Neurodynamic state fluctuations in the epileptic brain

Catherine Stamoulis (Childrens Hospital Boston /Harvard Medical School), Bernard S. Chang (Beth Israel Deaconess Medical Center /Harvard Medical School)

RATIONALE: Dynamically, the brain is never at rest. In the absence of task-induced neuromodulations and behavior- and/or stimulus-related state transitions, neural dynamics may fluctuate transiently and spontaneously between resting states, associated with arousal, self-reflection, emotion, or memory retrieval, among other processes. These may be represent short-term, but stable states, and may involve specific networks, as as the Default Mode Network (DMN), and distinct frequency ranges. Brain dynamics are also modulated by physiological states at longer temporal scales, such as sleep. Furthermore, abnormal neuronal networks dynamics may also be associated with distinct, pathological states and related transitions. For example, neurodynamic fluctuations and local and/or global network transitions to unstable states may be mechanisms of seizure initiation in patients with epilepsy. To date, these mechanisms remain only partially understood,  which has limited the development of efficient treatments for medically intractable seizures.

METHODS: We investigated the spatio-temporal variation and stability of dynamic states encoded in non-invasive, continuous (recorded over multiple days) electroencephalograms (EEGs), from 5 patients with multiple focal seizures of temporal origin. EEGs were analyzed in the range 1-250 Hz. We hypothesized that latent dynamic state fluctuations are encoded in these recordings, but are not directly measurable. We, therefore, developed a dynamic non-linear state-space dynamic model and estimated state fluctuations from i) scalp EEGs covering the entire brain, and ii) subsets of signals that are part of the DMN. In this preliminary study, we investigated and characterized the stability of estimated states through a simple analysis of the time-dependent eigenvalues of the state-space matrix.

RESULTS: Dynamic state fluctuations were identified at distinct spatio-temporal scales and frequency ranges. In particular, spatially global transitions to distinct, yet stable states, not associated with measurable changes in any particular physiological state, e.g., sleep, were estimated at least 45 min prior to clinical seizure onset in all patients (and on average 70 min prior to ictal onset in two patients), at frequencies <=12 Hz. Furthermore, transitions of shorter duration, from a stable to an unstable state were estimated at higher frequencies (typically 30-50 Hz) on average 5-12 min prior to seizure onset, and at different rates. Rapid transient fluctuations to unstable states were specifically estimated from EEGs covering temporal regions and slower transitions were estimated in frontal channels and occipital channels. Dynamic instability extended into the ictal interval. On average 20-40 s into that interval, transitions to one or multiple states (depending on the patient) were estimated, which, in turn, extended into the post-ictal interval. Finally, the brain returned to a dynamic state of the same type of stability as estimated at baseline on average 70-90 min following seizure termination. No specific differences in state variation and stability were observed in elements of the DMN.

CONCLUSION: There preliminary results suggest dynamic state fluctuations and possibly pathological transitions to both stable and unstable states at different frequencies, and different rates, may represent at least one mechanism of neural instability that may trigger seizure initiation in the epileptic brain. Based on the type of their (in)stability, these fluctuations appear to be distinct from normal resting state variations encoded in normal baseline EEGs. 

Preferred presentation format: Poster
Topic: Computational neuroscience

Document Actions