Dear Ikko,


yes, Freesurfer uses some ROIs as well, but during preprocessing you create some 3D mesh of the cortex, which takes a lot of time ;-) and results in surface area and cortical thickness. This is not possible with SPM. With SPM you just get a sort of probability/proportion value for each of the voxels. Hutton et al. (Neuroimage, 2008) "Voxel-based cortical thickness measurements in MRI" have described a cortical thickness measurement based on segmentations with SPM, but as far as I know no toolbox has been published. So if you're really interested in cortical thickness you will have to turn to Freesurfer or Brainvoyager.


If GM volume/density is sufficient for you: With Marsbar you can indeed look at multiple ROIs. Actually you could do the same with some simple Matlab scripts, but if you like to work with GUIs (like me), then Marsbar is perfect. Basically you put the marsbar folder into the SPM toolbox folder (make sure the folder is called marsbar and not something including the version number), then you start SPM, start Marsbar via the SPM GUI / Toolbox drop-down option.

In the first step you have to create the ROI files. In case you're interested in several ROIs and want to use brain atlases like the AAL or the LPBA40, then you would propably rely on the single nifti file in which all the different ROIs are coded by different numbers:
-----
ROI definition -> Import -> Number labelled ROI image -> creates mat files for each of the ROIs
-----

In the second step you run the analysis:
-----
Design -> Set design from file -> select the SPM.mat from your group analysis
Data -> Extract ROI data (default) -> select all the ROI mat files you're interested in
Plot data (simple) -> this just plots the average values for different ROIs for different subjects, it's quite nice to get an overview whether there are huge outliers
Export data -> saves the data (use the summary option), for example text as a text file, with rows for subjects and columns for ROIs
-----

Two aspects to consider:
1) In a whole-brain VBM group model you typically define some intensity threshold like 0.1 or 0.2 as you want to restrict the analysis to voxels with GM volume or density above a certain threshold. In contrast, for ROI analyses you might be interested in the average GM values in particular ROIs, taking into account voxels with a low volume/density. In that case you would have to rerun the whole-brain group model without any thresholds specified and then rely on this model for ROI analyses.
2) If you used unmodulated GM files then GM values correspond to density "units". In that case you would probably go with the average density for ROI analyses, no additional step necessary. If you used modulated GM files, then GM values represent volume, with a range from 0 to values above 1 (due to the modulation the GM volume within a particular voxel can be larger than 1). If you want to convert this into mm^3 you have to take into account the size / number of voxels of the ROIs. One option is to load the ROI mat files with marsbar via ROI definition -> View, the volume is plotted then. In case you used an intensity threshold for the group model you might have to take into account that only a subvolume of the ROI was analysed.


Best,

Helmut