Dear Daniel,
Ignore the comment about the nifti format, as this is *not* the case for
rigid-body or affine transformations. For within-subject registration
(without fieldmaps) you should be using rigid-body registration, which
has 6 dof. Therefore the flirt call should contain "-dof 6". Also, it is
often better to register using the structural image as the reference
image and then invert the transformation (as using the structural as
the reference tends to work better for correlation ratio). It is then
*essential* to check whether the registration has worked by looking
at the output from the registration. I see that in your command you
didn't even generate this, which is not a good idea.
So I recommend the following:
flirt -ref xxxx/T1_c_brain -in xxxx_dwi.nii -omat diff2struc.mat -out diff2struc
convert_xfm -omat struc2diff.mat -inverse diff2struc.mat
Look at the image diff2struc and compare it with T1_c_brain. It is only
worth proceeding if this registration looks good.
If the registration is good then you can run:
applywarp -i xxxx/StructuralMask.nii.gz -o xxxx/DiffusionOutput -r xxxx_dwi.nii --premat=xxxx_struc2diff.mat -s
You are correct that -s and --super are equivalent so -s is fine.
This should then work OK.
All of this is covered in more detail in the FSL course.
If you continue to have problems then recommend that you look
through this first if you are still having trouble after that then
feel free to email the list.
Note that if you have fieldmaps then the required steps will be different.
We do recommend the use of fieldmaps whenever possible as they can
significantly improve the registration.
All the best,
Mark
On 5 Oct 2011, at 14:38, Daniel Cole wrote:
> Hi Matt,
>
> I was under the impression that the " -s " flag was equivalent to " --super "? I tried it again using --super instead but got the same result. Any ideas?
>
> Thanks,
>
> Daniel
>
> On Wed, Oct 5, 2011 at 2:09 PM, Matt Glasser <[log in to unmask]> wrote:
> Nope that isn’t it. If he uses --premat it accepts affine matrices. Please use the --super flag in applywarp as I suggested in the first e-mail.
>
>
>
> Peace,
>
>
> Matt.
>
>
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Christine Zakrzewski
> Sent: Wednesday, October 05, 2011 1:00 PM
>
>
> To: [log in to unmask]
> Subject: Re: [FSL] FLIRT - Nearest Neighbor option failing
>
>
>
> Hey Daniel. Your linear transform probably has to be in NifTI format—ie, not <struc2diff.mat>, but < struct2diff.nii>.
>
>
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Daniel Cole
> Sent: Wednesday, October 05, 2011 1:01 PM
> To: [log in to unmask]
> Subject: Re: [FSL] FLIRT - Nearest Neighbor option failing
>
>
>
> Hello again,
>
>
>
> I figured out how to generate the matrix using flirt but am now having trouble using applywarp. I decided it was better to get the seed masks into diffusion space so that my tracking outputs in diffusion space.
>
>
>
> My code looks like this:
>
>
>
> "flirt -in xxxx/T1_c_brain -ref xxxx_dwi.nii -omat struc2diff.mat"
>
>
>
> "applywarp -i xxxx/StructuralMask.nii.gz -o xxxx/DiffusionOutput -r xxxx_dwi.nii --premat=xxxx_struc2diff.mat -s "
>
>
>
> But when I open the output of applywarp I get nothing, 'fslstats output -M' gives me 0.
>
> Is there something simple I'm missing?
>
>
>
> Many thanks,
>
>
>
> Daniel
>
>
>
> On Tue, Oct 4, 2011 at 2:46 PM, Daniel Cole <[log in to unmask]> wrote:
>
> Why didn't I think of this! For some reason I kept thinking you need the seed to be in diffusion space for probtracking...
>
>
>
> My bedpost directory doesn't have a structural2diff.mat - is there an easy way to create this?
>
> I know FLIRT creates one but I'm sure there's a better way.
>
>
>
> Thanks so much!
>
>
>
> Daniel
>
>
>
> On Tue, Oct 4, 2011 at 2:37 PM, Matt Glasser <[log in to unmask]> wrote:
>
> You should use applywarp --premat=<yourmat> with the --super flag set instead of FLIRT to resample the ROIs. However, why not just make use of the --xfm option of probtrackx to do this for you?
>
>
>
> Peace,
>
>
> Matt.
>
>
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Daniel Cole
> Sent: Tuesday, October 04, 2011 1:24 PM
> To: [log in to unmask]
> Subject: [FSL] FLIRT - Nearest Neighbor option failing
>
>
>
> Hello FSL list,
>
>
>
> I have a rather odd issue that I was unable to find a solution to in the archives. I am attempting to linear register a structural ROI into a subject's native space for use in probabilistic tracking. However, when I attempt to use nearest neighbor interpolation my ROI disappears!
>
>
>
> Some things you should know:
>
> • My ROI is only 1 voxel thick.
> • 157 voxels total (for this particular ROI)
> • Registration with other subject's is fine.
>
>
> This doesn't happen with all of my subject's but a good chunk of them are problematic. I'm guessing that somehow the voxels I've chosen for my ROI are being lost in the transformation from the high res to the diffusion image?
>
> Is there a way to know this for sure?
>
> Also, what would be the best way to account for this... make my ROIs thicker for all subjects?
>
>
>
> Thanks so much!
>
>
>
> Daniel
>
>
>
>
>
> P.S. The code for trilinear/nearest neighbor is as follows.
>
>
>
> Tri
>
> "/RCBI/fsl/bin/flirt -in /Volumes/rcbiUsers/dcole/Broca/${sub}/t1/T1_c_brain.nii.gz -ref /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS/${sub}/${sub}_dwi.nii -out /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperTriTest -omat /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperTriTest.mat -bins 256 -cost corratio -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 12 -interp trilinear
>
>
>
> /RCBI/fsl/bin/flirt -in /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/Structural/${sub}_LOper_struc.nii.gz -ref /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS/${sub}/${sub}_dwi.nii -out /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperTriTest_shadowreg_${sub}_LOper_struc.nii -applyxfm -init /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperTriTest.mat -interp trilinear"
>
>
>
> NN
>
> "/RCBI/fsl/bin/flirt -in /Volumes/rcbiUsers/dcole/Broca/${sub}/t1/T1_c_brain.nii.gz -ref /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS/${sub}/${sub}_dwi.nii -out /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperNNTest -omat /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperNNTest.mat -bins 256 -cost corratio -searchrx -90 90 -searchry -90 90 -searchrz -90 90 -dof 12 -interp nearestneighbour
>
>
>
> /RCBI/fsl/bin/flirt -in /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/Structural/${sub}_LOper_struc.nii.gz -ref /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS/${sub}/${sub}_dwi.nii
>
> -out /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperNNTest_shadowreg_${sub}_LOper_struc.nii -applyxfm
>
> -init /Volumes/rcbiUsers/dcole/FSLProcessing/TBSS4.1/Probabilistic/Masks/Handdrawn/${sub}LOperNNTest.mat -interp nearestneighbour"
>
>
>
> --
>
> Daniel Cole
>
> University of Rochester
>
> Brain and Cognitive Sciences
>
> [log in to unmask]
>
>
>
>
>
>
>
>
> --
>
> Daniel Cole
>
> University of Rochester
>
> Brain and Cognitive Sciences
>
> [log in to unmask]
>
>
>
>
>
>
>
>
> --
>
> Daniel Cole
>
> University of Rochester
>
> Brain and Cognitive Sciences
>
> [log in to unmask]
>
>
>
>
>
>
>
>
> The information in this transmission is intended for official use of the Kessler Foundation. It is intended for the exclusive use of the persons or entities to which it is addressed. If you are not an intended recipient or the employee or agent responsible for delivering this transmission to an intended recipient, be aware that any disclosure, dissemination, distribution or copying of this communication, or the use of its contents, is strictly prohibited. If you received this transmission in error, please notify the sender by return e-mail and delete the material from any computer.
>
>
>
>
> --
> Daniel Cole
> University of Rochester
> Brain and Cognitive Sciences
> [log in to unmask]
>
|