I think the thing with VBM is that it is hard to get your head around what is actually being measured. The quantity is called "gray matter density" which is a number between 0 (no GM) and 1 (full GM). This is used in two overlapping contexts. At the voxel level, the density is determined by the voxel intensity. Eg, if GM in the atlas is typically 100 and WM is 80 and the voxel is 90, then it may get a density of 0.5. This can happen because the voxel is on the boundary between full GM and full WM (ie, a partial volume effect) or it may be that it is in a structure that has a large mixture of GM and WM. Eg, pallidum will have a GM density much less than 1. The second context appears when you smooth the data. This replaces the voxel-wise GM density with an average in a given neighborhood. This also gives a number between 0 and 1 and can more-or-less be thought of as the amount of GM in the sphere of the smoothing kernel. For example, when someone is young, they have a nice plump cortex where the gyri nearly press into each other (small sulcal spaces). As one gets older, the brain loses tissue in a complicated spatial pattern. Some of it GM, some of it WM. As the brain atrophies, it starts to unfold a bit, meaning that the sulcal spaces get larger. This may be due to loss of GM or WM near that sulcus or loss of tissue remotely. In either case, the GM density in this area will drop regardless of whether there is actual loss of GM there or not. This makes VBM hard to interpret.
With surface-based thickness analysis, we are only looking at GM and the effect can only be local. Furthermore, the intersubject registration is driven by the folding patterns of the white surface. These are independent of the thickness measurement. In VBM, the intersubject registration is driven by the GM density itself, meaning that the metric is confounded by the fit to the atlas. This further complicates the interpretation. I don't mean to say that VBM analysis is not useful, just that it is hard to interpret and that differences with surface-based analysis should not come as a surprise.
doug
On 10/30/2015 01:09 PM, O'Shea,Andrew wrote:
Hello John, I would highly recommend you take a look at this paper http://cds.ismrm.org/protected/11MProceedings/files/ISMRM2011-8410.pdf (Greve, 2011. An Absolute Beginner’s Guide to Surface- and Voxel- based Morphometric Analysis). I have found it to be very helpful for my understanding. -Andrew
From: <freesurfer-bounces@nmr.mgh.harvard.edu mailto:freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of John Anderson <j.anderson@publicist.com mailto:j.anderson@publicist.com> Reply-To: "freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu> Date: Friday, October 30, 2015 at 10:54 AM To: "freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu> Subject: [Freesurfer] Voxel based vs surface based
Hi Doug and Bruce, Kindly, I have question regarding the difference between voxel based ( VBM/ FSL pipeline) and surface based analysis ( Freesurfer ).
I have two groups of subjects Group A and Group B I ran morphometric analysis in Freesurfer ( recon-all then manual edits) to check the difference in cortical thickness between the groups. Then I used Qdec to correct the results for multiple comparisons and to visualize the significant clusters. Then I ran VBM analysis between the groups ( the same subjects) to study the difference in gray matter volume between the groups.
I got significant difference between the groups in both methods but in different areas of the brain ( the significant clusters are not in the same location ).
What is the difference between surface based analysis and voxel based analysis technically. Is it right to get different results?
Thanks in advance Bests, John
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer