DS_Post_Fac
DS_Post_Fac_T2
DS_Post_Sup
DS_Pre_Fac
DS_Pre_Sup
Variations
Direction Selective, Pre-Synaptic, Facilatory
Complex direction selective (DS) cells are created within a spiking network model from pair-wise interactions of spiking inputs from non-DS, orientation-tuned simple cells. The DS interaction involves (i) cell pairs with spatial RFs that are phase-offset by about 90 deg, (ii) a temporal delay that is implemented pre-synaptically (relative to the synapse at the transformation from non-DS to DS), and (iii) a facilitatory interaction. This is a hierarchical model containing four distinct populations of spiking units: LGN (ON and OFF center), V1 simple inhibitory, simple excitatory, and complex DS. The spiking cells are conductance-driven integrate and fire units modeled on those of Troyer et al. (1998).
Results
References
- Baker PM, Bair W (2010) Cross-correlation analysis reveals
circuits and mechanisms underlying direction selectivity. Conference
Abstract: Computational and Systems Neuroscience 2010. doi:
10.3389/conf.fnins.2010.03.00332
- Baker PM, Bair W (in prep) Cross-correlation analysis reveals
circuits and mechanisms underlying direction selectivity.
- Troyer TW, Krukowski AE, Priebe NJ, Miller KD (1998) Contrast-invariant orientation tuning in cat visual cortex: thalamocortical input tuning and correlation-based intracortical connectivity. J Neurosci 18:5908--5927.
(A) Organization of the four populations of spiking units: LGN-lateral geniculate nucleus, IN-inhibitory V1 simple cells, EX-excitatory V1 simple cells, DS-direction selective V1 cells.
(B) A population of four V1 DS complex cells.
(C) Within a 12,12,4 (x,y,z) lattice of V1 inhibitory simple cells (IN), cells in the 3rd z-layer that are presynaptic to the white-circled cell in (D) are shown in color. Colors indicate preferred orientation (see orientation key between panels C and D).
(D) The 12,12,4 lattice of V1 excitatory simple cells (EX) is shown where color indicates orientation (see orientation key). The white-circled cell gets IN inputs as marked in (C) and LGN inputs as marked in (E). The spatial LGN-driven excitatory RF is shown at the bottom of F (red square). This cell contribute input to the DS cell in (B, red-circle).
(E) The LGN lattice is 32,32,2, where the two z-layers represent ON and OFF cells. The 12,12,4 lattices of EX and IN cells align to the central 12,12 sub-grid of cells in the LGN, in terms of their RF centers within the field of view.
(F) Approximate spatial RFs for the four EX cells within the z-stack indicated by the white circle in (D). Note that the spatial phase of the RF shifts by 90 deg between z-layers.
(G) Each DS cell in (B) receives an excitatory conductance that is the sum of signals from subunits that operate on pairs of EX inputs. The RFs of the inputs pairs are approximately in quadrature (90 deg phase shift).
(H) The spike trains from one EX unit in a pair is convolved by a causal mask that peaks at DT 20 ms. The resulting mask signal, which delays and integrates the spikes from one unit, is then used to multiply the amplitude of the spikes from the second unit in the pair. The summed conductance of the post-synaptic DS unit (with spontaneous background inputs, i.e. noise, turned off for clarity), shows the influence of the EX spikes (red asterisks). Inputs from other subunits are marked with gray asterisks.