Philosophy
This tutorial will walk you through how to run a VBM study using SPM8. It has 3 parts: Preparation, Processing and Post-Processing. This tutorial also has 4 files to help you through. The first is the SPM8 manual where you can find much of the "Processing" section of this tutorial outlined. The next file is located in the Processing section and is an example SPM8 file, and the last two are also in the Processing section and are articles by Ashburner explaining the science. VBM requires that all the subjects' MPRAGEs have been dicomed.
Section 1: Preparation
Before the VBM can be performed, it is important to make sure that the MPRAGE is in the proper orientation and alignment with the standard MNI template.
1. Start SPM from the group results directory
-- cd into directory, type matlab. File -> Set Path -> SPM8 -> Move to Top
-- type spm pet in Matlab command window that opens up
2. SPM8 Window
-- Click Display
-- Choose subjects copied MPRAGE (NOTE: It is a good idea to make a copy of the original MPRAGE, and perform this step on that copy)
-- Roll: -pi/2
-- Yaw: pi/2
-- Click Reorient Image
-- Choose subjects copied MPRAGE
3. Repeat for all subjects.
4. Copy reoriented images into VBM analysis directory.
NOTE: It is also a good idea to check the quality of registration for each MPRAGE to the MNI template during this step and to correct for any gross misalignments and fixing them before moving on.
Section 2: Processing
Now that the data has been prepared, this following will set up the script to run VBM on the subjects. A batch of what the script should look like, with a single fake MPRAGE, can be found HERE.
1. Start SPM from the group results directory
-- cd into directory, type matlab. File -> Set Path -> SPM8 -> Move to Top
-- type spm pet in Matlab command window that opens up
2. Graphics Window
-- Click TASKS -> Batch
3. SPM -> Spatial
-- Add Coreg: Estimate
-- Reference Image: click Select Files and choose /fmri/spm8/tpm/white.nii
-- Source Image: click Select Files and choose first subjects MPRAGE
-- Other Images: click Select Files and choose all other MPRAGEs
4. SPM -> Tools
-- Add New Segment
-- Volumes: click Dependency and choose Coreg: Estimate: Coregistered Images
-- Tissue 1: Change Native Tissue to DARTEL Imported
-- Tissue 2: Change Native Tissue to DARTEL Imported
-- Tissue 3: Change Native Tissue to DARTEL Imported
-- Tissue 4: Change Native Tissue to None
-- Tissue 5: Change Native Tissue to None
5. SPM -> Tools -> DARTEL Tools
-- Add Run DARTEL (create Templates)
-- Images: click New Images (do this twice)
-- New Images 1: click Dependency and choose New Segment: rc1 Images
-- New Images 1: click Dependency and choose New Segment: rc2 Images
6. SPM -> Tools -> DARTEL Tools
-- Add Normalize to MNI space
-- DARTEL Template: click Dependency and choose Run DARTEL (create Templates): Template (Iteration 6)
-- Select according to: click Many Subjects
-- Run DARTEL (create Templates): Flow Fields
-- Images: click New Images (do this twice)
-- New Images 1: click Dependency and choose New Segment: rc1 Images
-- New Images 1: click Dependency and choose New Segment: rc2 Images
-- Preserve: click Preserve Amount ("modulate")
Section 2: Post-Processing
This previous section was the "bulk" of VBM. After the script has finished you will end with files that begin with smwrc1 (smoothed, modulated, normalized, realigned GREY matter) and smwrc2 (smoothed, modulated, normalized, realigned white matter). They are Statistical Parameter Maps where every voxel contains the probability value of being grey (or white) matter, from zero to one. To prevent edge effects from dominating your group t-tests, you must threshold the data, zero-ing any value below a certain probability. Most literature uses values between 10% (0.1) to 30% (0.3). Its best to test different values. Usually 20% works well. The easiest way to do this is with FSL4.1 using the following commands.
-- fslmaths [smwrc1-file] -thr .2 [output file name]
-- gzip -d [output from above]
The second step decompresses the file to a normal .nii (nifti format) as fslmaths does its math in compressed format.