ALE Meta-Analysis Using CSL Software


(1) Make input text file. Make a text file that has one row for each XYZ coordinate (input focus). Each row should have FOUR numbers: X-coordinate, Y-coordinate, Z-coordinate, and Study Code. XYZ coordinates should be in MNI space. The study codes should be positive consecutive integers, starting from 0 (not starting from 1!). Here is an example . Let's assume for the sake of example that your text file is named Foci.txt.


 

(2) Check input text file. Run the command

 

wc Foci.txt

 

This should return three numbers. They are, in order: number of lines, number of words, and number of characters in the text file. The number of lines should equal the number of input foci. The number of words divided by the number of lines should be four, since there are three coordinates (XYZ) plus a study code on each line. If not, something is wrong with the text file; examine it closely for errors.


 

(3) Create ALE Map. Log onto the machine named medx (or any machine on which MEDx is installed), perhaps using VNC, get into a csh or tcsh shell, then at a UNIX prompt type

 

aleMake.csh step1 <FourColumnTextFile> <FWHM>

 

e.g.

 

aleMake.csh step1 Foci.txt 12

 

if you want to do the ALE analysis with a FWHM of 12 mm. (This step makes a call to the software MEDx, which is installed only on the machine named medx. This is why you must log onto the machine named medx for this step.)


 

(4) Perform inference. Even though you have now created an ALE map in step (3), you still need to figure out which voxels, if any, are to be called "statistically significant." Log onto one of the app machines (app1-app7, or node018 if you're in CFMI), go to the same directory, and then type

 

aleMake.csh step2 <FourColumnTextFile> <FWHM> <NumberOfPermutations> <FDRThreshold>

 

e.g.

 

aleMake.csh step2 Foci.txt 12 10000 0.05

 

The FWHM you specify for this step must be the same FWHM you specified in the previous step. Otherwise the p-values and therefore the inference will be thrown off.

 

This second step can take a long time, perhaps several days, depending on how many input foci you have. It will generate Overlay images which you can load into MEDx with colors. If you have time, it may be worthwhile to cross-check with the results you get using Search&View's implementation of the ALE method. (Since this step can take a long time, and since it doesn't make a call to the software MEDx, it is preferable that you use one of the app machines rather than the machine named medx for this step. When possible, please reserve use of the machine named medx only for when you're using the software MEDx.)


 

Here is something that I wrote up a while ago that may help provide background information.