Abstracts

Rationale: Detection of faciobrachial dystonic seizures (FBDS) in patients with Leucine-rich glioma inactivated-1 (LGI1-IgG) associated autoimmune encephalitis can be challenging due to high event frequency and lack of a robust EEG correlate. Wearable devices may provide a convenient, noninvasive means to measure these events and provide an objective seizure diary in this condition. This study sought to assess whether FBDS can be identified and differentiated from normal arousals reliably using a wrist-worn device.

Methods: Two LGI1-IgG seropositive patients and four control subjects, wearing a wrist-worn device (Empatica E4, Empatica Inc., Italy), were recruited for the study. Two eight-hour sleep periods of each patient, one pre- and one post-treatment, were recorded. The wearable device collected accelerometry (ACC), electrodermal activity (EDA), and other physiological signals. An algorithm was developed based on ACC signals to detect events of interest. After event identification, we calculated the ACC magnitude and duration of the detected events, as well as EDA characteristics.

Results: Statistically significant differences in wearable signal characteristics (example in Figure 1) were found before and after treatment, and between FBDS events and nocturnal arousals in control subjects.  We estimated the frequency of actigraphy events 13.93±3.82 ev/h in pre- and 9.66±3.37 ev/h in post-treatment sleep period, for the first LGI1 patient, and 26.21±13.16 ev/h in pre- and 29.06±9.79 ev/h in post-treatment for the second patient, with 3.47±2.77 ev/h for the control group. For ACC magnitude of the detected events (Figure 2, Panel A), significant differences of means between controls and the FBDS events were observed before treatment (Pre1 and 2; Pre: pre-treatment). While no significance was found between controls and Post1, it was found between controls and Post2. Significant differences were also found between Pre2 and Post2. For event duration (Figure 2, Panel B), we also found significant differences between controls and the FBDS events before the treatment (Pre1 and 2). Significant differences were found between controls and Post1, but not Post2. Of note, for both ACC magnitude and event duration, the medians were lower after treatment. For EDA activities, both tonic and phasic (Figure 2, Panel C and D) components for controls were significantly lower than those for patients before treatment (Pre1 and 2), but not all of them were significant between controls and post treatment (Post1 and 2; Post: post-treatment). We also observed both means and medians decreased after treatment in both patients. Put together, the results of actigraphy and EDA indicate changes of the measurements following treatment.

Conclusions: Robust characteristics were demonstrated in signals recorded with a wearable device during FBDS events, supporting the idea that an accurate detector can be developed to automatically identify FBDS to provide objective evidence for treatment guidance and clinical decisions. Future work will expand the number of subjects involved in the study and further validate detected events using video EEG.

Funding: EFA Seizure Gauge and Mayo Kern Center for the Science of Health Care Delivery