Oh, I see. This seems to clear up things quite a bit. In fact, for these
3D images I have been using the DICOM toolbox. For others, I used
medcon, but it didn't work either.
avwinfo gives the following:
dim1 128
dim2 512
dim3 512
dim4 1
datatype 4
pixdim1 -1.4999920130
pixdim2 0.4687500000
pixdim3 0.4687500000
pixdim4 1.0000000000
cal_max 106.0000
cal_min 0.0000
glmax 0
glmin 0
origin1 64
origin2 256
origin3 256
file_type ANALYZE-7.5
Is there anything I could change in the header? I also noted that when I
looked at the images with MRICro, the look somewhat flattened in the x
dimension - Fslview did not complain, though...
I'll give the patch a try on my local machine (I have no writing rights
on our main cluster) - thanks again!
Cheers,
Cornelius
On Mon, 2005-06-06 at 10:38 +0200, Andreas Bartsch wrote:
> Good morning,
>
> Just to add to Steve's response:
> If you have, for example, acquired your data at a matrix of 512 but only sample a matrix of 256 you're effective resolution is FoV / 256 but some conversion routines will "pretend" that you have sampled by FoV / 512. I think it is possible that you have in fact sampled a resolution of ~ 1mm in plane but that due to the above the analyze or whatever format files are stored at ~ 0.5mm in-plane resolution. Are you, by any chance, using SPM's DICOM toolbox to convert Siemens data? What does avwinfo tell you?
>
> Cheers-
> Andreas
>
> -----Ursprüngliche Nachricht-----
> Von: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Im Auftrag von Stephen Smith
> Gesendet: Montag, 6. Juni 2005 10:22
> An: [log in to unmask]
> Betreff: Re: [FSL] FLIRT trouble
>
>
> Hi - I just ran FLIRT with the default settings - ie 12 DOF and normal search, and the epi->T1 registration ran fine. Your T1 does have the x and y voxel dimensions < 0.5mm which does probably mean that you're getting bitten by that old bug - I think this was patched in one of the patch releases after the original 3.2b FSL release, so you can either get this by taking the latest patch or just re-downloading FSL.
>
> Cheers, Steve.
>
>
> On Mon, 6 Jun 2005, Cornelius Werner wrote:
>
> > Hello Steve,
> >
> > thanks for your effort! I don't think my T1's are of very high
> > resolution - we usually do not have enough time for that :-) Besides,
> > I'm pretty sure I used the most recent version of FSL. Does FLIRT
> > depend on any external libraries that might not be in order?
> > Anyway: could you be so kind to tell me which parameters you provided
> > FLIRT with (or send me the log-file)? That would be greatly
> > appreciated.
> >
> > Cheers
> > Cornelius
> >
> >
> > On Sat, 2005-06-04 at 11:23 +0100, Stephen Smith wrote:
> > > Hi Cornelius, thanks for sending the data. Registering
> > > example_func01 to
> > > T1_bet_struct01 works fine for me - I think what's going on for you is an
> > > old FLIRT bug when one of the images is of very high resolution (<.5mm in
> > > any direction). If you take the latest FSL (or patch) download hopefully
> > > all should be ok.
> > >
> > > The patches are available from:
> > > http://www.fmrib.ox.ac.uk/fsldownloads/patches/3.2/
> > >
> > > Cheers, Steve.
> > >
> > >
> > > On Fri, 3 Jun 2005, Cornelius Werner wrote:
> > >
> > > > Hello all,
> > > >
> > > > I know, I posted this problem some time ago, but I am still having
> > > > problems with my T1 structurals. Neither can I register my EPIs
> > > > onto them, nor can I register the structurals to any template
> > > > image (avg152, colin27, whatever). Thus, I'm stuck with
> > > > registering my EPIs directly onto standard brains, which is far
> > > > from optimal. The problem is that FLIRTing to/from structurals
> > > > results in considerable offsets in all directions. Can this be due
> > > > to the origin in the structurals not being AC?
> > > >
> > > > I created the structurals as follows:
> > > >
> > > > SIEMENS Sonata DICOMs -> xmedcon -> ANALYZE -> avwchfiletype ->
> > > > NIFTI.GZ
> > > > -> Bet -> skull stripped structurals.
> > > >
> > > > In FSLView, they look ok, however...!
> > > >
> > > > Does anyone else also suffer from this problem? Anyone having a
> > > > solution?
> > > >
> > >
> >
>
--
Cornelius Werner
Institute of Medicine (IME)
Cognitive Neurology Group
Research Centre Juelich
52425 Juelich
Germany
Phone: +49-(0)2461-61-8609
Mail: [log in to unmask]
|