Abstracts

Rationale:
Focal and generalized seizures are often associated with changes in autonomic function that lead to changes in electrodermal activity (EDA). For this reason, EDA recordings have been evaluated for seizure detection. Despite the relevant contribution, specificity of EDA changes is low. We aimed to characterize fluctuations in EDA throughout a 24-hour cycle and hypothesized that diurnal and nocturnal rhythms may affect EDA activity.
Method:
Patients from Boston Children’s Hospital who had either focal impaired awareness (FIAS) or generalized/focal to bilateral tonic-clonic seizure (GTCS) while wearing an E4 device (Empatica®, Milan, Italy) during continuous video-EEG monitoring were included. Continuous electrodermal activity recordings were low pass filtered (cut of 0.4 Hz), smoothed, and had spikes eliminated. EDA level and EDA spectral power (0.01 to 0.3 Hz) were then averaged by hour, excluding hours in which seizures occurred. We performed a nonlinear mixed effects analysis (see Figure 1 A) for formula). In the first step, we run the model for the group with k 1 and 2, and select the best model according the Bayesian information criterion (BIC) values. To compare seizure induced EDA responses, we calculated EDA relative to the mean of the pre-ictal period (30 min) and detected the peak height and time within 30 minutes post-ictally. We compared night (10pm to 5:59am), morning (6am to 1:59pm) and afternoon (2pm to 9:59pm) seizures using the Kruskal-Wallis-test.
Results:
We included data recordings from 43 patients (GTCS: sex:13 males (62%) and 8 females (38%); age as median: (p25-p75), 13.2 (11.1-15.4) years; FIAS: sex: 15 males (68%) and 7 females (32%); age as median: (p25-p75), 9.9 (7.2-13.2) years). Indicated by lower BIC values, the model with k = 2 was the best fitting model for both EDA level and EDA power. The resulting 24-hour pattern is illustrated in Figure 1 B). Group average and standard error values for EDA level and EDA power per hour of the day. Solid lines show best model fit. Seizure induced EDA peak characteristics were analyzed for 59 seizures and are shown in Figure 2. For EDA level, time of occurrence was not significantly different (p = 0.81), but the height differed (p = 0.01) between different times of day. For EDA power, the peak time was different (p = 0.04), while peak heights showed no significant differences (p = 0.67). See Figure 2 for median curve of EDA level and power responses for seizures occurring during afternoon, night, or morning.
Conclusion:
Twenty-four hour recordings of EDA levels show an overarching pattern of modulation among patients. Therefore, seizure occurrence at different times of day might affect the electrodermal activity system differently depending on the initial EDA level based on differences in EDA responses. These findings might help refine the baseline for seizure detection and prediction, including electrodermal activity. Next steps encompass evaluation of confounders, such as wakefulness and medication, as well as comparing results to a control group.
Funding:
:German Research Foundation (VI 1088/1-1) Epilepsy Research Foundation