Hi,
What exactly do you mean by "off"?
Is it just a simple translation?
Have you tried using the translation only option for flirt (one of the
GUI options or using a schedule file at the command line)?
I'm not sure why you cannot get the registration working with your
images.
If you upload them to our site then I'll have a look at them.
It should be possible to get things registered - especially if you
think they
already are in minc format.
All the best,
Mark
On 12 Mar 2008, at 23:32, Roman Loonis wrote:
> I used fslview to check the result and the files were off with
> respect to each other. I originally did not check with other
> software. I knew the files were perfectly registered when the files
> are in minc format. I used a program called register which uses the
> center of the volume as the origin. If I understand what you are
> saying, then this means that the files are not properly registered
> for FSL. How can I do the extra registration? I tried employing the
> FSL registration procedure (which I have successfully down for
> human brains), but for this set of monkey brains I am unable to get
> proper registering with FSL. So what should I do?
>
> Thanks for the help,
>
> Roman
>
>
> On 3/12/08, Mark Jenkinson <[log in to unmask]> wrote: Hi,
>
> Just look at diff2anat and anatomical in the same fslview.
> If the ident.mat really was the correct transformation then these
> should align.
>
> However, this seems to be what I suggested before in this rather
> longish thread
> and you were unhappy with the result. But did you use fslview to
> check it or
> some other software? You need to beware of checking with other
> software.
> It also might relate to how you know they are "perfectly registered",
> as if this
> is based on assuming the centre of volume is the origin, then for FSL
> these
> images will not be registered at all (if the FOV is different) and
> will require some
> extra registration.
>
> Let me know what the result is of looking at these files in fslview,
> and if they
> are not right, how you determine that the original files are properly
> registered.
>
> All the best,
> Mark
>
>
> On 12 Mar 2008, at 20:54, Roman Loonis wrote:
>
> > I inputed the following command:
> >
> > flirt -in diffimage -ref anatomical -init $FSLDIR/etc/flirtsch/
> > ident.mat -applyxfm -out diff2anat
> >
> > The images were off in a manner which would seem to be consistent
> > with the fact that a corner was chosen for the alignment centre
> > instead of the center. Does this mean everything is alright? How
> > can I be sure?
> >
> > Roman
> >
> > On Wed, Mar 12, 2008 at 1:50 PM, Mark Jenkinson
> > <[log in to unmask]> wrote:
> > Hi,
> >
> > What reference image did you use when applying flirt?
> >
> > It is the reference image that determines both the resolution and
> > the Field Of View. When checking if it is "aligned" to another
> image
> > with a different FOV, then the choice of where to put the
> "centre" is
> > arbitrary and each package may do a different thing. We choose the
> > corner as the alignment centre, but quite often the centre of the
> > volume
> > is chosen instead. If your viewing program chose this, then the
> > images
> > would appear to be non-aligned even when they would be considered
> > to be aligned within the FSL pipeline.
> >
> > All the best,
> > Mark
> >
> >
> > Roman Loonis wrote:
> > > Just to repeat. I applied an identity transformation, using the
> > flirt
> > > command as requested, and compared the image to itself (using a
> > > program which can look to see if pictures are registered without
> > > needing to have the same image dimensions). The transformed image
> > was
> > > off.
> > >
> > > Roman
> > >
> > > On Wed, Mar 12, 2008 at 11:25 AM, Roman Loonis
> > <[log in to unmask]
> > > <mailto:[log in to unmask]>> wrote:
> > >
> > > I did output an image with the transformation applied to
> > it... "I
> > > applied the identity transformation just as you suggested
> to the
> > > diffusion image." My data is not in neurological orientation.
> > The
> > > problem I believe occurred while using flirt. It may not have
> > > dealt properly with the changing of resolutions.
> > >
> > > Roman
> > >
> > > On Thu, Mar 6, 2008 at 5:30 PM, Matt Glasser <[log in to unmask]
> > > <mailto:[log in to unmask]>> wrote:
> > >
> > > Simply finding a transformation between your two datasets
> > with
> > > FDT is not the same as outputting an image which has had
> > this
> > > transformation applied to it and is thus physically in the
> > > same space as the reference image. Also, make sure your
> > data
> > > is not in neurological orientation, as some FSL programs
> > have
> > > a problem with neurologically oriented data (they flip it
> > over
> > > the y axis so that left becomes right and right becomes
> > left).
> > >
> > >
> > >
> > > Peace,
> > >
> > >
> > >
> > > Matt.
> > >
> > >
> > >
> > >
> >
> ----------------------------------------------------------------------
> > --
> > >
> > > *From:* FSL - FMRIB's Software Library
> > > [mailto:[log in to unmask] <mailto:[log in to unmask]>]
> *On
> > > Behalf Of *Roman Loonis
> > > *Sent:* Thursday, March 06, 2008 4:04 PM
> > > *To:* [log in to unmask] <mailto:[log in to unmask]>
> > > *Subject:* Re: [FSL] Registration for Diffusion Data
> > >
> > >
> > >
> > > I applied the identity transformation just as you
> > suggested to
> > > the diffusion image. However, the resulting diffusion
> image
> > > was not properly registered. With respect to the
> anatomical,
> > > the diffusion image just appeared to be off by a few
> voxels
> > > (however in the correct orientation). I knew, nonetheless,
> > > that the two images are perfectly registered. As a
> > > consequence, I took the original diffusion image and
> > > superimposed the transformed diffusion image by converting
> > > them into mincs and subsequently using the minc program
> > > register to compare the two. The exact same problem was
> > > observed. The transformation using the identity matrix in
> > > flirt altered the image. What should I do?
> > >
> > >
> > >
> > > Roman
> > >
> > > On Sat, Feb 23, 2008 at 12:32 PM, Mark Jenkinson
> > > <[log in to unmask] <mailto:[log in to unmask]>> wrote:
> > >
> > > Hi,
> > >
> > > I would first check that the identity transformation truly
> > > does relate
> > > your images. Do this by running flirt:
> > > flirt -in diffimage -ref anatomical -init
> > > $FSLDIR/etc/flirtsch/
> > > ident.mat -applyxfm -out diff2anat
> > > then load diff2anat and anatomical into fslview. If
> you are
> > > happy with
> > > this then you can copy $FSLDIR/etc/flirtsch/ident.mat
> > into the
> > > place
> > > where the diffusion to structural registration file would
> > normally
> > > be saved and everything should work from then on.
> > >
> > > Alternatively, you can use the above flirt command to
> > manually
> > > resample
> > > masks or other images between the two resolutions (in
> either
> > > direction).
> > > Remember to rethreshold and binarise your mask image after
> > > transformation,
> > > as voxels may only partly overlap and hence generate
> values
> > > between 0
> > > and 1.
> > >
> > > All the best,
> > > Mark
> > >
> > >
> > >
> > >
> > > On 23 Feb 2008, at 14:32, Roman Loonis wrote:
> > >
> > > > Yeah I understand that. So does that mean if two images
> > which are
> > > > already registered, but have two different resolutions,
> > are
> > > > considered in the same space for probtrackx? How can I
> > get a mask
> > > > on one be the same on the other despite the reduction in
> > > resolution?
> > > >
> > > > Thank you,
> > > > Roman
> > > >
> > > > On Sat, Feb 23, 2008 at 2:24 AM, Steve Smith
> > > <[log in to unmask] <mailto:[log in to unmask]>>
> > > > wrote:
> > > > Hi - it doesn't quite work like that. If you use the
> > > registration part
> > > > of the FDT GUI then the transform between diffusion and
> > > structural
> > > > data will get computed and saved.
> > > >
> > > > Then when you specify a mask in structural space you
> just
> > > tell FDT
> > > > that this is the space it's in, and it will use the
> > appropriate
> > > > transform for you, when running the tractography.
> > > >
> > > > Cheers.
> > > >
> > > >
> > > >
> > > > On 21 Feb 2008, at 19:08, Roman Loonis wrote:
> > > >
> > > > > I have already registered my images (the anatomical
> > to the
> > > diffusion
> > > > > data) prior to using bedpostx
> > > > > and I have modified correspondingly the resulting
> > changes
> > > in the
> > > > > vector directions for my bvecs file.
> > > > > I had to do so because I am looking at monkey
> diffusion
> > > data. My
> > > > > question is the following...
> > > > > Although the files are properly registered and
> therefore
> > > need no
> > > > > transformation file, they still have
> > > > > different resolutions. When I input my masks which I
> > have
> > > drawn on
> > > > > the anatomical, do I need to
> > > > > indicate in probtrackx that the seed space is not the
> > diffusion
> > > > > space? How can I make sure that there
> > > > > is a proper translation of the higher resolution mask
> > into
> > > a lower
> > > > > resolution one?
> > > > >
> > > >
> > > >
> > > >
> >
> ----------------------------------------------------------------------
> > > > -----
> > > > 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] <mailto:[log in to unmask]>
> > > http://www.fmrib.ox.ac.uk/~steve
> > > <http://www.fmrib.ox.ac.uk/%7Esteve>
> > > >
> >
> ----------------------------------------------------------------------
> > > > -----
> > > >
> > >
> > >
> > >
> > >
> > >
> >
>
|
