Seizure detection accuracy



We are constantly validating and improving our algorithm, to test its accuracy we collaborate with top hospitals.


On July 7  2017, at the 1st International Congress on Mobile Health Devices and Seizure Detection in Epilepsy in Copenhagen,  we presented a study that we conducted with a beta tester, who was monitored for one year (5968 hours: 249 days) with our algorithm set at 98% sensitivity. The results showed that 45 out of 46  seizures were detected, with a false alarm rate of  0.14 per day, which amounts to around 1 every 6 to 7 days.


Two months later, at the 32nd International Epilepsy Congress we presented another study in which 27 patients were monitored for 95 days (2286 hours), they experienced 111 seizures in total (65 during periods of rest). With a sensitivity of 89%, the algorithm detected 99 seizures with a false alarm rate of  0.58 per day. When we increased the sensitivity to 93% (Rest Mode ON), 103 seizures were detected maintaining the same number of false alarms.


The full details of this work are included in these articles:


Regalia, G., Caborni, C., Migliorini, M., Onorati, F., Picard, R. (2017).  Real-time seizure detection performance with Embrace alert system: one-year real-life setting case study. International Congress on Mobile health devices for seizure detection 2017.


Caborni, C., Migliorini, M., Onorati, F., Regalia, G., Picard, R., (2017).  Tuning decision thresholds for active/rest periods significantly improves seizure detection algorithm performance: An evaluation using Embrace Smartwatch on outpatient settings. 32nd International Epilepsy Congress 2017.


Have more questions?