Dear Alireza,
> To correct my previous post, I have to claim that it is not really precise to use JHU atlas for ROI analysis when one normalizes to the most typical subject in the study. I guess, the best way is to compute the deformation that requires to transform the atlas to the most typical subject in the study following by affine-alignment to MNI might be required to achieve the precise results.The following would be the way that I gonna go:
>
> #linear registration to provide a good starting point for non-linear registration(fnirt)
>
> fsl4.1-flirt -in JHU-ICBM-FA-1mm -ref target -dof 12 -cost corratio -out flirted-JHU-ICBM-FA-1mm -omat JHU-ICBM-FA-1mm-to-target.mat
>
> # non-linear registration
>
> fsl4.1-fnirt --ref=target --in=JHU-ICBM-FA-1mm --aff=JHU-ICBM-FA-1mm-to-target.mat --cout=coef_JHU-ICBM-FA-1mm-to-target --iout=fnirted_JHU-ICBM-FA-1mm --jout=jac_JHU-ICBM-FA-1mm-to-target --jacrange=0.1,10
I would strongly recommend you use an appropriate config-file for calling fnirt. Add for example --config=FA_2_FMRIB58_1mm.cnf to the call above.
You should also note that an alternative is to register the subject to the atlas and use the inverse of those warps (which you get from invwarp). In general it is a better strategy to have the higher quality image as your --ref image. In particular in this case the configuration file I suggested would work better that way.
>
> #apply the computed deformation map, to transform atlas to target, to the atlas labels
>
> fsl4.1-applywarp -i JHU-WhiteMatter-labels-1mm -r target -o fnirted_JHU-WhiteMatter-labels-1mm -w coef_JHU-ICBM-FA-1mm-to-target
>
> #apply linear registration(from target to MNI) to align the labels (already in the typical subject space) onto MNI space
>
> fsl4.1-flirt -in fnirted_JHU-WhiteMatter-labels-1mm -ref MNI152_T1_1mm_brain -out flirted_fnirted_JHU-WhiteMatter-labels-1mm -interp nearestneighbour -applyxfm -init target_to_MNI152.mat
It is always better to resample as few times as possible. So you would be better off using the convertwarp utility to combine your non-linear and your affine transform into a single transform that you can apply using applywarp.
Good luck Jesper
>
>
>
> --- On Thu, 4/7/11, Alireza Salamy <[log in to unmask]> wrote:
>
> From: Alireza Salamy <[log in to unmask]>
> Subject: Re: [FSL] JHU atlas non-linear normalization to the most typical subject space
> To: [log in to unmask]
> Date: Thursday, April 7, 2011, 5:15 PM
>
> I am also interested in this topic and I guess one might not need to apply FNIRT to the atlas as long as the target image is affine-aligned onto MNI space. So my guess would be that if one use the most typical subject in the study as a target and go through the stream of TBSS, as it is described by the manual, it'll still be possible to use JHU atlas for ROI analysis since all your FA images have been non-linearly registered to the target following by affine-alignment onto the MNI.
> I will appreciate if somebody can correct me if I am wrong?
>
> Best
> /Alireza
>
> --- On Thu, 4/7/11, Stephen Smith <[log in to unmask]> wrote:
>
> From: Stephen Smith <[log in to unmask]>
> Subject: Re: [FSL] JHU atlas non-linear normalization to the most typical subject space
> To: [log in to unmask]
> Date: Thursday, April 7, 2011, 11:43 AM
>
> HI - if you're wanting to add additional steps that TBSS doesn't do by default, you'll need to get inside the scripts and work that out - though the scripts are pretty straightforward.
> It sounds like maybe you're wanting to add a nonlinear registration stape (FNIRT) to the registration of the target subject into standard space - you could just avoid this by using the -T option instead in the first place.
>
> Cheers.
>
>
> On 6 Apr 2011, at 14:02, kambiz rakhshan wrote:
>
>> Dear Fsl experts,
>>
>> I have normalized my FA maps using tbss-2-reg -n to normalize each subject to the most representative subject in the study and then affine-aligned them into MNI space. Now, I would like to use JHU atlas to compute mean FA across some certain tracts.Doing so, one needs to non-linearly align JHU to the typical subject brain.Additionally, image containing labels for individual tracts should be transformed to the typical subject space using some interpolation.
>> Can anybody tell me what steps I should take to get this procedure accomplished?I assume one should use FNIRT or ....? but how?
>> How can I also find about which subject was the most typical one?
>>
>> Many thanks
>> /Kambiz
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
>
>
|