Uni-Logo
You are here: Home News Publications Heterogeneous networks underlie functional heterogeneity
Document Actions

Heterogeneous networks underlie functional heterogeneity

Simulation and analysis of neuronal networks explain distribution of orientation selectivity in the visual cortex

Heterogeneous networks underlie functional heterogeneity

Click to enlarge. Description: see bottom of page

Neurons in the primary visual cortex (V1) of mammals respond differently to different orientations of a moving bar. This orientation selectivity is believed to be one of the basic filtering operations that the visual cortex performs in order to process an image. Since this is an emergent property that appears at the level of the cerebral cortex for the first time, and essentially is absent in structures further along the pathway of visual processing like the lateral geniculate nucleus (LGN), this feature of the brain has been used for decades as a window to study cortical function. Its underlying circuitry, however, and the necessary and sufficient mechanisms contributing to its emergence, have remained elusive up to now.

Although originally studied in the cortex of cats and monkeys, many studies have recently focused on orientation selective neurons in rodents, thanks to the development of new techniques for the recording and manipulation of orientation selective responses in these animals. One major difference between the visual cortex of cats and monkeys and that of rodents is, however, the spatial organization of orientation preferences. In contrast to the smooth organization of orientation selectivity in the former species, where neighboring neurons typically prefer a similar orientation, the visual cortex of rodents is “disordered” and neurons with different preferred orientations are mixed, in a salt-and-pepper fashion. Nevertheless, in both species a similar level of orientation selectivity and a broad distribution of it have been reported.

In a new paper published in the journal PLOS ONE, Sadra Sadeh and Stefan Rotter from the Cluster of Excellence BrainLinks-BrainTools and the Bernstein Center at the University of Freiburg studied the emergence of orientation selectivity in rodent-like recurrent networks. By simulating large-scale neuronal networks that possessed different connection topologies, and mathematical analysis of them, the authors could show how orientation selectivity and its distribution emerge in networks with heterogeneous connectivity. The theoretical analysis shows that, as long as the network responds to signals in a linear fashion and does not show abrupt changes in this behavior, it is possible to predict the output distribution of orientation selectivity in model cortical networks. The results of this study show that, in otherwise statistically homogeneous networks, the functional heterogeneity of recurrent connections  can account for the broad distribution of output orientation selectivity. Such heterogeneity was already reported experimentally. How this heterogeneity might benefit sensory processing is an interesting question that awaits further theoretical and experimental studies.

 

Original publication:

Sadeh S, Rotter S (2014) Distribution of Orientation Selectivity in Recurrent Networks of Spiking Neurons with Different Random Topologies. PLoS ONE 9(12): e114237. doi:10.1371/journal.pone.0114237



Figure caption:

Internal structure of the recurrent network dominates the pattern of orientation selective responses in linearly unstable regimes. Firing rate of neurons (put on a 100×100 grid) in response to different stimulus orientations (Ө) are shown for two networks with (A) local inhibitory and global excitatory connectivity, and (B) local excitatory and global inhibitory connectivity. Pseudo-color shows the firing rate of each neuron. The preferred orientation of the input to neurons is assigned randomly, independent of the spatial position. Non-random spatial patterns of activity, thus, emerge when the network shows linearly unstable dynamics. This is more clear for the network with local excitatory connectivity (B), and leads to very large firing rates of “winner” neurons. In this regime, the prediction of a theory based on linearization of dynamics does not hold anymore (not shown).

Personal tools