Yes, I think you have transformed it twice. It is probably best that you
transform your masks into native diffusion space (if you are going to use
that nonlinear transform) with nearest neighbor interpolation (using
applywarp), track from them, and then transform the results back into
standard space (using applywarp again with the inverse transformation
matricies and warps).
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Vanessa Lim
Sent: Monday, September 21, 2009 7:30 AM
To: [log in to unmask]
Subject: Re: [FSL] averaging prob tracks DTI
Hi,
When I did the tracking I used standard2diff.mat. We have 2 masks L&R A1s
which were in standard space.
Have I transformed it twice because I've flirted the FA images in the below
code and as well as in the registration part of the fdt box?
Thanks,
Vanessa
When you did the tracking, did you do it in native space or did you supply
some xfm?
Peace,
Matt.
On 9/16/09 1:48 PM, "Vanessa Lim" <[log in to unmask]> wrote:
> Hi Matt,
>
> We are interested in the standard space, and I did the FLIRT and then
FNIRT
> using:
> flirt -ref ${FSLDIR}/data/standard/FMRIB58_FA_1mm -in dti_FA -omat
> dti_FA_AffineTrans.mat
> fnirt --in=dti_FA --aff=dti_FA_AffineTrans.mat
--cout=dti_FA_NonlinearTrans
> --config=FA_2_FMRIB58_1mm
> applywarp --ref=${FSLDIR}/data/standard/FMRIB58_FA_1mm --in=dti_FA
> --warp=dti_FA_NonlinearTrans --out=warped_FA
>
> This gives a nice warped FA but, when I tried to apply the
> dti_FA_NonlinearTrans to the paths that were previously generated from the
> FDT registration and probtrackX, it doesn't seem to do much but it moved
the
> paths so that it wasn't aligned with the FMRIB58_FA image.
>
> Am I supposed to apply the transform to the xfms files that bedpostX has
> already generated and then redo the probtrackX or am I supposed to redo
the
> registration (which I can't figure out in order to make the files in the
> xfms directory).
>
> Many thanks,
> Vanessa
>
> On 9/14/09 6:12 PM, "Matt Glasser" <[log in to unmask]> wrote:
>
>> I just sum the pathways and the waytotal numbers; it is not an average it
is
>> a sum.
>>
>> You will have better alignment between your diffusion and structural
images
>> if you unwarp the EPI. This is always the case as even at 1.5T there is
>> some distortion of the EPI. If you don't have a field map and don't care
as
>> much about registration between your diffusion and structural images (but
>> care more about registration between your diffusion image and standard
>> space), you could register the FA to the FA template included with FSL
using
>> first FLIRT with 12 DOF and then FNIRT. If diffusion to structural
>> alignment is important for what you are doing I would do it as I
suggested
>> in that thread.
>>
>> Peace,
>>
>> Matt.
>>
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf
>> Of Vanessa Lim
>> Sent: Monday, September 14, 2009 1:30 PM
>> To: [log in to unmask]
>> Subject: Re: [FSL] averaging prob tracks DTI
>>
>> Thanks Matt,
>>
>> Yes, we are trying to track the callosal pathways between the A1s. Do
you
>> average the tracks as well? i.e. In fslmaths -add then -div by 2? Or do
>> you just do the add for summing?
>>
>> I used the standard registration in FDT but we thought about doing nfirt
to
>> register the DTI, I've been reading the T1 to b-zero thread and averaging
>> pathways and another thread on averaging the tracks where I think you
>> suggested "unwarp your dti data (using a filedmap), then use FLIRT to
>> register them to the T1 (6DF) and use FNIRT to warp the subjects T1 to
MNI
>> space and finally apply this whole transformation to your paths." but I
>> wasn't sure if we should do this when the registration seemed fine?
>>
>> Thanks,
>> Vanessa
>>
>> On 9/14/09 1:03 PM, "Matt Glasser" <[log in to unmask]> wrote:
>>
>>> I am unclear on which pathways you are attempting to find. Are you
trying
>>> to track callosal pathways between primary auditory cortices in both
>>> hemispheres? It is certainly likely when using a local tractography
>> method
>>> like that in probtrackx that you will not necessarily get exactly the
same
>>> results when tracking in one direction vs another. If you want to get
the
>>> common tracts in both methods, you could multiply the two results
together
>>> (this is perhaps not exactly what you want, but it will enhance common
>>> pathways relative to noncommon pathways). If you just want to get all
of
>>> the pathways, you can sum the two results. I would not say that you can
>>> assume that pathways found with one direction but not the other are
>>> necessarily wrong or invalid, though they may be less likely to be
correct
>>> than pathways found with both directions. In my own work, I tend to
track
>>> in both directions and then sum the results.
>>>
>>> Peace,
>>>
>>> Matt.
>>>
>>> -----Original Message-----
>>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf
>>> Of Vanessa Lim
>>> Sent: Monday, September 14, 2009 9:36 AM
>>> To: [log in to unmask]
>>> Subject: [FSL] averaging prob tracks DTI
>>>
>>> Hi,
>>>
>>> I have a left and right primary auditory cortex (A1) masks which I used
as
>>> seeds going
>>> from for example: mask image Left A1 and using the waypoint mask of the
>>> right A1 and I
>>> did the same analyses on the same Ss but reversing the left and right
>> mask,
>>> with the left
>>> A1 now the waypoint mask.
>>>
>>> I get different tracks for LA1 to RA1 compared RA1 to LA1 for all the
>>> subjects. Is this
>>> correct? If so, is it valid to try and get only the tracks that are
>> common
>>> to both left to right
>>> and right to left? If it is valid, can anyone please tell me how I
might
>>> accomplish this
>>> please?
>>>
>>> Many thanks,
>>> Vanessa
|