Nonperiodic Synchronization in Heterogeneous Networks of Spiking Neurons

Abstract

Neural synchronization is of wide interest in neuroscience and has been argued to form the substrate for conscious attention to stimuli, movement preparation, and the maintenance of task-relevant representations in active memory. Despite a wealth of possible functions, the mechanisms underlying synchrony are still poorly understood. In particular, in vitro preparations have demonstrated synchronization with no apparent periodicity, which cannot be explained by simple oscillatory mechanisms. Here, we investigate the possible origins of nonperiodic synchronization through biophysical simulations. We show that such aperiodic synchronization arises naturally under a simple set of plausible assumptions, depending crucially on heterogeneous cell properties. In addition, nonperiodicity occurs even in the absence of stochastic fluctuation in membrane potential, suggesting that it may represent an intrinsic property of interconnected networks. Simulations capture some of the key aspects of population-level synchronization in spontaneous network spikes (NSs) and suggest that the intrinsic nonperiodicity of NSs observed in reduced cell preparations is a phenomenon that is highly robust and can be reproduced in simulations that involve a minimal set of realistic assumptions. In addition, a model with spike timing-dependent plasticity can overcome a natural tendency to exhibit nonperiodic behavior. After rhythmic stimulation, the model does not automatically fall back to a state of nonperiodic behavior, but keeps replaying the pattern of evoked NSs for a few cycles. A cluster analysis of synaptic strengths highlights the importance of population-wide interactions in generating this result and describes a possible route for encoding temporal patterns in networks of neurons.

Tradeoffs and Constraints on Neural Representation in Networks of Cortical Neurons

“…Mature networks (14 -21 days in vitro) were chosen for experimentation based on their ability to reliably respond to more than one source of low-frequency (0.1 Hz) stimulation. Reliability is defined as above 50% success in evoking a synchronous population response, denoted "network spike" (NS) as explained in Results (Eytan and Marom, 2006;Thivierge and Cisek, 2008). The first 10 ms following each stimulus were removed from the data, to exclude spikes that were directly evoked by the stimulus itself; this point is further elaborated on in Results, General considerations.…”

“…The main mode of activity in these networks, both spontaneous (van Pelt et al, 2004;Chiappalone et al, 2007) and in response to electrical stimulation is the network spike (NS)-an event of synchronous network activity (Eytan and Marom, 2006;Thivierge and Cisek, 2008;Thiagarajan et al, 2010) lasting tens to hundreds of milliseconds. At the system level (in vivo), the network spike is a universal phenomenon that characterizes responses to sensory objects, regardless of the stimulus modality, stimulus complexity or cortical area involved (Meister et al, 1991;Riehle et al, 1997;Usrey and Reid, 1999); behaviorally relevant objects are believed to be represented by the activity of neurons within the network spike time-amplitude envelope (Keysers et al, 2001;Wesson et al, 2008;Foffani et al, 2009).…”

“…Activity then reverberates across the network and eventually dies out (Eytan and Marom, 2006;Thivierge and Cisek, 2008), creating the characteristic shape of the network spike. An example of a stimulus-evoked NS, with the immediate response, the recruitment phase, followed by a reverberating phase that lasts a couple hundreds of milliseconds (Jimbo et al, 2000;Eytan and Marom, 2006), is provided in Figure 1.…”

The limits of brain determinacy

“…(b) Sensitivity to spike timing The interaction between neurons in complex structures like the cerebral cortex has been likened to that between musicians in an orchestra; they must precisely orchestrate their neural activity to function adequately [68]. This need exists also at the level of individual neurons, because most neurons need to be activated almost simultaneously by several inputs to produce an action potential, and they are very sensitive to tiny variations in the synchrony of the inputs.…”

A Computational Model for Conscious Visual Perception and Figure/Ground Separation

“…The change of the activation potential V i of neuron i is described by the following equation (Thivierge and Cisek, 2008) …”