Exercise 4 - Retino-Cortical Model - Responses to Equiluminance

Objectives:

Use a network-based retinal model in place of linear filter LGN units.
Gain experience using chromatic stimuli.
Think about the meaning of equiluminance at the level of single neurons and networks.
Other documentation:

See the models ELC_V1_Simp and Ret_Mesh at www.iModel.org.
Download the parameter files

ELC_V1_Simp.0.moo - population model including a color-sensitive retina and orientation-tuned cortical units

sine_dir_30c.stm - achromatic direction tuning at 30 degree increments

sine_dir_30_eql.stm - red-green equliluminant direction tuning at 30 degree increments

sine_g_var.stm - vary the green component amplitude of the red-green grating

t4.rsp - response to be saved

Examine the contents of the parameter files
Open the model file in a text editor and observe that the main mod_type parameter is pop, indicating that this is a population model. It contains five populations of units (< pop > objects), named as follows,

in_rg - Inhibitory, orientation-tuned, reg-green sensitive cells
in - Inhibitory, orientation-tuned, achromatic cells.
ex_rg - Excitatory, orientation-tuned, red-green sensitive cells
rgc - retinal ganglion cells (midget parvocellular units)
bipolar - midget bipolar cells, which are driven from single cones

There is also a sixth large object <retina0>, which defines a color retina, specifying the cone density, proportions of L, M, and S cones, and the sensitiviy of those cones to the RGB (red, green, blue) channels of the target video display (that which would be used to collect data in the experimental lab).
Run the model and open the GUI (graphical user interface)
This command will first contruct the model architecture, and then open the GUI. It may take a minute for the GUI to appear,
  wm mod ELC_V1_Simp.0.moo sine_dir_30c.stm t4.rsp gui_flag 1
In one window, you should see the cortical population in_rg, and in the other window, the retinal population rgc. If you click on one of the cortical units (e.g., unit 5,5,0), the presynaptic input map appears. (For simplicity in this model, there is no LGN and the RGCs project directly to the cortical cells.) Within the pre-synaptic map, filled circles represent ON center cells, whereas open circles represent OFF center cells.
For unit 5,5,0 (in_rg), you can see that there are three vertically oriented clusters of inputs, where the central cluster contains only RED (L) OFF and GREEN (M) ON inputs. Therefore, This is a "green" sensitive sub-unit of a simple cell receptfive field (RF). The flanking sub-regions consist of only RED ON and GREEN OFF inputs. Thus, the in_rg population should be sensitive to the orientation of red-green patterns.
Compare this to the in population (click on the 2nd icon from the left at the upper left corner of the main window). Unit 5,5,0 (in) also has three RF subregions, but the inputs are segregated on the basis of ON and OFF (filled and open symbols), and not on the basis of cone type (L or M). Therefore, this cells should be sensitive to the orientation of luminance edges, but less sensitive to color-defined edges.
The third population, ex_rg (star shaped icons), are excitatory units that are built in the same way as the in_rg units, except that they get inhibitory inputs from the two inhibitory populations. You can observe this by right-clicking on the icon legend (upper left) for the in units (deep blue icons, 2nd from left), and then (left) clicking on the ex_rg units to see their pre-synaptic inputs. The fact that these excitatory units receive inhibition from the luminance sensitive in population means that they should have a diminished response to luminance defined stimuli.
The next population, rgc (click the gray circle icon at the upper left), gets single inputs from the bipolar layer (right click the yellow circle icon, to put the bipolar layer in the secondary window). By clicking on an RGC unit, you can see that the it receives input from only one bipolar cell. We will discuss in lecture how the bipolar responses are formed in the retina model.
Compare direction tuning for achromatic and chromatic stimuli
First, we examine how the model units respond to an achromatic direction tuning stimulus. The response file that we use here will cause responses from several units in each population to be saved.
The following simulation will take about 10 minutes to run (depending on your hardware). There are 3 repeats of 12 stimuli. The value of tn is set low (512 time units, which is 1024 ms) to reduce the overall run time. (To avoid waiting, you can download the result, ex4_ach.nd).
  wm mod ELC_V1_Simp.0.moo sine_dir_30c.stm t4.rsp tn 512 stim_nrpt 3 outfile ex4_ach contrast 0.15
View the responses,
  nd ex4.nd
Use the Channel ComboBox to select units from different populations. Which populations show clear orientation tuning? Is oriention tuning present in the F1 or DC response? Note that the responses to the ex_rg units are suppressed because of their inhibitory inputs.
Now, use a chromatically defined grating stimulus to run a direction tuning curve. (The following command will take about 10 mintues to run, but the resulting file can be downloaded: ex4_eql.nd).
  wm mod ELC_V1_Simp.0.moo sine_dir_30_eql.stm t4.rsp tn 512 stim_nrpt 3 outfile ex4_eql
View the responses,
  nd ex4_eql.nd
Compare these equiluminant direction tuning curves to those for the achromatic stimulus for each cortical population.
Verify the equiluminant point for the red-green grating
To determine the equiluminant point for a neuron, we will vary the luminance of the green component of a drifting red-green grating at the optimal SF and having direction 0. (Or, download the result: ex4_gvar.nd)
  wm mod ELC_V1_Simp.0.moo sine_g_var.stm t4.rsp tn 512 stim_nrpt 3 outfile ex4_gvar


  nd ex4_gvar.nd
When viewing these results, keep in mind that you should look at units in the model that have a preferred direction of 0, because this is the direction of the grating stimulus used here. For example, look at the tuning curves for unit in 5,5,1 and ex_rg 5,5,1. Note the locations of the peaks and dips in the tuning curves.
What conditions would have to hold for an "equiluminant" stimulus to be undetected by cortical cells such as the in units here?
End of Exercise 4