SpinozaWiki:Voxel Based Morphometry

See also the FSL site on VBM [1].

Voxel Based Morphometry

With voxel based morphometry (VBM) we investigate differences in voxel-wise gray matter changes between populations or investigate the gray matter anatomy with relation to one score (IQ, age). Alternatively VBM is interesting if you have done a fMRI study and want to know whether a BOLD-MRI effect is caused by structural gray matter differences or a ‘pure’ difference in function.

VBM consists of the following steps

1. Apply BET to a T1 MRI scan. In this step the non-brain tissue is removed from the brain (see figure below -> bet).
2. Placing the T1 MRI scans in a new directory called VBM
3. Generate a template list if necessary (see below).
4. Tissue-type segmentation in which the gray matter is segmented from the rest of the brain (see figure below -> segment).
5. Affine registration of this image to the MNI152 brain.
6. Average the brains obtained in step 3 to create a study specific template.
7. Check the generated template image (see below).
8. Limited non-linear registration of the gray matter images from step 3 to the study specific template (see figure below -> normalise).
9. Correction in the intensity of the voxels for contraction or expansion during the non-linear warping (this makes the voxel intensity in the normalised space represent the absolute amount of gray matter volume). See figure below -> modulate).
10. Smoothing of images.
11. Evaluating differences with permutation testing.

These steps can be automated with the following functions

1. BetCrawler (step 1).
2. PrepareVBM (step 2).
3. fslvbm_2_template –n (step 4 to 6).
4. fslvbm_3_proc (step 8 to 10).

Generating a study specific template list

When you want to compare two groups with each other it is important that the mask that we will generate for non-linear warping is unbiased. That is, that this mask contains as many subjects from the first group as the second group. This is done by making a file that contains the filenames of the scans that should be included in the mask. If all subjects need to be used it is not necessary to make this list explicitly. This list can be made with kwrite. Enter on each of the lines the name of one of the anatomy files that you want to include in the mask and finish the file with a new line. Save the file under the name template_list.

Check template file

Open the file template_GM.nii.gz. Check whether it looks normal (eg that the brain stem is not on top of the brain). After step 3 has finished, look at GM_Mod_merg to see if each of the scans in the template look normal (that is do not look extremely abnormal). You can do this by playing the movie; each volume represents a different participant.

Input for permutation testing

In the last step (9) a number of smoothed versions of the transformed white matter images are generated (smoothing kernel of 2, 3 and 4 mm). These are called GM_mod_merg_s*.nii.gz. The most robust results are probably going to be obtained with a smoothing kernel of 3 or 4 mm. These files serve as input for SpinozaWiki:Permutation testing.