VBM

Preprocessing

Create the mean T1, coregister to the MNI T1 template.

Use the SPM5 "Segment." Data will be your T1 scans (you can enter multiple subjects in different folders, but make sure you only have one T1 per subject). Output files for gray and white matter should be at least "Modulated + Unmodulated Normalised" and I would recommend "Native + Modulated + Unmodulated" (native files can be used to check the segmentation). For CSF you don't need the Modulated files (the native and unmodulated files can interesting for seeing where CSF is. Save bias corrected, and use thorough clean-up.

Smooth (typically 12 FWHM, but can use 10) the "density" or 'modulated files: "density" files are prepended with "wc1" (gray matter) or "wc2" (white matter). The modulated files are prepended with "mwc1" or "mwc2."

Statistics

The files used for analysis can be the smoothed density or modulated images (prepended with "swc1" or smwc1").

So set up a model, go to SPM Statistical Analysis.

Other modalities

Hi,

Just a thought - with VBM and modulation - as far as I understand for the modulation it takes into account the intensity of the voxel - is that correct? If so does it matter whether you use a T1 or a T2 image to run the modulated VBM?

I assume that you then cannot use for example a FA image to run the VBM?

Cheers

Luke

Actually no to both your questions:

Modulation takes account the volume of the original voxel.

Consider say hippocampus – lets assume native and template space are 1 mm resolution

GM density calculated across brain

GM normalized: Native space 1200mm3 => template space 1000mm3

Each voxel in template space is 1mm3; however it represents a part of the brain that was more volume in native space

To account for the larger volume, the GM density value of template space voxels are multiplied by the volume change (which is the Jacobian “j…” files)

Perhaps the voxel in the template represents a brain are that was 1.2mm3 in native space, so the jacobian at that voxel would be a value of 1.2.

In any case, that multiplication is termed modulation in VBM world

If you only want volume and don’t care about GM or WM density, you can just analyse the jacobian – that’s deformation-based morphometry (DBM), which is an option in CSPM.

The GM and WM density are what takes account of the intensity of the anatomical.

Re running VBM on FA, it depends:

You can use FA/b0 to create new GM and WM maps and do a new segmentation, then you get basically a low res VBM – but this would indeed be VBM. The main requirements are that there are separate intensities across the different tissue/compartment types. So GM should 1) have fairly consistent intensity and 2) look different (lighter/darker) to WM and look different to CSF.

You could also take the normalized FA and “modulate it by the jacobian, which would give you a slightly odd FA- volume measure…

Finally – I’m implementing CAT12 to be newer/better/more VBM….