Seizure detection algorithms based on EMG signals

Background: the currently used non-invasive seizure detection methods are not reliable. Muscle fibers are directly connected to the nerves, whereby electric signals are generated during activity. Therefore, an alarm system on electromyography (EMG) signals is a theoretical possibility. Objective: to show whether medical signal processing of EMG data is feasible for detection of epileptic seizures.

Methods: EMG signals during generalised seizures were recorded from 3 patients (with 20 seizures in total). Two possible medical signal processing algorithms were tested. The first algorithm was based on the amplitude of the signal. The other algorithm was based on information of the signal in the frequency domain, and it focused on synchronisation of the electrical activity in a single muscle during the seizure.

Results: The amplitude-based algorithm reliably detected seizures in 2 of the patients, while the frequency-based algorithm was efficient for detecting the seizures in the third patient. Conclusion: Our results suggest that EMG signals could be used to develop an automatic seizuredetection system. However, different patients might require different types of algorithms /approaches.