBEC) and Proc. of
the Nordic-Baltic Conf. on Biomedical Engineering and Medical Physics
(NBC), pages 145–148. Springer, 2017.
[78] Jorge Blasco and Pedro Peris-Lopez. On the feasibility of low-cost wear-
able sensors for multi-modal biometric verification. Sensors, 18(9):2782,
[79] Edgar Galido, Ma Carmina Esplanada, Christine Joy Estacion,
Jim Patrick Migri˜no, Joellah-Keren Rapisora, Joyce Salita, Timothy
Amadoa, Romeo Jorda, and Lean Karlo Tolentino. EMG speed-
controlled rehabilitation treadmill with physiological data acquisition
system using BITalino kit. In Proc. of the IEEE Int’l Conf. on Hu-
manoid, Nanotechnology, Information Technology, Communication and
Control, Environment and Management (HNICEM), pages 1–5, 2018.
[80] Tyler Przybylski, Patrick Froehle, Christopher McDonald, Milad
Mirzaee, Sima Noghanian, and Reza Fazel-Rezai. Wearable wireless
body area network for aeronautical applications. In Proc. of the IEEE
Int’l Conf. on Electro/Information Technology (EIT), pages 563–568,
[81] Tae-Yang Han, Se-Dong Min, and Yunyoung Nam. A real-time sleep
monitoring system with a smartphone. In Proc. of the Int’l Conf. on In-
novative Mobile and Internet Services in Ubiquitous Computing (IMIS),
pages 458–461, 2015.
[82] Krzysztof Kutt, Wojciech Binek, Piotr Misiak, Grzegorz J Nalepa, and
Szymon Bobek. Towards the development of sensor platform for process-
ing physiological data from wearable sensors. In Proc. of the Int’l Conf.
on Artificial Intelligence and Soft Computing (ICAISC), pages 168–178.
Springer, 2018.
[83] Carre Technologies Inc (Hexoskin). Health research - hexoskin bio-
metric shirt for remote monitoring.
moodmetric clinical research/, 2019 (accessed August 6, 2019).
[84] Moodmetric. Moodmetric technology shows great promise in identify-
ing stress levels in a work environment.
moodmetric clinical research/, 2019 (accessed August 6, 2019).
[85] Moodmetric. Moodmetric smart ring.,
2019 (accessed August 6, 2019).
[86] Jari Torniainen, Benjamin Cowley, Andreas Henelius, Kristian Lukan-
der, and Satu Pakarinen. Feasibility of an electrodermal activity ring
prototype as a research tool. In Proc. of the Annual Int’l Conf. of
the IEEE Engineering in Medicine and Biology Society (EMBC), pages
6433–6436, 2015.
[87] Healbe. Smart band healbe gobe2., 2019
(accessed August 6, 2019).
[88] Schaefer. Accuracy of gobe2tm smartband in estimating the
calorie intake of food.
1Pu1xFXp6K7fXS4D0IDksyD9VeB3YYiUW/view, 2018 (accessed Au-
gust 6, 2019).
[89] SentioSolutions. World’s first emotion sensor mental health advisor., 2019 (accessed June 1, 2019).
[90] PIP. Stress management device., 2019 (accessed
June 1, 2019).
[91] Massimiliano de Zambotti, Leonardo Rosas, Ian M Colrain, and Fiona C
Baker. The sleep of the ring: comparison of the ¯oura sleep tracker against
polysomnography. Behavioral sleep medicine, 17(2):124–136, 2019.
[92] Massimiliano de Zambotti, Aimee Goldstone, Stephanie Claudatos,
Ian M Colrain, and Fiona C Baker. A validation study of Fitbit charge
2TM compared with polysomnography in adults. Chronobiology interna-
tional, 35(4):465–476, 2018.
[93] Oura. Oura ring., 2019 (accessed August 6,
[94] SpireHealth. Spire health: Clinical-grade health monitoring and insights., 2019 (accessed August 6, 2019).
[95] Inc FitBit. Fitbit charge 3. https://
www.\MakeUppercase{F}it\MakeUppercase{b}, 2019
(accessed August 6, 2019).
[96] Kiana Lewis et al. Validation of Wearable Biofeedback Technology for
Heart Rate Tracking Via Reflective Photoplethymorgraphy. PhD thesis,
California State Polytechnic University, Pomona, 2017.
[97] Mark Holt, Ben Yule, Dylan Jackson, Mary Zhu, and Neema Moraveji.
Ambulatory monitoring of respiratory effort using a clothing-adhered
biosensor. In Proc. of the IEEE Int’l Symp. on Medical Measurements
and Applications (MeMeA), pages 1–6, 2018.
[98] PulseOn. Ohr tracker — pulseon.
tracker, 2019 (accessed August 6, 2019).
[99] Adrian Tarniceriu, Jarkko Harju, Zeinab Rezaei Yousefi, Antti Vehkaoja,
Jakub Parak, Arvi Yli-Hankala, and Ilkka Korhonen. The accuracy of
atrial fibrillation detection from wrist photoplethysmography. a study on
post-operative patients. In Proc. of the Annual Int’l Conf. of the IEEE
Engineering in Medicine and Biology Society (EMBC), pages 1–4, 2018.
[100] Ricard Delgado-Gonzalo, Jakub Parak, Adrian Tarniceriu, Philippe
Renevey, Mattia Bertschi, and Ilkka Korhonen. Evaluation of accu-
racy and reliability of pulseon optical heart rate monitoring device. In
Proc. of the Annual Int’l Conf. of the IEEE Engineering in Medicine
and Biology Society (EMBC), pages 430–433, 2015.
[101] Qardio. Smart wearable ecg ekg monitor - qardiocore. https:
iphone/, 2019 (accessed August 6, 2019).
[102] Qardio. Qardiocore: Wireless ecg monitor receives tga ap-
proval to launch in australia.
launch-australia/, 2019 (accessed August 6, 2019).
[103] KomodoTechnologies. Aio smart sleeve - ecg wearable heart rate vari-
ability monitor., 2019 (accessed August 6,
[104] VitalConnect. Vitalpatch - vitalconnect.
solutions/vitalpatch/, 2019 (accessed August 6, 2019).
[105] Scosche. Waterproof armband heart rate monitor — rhythm24.
rate-monitor, 2019 (accessed August 6, 2019).
[106] EmotiBit. Emotibit., 2019 (accessed Au-
gust 19, 2019).
[107] mPath. The moxo sensor - mpath.
the-moxo-sensor, 2019 (accessed August 6, 2019).
[108] OmronHealthcare. Heartguide — wearable blood pressure monitor
— omron.
wearable-blood-pressure-monitor-bp8000m/, 2019 (accessed Au-
gust 6, 2019).
[109] GoodParentsInc. Kiddowear., 2019 (ac-
cessed August 6, 2019).
[110] Bitbrain Technologies. Versatile bio.
equipment/products/versatile-bio, 2019 (accessed August 6, 2019).
[111] TaisonDigital. Bracelet - emotion sensing technology - mood app — up-
mood., 2019 (accessed August 6, 2019).
[112] Myant. Skiin connected apparel., 2019 (accessed
June 1, 2019).
[113] SkyLabs. Product — skylabs.,
2019 (accessed August 6, 2019).
[114] OmronHealthcare. Omron healthcare officially launches heart-
2019 (accessed August 6, 2019).
... Recently in 2020, Abreu, Fred et al [42] did a significant exploration on, wearables and related devices, that can be utilised for epilepsy prediction, the findings presents devices, some with multiple sensors, characterised with respect to their applicability to research, validation status, form factor or body positioning, battery duration, method to access the data, measured signals and, their applicability to epilepsy prediction (EP) [42]. This is a vital study since the devices have already been validated, and connectivity options identified, and since they are beneficial for epilepsy they can be proposed in the IoT based Epilepsy monitoring model. ...
... Recently in 2020, Abreu, Fred et al [42] did a significant exploration on, wearables and related devices, that can be utilised for epilepsy prediction, the findings presents devices, some with multiple sensors, characterised with respect to their applicability to research, validation status, form factor or body positioning, battery duration, method to access the data, measured signals and, their applicability to epilepsy prediction (EP) [42]. This is a vital study since the devices have already been validated, and connectivity options identified, and since they are beneficial for epilepsy they can be proposed in the IoT based Epilepsy monitoring model. ...
... The IoT based Epilepsy monitoring model proposal in Fig.9 shows areas on the body where parameters are measured, each area is indicated with a colour matching the parameter. The chosen devices have been proposed based upon the factors in the study by Abreu, Fred et al [42] for the best battery life, validity, and connectivity options selected for ease of connection to the cloud platform in the IoT based Epilepsy monitoring model. Since Embrace2 device uses its own onboard processing it is not adaptable for the model proposed in this study. ...
... This requires non-intrusive monitoring techniques that can be implemented in ambulatory and real-life environments. Arguably, wearable biomedical sensors are a distinct answer to this prominent need, designed to be seamlessly integrated in everyday use garments or as body-worn accessories [3]. Moreover, several of these devices are already pervasive in our everyday lives, in the form of smartwatches that monitor Photoplethysmography (PPG) and Electrocardiography (ECG) in every-day activities, Electroencephalography (EEG)headbands, and numerous other devices, as explored in [3]. ...
... Arguably, wearable biomedical sensors are a distinct answer to this prominent need, designed to be seamlessly integrated in everyday use garments or as body-worn accessories [3]. Moreover, several of these devices are already pervasive in our everyday lives, in the form of smartwatches that monitor Photoplethysmography (PPG) and Electrocardiography (ECG) in every-day activities, Electroencephalography (EEG)headbands, and numerous other devices, as explored in [3]. ...
Conference Paper
Epilepsy is a neurological disease that affects about 50 million people worldwide. It is a disorder of the central nervous system, characterized by recurrent seizures that can have a massive impact in the physical and mental health of the people who suffer from it, as well as their loved ones. Long-term monitoring of epilepsy in uncontrolled environments is key to provide accurate characterization of the disease, and to create tools that improve the patients’ lives. Although some wearable devices (particularly with motion and cardiac-based sensors) are quickly gaining ground as everyday use monitoring devices, in the scope of epilepsy, electroencephalography (EEG) remains the gold-standard. Therefore, it is of utmost importance to include this modality in ambulatory settings, leveraging its extensive presence in literature and available databases. Nevertheless, in long-term recordings, having information about the onset of the seizure is of utmost importance for effective analysis of the collected data. Hence, this work explores the use of a single-channel, non-intrusive, EEG configuration in automatic seizure detection, with the purpose of event annotation in long-term recordings. This is a key element to the creation of multimodal datasets that can be used in seizure detection and, eventually, prediction, as well as towards comprehensive multimodal epilepsy monitoring techniques. A seizure-specific Support Vector Machines (SVM) classifier was designed for labeling eight different types of seizure, using a limited-channel configuration (Fp1-Fp2). Our work uses the TUH EEG Seizure Corpus, for which encouraging results were achieved for tonic-clonic and myoclonic seizures, with sensitivities of 98.9% and 98.2%, as well as precisions of 100% and 99.8%, respectively.
... Contentment, Sadness or Relief) [18]. Overall, the EDA and PPG sensors are predominant in wearable systems for research in Affective Computing [11,19,20]. ...
Full-text available
The availability of low-cost biomedical devices has driven a growing interest in the use of physiological signals for mental and emotional health research. Due to their potential for integration in discrete wearable form factors, Photoplethysmography (PPG) and Electrodermal Activity (EDA) are particularly popular, especially in out-of-the-lab experiments. Although high-resolution data acquisition should be a priority, the sampling rate can greatly affect the power consumption and memory storage of the devices in long-term recordings. Moreover, systems with shared computational resources that simultaneously monitor different signals, can also have communication channel bandwidth constraints that limit the sampling rate. This work seeks to evaluate how the sampling rate and interpolation affect the signal quality of PPG and EDA signals, in terms of waveform morphology and feature extraction capabilities. We study the minimum sampling rate requirements for each signal, as well as the impact of interpolation methods on signal waveform reconstruction. Using a previously recorded dataset with signals originally recorded at 1 kHz, we simulate multiple lower sampling rates. Results show that for PPG a 50 Hz sampling rate with quadratic or cubic interpolation can achieve a temporal resolution identical to that of a 1 kHz acquisition, while for EDA the same can be said but with a 10 Hz sampling rate. Other recommendations are also proposed depending on the signal application.
... Devices found in the state of the art, characterised with respect to validation status, body positioning, battery duration, communication protocol, measured biosignals, and maximum sampling rate[13] ...
Full-text available
In the latter years, we have been observing a growth in wearable technology for personal use. However, an analysis of the state of the art for wearable technology shows that most devices perform data acquisition from individual subjects only, relying on communication technologies with drawbacks that prevent their use in collective real-world scenarios (e.g. a cinema, a theatre, and related use cases). When analysing the emotional response in groups, two types of emotions appear: individual (influenced by the group) and group-based emotions (towards the group as an identity). To fill the existing gap, we propose a biocybernetic engine for real-time data acquisition of multimodal physiological data in real-world scenarios. Our system extends the state of the art with: (1) real-time data acquisition for the signals being acquired (20 devices at 25 Hz; 10 devices at 60 Hz); (2) creation of a standalone local infrastructure with end-user interface for monitoring the data acquisition; (3) local and cloud-based data storage. We foresee that this platform could be the basis for the creation of large databases in diverse real-world scenarios, namely health and wellbeing, marketing, art performances, and others. As a result, this work will greatly contribute to simplify widespread biosignals data collection from unobtrusive wearables. To evaluate the system, we report a comprehensive assessment based on a set of criteria for data quality analysis.
... c sensors (Bota et al., 2019). Although many have been developed and validated (Abreu, 2020), for group settings, such as theatre or cinema, there is a need in the literature for a device able to perform collective data collection from several devices synchronously. Therefore, in this work: (1) We introduce the Xinhua Net Future Media Convergence Institute (FMCI) device containing a Galvanic Skin Response (GSR) embedded sensor, describing the hardware development path throughout its various versions. ...
Conference Paper
Full-text available
In the recent years, we have been observing an increase of research work involving the use of biomedical data in affective computing applications, which is ever more dependent on data and its quality. Many physiological data acquisition devices have been developed and validated. However, there is still a need for pervasive and unobtrusive equipment for collective synchronised acquisitions. In this work, we introduce a novel system, the Electrodermal Activity (EDA) Xinhua Net Future Media Convergence Institute (FMCI) device, allowing group data acquisitions, and benchmark its performance using the established BITalino as gold standard. We developed a methodical experimental protocol in order to acquire data from the two devices simultaneously, and analyse their performance over a comprehensive set of criteria – Data Quality Analysis. Additionally, the FMCI data quality is assessed over five different setup scenarios towards its validation in a real-world scenario – Data Loss Analysis. The experimental results show a close similarity between the data collected by both devices, paving the way for the application of the proposed equipment in simultaneous, collective data acquisition use cases.
Full-text available
Brain-machine interfaces (BMIs) hold promise for the restoration of sensory and motor function and the treatment of neurological disorders, but clinical BMIs have not yet been widely adopted, in part because modest channel counts have limited their potential. In this white paper, we describe Neu-ralink's first steps toward a scalable high-bandwidth BMI system. We have built arrays of small and flexible electrode "threads", with as many as 3,072 electrodes per array distributed across 96 threads. We have also built a neurosurgical robot capable of inserting six threads (192 electrodes) per minute. Each thread can be individually inserted into the brain with micron precision for avoidance of surface vasculature and targeting specific brain regions. The electrode array is packaged into a small implantable device that contains custom chips for low-power on-board amplification and digitiza-tion: the package for 3,072 channels occupies less than (23 × 18.5 × 2) mm 3. A single USB-C cable provides full-bandwidth data streaming from the device, recording from all channels simultaneously. This system has achieved a spiking yield of up to 85.5 % in chronically implanted electrodes. Neu-ralink's approach to BMI has unprecedented packaging density and scalability in a clinically relevant package.
Full-text available
Background and Objective: Heart rate variability parameters are studied by the research community as potential valuable indices for seizure detection and anticipation. This paper investigates heart activity abnormalities during focal epileptic seizures in childhood. Methods: Seizures affect both the sympathetic and parasympathetic system which is expressed as abnormal patterns of heart rate variability (HRV) parameters. In the present study, a clinical dataset containing 42 focal seizures in long-term electrocardiographic (ECG) recordings from drug-resistant pediatric epileptic patients (with age 8.2±4.3 years) was analyzed. Results: Results indicate that the time domain HRV parameters (heart rate, SDNN, standard deviation of heart rate, upper envelope) and spectral HRV parameters (LF/HF, normalized HF, normalized LF, total power) are significantly affected during ictal periods. The HRV features were ranked in terms of their relevance and efficacy to discriminate non-ictal/ictal periods and the top-ranked features were selected using the minimum Redundancy Maximum Relevance algorithm for further analysis. Then, a personalized anticipation algorithm based on multiple regression was introduced providing an “epileptic index” of imminent seizures. The performance of the system resulted in anticipation accuracy of 77.1% and an anticipation time of 21.8 sec. Conclusions: The results of this analysis could permit the anticipation of focal seizures only using electrocardiographic signals and the implementation of seizure anticipation strategies for a range of real-life clinical applications.
Conference Paper
Full-text available
EMG speed-controlled rehabilitation treadmill (BitAid) uses the muscle activity at lower extremities to determine the speed of the treadmill necessary for the rehabilitation of the patient. It targets on aiding patients with lower extremity problems and neurologic cases like after-stroke. The biomedical equipment used is a plugged type BITalino Kit which can determine electrocardiogram (ECG), electrodermal activity (EDA), electromyogram (EMG) and balance using accelerometer (ACC). The data from these sensors are displayed in the tablet attached in front of the patient allowing them to see the reaction of their body during rehabilitation. A suit was designed for placing the electrodes to several points on the patient's body. The overall performance of the device has been evaluated and shown excellent ratings on its necessity, reliability, stability, user-friendliness and ease of operation.
Full-text available
The low-cost multimodal platform BITalino is being increasingly used for educational and research purposes. However, there is still a lack of well-structured work comparing data acquired by this toolkit against a reference device, using established experimental protocols. This work intends to fill said gap by benchmarking the performance of BITalino against the BioPac MP35 Student Lab Pro device. We followed a methodical experimental protocol to acquire data from the two devices simultaneously. Four physiological signals were acquired: Electrocardiography, Electromyography, Electrodermal Activity and Electroencephalography. Root mean square error and coefficient of determination were computed to analyze differences between BITalino and BioPac. Electrodermal activity signals were very similar for the two devices, even without applying any major signal processing techniques. For electrocardiography, a simple morphological comparison also revealed high similarity between devices, and this similarity increased after a common segmentation procedure was followed. Regarding electromyography and electroencephalography data, the approach consisted of comparing features extracted using common post-processing methods. The differences between BITalino and BioPac were again small. Overall, the results presented here show a close similarity between data acquired by the BITalino and by the reference device. This is an important validation step for all researchers working with this multimodal platform.
Full-text available
The goal of real-time feedback on physiological changes, stress monitoring and even emotion detection is becoming a technological reality. People in their daily life experience varying emotional states, some of which are negative and which can lead to decreased attention, decreased productivity and ultimately, reduced quality of life. Therefore, having a solution that continuously monitors the physiological signals of the person and assesses his or her emotional well-being could be a very valuable tool. This paper aims to review existing physiological and motional monitoring devices, highlight their features and compare their sensing capabilities. Such technology would be particularly useful for certain populations who experience rapidly changing emotional states such as people with autism spectrum disorder and people with intellectual disabilities. Wearable sensing devices present a potential solution that can support and complement existing behavioral interventions. This paper presents a review of existing and emerging products in the market. It reviews the literature on state-of-the-art prototypes and analyzes their usefulness, clinical validity, and discusses clinical perspectives. A small number of products offer reliable physiological internal state monitoring and may be suitable for people with Autism Spectrum Disorder (ASD). It is likely that more promising solutions will be available in the near future. Therefore, caregivers should be careful in their selection of devices that meet the care-receiver’s personal needs and have strong research support for reliability and validity.
Background So far, only generalized tonic-clonic seizures can be reliably detected with non-invasive wearable devices. We aimed to develop an automated seizure detection algorithm using a wearable ECG-device for detecting both GTC and focal seizures. Material and methods We recorded ECG using a dedicated wearable device (ePatch®) during long-term video-EEG monitoring. In this phase-2 clinical study, 100 patients were prospectively recruited; 43 of the patients had 126 seizures (108 focal, 18 GTC) of >20 s duration during recording (941 h training data, 2238 h test data). We analyzed 20 heart rate variability (HRV)-parameters and 6 combinations of these using either 50 or 100 R-R intervals sliding window with maximum overlapping. Each HRV-parameters cut-off value for seizure-alarm was set to 105% of the highest non-seizure period during training data of the same patient. Positive responders of seizure detection were defined, for each HRV-parameter, as patients with >66% of seizures detected. Results In total, 53.5% of the patients were responders for the best performing algorithm. In these patients, the method achieved a sensitivity of 93.1% and false detection rate of 1.1/day. An average of >50 beats/minute HR increase or decrease during seizure(s) is a positive predictor of being a responder of seizure detection (PPV: 87.0%, NPV: 90.0%), making it easy to define for which patients a reliable seizure alarm is feasible. Conclusions High sensitivity and low false positive alarm rates can be achieved with our algorithm analyzing ECG-signals using the wearable device in persons with average HR changes >50 beats/min during seizures.
The goal of this study was to evaluate and summarize the current literature on multimodal changes of the autonomic nervous system (ANS) in people with epilepsy (PWE). We included studies reporting ANS measures of at least two modalities and with a minimum of one group of people with epilepsy. We screened two hundred eighty-three abstracts and sixty-six full texts, of which twenty-two met our inclusion criteria. Eleven studies reported ictal and interictal cardiac and respiratory changes. Three studies investigated the correlation between cardiac and respiratory markers, whereby two found no correlation and one showed a relation. Six studies evaluated electrodermal and cardiac parameters and showed effects on both ANS subsystems that jointly indicate a shift toward increased sympathetic activity for people with epilepsy during rest and during activity. Two studies assessed three modalities and reveal epilepsy-related alterations within the ANS. In summary, there is a growing interest in multimodal monitoring approaches, such as combining at least two ANS modalities, to describe epilepsy-related changes in ANS activity and to test for the potential to use ANS markers for seizure detection and prediction. Most studies report multiple unimodal analyses while only few studies analyzed multimodal patterns. Patterns of changes depend on the type of epilepsy and differ on an individual level; therefore, a multimodal approach might offer an approach to more individualized monitoring and, ultimately, management.
Background: Emotion is composed of cognitive processing, physiological response and behavioral reaction. Heart rate variability (HRV) refers to the fluctuations between consecutive heartbeat cycles, and is considered as a non-invasive method for evaluating cardiac autonomic function. HRV analysis plays an important role in emotional study and detection. Objective: In this paper, the physiological foundation of HRV is briefly described, and then the relevant literature relating to HRV-based emotion studies for the performance of HRV in different emotions, emotion recognition, the evaluation of emotional disorders, HRV biofeedback, as well as HRV-based emotion analysis and management enhanced by wearable devices, are reviewed. Significance: It is suggested that HRV is an effective tool for the measurement and regulation of emotional response, with a broad application prospect.
Introduction Photoplethysmography (PPG) is used as a surrogate of electrocardiograms (ECG) for heart rate variability (HRV) analysis or respiratory rate monitoring. PPG is a more convenient way to measure HRV than ECG at rest, since respiration could be a confounding factor in HRV evaluation. However, it remains unclear whether or not controlled breathing affects breath-volume and breathing rate when HRV and pulse rate variability (PRV) are measured in different situations. Consciously controlled breathing was performed to alter the autonomic nervous states of subjects caused by respiratory sinus arrhythmia (RSA). The aim of this study was to investigate the coherence between parameters derived from ECG and PPG on healthy subjects with/without controlled breathing. Method With 30 healthy volunteers, we measured their respiratory frequency and recorded their ECG and PPG signals during spontaneous breathing and controlled breathing, including natural paced breathing, rapid and deep breathing, slow and deep breathing, rapid and shallow breathing, and slow and shallow breathing. Results Obvious coherence was observed between pulse rate and heart rate in both spontaneous breathing and controlled breathing tasks. However, a comparison of PRV and HRV indices demonstrated significant differences during controlled breathing. The results based on time domain and nonlinear method analysis showed that the frequency-dependent changes have more of an impact. The results also indicated that breathing corresponded well in ECG-derived parameters comparing with PPG-derived ones. Conclusion We concluded that PPG-based devices cannot be applied as a precision screening tool to detect HRV, particularly during the cardiopulmonary analysis for the controlled breathing maneuver.
Wearable automated seizure detection devices offer a high potential to improve seizure management, through continuous ambulatory monitoring, accurate seizure counts, and real-time alerts for prompt intervention. More importantly, these devices can be a life-saving help for people with a higher risk of sudden unexpected death in epilepsy (SUDEP), especially in case of generalized tonic-clonic seizures (GTCS). The Embrace and E4 wristbands (Empatica) are the first commercially available multimodal wristbands that were designed to sense the physiological hallmarks of ongoing GTCS: while Embrace only embeds a machine learning-based detection algorithm, both E4 and Embrace devices are equipped with motion (accelerometers, ACC) and electrodermal activity (EDA) sensors and both the devices received medical clearance (E4 from EU CE, Embrace from EU CE and US FDA). The aim of this contribution is to provide updated evidence of the effectiveness of GTCS detection and monitoring relying on the combination of ACM and EDA sensors. A machine learning algorithm able to recognize ACC and EDA signatures of GTCS-like events has been developed on E4 data, labeled using gold-standard video-EEG examined by epileptologists in clinical centers, and has undergone continuous improvement. While keeping an elevated sensitivity to GTCS (92-100%), algorithm improvements and growing data availability led to lower false alarm rate (FAR) from the initial ˜2 down to 0.2-1 false alarms per day, as showed by retrospective and prospective analyses in inpatient settings. Algorithm adjustment to better discriminate real-life physical activities from GTCS, has brought the initial FAR of ˜6 on outpatient real life settings, down to values comparable to best-case clinical settings (FAR < 0.5), with comparable sensitivity. Moreover, using multimodal sensing, it has been possible not only to detect GTCS but also to quantify seizure-induced autonomic dysfunction, based on automatic features of abnormal motion and EDA. The latter biosignal correlates with the duration of post-ictal generalized EEG suppression, a biomarker observed in 100% of monitored SUDEP cases.