Feed-forward Propagation Mechanism Of Neural Signals

The processing and propagation of neural information in an excitable neural system are the key issues in neurosciences.Sensory and motor neural systems are characterized by modular and hierarchical structures,which determine neural signals to be propagated in a feed-forward manner.The most attractive computation model for studying signal propagation in neuroscience is the multilayered feed-forward network(FFN),where each layer is related to a functional group of neurons,and neural information is transmitted from one layer to another.Combining the study of FFN with some experimental observations,this paper provides the insights into the experimental findings.A recent experiment on primate parietal cortex shows that spike-time regularity increases consistently from primary sensory to higher cortical areas.Considering the obviously hierarchical structure of parietal areas in anatomy,we construct a multilayered FFN mimicking neural information transmission pathway.The first layer is subjected to Poisson input and the evolution of spike-time regularity through downstream layers is investigated.Numerical results reveal that despite the obviously irregular spiking patterns in previous several layers,neurons in downstream layers can generate rather regular spikes.This largely depends on the network topology.Specifically,it is found that the optimal topology parameter(i.e.,synaptic connection probability and synaptic weight)maximizes the spike-timing regularity in deeper layers.Furthermore,it is demonstrated that synaptic properties,including inhibition and synaptic transient dynamics have significant impacts on spike-timing regularity propagation.The emergence of increasingly regular spiking patterns in higher parietal regions can thus be viewed as a natural consequence of spiking activity propagation between different brain areas.We validate an important function served by increased regular spiking: promoting reliable propagation of spike-rate signals across downstream layers.It is found that during waking compared with being asleep or anesthesia,neuronal responsiveness to external stimulus can be more reliable.In terms of this phenomenon,we study how the local network(layer)state would determine the ability of FFN to transmit neural information.Each layer is modeled as a random sub-network,where excitatory and inhibitory neuronal populations interact with each other and then generate intrinsic network dynamics.Local network dynamics can be continually modulated by the excitation-inhibition balance.It is demonstrated that for both synfire propagation and firing rate propagation,an optimal local excitability state is found respectively to maximize the performance of signal propagation.Finally,a simple mean field approach that bridges response properties of long-range connectivity and local sub-networks is utilized to reveal the underlying mechanism..The results in this paper are useful for understanding the mechanism of neural information processing and propagation,and more importantly,provide the basis for the designs control strategies for modulating neural coding.