Abstracts

Rationale: It is estimated that 25% of patients previously diagnosed with epilepsy who are not responding to anti-epileptic drugs are found to have PNES[1]. The Brain Sentinel® Monitoring and Alerting System is an FDA cleared device that records surface electromyography (sEMG) that may be related to seizure activity, and is designed to alarm for generalized tonic-clonic seizures (GTCS). Using frequency and amplitude analyses, the utility of this device can be expanded to provide diagnostic information that may help in the differentiation between captured GTCS and PNES recordings.[1] - Benbadis SR. “The problem of psychogenic symptoms: is the psychiatric community in denial?”. Epilepsy and Behav 2005;6:9-14 Methods: In a multicenter, phase III trial of the sEMG monitoring device with concomitant vEEG recording, 29 GTC seizures and 11 motor PNES were classified by three ABPN certified epileptologists (via vEEG) while patients were wearing the sEMG recording device. Of these events, 1 of the motor PNES events had electrode issues that affected the sEMG recording. The sEMG for the remaining 10 PNES and 10 randomly selected GTCS were divided into 2 minute epochs for feature extraction and analysis.The primary analysis technique involved the transformation of the sEMG data into frequency-driven traces. The two traces were summated wavelet-transform output coefficients from high frequency, 150-260Hz, and low frequency, 6-70Hz, ranges (Figure 1).An algorithm was developed for extracting features from the traces, most notably the area under the curve (AUC). The AUC of each the high and low frequency traces, as well as the ratio of high/low AUCs, were calculated for the regions within each epoch that met the criteria of being above a threshold. The threshold was determined using two methods: 1) calculating the standard deviation of a baseline region within each epoch and multiplying by 3; 2) using a set value (1000 a.u.) that was selected by visual inspection. The high and low AUC ratios were calculated for both thresholds. These values were compared using unpaired student’s t-test with bonferoni correction and found to be significant if p < .05. Results: The AUC for the ratio between the high/low frequency traces for GTCS and PNES were calculated for two different threshold settings. Using threshold method 1, the AUC ratios were not significantly different between GTCS (0.26 ± .09 a.u.) and PNES (.06 ± .02 a.u.) events (p = .056). Using method 2, the AUC ratios were found statically different between GTCS (.26 ± .09 a.u.) and PNES (.002 ± .002 a.u., p = .018, Figure 2). Conclusions: The sEMG data from the Brain Sentinel®’s device adequately expresses energy from muscle activity during GTCS and PNES events. The most impactful attribute was the absence of high frequency data in PNES events, with a strong presence in GTCS, with predominance in the tonic phase. This feature, along with others such as duration, will be placed into an artificial neural network for enhanced differentiation. Funding: Funded by Brain Sentinel