Benjamin,

Given the ringing on the glass brain, I wonder if what you are seeing is 
some type of edge effect (eg from dura). I believe that DARTEL's warps 
tend to smear outside the brain, so maybe that is the culprit. It's hard 
to tell though because I'm not sure what image you are overlaying on. If 
that's just some brain in MNI space, like ch2, then the blobs you are 
seeing are probably in different areas than they appear to be (default 
DARTEL will not give you results in MNI space!!!). If the funny areas 
are indeed outside the gray matter, then your problem is solved simply 
by using a good mask (eg mean(c1)>.5 or something of that nature). 
Here's a quick way to check:

1) Use Create Warped to warp all your T1 images just like you did with 
the GM
2) Use ImCalc to make a mean image
3) Overlay your results on the mean image

Also, if you see ghosting on your mean image, then something went wrong 
with a subject. You can easily make mean c1 and c2 images as well for a 
quick quality check on your segments.

On a separate note, as Philipp suggests, adding TIV and age covariates 
may help, though I suspect that neglecting those would not give the type 
of ringing artifact that you are seeing.

Regards,
Neil

Dana Perantie wrote:
>
> Hi Benjamin,
>
> I suggest checking the mask to see what voxels have been included in 
> the analysis. If the mask looks odd, if you haven’t already, check all 
> of the images that you input to make sure they are all OK (e.g. on the 
> same scale and the segment looks reasonable). One funny image can ruin 
> the masking for the whole analysis… this has happened to me before 
> (e.g. when a segment failed for one subject, or when accidentally 
> including a white matter image in a gray matter analysis).
>
> I believe changing the voxel extent only alters which results are 
> viewed - it does not change the statistics or significance.
>
> I would suggest trying to view Results without FWE correction. This 
> will apply the voxel-level significance threshold that you specify 
> (defaults to .001). When you examine the “whole brain” statistics, you 
> may see the FWE voxel or cluster corrected p values. You may also want 
> to look into non-stationary smoothness correction.
>
> Also when you have the “whole brain” statistics, check out your volume 
> information (at the bottom) – knowing how smooth your data are and how 
> many resels are included may be helpful in determining what’s going on.
>
> -Dana
>
> *From:* SPM (Statistical Parametric Mapping) 
> [mailto:[log in to unmask]] *On Behalf Of *Ben Pruce
> *Sent:* Monday, January 11, 2010 10:39 AM
> *To:* [log in to unmask]
> *Subject:* [SPM] DARTEL and VBM estimation and results
>
> Hello SPM'ers
>
> I have performed a VBM analysis of two groups using DARTEL in SPM8. 
> The images are from a Siemens 3T trio MPRAGE sequence. I have followed 
> the DARTEL steps using most default features and have a question 
> regarding the estimation and results. Both gray and white matter maps 
> were constructed and when I perform my 2 sample t-test the grey matter 
> looks very odd. Everything was showing up on the brain, so I did a 
> search and found it is suggested to mask threshold the images during 
> estimation so I re-performed the analysis with a relative threshold of 
> 0.8 and the GM maps still look odd. Is there a common mistake I have 
> made in this analysis based on the result? The images are Jacobian 
> modulated images (smwrc1.nii) and are FWE corrected at p>0.0001 with 
> 20 voxel extent. I have attached a screenshot. The white matter looks 
> reasonable when I threshold relative with 0.8, but a second question 
> comes to mind. Does DARTEL or initial threshold affect the FWE 
> correction? To see any result in the WM, I have to use FWE p>0.0001 
> with 20voxel extent, which seems odd to me. I have attached a 
> screenshot of this as well. Please excuse the resolution has been 
> degraded in the images to conserve space. Any help or direction to 
> readings would be much appreciated. Once again thank you for your kind 
> direction.
>
> Best,
> -- 
> Benjamin Pruce, R.A.
> Psychological and Brain Sciences
> Indiana University
> Bloomington, IN 47401
>