Abstracts

Rationale: The administration of short acting barbiturates has been widely used for the control of status epilepticus (SE) but the optimal treatment regimen remains unclear. The infusion rate of the anesthetic drug is continuously titrated to achieve and maintain a specified level of burst suppression in the electrographic (EEG) signal. Here, we present an automated EEG-based system that identifies the burst suppression pattern (BSP) and quantifies its power and temporal characteristics. To our best knowledge, such an automated system is not available in the market. Methods: An off-line EEG-based algorithm was developed for the detection and quantification of the temporal and spatial characteristics of BSP. The algorithm combines weighted information from the (i) amplitude, (ii) frequency content, and (iii) entropy of EEG. The weighted measures were used to classify EEG segments of 500 ms duration as bursts (B) or suppressions (S). We validated the algorithm's performance to identify the BSP by using the EEG data (mean duration: 100.7 minutes (IQR: 6.4-186.3)) recorded from our 11 pediatric SE patients (age: 7.63 yrs ± 4.47).The patients were treated mainly with pentobarbital (0.08-7 mg/kg/h) and isoflurane (0.1-1.5%). Unanimous detections by two independent neurophysiologists were used as the gold standard in estimating the performance of the algorithm. Results: The algorithm's sensitivity and specificity to identify the BSP was found to be 97.98% and 96.38% respectively; its positive predicted value and negative predicted value was 99.48% and 88.32% respectively. The algorithm requires minimal memory and computation time (< 1 s). The software depicts in three displays the following information: in the first display, the signal of two bipolar channels, the power of EEG frequency bands, the relative power ratio for the EEG frequency bands, and the Shannon entropy for these two channels (see Figure 1). In the second display, the algorithm displays (Figure 2) the ratio of (burst duration)/(suppression duration), the total duration of bursts, the total duration of suppression, and the ratio of the EEG signal power for burst and suppression. The third display depicts these four quantitative EEG measures on topography in order to provide spatial information of the sedation effect on the entire brain. Conclusions: Our algorithm identifies the BSP in the EEG signal of pediatric SE patients with an extremely high sensitivity and specificity. Compared to previously described methods for automated segmentation of EEG data into burst and suppression epochs (Thomsen et al., 1991; Lipping et al., 1995; Arnold et al.,1996; Griessbach et al., 1997; Sherman et al., 1997; Leistritz et al.,1999; Atit et al., 1999; Särkelä et al., 2002; Brandon Westovera et al., 2013), our system adds new value in the following aspect: it is the first algorithm employed on EEG data recorded from pediatric SE patients many of whom had BSPs different in character that those seen in healthy patients undergoing anesthesia for elective surgery.