I don't follow you.  You are saying that you have tracts outside the MNI
brain for all subjects that the FOV got cropped?  Does your twice
transformed FA line up with the original?

Peace,

Matt.

-----Original Message-----
From: Francis Wong [mailto:[log in to unmask]] 
Sent: Thursday, October 14, 2010 4:50 PM
To: [log in to unmask]
Cc: [log in to unmask]
Subject: Re: averaging prob tracks DTI

I think it might be related to this issue I have had:


>> 2) I performed the applywarp twice to check if the transform works:
>> 
>> applywarp --ref=FMRIB58_FA_1mm.nii.gz --in=dti_FA.nii.gz
>> --warp=FA_to_target_warp.nii.gz --out=warpped_FA
>> 
>> applywarp --ref=FMRIB58_FA_1mm.nii.gz --in=warpped_FA.nii.gz
>> --warp=target_to_FA_warp.nii.gz --out=re-warpped_FA
>> 
>> while the warpped_FA matches the FMRIB58 very well, the re-warpped_FA
does
>> not match the subject's native FA image that a large portion of it was
>> cropped. also, the dimension of the re-warpped_FA does not match the
>> original dti_FA.  Something must be wrong in the process.

I set the correct reference image (dti_FA) and perform the above steps
again.  The result is in the attachment.  The right most image is the
original dti_FA, the middle is the transformed one and the left most image
is the one that was transformed back to the native space.  I have the same
problem for all subjects that the left part got cropped.

Let me know you would like me to attach all the images and warps.

Thanks,
Francis


On Oct 14, 2010, at 4:34 PM, Matt Glasser wrote:

> I think some of your subjects are not registered correctly in that case,
as
> that is quite far off.
> 
> Peace,
> 
> Matt.
> 
> -----Original Message-----
> From: Francis Wong [mailto:[log in to unmask]] 
> Sent: Thursday, October 14, 2010 4:05 PM
> To: [log in to unmask]; Matt Glasser
> Subject: Re: averaging prob tracks DTI
> 
> It looks like this (in the attachment), although the values are small.
> 
> Thanks,
> Francis
> 
> On Tue, 12 Oct 2010 11:15:06 -0500, Matt Glasser <[log in to unmask]> wrote:
> 
>> That isn't the cause of that.  If you want to restrict tracts to be
within
>> the brain you need to use a stop mask of non-brain tissues (see previous
>> posts on the list about this).
>> 
>> Peace,
>> 
>> Matt.
>> 
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf
>> Of Francis Wong
>> Sent: Tuesday, October 12, 2010 11:07 AM
>> To: [log in to unmask]
>> Subject: Re: [FSL] averaging prob tracks DTI
>> 
>> Hi Jesper,
>> 
>> Yes it helps! thanks.
>> 
>> About the transformation in probtrackx, is there a way to specify the
>> interpolation method used to transform the tracking results back to MNI
>> space?  There are some tracks spilled outside the brain, and I wonder if
>> this is due to a sub-optimal interpolation.
>> 
>> Thanks,
>> Francis
>> 
>> On Tue, 12 Oct 2010 12:30:28 +0100, Jesper Andersson
> <[log in to unmask]>
>> wrote:
>> 
>>> Hi Matt and Francis,
>>> 
>>>> Okay, I don't know how FNIRT handles an --aff argument, if it
>>>> automatically
>>>> incorporates that into the resulting warp or not.  Perhaps Jesper or
>>>> Mark
>>>> knows?
>>> 
>>> I'm sorry I missed the initial part of this thread. fnirt includes the
>>> initial affine transform into its output (-cout) file. Applywarp knows
>>> this and extracts the affine warp from the header and combines that
>>> with the non-linear warps given by the coefficients.
>>> 
>>> Hope that answers your question?
>>> 
>>> Jesper
>> 
>> 
>> Hi Matt,
>> 
>> I am trying to perform what you have suggested, to apply the non-linear
>> transform obtained from TBSS to transform an ROI defined in MNI space to
a
>> subject's native diffusion space for probtrackx.  I would like to make
sure
>> if I am doing it right.  Many thanks in advanced.
>> 
>> From the outputs of TBSS, for each subject we have the warp field
>> (FA_to_target_warp.nii.gz) and the affine transform
>> (target_to_FA.mat)
>> 
>> then I did
>> 
>> invwarp --warp=FA_to_target_warp.nii.gz --out=target_to_FA_warp
>> --ref=FMRIB58_FA_1mm.nii.gz
>> 
>> to get the non-linear transformed needed to put the ROI to the subjects'
>> native diffusion space.
>> 
>> and here are my questions:
>> 
>> 1) Do we need to care about the affine transform anymore?  That is, when
>> doing the applywarp, do we need to put it as --premat or --postmat?
>> 
>> 2) I performed the applywarp twice to check if the transform works:
>> 
>> applywarp --ref=FMRIB58_FA_1mm.nii.gz --in=dti_FA.nii.gz
>> --warp=FA_to_target_warp.nii.gz --out=warpped_FA
>> 
>> applywarp --ref=FMRIB58_FA_1mm.nii.gz --in=warpped_FA.nii.gz
>> --warp=target_to_FA_warp.nii.gz --out=re-warpped_FA
>> 
>> while the warpped_FA matches the FMRIB58 very well, the re-warpped_FA
does
>> not match the subject's native FA image that a large portion of it was
>> cropped. also, the dimension of the re-warpped_FA does not match the
>> original dti_FA.  Something must be wrong in the process.
>> 
>> 3) Does probtrackx really accept the warp fileds as --xfm and --invxfm?
> Cos
>> in fdt_gui, there is no such option to perform a non-linear transform.
> 
>