Abstracts

Rationale: Impaired consciousness is one of the most disabling aspects of epilepsy. However, mechanisms underlying consciousness awareness remain poorly understood. Many theories of consciousness are based on changes in functional connectivity between different brain regions. These changes may fall into two major categories: 1) alterations in the overall dynamic state of brain regions. An alteration in the dynamic state of brain regions may be reflected by a “global ignition” leading to a conscious state, such as that suggested in the Global Neuronal Workspace model1.  2) changes in directional flows of information across regions. Change in directional flow of information has been suggested as a biomarker for states of consciousness in several studies, such as increased feedback (anterior-to-posterior) flow in awake states and feedforward (posterior-to-anterior) flow in an anesthetized state2. The purpose of this study was to determine the role of re-entrant processing in different natural states of consciousness (awake and asleep) as well as seizures.  Methods: Scalp EEG was recorded from subjects with or without (“normal”) history of seizures undergoing video EEG monitoring as part of their standard clinical care. EEG was acquired with electrodes conforming to the international 10-20 system for electrode placement, using a Nihon Kohden EEG system (Nihon Kohden America, Irvine CA), sampling at 200 Hz. A reduced bipolar montage was utilized, consisting of the following electrode pairs: F3-F7, P3-T5, F4-F8, and P4-T6, to capture anterior-posterior (frontal-parietal) information flow. Granger causality was calculated over sequential 2s non-overlapping sliding windows, excluding artifacts such as eyeblinks, using the MVGC toolbox4 in Matlab (Mathworks, Natick, MA), and visualized using Microsoft Excel (Redmond, WA). Granger causality is a measure of directionality of information flow. It is based on the concept that if there are two processes X(t) and Y(t), if the combination of X(t-1) and Y(t-1) predicts X(t) better than X(t-1) alone, Y(t) must have contain some information pertinent to X(t), and Y(t) “Granger-causes” X(t)3. Results: In all subjects, anterior-posterior information flow was dynamic (i.e., fluctuated over time). Feedforward and feedback flows were not entirely inversely correlated - that is, more feedforward did not translate to less feedback flow and vice versa. Furthermore, information flow in the left and right hemispheres were not tightly correlated. In both the awake and asleep states, net flow could be net feedforward or feedback at different time points. However, sleep states were generally associated with a net feedforward of information flow compared to awake state. This pattern seen in 'normal' subjects was altered in subjects with epilepsy interictally and depended on the location of the seizure focus. In addition, information flow was away from the seizure focus during focal seizures. Conclusions: Brain network characteristics are dynamic. Seizures can be associated with changes in directional information flow. These changes can differ from those in other altered consciousness settings such as sleep and general anesthesia. Network characteristics altered by seizures can also affect baseline directional information flow in different levels consciousness. These abnormalities may depend on the location of seizures. Re-entrant processing as captured by anterior-to-posterior directional connectivity may not be a robust biomarker for capturing state of consciousness and breaks down in the setting of neurological disorders such as epilepsy. Funding: No funding