Dear João,
On Mon, 11 Apr 2011 16:29:08 +0100, João Duarte <[log in to unmask]> wrote:
>Dear SPMers,
>
>I have some questions about the integration of VBM8 toolbox with DARTEL. I
>hope these are easy short-answer questions...
>
>I'm running an analysis using VBM8 with the following parameters
>
>"Estimation options"
> Tissue Probability Map: *spm8\toolbox\Seg\TPM.nii
>
>*Question: This TPM is only used to remove non-brain tissue from the image,
>right?*
Yes, I use the initial SPM8 segmentation (New Segment) only for removing non-brain tissue. The subsequent VBM8 segmentation does not use any prior information from tissue probability maps.
>
>"Extended options"
> Spatial normalization: High-dimensional: Dartel
>
>*Question: This option uses *spm8\toolbox\vbm8\Template_6_IXI550_MNI152.nii
>as template, which is in MNI space, right?*
Yes, this is a Dartel-template based on 550 healthy control subjects from the IXI-database. This Dartel-template was spatially normalized to MNI space to skip the MNI normalization that is still necessary in the original Dartel approach.
>
>"Writing options"
> Grey matter (the same for WM and CSF)
> Native space: yes
>
>*Question: the *_seg8.mat file contains the parameters of the segmentation,
>so I could segment another scan of the same subject using this file?
Yes, the seg8-parameters can be used with the option "write already estimated segmentations". This will skip the time-consuming estimation of the segmentations using New Segment.
>
>Question: The native space GM segment p1*.nii is registered with the*
>**spm8\toolbox\vbm8\Template_6_IXI550_MNI152.nii,
>in MNI space, right? Are these still probabilities of each voxel to be GM
>(or WM or CSF)?
Yes, the native segmentations are normalized to Dartel-MNI space. The results are GM probabilities. However, becasue I am using a partial volume estimation (PVE), the interpretation is a little bit different. The segmentations express the fraction of GM, that was estimated in that voxel. A value of 0.7 in the thalamus means that 70% of this voxel belong to GM and the remaining 30% belong to WM. At the border between GM and CFS the values express the probability to belong to GM and CSF. Two mixed classes are modelled.
>
>Question: The *_seg8.txt contains the volumes of GM, WM and CSF. Are these
>the number of voxels in each segment multiplied by the voxel size or the sum
>of all "voxel value x voxel size" (accounting for voxels with probability
>values < 1) ?*
These values are the sum (integral) of all voxels in native space (without any normalization) and are given in ml.
>
>"Writing options"
> Deformation fields: forward + inverse
>
>*Question: Do these fields map from the individual to the template, in this
>case the ***spm8\toolbox\vbm8\Template_6_IXI550_MNI152.nii? If I want to
>know the volume of a certain structure (say the left Thalamus for example),
>I would warp a mask of the left Thalamus (from marsbaar-aal for example,
>after exporting it to the space of **
>*spm8\toolbox\vbm8\Template_6_IXI550_MNI152.nii**) to the individual using
>the inverse deformation field? And on which image should I calculate the
>volume (native GM 'p1*.nii', normalized GM 'wrp1*.nii', modulated normalized
>GM 'm0wrp1*.nii' ?)? Now, once again, is it sufficient to count the number
>of voxels of the "individual mask" and multiply by the original voxel size
>or should I get the values of each voxel in the GM segment, multiply by
>voxel size and sum all of them up?
Forward deformations can be used to warp an image from native space to Dartel MNI space (if Dartel normalization is used) or to the MNI space of the TPM image (for low-diemsnional warping). The backward deformations can be used to warp an image from Dartel MNI space or TPM MNI space to native space. As you suggested these deformations can be used to warp predefined atlases to native space. Another application is to warp coregistered fMRI data to Dartel MNI space.
Volumes should be calulated rather based on native space images than modulated normalized images (although the values should be similiar). Again, probabilities can be summed up to consider values >1.
>
>**After running "VBM8: Estimate & Write" I run the module "Run DARTEL
>(create Templates)"*, using rp1*_affine.nii, the GM segments after an affine
>registration to the MNI template (as I chose high-dimensional normalization,
>is this the *spm8\toolbox\vbm8\Template_6_IXI550_MNI152.nii template?).
>This creates a flow field for each subejct:
>'u_rp1*_affine_DARTEL_template_name.nii', which I then use to 'Run DARTEL
>(create Warped)', with modulation, and get a modulated GM segment of each
>subject, in the space of the group-DARTEL-template.
>
>*Question: What are the ends (group-DARTEL-template to individual?) of these
>flow field mappings? Can these flow fields be used to compute volumes of
>structures? How can I do these and on which images (modulated DARTEL warped
>GM 'mwrp1*_affine.nii', native GM 'p1*.nii', DARTEL exported
>'rp1*_affine.nii' ?) should I do it?*
Again you have both options. Either you calculate volumes in native space or in modulated normalized images. This depends on the masks you want to use. However, in any case you have to consider affine spatial normalization depending on whether you want to estimate volumes corrected for ICV or not.
>
>*Question: The Marsbar ROIs were not created in my group-DARTEL-template but
>in MNI space. Is this a problem? Should I use the default VBM8 deformation
>fields and segment images or can I use the DARTEL flow fields and warped
>segments to calculate volumes of structures?*
If your ROIs are in MNI space (but not using Dartel) the Dartel normalization in VBM8 can be skipped and the SPM default normalization (low-dimensional) can be used. There might be still some discrepancies between the different spaces.
Regards,
Christian
____________________________________________________________________________
Christian Gaser, Ph.D.
Departments of Psychiatry and Neurology
Friedrich-Schiller-University of Jena
Jahnstrasse 3, D-07743 Jena, Germany
Tel: ++49-3641-934752 Fax: ++49-3641-934755
e-mail: [log in to unmask]
http://dbm.neuro.uni-jena.de
>
>
>Sorry for the long e-mail. I hope you can help me.
>
>Many thanks in advance.
>
>Regards,
>
>Jo�o Duarte
>
|