Statistical analysis of FA values

This page is a guide on how to statistically analyze your DTI data.

Here is a quick overview of what we will be doing:

  • Learning how to use randomise, part of the FSL suite of programs
    • randomise is a way to statistically compare DTI images between groups of participants

What you will need:

  • FSL (with randomise)
  • Your standard FA map (created from TBSS through the tbss_4_prestats command called "all_FA_skeletonised" in the stats subdirectory)

See the TBSS homepage for full details on the following process.

Starting out

  • First you will need to decide what kind of statistical testing you want to do on your data. You will find your choices here.
  • Next you will need to open a new command window and navigate to the folder with your tbss FA data files
  • The command you will use looks like this:
[bash]randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 --T2 -V[/bash]
  • As you can see, you need a design.mat and design.con file. How these look depends on the kind of statistical test you want to use

Creating the design.mat and design.con files for a simple two-group comparison

  • The design.mat files looks like this:
/NumWaves 2
/NumPoints 4
/PPheights 1 1
/Matrix
0 1
1 0
1 0
0 1

NumWaves = the two columns in the matrix. This would change if you were to try an ANOVA design
NumPoints = number of participants in your design
PPheights = signifies the min/max range of the design. You probably don't need to change this
Matrix = how you want to divide your participants. So if you have a "vulnerable"(V) and "non-vulnerable" (NV) group, you could assign all V participants a '1' in the first column and all NV participants a '1' in the second column. (Therefore, V would appear: 1 0 & NV would appear: 0 1)

  • You can simply open a new text file and create this file yourself by copying and pasting the format I have shown just above. Make sure to save it with the extension .mat and not .txt
  • NOTE: it is VERY important that you keep the order of the matrix consistent with the order that TBSS processed your data. To check the order that TBSS used, navigate to the FA folder created by TBSS and type the command imglob *_FA.* This is the order you'll want for your matrix
    • When you are finished making the design.mat file, place it in the stats subdirectory of your tbss directory
  • The design.con file will look like this:

/NumWaves 2
/NumContrasts 2
/PPheights 1 1
/Matrix
1 -1
-1 1

  • You can keep this file "as is" and simply create a text file with this information and save it as design.con in the stats folder. You would, however, want to change this for an ANOVA design

Running randomise

  • Now that you have your design files, we are ready to run randomise to compare your two groups
  • Navigate to the stats subdirectory and run the following command:
[bash]randomise -i all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 --T2 -V[/bash]

Viewing your results

  • Once the program has finished running, you can look at your results with the following command while in the stats directory:
[bash]fslview $FSLDIR/data/standard/MNI152_T1_1mm mean_FA_skeleton -l Green -b 0.2,0.7 tbss_tfce_corrp_tstat1 -l Red-Yellow -b 0.95,1[/bash]
  • Everything in red/yellow will be where these groups differ siginificantly at the p < .05 level (denoted by the 0.95,1 setting. You can, of course, change this threshold)
  • Notice that this command uses the "tstat1" image. This is where group1 > group2. If you want to see where group2 > group1, insert "tstat2" instead.

What now?

  • Now that you know where your groups differ, you can extract the FA values for these areas for each participant and run other tests on them in a statistics program like SPSS
    • More on this in the future