Ah - sorry I didn't read far enough to get to the screenshots. Dumbass!
The reason for this will be that the data is stored in neurological
format -
There is currently an inconsistency between fslview and fdt wrt
neurological data unfortunately - this will be fixed in the next
release. Sorry.
At present The safest thing to do is to change the data ordering to
radiological at the very first stage of processing - then to run
dtifit on the new radiological data check the arrows. If the arrows
are wrong, flip your vectors in x (change the sign of the x elements).
fslorient -forceradiological myimage
make sure you do this on data, nodif, and nodif_brain_mask
The alternative is to get the arrows looking correct in neurological
format (as you have done), and _then_ to flip your bvecs in x (again
change the sign of every x element). So that if you run dtifit again,
the new arrows will look wrong - however the internal processing for
tractography will work.
Sorry about this - it is only very recently that we have started
supporting neurological data, and this inconsistency slipped through
the net!
Cheers
T
On 25 Jun 2008, at 01:09, Matt Glasser wrote:
> Well, the tensors in his screenshots look okay, but perhaps the
> bvecs that
> was used for bedpostX is wrong. Can you post tensors of dyads1?
> Also, what
> images have you loaded in the screenshots (other than the pathways).
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf
> Of Tim Behrens
> Sent: Tuesday, June 24, 2008 4:10 PM
> To: [log in to unmask]
> Subject: Re: [FSL] Strange probtrackx results
>
> Almost certainly your bvecs are defined wrong (there will be a minus
> sign you have to change)
>
> Load up dti_FA, dti_V1 into fslview and look at dti_V1 in lines mode.
>
> If you stick an image of the three orthogonal views on the web, some
> kind soul will tell you what to do.
>
> T
>
> On 24 Jun 2008, at 18:03, Ted Yanagihara wrote:
>
>> Hi Ting,
>>
>> I hope I can provide some help until the experts come on. Since
>> these voxel coordinates are in standard space, did you register all
>> of your diffusion data (with the fdt registration toolbox) to the
>> MNI brain and then run probtrack? When you load the
>> data_brain.nii.gz that has been moved to standard space in FSLView,
>> does it look like there was good overlap with the MNI brain?
>>
>> ted
>>
>>
>> On Jun 24, 2008, at 4:59 AM, SUBSCRIBE FSL Ting-Shuo Yo wrote:
>>
>>> Hi,
>>>
>>> I am tring to use the PROBTRACKX in FDT, and I have encountered
>>> some strange
>>> results. I hope some experienced FSL users can help me figure out
>>> the problem.
>>>
>>> I have prepared my diffusion dataset with 7 b0 images in the
>>> begining,
>>> following by 60 gradient images. The "bvecs" and "bvals" are
>>> specified in
>>> the same order, and the results of DTIFIT seem fine (compared to
>>> other DTI
>>> tools).
>>>
>>> After running BEDPOSTX, I use two voxel points (one in gcc and the
>>> other in
>>> scc) as the seeds, but the results seems a bit strange: the
>>> connectivity
>>> distribution do not follow exactly the tensor orientation (to bend
>>> forward
>>> in gcc and backward in scc), but go across the ventricles. The
>>> results
>>> contradict to other fiber tracking tools, and I hope someone can
>>> help me
>>> find out the mistake I have made.
>>> (please see the pictures at:
>>> http://lh5.ggpht.com/tingshuo.yo/SGCvyHfEQyI/AAAAAAAACkM/
>>> WX436W1SvNc/fsl_probtrackx_01.jpg
>>> http://lh5.ggpht.com/tingshuo.yo/SGCvyLtIahI/AAAAAAAACkU/
>>> Z0dtkdZrBLc/fsl_probtrackx_02.jpg)
>>>
>>> Following is the command I used:
>>>
>>>> probtrackx --mode=simple --seedref=data.bedpostX/
>>>> nodif_brain_mask -o
>>> results -x seed.cc -l -c 0.2 -S 2000 --steplength=0.5 -P 5000 --
>>> forcedir
>>> --opd -s data.bedpostX/merged -m data.bedpostX/nodif_brain_mask --
>>> dir=results
>>>
>>> In the file seed.cc, only two lines specifying the voxel position
>>> of gcc(46
>>> 82 46) and scc(46 45 46).
>>>
>>> Thanks in advance.
>>>
>>>
>>> Ting
>>>
>>> =============================================================
>>> Ting-Shuo Yo
>>>
>>> Max Planck Institute for Human Cognitive and Brain Sciences
>>>
>>> E-mail: [log in to unmask]
>>> =============================================================
>>
>
|