help
help > RE: choosing template as structural image
Mar 20, 2013 06:03 AM | Alfonso Nieto-Castanon - Boston University
RE: choosing template as structural image
Hi Felix,
The segmentation of each subject anatomical volume is used in conn to define the gray matter / white matter / csf ROIs. Signals within the white matter and CSF areas are then used as a 'noise template' in order to remove potential physiological and movement related artifacts from the BOLD signal (compCor strategy, Behzadi et al. 2007 'A component based noise correction method (CompCor) for BOLD and perfusion based fMRI') before continuing to the connectivity analyses. Even after normalization to a structural template there still remain significant anatomical differences between subjects. Using a template to define the grey/white/csf areas instead will increase the risk of the resulting white matter / csf areas overlapping with the grey matter voxels within each subject, and this in turn is problematic because including signals from grey matter voxels in the 'noise template' can be shown to result in biases in the observed distribution of functional connectivity estimates (for details you may want to see Murphy et al. 2009, 'The impact of global signal regression on resting state correlations: are anti-correlated networks introduced?', as well as Chai et al. 2011, 'Anticorrelations in resting state networks without global signal regression'). Because of this, if you have the data, I would recommend to enter each subject anatomical volume (instead of a template) as the structural image for each subject (it should not add more than a few minutes per subject to perform the segmentation after all)
Hope this helps
Best regards
Alfonso
Originally posted by Felix Mueller:
The segmentation of each subject anatomical volume is used in conn to define the gray matter / white matter / csf ROIs. Signals within the white matter and CSF areas are then used as a 'noise template' in order to remove potential physiological and movement related artifacts from the BOLD signal (compCor strategy, Behzadi et al. 2007 'A component based noise correction method (CompCor) for BOLD and perfusion based fMRI') before continuing to the connectivity analyses. Even after normalization to a structural template there still remain significant anatomical differences between subjects. Using a template to define the grey/white/csf areas instead will increase the risk of the resulting white matter / csf areas overlapping with the grey matter voxels within each subject, and this in turn is problematic because including signals from grey matter voxels in the 'noise template' can be shown to result in biases in the observed distribution of functional connectivity estimates (for details you may want to see Murphy et al. 2009, 'The impact of global signal regression on resting state correlations: are anti-correlated networks introduced?', as well as Chai et al. 2011, 'Anticorrelations in resting state networks without global signal regression'). Because of this, if you have the data, I would recommend to enter each subject anatomical volume (instead of a template) as the structural image for each subject (it should not add more than a few minutes per subject to perform the segmentation after all)
Hope this helps
Best regards
Alfonso
Originally posted by Felix Mueller:
Hello,
I'm trying to perform a multisubject resting state analysis with CONN and I do the preprocessing with SPM. After coregistration of the functional data of each subjects with its structural image, i normalise with the MNI152_T1_2mm template and smooth. These preprocessed data I want to analyse in CONN. As far as i understood, I need a spatially normalised structural image for each subject in CONN.
What would happen if I used the MNI152 template as this strucural image for all subjects instead? Which information will I loose or which confounds will be increased? The reason I'm asking is that the segmentation of each structural image takes very long in comparison to performing just one segmentation for my whole analysis. Is there still a huge difference between the template and a structural image, which was normalised to this template?
Sorry, if this question sounds wrong or trivial, but I just started using SPM and CONN a few weeks ago.
Regards,
Felix
I'm trying to perform a multisubject resting state analysis with CONN and I do the preprocessing with SPM. After coregistration of the functional data of each subjects with its structural image, i normalise with the MNI152_T1_2mm template and smooth. These preprocessed data I want to analyse in CONN. As far as i understood, I need a spatially normalised structural image for each subject in CONN.
What would happen if I used the MNI152 template as this strucural image for all subjects instead? Which information will I loose or which confounds will be increased? The reason I'm asking is that the segmentation of each structural image takes very long in comparison to performing just one segmentation for my whole analysis. Is there still a huge difference between the template and a structural image, which was normalised to this template?
Sorry, if this question sounds wrong or trivial, but I just started using SPM and CONN a few weeks ago.
Regards,
Felix
Threaded View
| Title | Author | Date |
|---|---|---|
| Felix Mueller | Mar 14, 2013 | |
| Alfonso Nieto-Castanon | Mar 20, 2013 | |