processing-scripts
28 Subscribers 
processing-scripts > Finding overlap with priors
Showing 1-2 of 2 posts
Apr 27, 2010 06:04 PM | Lisa Eyler
Finding overlap with priors
Hi,
First, thanks for the clearly-coded and very useful scripts! I have some questions about the 4_segment.sh script. Although the script ran well, my resulting wm_mask and csf_mask are much more limited than I would expect and I also am having trouble examining how they overlap with the images that they are based upon using AFNI. A general question is about the value of finding the overlap with priors. It seems that this is the step where things are going strange, so I was curious why this is recommended as opposed to just jumping right to the binarizing step. A more specific question has to do with the appropriateness of the priors for my population (very young children). I assume the priors are from 152 adults? The most specific question has to do with my attempts to examine all the different outputs along the segmentation pathway. For some reason, AFNI doesn't seem to think that the priors or the _masked and _mask datasets are in standard space because the GUI will not allow me to overlay these onto func2standard. This last question may be one for the AFNI messageboard, but I would appreciate any thoughts about these issues you may have.
Lisa
First, thanks for the clearly-coded and very useful scripts! I have some questions about the 4_segment.sh script. Although the script ran well, my resulting wm_mask and csf_mask are much more limited than I would expect and I also am having trouble examining how they overlap with the images that they are based upon using AFNI. A general question is about the value of finding the overlap with priors. It seems that this is the step where things are going strange, so I was curious why this is recommended as opposed to just jumping right to the binarizing step. A more specific question has to do with the appropriateness of the priors for my population (very young children). I assume the priors are from 152 adults? The most specific question has to do with my attempts to examine all the different outputs along the segmentation pathway. For some reason, AFNI doesn't seem to think that the priors or the _masked and _mask datasets are in standard space because the GUI will not allow me to overlay these onto func2standard. This last question may be one for the AFNI messageboard, but I would appreciate any thoughts about these issues you may have.
Lisa
Apr 27, 2010 09:04 PM | Maarten Mennes
RE: Finding overlap with priors
Hi Lisa,
Glad that to hear that the scripts have been working out for you!
Yes, the wm and csf mask can be quite limited. We deliberately set a pretty high treshold to avoid inclusion of any other signal in the mask. (this might be important if one would wish not to do global correction). We included the overlap with priors to make sure that the tissue that is selected is generally (i.e., in the template) regarded as csf or wm. That way we try to avoid possible errors due to non-optimal segmentation or registration.
We've also been using this procedure successfully with kids before. The priors are indeed for adults, but then again most people are using an adult atlas for children (although this is a much debated issue).
To see the overlap you have to adjust the header of the .nii file to tailarach view. This is best done by 3dcopy the file (e.g., "3dcopy test.nii.gz test2"), this will recreate your file in AFNI format. Afterwards you can adjust the view with "3drefit -view tlrc test2"
Alternatively, you should be able to overlay the two images in fslview.
Hope this helps!
Also, keep an eye on the site as we will be releasing some developmental data (including kids) in the near future.
Maarten
Glad that to hear that the scripts have been working out for you!
Yes, the wm and csf mask can be quite limited. We deliberately set a pretty high treshold to avoid inclusion of any other signal in the mask. (this might be important if one would wish not to do global correction). We included the overlap with priors to make sure that the tissue that is selected is generally (i.e., in the template) regarded as csf or wm. That way we try to avoid possible errors due to non-optimal segmentation or registration.
We've also been using this procedure successfully with kids before. The priors are indeed for adults, but then again most people are using an adult atlas for children (although this is a much debated issue).
To see the overlap you have to adjust the header of the .nii file to tailarach view. This is best done by 3dcopy the file (e.g., "3dcopy test.nii.gz test2"), this will recreate your file in AFNI format. Afterwards you can adjust the view with "3drefit -view tlrc test2"
Alternatively, you should be able to overlay the two images in fslview.
Hope this helps!
Also, keep an eye on the site as we will be releasing some developmental data (including kids) in the near future.
Maarten