Regress out covariates

Submitted by ocell on
Dear REST team:

  I met difficulty regressing out the nuisnace covariates. Using REST, I extract time series of WM, CSF and global mean signal. The masks files inherent to the REST program has been resliced to 53x63x46. Then I put the 3 txt. files as covariates respectively, but the resultant functional images become inhomogenous by visual inspection. When they are put into SPM for estimation, the following error messages showed up: "no inmask voxels-empty analysis". I wonder if I have removed too many image data? How can I solve this problem? Thank you very much.  

YAN Chao-Gan

Mon, 06/14/2010 - 02:50

Permalink

.img need .hdr to be open. When the covariates are regressed out, the image looked a little different. I do not know what kind of analysis you are trying to do in SPM after regressing out the covariates, thus I do not know what this error message means.

Dear REST team: Thanks for you reply, I have put on the .hdr file. For the SPM estimation, I put time series of a ROI and 6 head motion parameters as the regressors, but the preprocessed image (smoothed, detrended, filtered, and nuisance signals from WM, CSF and global mean signals removed) couldn't be estimated. It seems that I have removed too much information so it shows the message " no inmask voxels".

I do not know much about the cause.
Could you check the time series of your ROIs?
If the ROI is not appropriate, then it may be full of zeros.

ocell

Sun, 06/20/2010 - 02:59

Permalink

Dear REST team:

  Thanks for your reply.
  However, before using REST to regress out the CSF and WM signals, the GLM model runs well with the same ROI. but after regress out the CSF and WM signals (I used the templates attached in REST program and resampled them), the GLM model won't run, showing that "no inmask volume", that's why I think that removing the CSF and WM time series seems to remove too much signals! Is there any way to resolv this problem?

   Thank you.
Forums
Discuss