Conference PaperPDF Available

Improvement of generalized tonic-clonic seizure detection by Embrace wristband on inpatient and outpatient data

G. Regalia1, C. Caborni1, M. Migliorini1,R.Picard1,2, F. Onorati1
Materials and Methods
Results Conclusions
is an FDA-cleared wrist-worn device that senses
(ACM) and
electrodermal activity
(EDA) to detect and
generalized tonic-clonic seizures
settings, Embrace’s machine-learning algorithm has shown
higher than 90%
false alarm rates
(FAR) as low as
0.2/24 hrs
settings, algorithm adjustments taking into account
real-life movements
lowered the FAR to levels
for the majority of patients and caregivers
Doctors, patients and families can also benefit from an automated offline GTCSs detection algorithm for
treatment management
, for example to accurately and automatically fill in a
digital diary
offline medical review
Figure 1
For this application, where high specificity is necessary, we show that it is possible
to increase the specificity
reducing the FAR,
by leveraging on
offline processing
AIM: To quantify the increase in specificity attainable by a GTCS detection algorithm optimized
for offline applications, such as diagnostics or treatment management.
For the first time, the performance of GTCS detection algorithms
optimized for real-time alerting, have been compared with a new GTCS
detection algorithm optimized for
offline processing
, optimized for
treatment management
Increasing the quantity and the quality of the information, it was
possible to
reduce the FAR by 63%
data, and
by 55%
(Sens =91%)
44%(Sens =95%)
data, with respect to
the FDA cleared online/embedded DA2 (
Table 2
At Sens = 91%, the new DA3 achieved 0 false alerts for 136 out of 159
inpatients (85%). This is an improvement over the FDA-cleared DA2
that achieved zero FA for 108 out of 159 (68%) of patients (
Figure 3
ratio #FA/#SZ
, a metric related to the specificity of the algorithm,
also improves, droppping to 1 (DA3at Sens = 91%on TS1) and below 0.5
(0.44,DA3on TS2) (
Table 2
Future work
will focus on improving the offline algorithm to reach
FAR = 0
,in order to embed the algorithm into a diagnostics and
treatment management framework.
Prospective analyses
were conducted on
2testing sets
,labeled as follows:
TS1 (inpatient)
GTCSs were labeled by three independent epileptologists by video EEG inspection.
TS2 (outpatient)
User reported GTCSs were reviewed by 2 data experts. Real-life activities prone to
generate false alarms are numerous in this dataset.
3Detection Algorithms
(DAs) were used for performance comparisons:
-an online DA, which has showed promising ambulatory performances
-an online DA, cleared by FDA for monitoring GTCSs during rest for patients
ages 6 and up
- a novel offline DA.
Improvement of generalized tonic-clonic seizure detection by Embrace wristband
on inpatient and outpatient data
1. Empatica, Inc, Boston, MA, US; 2. Massachusetts Institute of Technology, Cambridge, MA, US.
Regalia et al. (2019). Multimodal wrist-worn devices for seizure detection and
advancing research: Focus on the Empatica wristbands. Epilepsy Res. 153:79-82.
Onorati et al. (2018) Performance of awrist-worn multimodal seizure detection
system for more than a year in real-life settings. European Congress on Epileptology.
510(k) Premarket Notification Embrace 12/20/2018.(n.d.). Retrieved July 28,2019,
Francesco Onorati
Rosalind Picard
Giulia Regalia
hours #Sbj
Sbj w
10,368 159 32 56 X X
1,502 32 8 53 X X X
Table 1
Testing sets evaluated using three detection algorithms.
Figure 3.
Comparison of
false alert related metrics
between DA2 (green) and
DA3 (purple) on TS1
(inpatient data).
Left Column
- histograms of
number of subjects with each
range of FAR;
Right column
histograms of number of
subjects with each range of
number of false alarms
experienced (#FA) and
average recording hours of
recording for each bin.
Top Row
-FAR, #FA and
average recording hours per
bin for DA2 (light green) and
DA3 (light blue) with Sens =
bottom row -
and average recording hours
per bin for DA2(dark green)
and DA3(dark blue) with
Sens = 95%.
51 53 51 53
0.12 0.3
115 246 51 127
4.6 12.4
91% 91%
48 48 48
1.30 0.83 0.31
88 56 21
1.83 1.16 0.44
Table 2
Performance comparisons on TS1(DA2 and DA3
only) and TS2(DA1, DA2 and DA3), at Sens = 91% - 95% and
at Sens = 91%, respectively (#SZ = number of seizures
detected, #FA = number of false alerts).
Figure 1.
Diagram of a diagnostic/treatment management system based on an offline GTCS detection algorithm.
Figure 2.
Sensitivity and FAR for the 3 detection algorithms (DAs) on the 2
separate testing sets used for performance evaluation. In orange, DA1, in green
DA2, in purple DA3. Circle markers for test set TS1, square markers for TS2.
... Cogan et al. [12] also analyzed several biosignals, including heart rate and blood oxygen level, in a clinical setting and concluded that these signals are sufficient for accurate seizure detection. Another study [27], focusing on generalized tonic seizures, asked 40 subjects with a risk of generalized seizures to wear the empatica wrist band. They used the data from the accelerometer and EDA to detect seizures. ...
Conference Paper
Full-text available
Purpose The Embrace smartwatch combines accelerometer and electrodermal activity sensors to alert to generalized tonic-clonic seizures (GTCSs). While a multi-site clinical study validated 94.5% sensitivity (Se) of the GTCS detector with a false alarm rate (FAR) typically less than 0.2/day (Onorati et al, Epilepsia 2017), longitudinal analysis of Embrace performance is needed to understand its reliability and robustness in outpatient settings. A first analysis on ~1 year of data from one user showed Se=98% and FAR per day worn=0.14 (Regalia et al, Mobil 2017, Copenhagen). Here, we report on new longitudinal data from three users. Method One male (P1, 29y) and two female (P2 and P3, 15y) patients used Embrace from 04/2016 to 12/2017 in real-life. Each Embrace alarm generated a call to patients’ caregivers, and it was logged on an online database. Patients and caregivers were instructed to carefully mark any false alarms and any missed seizures. Two seizure-data experts independently inspected all the recordings to verify the accuracy of users’ self-reports. Results Overall, 618 (P1), 437 (P2) and 554 (P3) recording days were collected (average hours per day=23, 19 and 21, respectively). For P1, all 201 GTCSs were detected (Se=100%), with FAR=0.11. For P2, all 64 GTCSs were detected (Se=100%), with FAR=0.32. For P3, 63 out 65 GTCSs were detected (Se=97%), with FAR=0.12. Nocturnal seizures represented 87%, 92% and 43%, respectively, of each patients’ GTCS and were all detected except for one GTCS experienced by P3. No false alarms happened during sleep. Conclusion For the first time, the real-life long-term performance of a wearable GTCSs detector, i.e. Embrace, was evaluated for multiple users, each collecting data for more than 1.5 years, while diligently cross-checking each alarm with self-reports. Embrace provided an excellent trade-off between Se and FAR for all the testers. Analyses on more patients are underway.
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.