Sorting out the files

The output of randomise are statistic, p value, and corrected p value files, for each contrast. The contrasts to which statistics files correspond are indicated by numbers, e.g., *_tstat1 (or _fstat1),  *_tstat2, *_tstat3, and so on. For a two-group design matrix, you will typically have four contrasts; these are listed in the *.con text file created when you create the design matrix with the Glm GUI. Below is an example:

3D View

Prior to starting, make sure you have installed the special script for displaying results.

You will now display the FIRST findings on a standard brain. The script simply loads the FIRST shape and statistical results into FSLView; we need to add the template brain and adjust visualization settings. The steps are described in the user manual under Viewing results. The current page simply presents these steps in more detail.

Note: when the instructions in the user manual say: press "i" they mean click the little "i" icon in the FSLView GUI, not press the keyboard "i"!

• Run the script, giving as input the mask (which defines the shape surface) and the p value to overlay; the files with "tfce_p_tstat..." are uncorrected p values, and "tfce_corrp_tstat.." are corrected. This will open FSLView, and after ~30+ seconds, open the default brain with the overlay. For example:

[user@localhost temp]$ /usr/local/fsl/bin/first3Dview  L_Hipp_vertexMNI_2sample_mask.nii.gz  L_Hipp_vertexMNI_2sample_rand_clustere_corrp_fstat1.nii.gz

Note: FSLView uses radiological convention, so left on right as in the above example.

• Select "basestruct" (highlighted above); click "i" and change the image type to Mask/Label.

• The p or corrp overlay image (above basestruct) should already be set to Image type "Statistic" but you can check by selecting the image and clicking "i". This image is in a "1-p" numerical value (to avoidn technical problems with very small numbers). Thus, a range of 0.95 to 1 corresponds to p = 0.05 to p = 0. To see everything with the slightest A.B (or B>A) difference, change the minimum to 0.0001 (sometimes 0 causes an error), so it may look like this:
  
• Under Tools menu, select 3D Viewer

• If you get a message ...vtkDecimatePro.cxx .... "No data to decimate!", this most likely means the thresholds are messed up. Chnage
• Turn Clipping on (check the box on the top-right); this will "cut out" a chunk of the brain. From ehre, you want to adjust this cutout so you see the structure of interest, and enough of the brain to orient you.

• Click in the 3D part and click "b" to toggle the sliders on and off; you adjust the cutout by dragging the knobs (see below) at the end of the sliders. Often, you have to rotate the brain to see the knobs.

Note: "b" toggles the sliders on and off; I usually turn them off to rotate the brain.