Phase-Amplitude Coupling and Neuroglial Networks of the Brain: Big Role for a Small Cell

摘要

Epilepsy affects over 45 million people worldwide and while treatment options exist, antiepileptic drugs don't work in up to 50% of the cases; furthermore, poorly-treated epilepsy may cause complications such as sudden unexpected death in epilepsy (SUDEP) and status epilepticus. Brain rhythms can characterize both normal and pathologic brain states and the coupling between these rhythms, namely between low frequency oscillations (LFOs, 30 Hz) and high frequency oscillations (HFOs, 30+ Hz), has been shown to be specific to particular phenomena in epilepsy. Thus, better understanding the nature of this cross-frequency coupling (CFC) can shed light on novel therapies and treatment options. The first half of this thesis shows the importance of CFC features in epileptic seizures. Specifically, a) features of both ictal and postictal state are consistently significantly different from baseline interictal state; b) CFC features can be used with a convolutional neural network to identify ictal and postictal substates; c) postictal CFC features when used with a hidden Markov model (HMM) classify postictal EEG suppression (PGES) state without relying on visual inspection; d) ictal CFC features are associated with the duration of PGES - a marker linked to the risk of SUDEP in patients; and e) CFC features can also be used on data from patients who died from SUDEP to differentiate those patients from the rest of the population of patients with epilepsy. To better understand potential mechanisms behind CFC generation, a series of computational models of neuroglial networks are discussed in the second half of this thesis. These models showed that a) astrocytic factors play an important role in hyperexcitability, leading to an increase in average duration of spontaneous electrographic discharge (SED) and emergence of CFC features similar to those seen in epileptic patients; b) microglial synaptic pruning increases inter-SED durations, while astrocytic manipulation led to increased SED durations; and c) glial effects modulate synaptic activity to achieve CFC feature variability seen in seizure episodes. Finally, the model is compared to other meso- and macro-level models, as well as to patient data - through HMM classification, gamma-fit of SED distributions, and beta-fit of ictal CFC variability.