Download Orientation and direction tuning

Neural data-analysis Workshop
Oren Shriki, Eyal Zakkay
Exercise #1 – Analysis of orientation & direction
selectivity in V1 cells
 You are given a single mat file to be loaded by your MATLAB script:
o When loading the file, you should see a structure array (SpikesX10U12D) in
your workspace.
 The given structure array contains simultaneous in vivo extra cellular recordings from 10
units (neurons) in the primary visual cortex (V1) of a monkey. It was recorded using a
10x10 electrode array that is permanently implanted in the monkey’s V1. The given
dataset is a chosen subset of all units extracted by a spike sorting procedure performed on
the full 10x10 electrode recordings.
 This dataset represents an experiment using full field visual stimuli of moving periodic
oriented gratings with the following features:
o 12 directions of grating movement – evenly spaced on the circle (30o apart).
o 200 repetitions (1.28 seconds duration) for each stimulus.
 Each structure in the array contains a single field (TimeList).
o This TimeList is a vector of spike events. It represents the activity of the
chosen unit during the recorded repetition of the chosen direction.
o SpikesX10U12D(UnitID,DirectionID,Repetition).TimeList
can be used to access each object.
o DirectionID is an integer between 1 and 12, Repetition is an integer
between 1 and 200.
1. PSTH calculations and display
 Methodological remark: for understanding of the PSTH procedure refer to the previous
Class exercise and its complementary slides.
 Create a 3 dimensional array (PSTH) – i.e., with dimensions for the unit number, the
stimulus direction, and the time bin of the PSTH.
o The size of the time bin should be a parameter that is set at the code beginning
o Play with the time bin parameter until you are satisfied with the resulted figure
o Use zeros or nan to allocate memory
 Choose a representative unit and display the 12 PSTH signals (for the 12 stimulus
directions) on the same figure;
o For the display command use waterfall(t,d,PSTHimage); where t is the
time bin vector, d is the direction vector and PSTHimage is a 2 dimensional
subset of the previously calculated PSTH.
Neural data-analysis Workshop
Oren Shriki, Eyal Zakkay
o “A picture is worth a thousand words” – below is an example of the above
command on unit 4 of the given dataset:
2. Orientation and direction tuning
 Methodological remark: for understanding of the fitting procedure and the von-Mises
angular distribution function refer to the lesson slides.
 For the purpose of this task, we shall use the spike counts of each TimeList in the
given array and ignore the temporal dimension of the recordings.
o To convert the given array to SpikesCounts use:
length(SpikesX10U12D(UnitID,Direction,Repetition).Time
List);
 To estimate the conditional response of each cell, use the mean and the standard
deviation over the different repetitions. e.g.:
ResponseM = mean (SpikesCounts(UnitID, :, :), 3);
ResponseSD = std (SpikesCounts(UnitID, :, :), 1, 3);
o To understand the syntax above use the help command to see how mean and
std work when the input is a higher-order array (not a vector).
o For the required display use the subplot and errorbar commands, e.g.:
figure ('Color', 'w', 'Units', 'centimeters', 'Position', [0 0 25 10]); hold on;
for UnitID = 1:10
subplot (2,5,UnitID); hold on;
errorbar (d, ResponseM, ResponseSD, 'o');
...
Neural data-analysis Workshop
Oren Shriki, Eyal Zakkay
 Here are the tuning curves of the first 5 units:
 For each unit, fit the mean response to a von-Mises angular distribution and plot the fitted
curve on the appropriate figure panel.
o Methodological remark: Use the help command to learn about the fit function
and try to execute the example given in the slides.
o Note, that units 4 and 5 are direction selective as opposed to the remaining units
which tend to orientation selectivity.
o You can use such code to differentiate between the 2 cases:
if (FitDir)
vonMises = fittype ('A*exp(k*cos(x-PO))','coefficients',{'A','PO','k'},'independent','x');
else
vonMises = fittype ('A*exp(k*cos(2*(x-PO)))','coefficients',{'A','PO','k'},'independent','x');
end;
[A,G] = fit (Directions', ResponseM', vonMises);
o To differentiate the display use:
if (FitDir)
plot (180/pi*dx, A.A * exp (A.k * cos (dx-A.PO)), 'r');
else
plot (180/pi*dx, A.A * exp (A.k * cos (2*(dx-A.PO))), 'r');
end;
o The vector dx is defined as: dx = 0:0.01:2*pi;
3. Exercise deliverable
 The deliverable should include the MATLAB code you used and two figures similar to
the ones presented above (with 10 units in the second figure).