I'm afraid I have never gotten any _hr.nii.gz files as output. (I had seen
mention to them in the archives, but never saw them in person, so to
speak.) Is something silently not finishing? The last line in my
tbss_2_reg.o#### file is "Final Transformation" ...
Gary
On Mon, 14 Jan 2008, Gwenaëlle DOUAUD wrote:
> Hi Gary,
>
> this is not the first time I come up with this problem
> with the "new" tbss: you are actually looking at the
> wrong output, *_to_target_nonlinear.nii.gz. This is
> the warpfield derived from the nonlinear registration
> and its size is indeed bigger as it's a 4D file with
> three components in the fourth dimension (one for the
> deformations along the x-axis, one for y and one for z
> -once again with my own, improper words). Everything
> seems to have gone fine but the output you want to
> look at is *_to_target_nonlinear_hr.nii.gz.
> In case you do not have these outputs in your FAi
> directory, just let me know...
>
> Hope tis helps,
> Gwenaëlle
>
>
> --- Gary Strangman <[log in to unmask]> a
> écrit :
>
>> Hi all,
>>
>> I used to use FSL3.3 version of TBSS (successfully)
>> on my centos4 linux
>> box. This past Tues I downloaded & installed
>> fsl-4.0.2-centos4_32.tar.gz.
>> I must be doing something seriously wrong, since
>> every attempt at making
>> it through tbss_2_reg results in a super-smooth
>> output file (as if a 100mm
>> smoothing kernel was passed over the entire trio of
>> volumes in the
>> tbss_2_reg output file). I've tried data DTI from
>> three different Siemens
>> machines (1.5 Sonata, 3.0 Trio, 3.0 Tim Trio), all
>> with the same result. I
>> didn't see anything in the archives, but I'm hoping
>> this is just a "duh" on
>> my part. The command sequence below, using all
>> defaults except for bet -f,
>> reproduces the problem:
>>
>>> cd dicom_datadir
>>> mri_convert 610000-8-1.dcm c8_dti.nii # raw dicom
>> data --> nii
>>> eddy_correct c8_dti.nii c8_dti_eddy.nii 0 # no
>> probs, 10 B0, 60 angles
>>> bet vol0000.nii.gz c8_nodif_brain -f 0.2 -m #
>> gives a rounded frontal lobe
>>> dtifit -k c8_dti_eddy -o c8 -m c8_nodif_brain_mask
>> -r bvecs -b bvals
>> [output from dtifit looks fine, FA included]
>>
>> # re-orient our Siemens data to match the FMRIB58_FA
>> template
>>> fslswapdim c8_FA x -y z c8_FA_swap
>>> fslorient -forceradiological c8_FA_swap
>>
>> # set up filesystem
>>> mkdir tbss
>>> mv c8_FA_swap.nii.gz tbss/
>>> cd tbss/
>>> fslstats c8_FA_swap.nii.gz -R
>> [output = 0 1.2247] ... so use scaling of 10000
>>> tbss_1_preproc -f 10000 c8_FA_swap.nii.gz
>>> tbss_2_reg -T
>>
>> The output of the last command,
>> c8_FA_swap_FAi_to_target_nonlinear.nii.gz,
>> is hyper-smooth (doesn't look like an FA image at
>> all). It's also 73Mb,
>> which seems rather larger than the template (at
>> 14Mb). If create the affine
>> transofrm volume using ...
>>
>> ${FSLDIR}/bin/DR/transformation c8_FA_swap_FAi.hdr
>> c8_FA_sw_FAi_affine.hdr \
>> -dofin c8_FA_swap_FAi_to_target_affine.dof
>> -target target.hdr
>>
>> the affine-reg result nicely co-registers with the
>> FMRIB58_FA template, so
>> it's not tbss_2_reg, tbss_reg, or areg ... pretty
>> much leaving nreg to
>> blame. If I use the affine-aligned output as the
>> source volume for
>> registration, as in "tbss_reg c8_FA_sw_affine.nii.gz
>> target" I get the same
>> hyper-smooth output after the nreg step.
>>
>> What am I missing?
>>
>> -best
>> Gary
>>
>
>
>
> _____________________________________________________________________________
> Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr
>
>
>
|