Evaluation of novel algorithm embedded in a wearable sEMG device for seizure detection

Abstract

We implemented a modified version of a previously published algorithm for detection of generalized tonic-clonic seizures into a prototype wireless surface electromyography (sEMG) recording device. The method was modified to require minimum computational load, and two parameters were trained on prior sEMG data recorded with the device. Along with the normal sEMG recording, the device is able to set an alarm whenever the implemented algorithm detects a seizure. These alarms are annotated in the data file along with the signal. The device was tested at the Epilepsy Monitoring Unit (EMU) at the Danish Epilepsy Center. Five patients were included in the study and two of them had generalized tonic-clonic seizures. All patients were monitored for 2-5 days. A double-blind study was made on the five patients. The overall result showed that the device detected four of seven seizures and had a false detection rate of 0.003/h or one in twelve days.