Dynamic functional networks in human seizure activity

Mark Kramer (Boston University), Uri Eden (Boston University), Eric Kolaczyk (Boston University), Sydney Cash (Harvard Medical School, Mass General Hospital)

Epilepsy, the condition of recurrent unprovoked seizures, is the world's most prominent serious brain disorder, affecting some 50 million people worldwide.  For an estimated 30% of these patients, seizures remain poorly controlled despite maximal medical management.  Moreover, control of epilepsy through medication and surgery often results in significant, sometimes debilitating, side effects.  Advancing the therapeutic management of epilepsy requires a detailed understanding of the focal initiation and subsequent spread of the seizure over a network of interconnected brain regions.

The human brain is naturally conceived as a network capable of generating rhythmic fluctuations of coordinated neuronal activity.  This “brain network” consists of two fundamental components:  nodes (e.g., a cortical column) and edges (e.g., synaptic connections) that connect node pairs. In neuroscience, brain networks are typically divided into two categories:  persistent structural networks based on anatomical connections between brain regions, and transient functional networks representing the coupling between dynamic activity recorded from separate brain areas.  Ongoing research suggests common topological properties emerge in these brain networks, including small-worldness, heirarchal organization, and the presence of densly connected hubs.  In epilepsy - perhaps best characterized as a disease of brain rhythms - the relationship between dynamic functional networks and the pathological brain rhythms of this disease remains unknown.

Here we present a general paradigm for the inference of dynamic functional networks from time series data, and apply this paradigm to the anaysis of multielectrode invasive voltage recordings made directly from cortex and deep brain regions of human subjects during seizure.  We first describe the implementation of this paradigm for a computationally efficient choice of coupling measure and consider the robustness of this measure for simulated data sets.  We then apply this procedure to the analysis of electrocorticogram (ECoG) data recorded from a population of human subjects during seizure.  We investiagate how the dynamic functional networks evolve during seizure, and propose three new insights:  1) At the spatial scale of ECoG recordings, brain regions decouple during seizure, 2) Network topologies coalesce at seizure onset and just before termination, while fragmenting during seizure, and 3) Similar functional network topologies appear from seizure-to-seizure.

To address the relationship between dynamic networks and the rhythmic neuronal activity characteristic of the seizure, we develop a simulation study of network dynamics.  At each node, we implement a mean-field model of cortical activity capable of producing the rhythmic fluctuations of seizure.  We then connect these nodes with topolgical structures consistent with the functional networks observed in the clinical data.  In this way, we examine how the changing network connectivity (i.e., dynamics of networks) affects the neuronal rhythms produced at each node (i.e., dynamics on networks).  Combined, a deeper understanding of the network dynamics and neuronal mechanisms of the seizure promises to provide new insights, and perhaps theraputic strategies, for the treatment of epilepsy.

Preferred presentation format: Poster
Topic: Computational neuroscience

Document Actions