Nia's Guide to FSL VBM

From Bangor Imaging Unit
Jump to navigation Jump to search

Download and install FSL: http://www.fmrib.ox.ac.uk/fsl

Reference for FSL VBM: http://www.fmrib.ox.ac.uk/fsl/fslvbm/index.html

  1. Data needs to be in NIfTI or Analyse format, to convert DICOM data use a program like MRIConvert, which is very easy to download and use, simply load your images
  2. Place all of your structural images in one folder e.g. myVBM
  3. Open X11, point it to the folder where the data is saved using the ‘cd’ command to change directory, note that ‘cd ~’ will point X11 to the ‘home’ directory
  4. Specify the design of the study that you want to test at the end (it is easier to specify the design at this stage). Type in ‘glm_gui’ in the command window and a Graphical User Interface will pop up. Change the option in the top left corner to ‘Higher level analysis’. You can use the ‘Wizard’ to help you set up the design. Specify the number of groups and the regressors and contrasts that you want to test. Save the design under the name ‘design’ so that later stages of the analysis know to use this.
  5. Type in the command 'fslvbm_1_bet -b', this will perform brain extraction on all the images (if your images have a lot of neck then use –n instead of –b at the end)
    Depending on the amount of data this will take approximately 10-15 minutes
  6. The next stage will create a study-specific template. If you have more than one group the template needs to be constructed with the same number of subjects in each group, create a file called ‘template_list’ containing a list of images to use for constructing the template e.g.
    CON_S1.nii
    CON_S2.nii
    PAT_S1.nii
    PAT_S2.nii
    For the larger group select data, creating a group of the same size as the smaller group, at random to create the template
  7. Type in fslvbm_2_template –n to produce the template. This step takes a few hours to run.
  8. The next stage will register all the study images to the template and concatenate the data.
    Data will be concatenated in alphabetical order so make sure that the naming of your data is appropriate so that the data order will match what you have specified in your design.
    Type in fslvbm_3_proc. Again this will take a few hours to run.
  9. Next, use the ‘randomise’ command to perform nonparametric statistics on the data. You will need to point the terminal window to the ‘stats’ folder. The suggested command to use is:
    randomise –i GM_mod_merg_s3 –o GM_mod_merg_s3 –m GM_mask –d design.mat –t design.com –n 5000 –T –V
    which will perform 5000 permutations on the data.
  10. In order to obtain cluster-level thresholding type in:
    randomise -i GM_mod_merg_s3 -m GM_mask -o GM_mod_merg_s3 -d design.mat -t design.con -c 2.3 -n 5000 -V
    which uses the images with smoothing of Sigma=3mm and applies a cluster threshold of 2.3. Then threshold the images and use it to mask the corresponding tstats map
    fslmaths GM_mod_merg_s3_clustere_corrp_tstat1 -thr 0.95 -bin mask_pcorrected
    fslmaths GM_mod_merg_s3_tstat1 -mas mask_pcorrected fslvbm_tstat1_corrected
    which thresholds the image at p=0.95.
  11. You can then view your images using
    fslview $FSLDIR/data/standard/MNI152_T1_2mm fslvbm_tstat1_corrected -l Red-Yellow -b 2.3,4