Segmentation

Submitted by hoptman on

Hi,

I'm confused as to the meaning of some of the files created during segmentation.  I ended up using traditional (non-Dartel) segmentation.  It mostly worked fine and I get files like:

c1coMPRAGE.nii  coMPRAGE_seg_inv_sn.mat  mwc2coMPRAGE.nii   wc2coMPRAGE.nii
c2coMPRAGE.nii  coMPRAGE_seg_sn.mat      mwc3coMPRAGE.nii   wc3coMPRAGE.nii
c3coMPRAGE.nii  mcoMPRAGE.nii            sym_wcoMPRAGE.nii  wcoMPRAGE.nii
coMPRAGE.nii    mwc1coMPRAGE.nii         wc1coMPRAGE.nii    wcoMPRAGE_sn.mat

However, in some cases, the files mwc1coMPRAGE.nii or mwc3coMPRAGE.nii are empty (but of normal size).  What is the consequence of this for my processing stream?  I'm using covariates as in the Yan et al., 2012 paper.  I know the CSF confound series would be generated from one of the *c3coMPRAGE.nii files, but which one?

Best,

Matt

 

YAN Chao-Gan

Fri, 11/15/2013 - 04:45

Hi Matt,

Here is a description (Normalize using T1 image unified segmentation):

c1: GM; c2: WM; c3: CSF; wc1: GM after spatial normalization; mwc1: modulated GM after spatial normalization.

wcoMPRAGE.nii: T1 image after spatial normalization; sym_wcoMPRAGE.nii: Normalizing T1 image (in MNI space) to a group specific symmetric template.

coMPRAGE_seg_sn.mat: information to transform from native space to MNI space

coMPRAGE_seg_inv_sn.mat: information to transform from MNI space to native space.

Only a few subjects have empty mwc1coMPRAGE.nii or mwc3coMPRAGE.nii, right? Did they fail spatial normalization also? If these files were empty, then they are abnormal. You may need to track back what happened there.

Currently, DPARSF is using SPM's priors (WM: >90%; CSF: >70%) to extract WM and CSF signals -- the masks will be more comparable across subjects.

In the next version, an option of masks generated from segmentation results will be provided (although I still concern about that the masks will be more variable across subjects then). In addition, CompCor method (principle components other than mean) will be incorporated.

Best,

Chao-Gan

Hi Chao-Gan,

It failed on a total of 10 subjects (out of 117).  For 8 of them, the file wc1*.nii looks fine, so the registration worked for those,  so  I'm a little confused.

Best,

Matt

Hi Matt,
If the wc1 files are normalized correctly, but the functional images were not, then there maybe a mis-coregistration between functional images and T1 images.

Solution 1: do interactively reorienting before hand.
Solution 2: Extract the brain via betting the T1 image and functional image, and then do coregistration. This solution will be available in the next version.

For those have empty mwc1 files, I am not sure what happened. But there should be something wrong.

Best,

Chao-Gan


On Fri, Nov 15, 2013 at 11:59 AM, The R-fMRI Network <rfmri.org@gmail.com> wrote:
[To post a comment, please reply to rfmri.org@gmail.com ABOVE this line]

Commented by Matthew Hoptman (hoptman)

Hi Chao-Gan,

It failed on a total of 10 subjects (out of 117).  For 8 of them, the file wc1*.nii looks fine, so the registration worked for those,  so  I'm a little confused.

Best,

Matt


Online version of this post: http://rfmri.org/comment/2955#comment-2955


<