Hi,
Distortion correction needs a fieldmap or blip-up-blip-down DWI (see "topup" for the latter).
Registration of the FA to the FA template would get you into standard space directly, rather than going via the T1.
If you cannot apply distortion correction then going directly with the FA will probably work much better.
All the best,
Mark

On 23 Aug 2013, at 15:40, Mojdeh Zamyadi <[log in to unmask]> wrote:

> Hi Mark,
> 
> Thanks so much for the reply. The reason I used nonlinear registration was because the linear registration was producing very bad results, although I haven't tried distortion correction, could please let me know how I can do distortion correction using FSL or any other tools? do I need any other images beside the DTI?
> 
> I don't really get the point of registering my FA images with the standard-space FA image. I want to do thalamic parcellation using DTI tractography and for that I need to segment (using a paediatric atlas, my subjects are 6 months-2yrs) the cortex to specific ROIs, and so for that I have to register the atlas to FA (I've tried B0 as well and FA seems to work better) via T1. Do you think there is a better way to achieve automatic segmentation of DTI images?
> 
> Thanks again for your help,
> -Mojdeh
> 
> 
> On 08/23/2013 05:09 AM, Mark Jenkinson wrote:
>> Hi,
>> 
>> There are a couple of things to point out here:
>> 
>> 1) You should always use a config file with fnirt (e.g. $FSLDIR/etc/flirtsch/T1_2_MNI152_2mm.cnf)
>> 
>> 2) We would not recommend non-linear registration within-subject, and would normally register the diffusion to the structural using linear registration plus distortion correction (since the distortion correction, either via fieldmaps or via blip-up-blip-down acquisitions is far more constrained than a general non-linear registration).  And, importantly, fnirt does not work correctly between modalities (i.e. structural and FA), so you should not do that.
>> 
>> 3) If you cannot do distortion correction then you might be better off registering directly between your FA and the standard-space FA image ($FSLDIR/data/standard/FMRIB58_FA_1mm) using fnirt (with config $FSLDIR/etc/flirtsch/FA_2_FMRIB58_1mm.cnf)
>> 
>> All the best,
>> 	Mark
>> 
>> 
>> 
>> On 21 Aug 2013, at 18:17, Mojdeh Zamyadi<[log in to unmask]>  wrote:
>> 
>>> Hi all,
>>> 
>>> I'm trying to segment FA images of paediatric subjects using a T1 template. What I've done so far is:
>>> 
>>> 1. nonlinear registration of atlas-to-T1:
>>> 
>>> fnirt --in=ATLAS_for_reg_bias_corrected.nii.gz --ref=T1_brain_use_for_reg.nii.gz --aff=ATLAS_to_T1.omat --cout=ATLAS_to_T1_FNIRT_warp
>>> 
>>> 2. nonlinear registration of FA-to-T1. I've then inverted the FA-to-T1:
>>> 
>>> fnirt --in=FA_NII_masked.nii.gz --ref=T1_brain_use_for_reg.nii.gz --aff=FA_2_T1.mat --cout=FA_NII_masked_2_T1_FNIRT_warp
>>> 
>>> 3. invert the FA-to-T1 warp to get a T1-to-FA:
>>> 
>>> invwarp -o T1_2_FA_FNIRT_INV_warp -w FA_NII_masked_2_T1_FNIRT_warp -r FA_NII_masked.nii.gz
>>> 
>>> my question now is, can I somehow concatenate the two "ATLAS_to_T1_FNIRT_warp" and "T1_2_FA_FNIRT_INV_warp" and the resample my atlas labels into the DTI (FA) space? or should I run two separate "applywarp" (for Atlas-to-T1 and T1-to-FA) commands?!
>>> 
>>> Thanks,
>>> -Mojdeh