Abstracts

Rationale: The purpose is to compare surface-EMG (sEMG) signals from electrodes placed on the biceps muscle to sEMG artifacts recorded in EEG electrodes during generalized tonic-clonic seizures (GTCS). sEMG signals recorded from the biceps have been used successfully by automated systems to detect GTCS. It is unknown if EMG signals in EEG electrodes can be used for automated detection of GTCS. Methods: These data were retrospectively sampled from 201 subjects that were enrolled in a prospective study while receiving routine clinical care in one of 11 NAEC level IV centers. The primary aim of the study was to evaluate the clinical effectiveness of the Brain Sentinel GTCS detection device. Video-EEG recordings were independently evaluated by three board certified epileptologists to identify GTCS and define the lengths of tonic and clonic phases. Automated processes were used to determine the length of tonic and clonic phases using sEMG and EEG recorded during events. To date, sEMG and EEG results have been compared to the epileptologists' results for six events. A full analysis of a greater number of events will be provided at the time of the conference. The Brain Sentinel device recorded sEMG data at at sampling rate of 990Hz. Various EEG amplifiers were used across sites and three sampling rates were used for EEG collection: 200Hz, 256Hz, and 1024Hz. In keeping with the Nyquist limit for frequency evaluation, the highest frequency we used for evaluation across all recorded signals was 100Hz. Clonic duration of a GTCS seizure was evaluated from signal amplitude increases in a range of 30-60Hz. The lower-frequency band cutoff was selected to avoid interference from EEG signals. To determine tonic duration, total seizure duration was quantified using the root mean square of the baseline signal, and the clonic seizure time was subtracted from the total seizure duration (Figure 1). A global threshold value was set to determine the start and stop of each phase. The continuous wavelet transform with the Morlet mother wavelet was used with varying coefficients to provide a 1-100Hz wavelet decomposition. Chebychev high pass filters (2Hz) were used to remove DC offsets in EEG signal. Statistical significance was tested using paired T-Test with Bonferroni correction. As a result of our phase calculation methods, tonic-clonic overlap could not be estimated for EEG or sEMG. Results: GTCS phase duration was evaluated from video review, EEG, and sEMG. There was no significant difference between the duration of the tonic and clonic phases as calculated by the automated wavelet analysis algorithm within the EEG versus the EMG signals. There was also no significant difference between the results of the automated wavelet analysis and the independent neurologist's review. Results are summarized in Figure 2. Conclusions: In the small number of EEG/EMG recordings analyzed, the total seizure time and clonic duration in both the EEG and sEMG data was not significantly different from the independent neurologist's review. Generally, this shows that the EMG signal recorded in EEG may be as useful as that recorded in sEMG of the biceps muscle for quantifying GTC seizure phase durations. These results suggest that EMG artifact in EEG may be useful for assessing some aspects of seizure semiology and for automated detection of GTCS. Funding: No funding was received in support of this abstract.