Hi Sushravya,
I've looked at your example. First, there is no such "truncation", at least on this image. It's true that the original FA was not betted properly (in the dtifit process), so there is quite a thick "rind" around it that the erosion introduced in tbss 1st step only partly removes. This is enhanced after the NL registration.
The problem here is that, quite possibly because of PVE together with the quality of the data, the external/extreme capsules don't appear clearly on the FA image. FNirt (NL algorithm) then struggles to match the external/extreme capsules of the FMRIB58 template with something that is absent from your image. If you look at the alignment of the white matter except in the putamen/insula region, fNirt is actually doing a good job (cf snapshot attached).
I'm afraid there is no simple way to solve this problem (if any). Using fLirt is certainly not an option. Hopefully this does not happen in too many subjects and you simply (though painfully) have to reject those subjects from the TBSS analysis.
If it does occur in too many FA images though, then get back to me and we (with Jesper) will try to come up with something to try to help the registration.
Cheers,
Gwenaelle
--- En date de : Mer 27.5.09, Sushravya Raghunath <[log in to unmask]> a écrit :
> De: Sushravya Raghunath <[log in to unmask]>
> Objet: Re: [FSL] Re : [FSL] registered images distorted - tbss_3_postreg step
> À: [log in to unmask]
> Date: Mercredi 27 Mai 2009, 17h31
> Hi Gwenaelle,
>
> Sorry about that. Please find all the files at
> 466544.
>
> Thanks,
> Sushravya
>
>
>
> On Wed, May 27, 2009 at 10:28 AM,
> Gwenaëlle DOUAUD <[log in to unmask]>
> wrote:
>
> Hi Sushravya,
>
> I would actually need **all** the files concerning this
> subject (to be sure that you put all of them, do an ls
> *C_20090417_35938* in your FA/ directory).
>
>
> Thanks,
> Gwenaelle
>
> --- En date de : Mer 27.5.09, Sushravya Raghunath <[log in to unmask]> a écrit :
>
>
>
> > De: Sushravya Raghunath <[log in to unmask]>
> > Objet: Re: [FSL] Re : [FSL] registered images
> distorted - tbss_3_postreg step
>
> > À: [log in to unmask]
> > Date: Mercredi 27 Mai 2009, 15h36
>
>
>
> > Hi Gwenaelle,
> >
> > I have uploaded files as you had asked. Please find
> it
> > at 986721.
> >
> > Thanks,
> > Sushravya
> >
> >
> >
> > On Wed, May 27, 2009 at 5:41 AM,
>
> > Gwenaëlle DOUAUD <[log in to unmask]>
> > wrote:
> >
> > Hi Sushravya,
> >
> > sorry I was not clear enough: could you please put
> *all*
>
> > the files and images corresponding to this
> > "C_20090417_35938" subject (_FA.nii.gz,
> > *_warp.nii.gz etc.) in a tgz file and upload it?
> >
> >
> > Thanks,
> >
> > Gwenaelle
>
> >
> >
> > --- En date de : Mar 26.5.09, Sushravya Raghunath
> <[log in to unmask]> a écrit :
> >
> >
>
> > > De: Sushravya Raghunath <[log in to unmask]>
> > > Objet: Re: [FSL] Re : [FSL] registered images
>
> > distorted - tbss_3_postreg step
> >
> >
> > > À: [log in to unmask]
> > > Date: Mardi 26 Mai 2009, 21h24
> >
> > > Hi Gwenaelle,
> > >
>
> > > The commands used were tbss_2_reg -T and
> > > tbss_3_postreg -S for tbss steps 2 and 3
> > respectively.
> > >
> > >
> > > Have uploaded an example. The reference number
> is
> >
>
> > > 129604.
> > > Thanks for the reply. Awaiting suggestions.
> > > Thanks,
> > > Sushravya
> > >
> > >
> > >
> > >
> > >
> > > On Tue, May 26, 2009 at 2:54 PM,
>
> >
> > > Gwenaëlle DOUAUD <[log in to unmask]>
> >
> >
> >
> > > wrote:
> > >
> > > Hi Sushravya,
> > >
> > > by no mean a flirt (or any affine) registration
> would
>
> > be
> > > enough to match the WM across the subjects,
> especially
> > if
> > > you're dealing with healthy elderly and MCI
> > patients!
> >
> > >
> > >
> > > What you might have found in the archives is
> that
>
> > flirt is
> > > working better than areg (present in an old
> version of
> > TBSS)
> > > for placing the best target into the MNI space...
> This
> > is
> > > obsolete now (we only use flirt and fnirt) and is
> not
>
> > valid
> >
> > > if you have used the tbss2 -T recommended option
> > anyway.
> > >
> > >
> > > What are the commands/option you're run
> exactly
> > for
> > > tbss2 and 3? Could you please maybe upload (http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi)
>
> >
> > > an example of one FA_to_target that has failed?
> > >
> > >
> > > Cheers,
> > > Gwenaelle
> > >
> > > --- En date de : Mar 26.5.09, Sushravya
> Raghunath
> > <[log in to unmask]> a écrit :
>
> >
> > >
> > >
> > > > De: Sushravya Raghunath <[log in to unmask]>
> > > > Objet: [FSL] registered images distorted -
>
> >
> > > tbss_3_postreg step
> > >
> > > > À: [log in to unmask]
> > > > Date: Mardi 26 Mai 2009, 19h34
> > >
> > >
> > >
>
> > > > Experts,
> > > >
> > > > I am running TBSS Analysis of Control versus
> MCI
> >
> > > human
> > > > subjects.
> > > >
> > > > The _FA_to_target.nii.gz images (resulting
> after
>
> > > > tbss_3_postreg step) are
> > >
> > > > distorted. In some cases the regsitered
> images
> > have
> >
> > > > truncated edges whereas
> > > > FA image is complete.
>
> > > >
> > > > I read in older posts that FLIRT
> registration
> > might
> > > be
> > > > better and registered the
> > >
> > > > FA image to MNI152_T1_1mm_brain usifn FLIRT.
> The
>
> >
> > > FLIRT
> > > > registered images
> > > > look better. The truncated edges are still
> seen
> > with
> > > FLIRT
> > > > registration but they
> > > > are better than tbss registered images.
>
> > >
> >
> > > >
> > > > Is this the right way of going about in
> such
> > > situation?
> > > >
> > > > If yes, how can I have FLIRT registration
> done
> > in
> > > > tbss_3_postreg step?
>
> > > >
> > > > Any suggestion would be helpful.
> >
> > > >
> > >
> > > > Thanks,
> > > > Sushravya
> > > >
> > >
> > >
> > >
> > >
>
> > >
> > >
> >
> >
> >
> >
> >
> >
>
>
>
>
>
>
|