Embrace, a wearable convulsive seizure detection and alert system
First performance report of a case study in real-life settings
R. Picard1,2, G. Regalia1, C. Caborni1, M. Migliorini1, F. Onorati1
1 Empatica, Inc, Cambridge, MA and Milan, Italy; 2 MIT Media Lab, Massachusetts Institute of Technology, Cambridge (MA), US
•Empatica’s Embrace is acomfortable wristband wearable convulsive seizure (CS) detection system (Figure 1).
•It is based on accelerometer (ACC) and electrodermal activity (EDA) data and works with asmartphone app, Alert.
•Embrace with Alert app is amedical device (CE) in the European Union.Alert is available only through aclinical trial in the
•In EMU settings, the system has achieved Sens =92-100%and false alarm rate (FAR) = 0.15-2.02 false alarms per
•EMU settings may differ from the real-life environment.
•We have recently shown that adding outpatient data to the EMU-based training set allows to attain better performance2.
1. Apatient with Dravet Syndrome (14y) wore the Embrace 113 days for 1,973 hours (daily worn time =17.2±5 hrs/day).
2. This patient’s data were not used to train the classifier.
3. As aground truth the patient’s caregiver was asked to meticulously annotate the occurrence of each CS.Alert without
aCS were labeled as afalse alarm (FA).
•The system detected 22 out of 24 recorded CSs (Sens =92%), with adelay from 15-64 seconds (38 sec avg)between
the onset and the alert, and 6-57 seconds (11 sec avg)between the alert and call.
•The 2missed seizures were characterized by alow EDA and amild brief clonic component (Figure 3A).
•FAR = 0.35 false alarms per day worn,with atotal of 39 FAs.In 88 days out of 113 there were no FAs.In only 2days
out of 113 were there more than 2FAs (Figure 2B).
•We reported the performance of Embrace+Alert system used for 3months in real-life setting.
•The performance,both Sens and FAR, mirrored the results obtained in EMU settings.
•All seizures during sleep were detected (Figure 2A).No FA occurred during night.
•Caregiver reported that FAs were generated by activities like hands shaking,car transport and dancing.
To present the first case study about long-term Embrace recordings and alerts in real-life settings.
Francesco Onorati: firstname.lastname@example.org
Rosalind Picard: email@example.com , firstname.lastname@example.org
Figure 3. EDA (top) and ACM (bottom) for two CSs recorded with Embrace.The violet areas mark the seizures.Left(A):missed event.Right(B):detected
Figure 1. Schematics of Embrace convulsive seizure (CS) detection and alert system: a wristband detects an event, transmits an alert to the smartphone,
which generates acall via acloud-based service to designated caregivers.Bottom:Empatica’s Embrace.
Figure 2. Left (A):Distribution of patient’s seizures based on hour of the day.Right (B):Histogram of FAR over the 113 days of recording.
1 Regalia et al. "An improved wrist-worn convulsive seizure detector based on accelerometry and electrodermal
activity sensors”, 69th AES, December 4-8, 2015, Philadelphia, PA.
2 Onorati et al. "Improving convulsive seizure detection by exploiting data from outpatient settings using the
Embrace wristband”,12th European Congress on Epileptology, September 11-15, 2016, Prague.