Brain Mapping Center

How To Plot Eigenvariate Timecourses

NOTE: This procedure plots an eigenvariate, which is some abstract combination of your data and your model. Therefore, for the same pixel or set of pixels, you may get different time courses for different models with which you select the pixels. So this just isn't all that useful.
 
Here's how you can plot the 1st eigenvariate of clusters using SPM:
1) Put the cursor on your cluster in your contrast of interest
2) Hit the V.O.I. button
3) Select the contrast of interest
4) Type in the name of the V.O.I. (eg. amygdala)
5) Enter the search radius in mm (eg 8)

 

 

It might say something about the # of voxels it can use ... see the note below.

This will store the results in a MATLAB environment variable called xY.Y, and it will save the results to a file called VOI_amygala.mat in this example.

To export it to something excel can process, type the following in the matlab window:
1) a = reshape(xY.Y, 116, 9);
Turn the subject matrix from linear to rectangular form for 116 images, 9 subjects.
Adjust these numbers according to the scan duration and number of subjects.
2) save LAmygdalaEigen1.txt a -ascii -tabs
This will save it to an ascii file suitable for excel.
The average time series across subjects can be computed using the mean command. You can plot it and save it using the following:
figure
ts=mean(reshape(xY.Y,116,9),2);
plot(ts)
save LAmygdalaEigen1Avg.txt ts -ascii -tabs

 

 
 
 
 
 

NOTE:
SPM does something that is not obvious. When you specify your design matrix, it performs an F-test to see how good the model fits the data. Under the Defaults menu option in SPM, there is a setting for the F-threshold. SPM will only save the timecourses of voxels that meet the F threshold (it saves them in the file Y.mad). If you get an error that a lower # of voxels can be used than your cluster size, then it means you are losing some voxels because they don't satisfy the F-threshold, even though they satisfy your T-threshold. Also, the coordinates of your center location may be changed because the center of your cluster is one of these sub-F-threshold voxels.

The only solution is to change the Default F-threshold (default 0.001 -> 0.01 or 0.05) and then completely respecify the model. (You may be able to get away w/ just estimating the model ... but make sure that the file Y.mad gets updated).
 

Thanks to Michael Zeineh for the info...
 
 

this page is maintained by fred8.31.00