Sensory Input Transformation in Layer 4 of Primary Somatosensory Cortex Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
  • Gong, Peng
    • Affiliation: School of Medicine, UNC/NCSU Joint Department of Biomedical Engineering
  • Neural representation of sensory information in the cerebral cortex undergoes a series of transformations, starting from its initial form at the level of thalamic neurons through a succession of cortical layers of multiple cortical areas. In the somatosensory system, the first such transformation takes place in the input layer, or Layer 4, of area 3b. This study explores several of its known properties: (1) the cortex is organized as a set of minicolumns, each a radial cord of cells 30-50 μm in diameter; (2) receptive fields of neighboring minicolumns occupy shuffled positions on the skin; (3) Layer 4 neurons possess more complex functional properties than the thalamic neurons from which they receive their inputs; and (4) neighboring neurons are decorrelated in their stimulus response behaviors. The neural mechanisms responsible for these properties were investigated in this study in a computational model of a field of minicolumns with self-organized Hebbian thalamocortical connections. A parametric study of this model optimized its performance on an “omnipotency” test, which measures the capacity of a set of Layer 4 neurons in the model to represent arbitrarily defined nonlinear functions. The maximal omnipotency was achieved in the model in which: (1) adjacent minicolumns had fixed inhibitory interconnections; (2) more widely separated minicolumns had anti-Hebbian inhibitory interconnections; and (3) each neuron was modeled as an electric circuit consisting of two serially connected electrical compartments, with thalamic and anti-Hebbian inhibitory connections terminating in the distal compartment, and the fixed inhibitory connections terminating in the proximal compartment. When optimized for omnipotency, such a model exhibited among its emergent properties the shuffled receptive fields, decorrelated stimulus-response behaviors, and higher-order functional properties characteristic of the real cortical networks. In conclusion, this modeling study suggests that stimulus information is transformed in Layer 4 to maximize its linear coding of higher-order stimulus features via (1) fixed inhibitory interactions among adjacent minicolumns, carried out by connections of chandelier cells on the initial axon segments of spiny-stellate cells; and (2) anti-Hebbian inhibitory interactions among more distant minicolumns, carried out by connections of basket cells on the somata and dendrites of the spiny-stellate cells.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Favorov, Oleg
  • Hsiao, Henry
  • Tommerdahl, Mark Allen
  • Whitsel, Barry
  • Kelly, Douglas
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2008
  • This item is restricted from public view for 2 years after publication.

This work has no parents.