A Dynamic Bayesian Network Model for UP/DOWN States of Spontaneous and Optogenetically Evoked Activity in the Primary Visual Cortex

Home › Groups › OWEISSLAB › Publications › A Dynamic Bayesian Network Model fo…
Author(s): 
Ali Mohebi
Jessica A. Cardin
Karim G. Oweiss
Associated Project: 
Mining Large-Scale Neural Ensemble Recordings
Abstract: 

Precisely-timed interaction between excitatory and inhibitory neurons within and between layers of the cerebral cortex is a key element in brain function. Simultaneous recording of the extracellular activity of these neurons using microelectrode arrays enables observing the dynamics of the underlying network activity. Once categorized, dynamic graphical models could be efficiently used to infer the functional connectivity between the observed cells from the recorded spike trains and characterize network state transitions as they pertain to stimulus dynamics. Identifying the type of neurons in these recordings, however, is not straightforward due to the variability in extracellular spike shapes and the irregularities often observed in their interspike intervals.

In this study, optogenetic tools were used to genetically target fast spiking interneurons in the mouse primary visual cortex. Spiking activity of interneurons was modulated by illuminating the region with very short pulses (<1 ms) of light (~470 nm wavelength). We used Dynamic Bayesian Networks (DBN) analysis to infer the effective connectivity between 17 simultaneously recorded neurons in the presence and absence of light stimuli with various frequencies. We compared neuronal interaction during cortical UP and DOWN states identified by the high and low gamma power, respectively, and found distinct patterns of network connectivity under each state. To quantify these differences, networks adjacency matrices were represented in a principal component feature space and using Fisher Discriminant Analysis (FDA), we found the between similarity of inferred networks for UP/DOWN states to be on average 30 times higher than their within similarity, suggesting distinguished functional networks for UP/DOWN states. We also found evidence for the presence of light-frequency-specific connectivity, suggesting that the dynamics of interaction depends on the rate at which the output of inhibitory drives affects excitatory neurons. Taken together, these methods provide a novel and unprecedented way to characterize the dynamics of computations in cortical neuronal circuits.
Year: 
2011-11
Conference/Journal Name: 
Society for Neuroscience Abstracts, No 626.21, Nov 2011