Odd behavior in Masks directory

Submitted by hoptman on

Hi,

I am seeing some strange behavior in my Masks directory.  I have images that look reasonable to being with, and reasonable as normalized, but do not look reasonable in the Masks directory (see attached for mask image superimposed on corresponding FunImgR).  Is this problematic?  What do the images in Masks represent?

Thanks,

Matt

YAN Chao-Gan

Mon, 03/31/2014 - 15:26

Hi Matt,

Sometimes FSL view can not handle reoriented images correctly. You can use MRIcroN to superimpose and check again?

Best,

Chao-Gan

Hi Chao-Gan,

As you can see, it looks better, but I'm still missing some dorsal prefrontal regions.  Thoughts?
Thanks,

Matt

 

Hi Matt,

This seems weird. Did you do interactive reorient? Could you paste a setting GUI here?

Best,

Chao-Gan

hoptman

Tue, 04/01/2014 - 13:21

In reply to by YAN Chao-Gan

HI Chao-Gan,

I didn't do any interactive work.  I also registered to the EPI template in part because of this kind of issue.

Here's the settings window.

Thanks,

Matt

Hi Matt,

If you are using EPI template to do spatial normalization. The masks will not be warped back to original space. Those were generated by another setting?

Thanks,

Chao-Gan

Hi Chao-Gan,

That was what I was asking.   I'm still not quite sure why the files were created.  I think they may not be necessary in any event, but I got them even when I deleted the Mask directory and reran with the settings you see.

Best,
Matt

Hi Matt,

According to this setting, the masks were warped by T1 image unified segmentation -- You have checked "Segment" but didn't use it to do spatial normalization. Is the normalization effect of DARTEL or T1 image unified segmentation not good? Possibly if you do "reorient T1" and "reorient Fun" will help to improve the normalization effect with T1 images.

Best,

Chao-Gan

Hi Chao-Gan,

That's correct.  A few of the normalizations just didn't work.  I can try the reorient steps.

Best,
Matt

Hi Chao-Gan,

Looks like it's  not going to work.  I think I'll stick with my original analysis and not worry about the files in the masks directory.  It might have to do with the fact that I have data from MRPAGEs that had fairly different contrasts.

--Matt

Hi,

I have a follow-up question.  I got decent segmentations using standard segmentation (not as good with DARTEL), but some of the mwc1* files did not come out.  The wc* files look good, so somehow the data are not getting appropriately modulated.   My next step is to do nuisance covariation.  I assume that just having the wc* files is good enough for that stage, but do I really need the mwc* files for the nuisance covariance?  I guess what I'm asking is:  what files go into the nuisance covariation stage when it is done right after the segmentation?

Thanks,

Matt

Hi Matt,

Sorry for the delayed response.

I am guessing that missing mwc* files may related to the version of SPM you are using. You can type in

[SPMversion, Release]=spm('Ver')

to check the version and release of SPM.

For the current version, nuisance regression is using the SPM a priori tissue masks to do nuisance regression. These masks will be warped back to native space if doing nuisance regression in native space. In a version currently under testing, segmented tissue and CompCor kind of methods are incorporated.

Thanks,

Chao-Gan

Hi Chao-Gan,

I'm using SPM8 version 5236, which I think is the most recent one.  However, I still have missing mwc1* images for some subjects, so I'm not sure what is going on here.

I went back to DARTEL and got reasonable registrations, although I get holes in the images, as another user mentioned.  I used the Smooth by DARTEL option in the DPARSFA call (see below).  If I do that, can I assume those holes will get filled in, as per your prior discussoin?

Thanks,

Matt

Hi Matt,

I am very sorry for the delayed response -- buried myself in a grant application.

Yeah, you can check this issue in a previous SPM post: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1202&L=spm&D=0&P=529502

To my understanding, Smooth by DARTEL is performing smooth in original space, thus the holes still exist. Thus personally I will use the "normal" smooth instead.

Best,

Chao-Gan

Attachment Size
dpars.jpg 68.76 KB