Abstracts

Rationale: After a seizure terminates, the ensuing electro-clinical changes that are often variable are collectively defined as the postictal state. This state is critically important as a cardio-respiratory compromise, and arousal dysfunction is often present and can contribute to mortality including SUDEP (Sudden Unexpected Death in Epilepsy). Prolonged loss of consciousness or confusion can contribute to poor quality of life.  EEG changes in the postictal state are time variant and seizure dependent. It can range from continuous suppression to different degree of slowing (delta/theta) or epileptiform discharges (Periodic lateralized epileptiform discharges). Clinically patient can be obtunded or minimally responsive to external stimuli to confusion/agitation or with no/minimal loss of awareness. Interventions like closed loop brain stimulation that can accelerate recovery from the postictal state are attractive and might provide survival benefit. To develop such therapeutic intervention, the first important step is to quantify the behavioral state in the post-ictal period using electrocorticogram. Here we propose a novel strategy where automated detection of seizure termination and postictal behavioral changes are estimated using a single channel electrocorticogram recorded from seizure onset zone (SOZ). Methods: Twelve seizures recorded from five patients with intractable partial epilepsy undergoing intracranial EEG investigation were included in the study. The EEG recording was downsampled to 500 Hz from 2 KHz. A single channel with a minimal artifact from the SOZ was identified, and analysis was performed on that electrocorticogram(ECoG). Multiresolutional Teager energy was estimated from the ECoG (MTEO(k)) for each 1 second, and an adaptive threshold was developed for detection of seizure onset and termination. Shannon Permutation entropy (SPE) and Tsalis Permutation entropy (TPE) was performed initially to maximally parse sleep and awake ECoG. The three classifiers were then used to detect seizure termination and estimate behavioral changes (awake but confused/minimally responsive/awake without confusion) in the post-ictal period. Results: From video EEG analysis and clinical examination during post-ictal state, the following behavioral states were identified- awake but confused after CPS (N=5), minimally responsive after sGTC (N=4), awake without confusion after SPS (N=3).  TPE and SPE were able to classify post-ictal behavioral changes accurately for the minimally responsive state, but there was overlap between two awake states. Conclusions: Estimating states of vigilance in the post-ictal state using single electrocorticogram is feasible. Further studies are required to develop this novel strategy. Funding: This study was supported by NSF grant OIA 1632891