Date: Thu, 7 Jun 2012 07:52:35 +0930
From:
[log in to unmask]Subject: Re: [FSL] using TFCE on tstat maps
To:
[log in to unmask]Dear All,
Just to clarify the situation here - there is no problem with the NIFTI standard,
the AFNI or FSL interpretations of NIFTI (which are the same) or AFNI/FSL
interoperability.
We therefore do not see a problem with using NIFTI files created with
to3d from AFNI, as long as you check that there is a valid sform or qform
(use fslhd and check that at least one of qform_code or sform_code is
non-zero - see below).
The previous problems that were being brought up here were to do with
images converted from ANALYZE versions and datasets with mixed
orientation (some radiological and some neurological). Although such a
mixed dataset is OK with FSL it does not work with AFNI and in general
should be avoided. The issue with converting ANALYZE images is more
serious. In such a case there is no information about whether the image
is radiologically or neurologically stored (equivalent to left-handed or
right-handed coordinate systems) and the resultant NIFTI image normally
has *no* orientation information set (both qform and sform codes are zero
and are therefore "Unknown"). Such images can be run through different
packages but will rely on those package defaults, which is dangerous as
these defaults may not be consistent and may not be correct with respect
to what the true data orientation is. The only good solution to this is to
determine somehow (from markers in the image or comparing to a
dicom viewer, or something like this) what the true orientation is and then
modifying the NIFTI image to include this information. This can be done
using the tool fslorient (where either the qform or sform code is set to
non-zero and the appropriate matrix entries are specified - mainly to set
the sign of the determinant). The image should then be checked in FSLView
to make sure that the labels are correctly placed with respect to the brain
anatomy. Once this is done the NIFTI image can be used safely. Without
doing this any NIFTI image that does not have a valid qform or sform (where
both codes are zero) should be avoided as it is likely to cause problems.
I personally would recommend avoiding ANALYZE altogether and
reconverting from the original source (e.g. DICOM) directly to NIFTI as
that is more likely to encode the orientation information correctly.
All the best,
Mark
On 5 Jun 2012, at 09:40, Christine Zakrzewski wrote:
MJ--Apparently, there is something about the NifTI standard (sform, qform) that AFNI & FSL interpret differently. This is the problem that I encountered last fall. I have talked with Bob about, and a consensus has supposedly never been reached. I would warn users against using NifTI files created with to3d with FSL programs, better to stick with dcm2nii if your preprocessing and higher-level analysis are to be performed with FSL tools. See you in Beijing--CZ
Date: Tue, 5 Jun 2012 09:09:59 +0930
From:
[log in to unmask]Subject: Re: [FSL] using TFCE on tstat maps
To:
[log in to unmask]Hi,
My understanding was that AFNI would write out valid NIFTI files and
that FSL would read valid NIFTI files correctly. I do not remember that
we have found a current situation in which there is an existing problem.
I was under the impression that this reflected a problem in old versions
of either FSL or AFNI and at the present could not be replicated. I have
not seen an example of a valid NIFTI file that FSL did not read correctly,
so until I see this example I would not be concerned about left-right
flipping as I am very confident that FSL reads valid NIFTI files completely
correctly.
All the best,
Mark
On 5 Jun 2012, at 00:54, Christine Zakrzewski wrote:
You misunderstand, Michael.
It is not that there is something about the nifti header that FSL reads incorrectly. It is a problem when creating nifti files with to3d and dicom and then working across platforms that creates the issue. When creating nifti files with afni programs, there is something about the sform and qform fields that fsl does not read correctly. Have talked about this with mj extensively, bob cox--not so much!
Christine
> Date: Mon, 4 Jun 2012 09:52:29 -0500
> From:
[log in to unmask]> Subject: Re: [FSL] using TFCE on tstat maps
> To:
[log in to unmask]>
> Hi Christine,
> Can you please elaborate what in the nifti is not read correctly by FSL,
> as I'm sure that Mark Jenkinson would be interested in this if it is
> indeed the case. If you have a properly specified nifti header, then
> FSL will read it properly and there need not be any concern about R/L
> flipping within the suite of FSL tools.
>
> cheers,
> -MH
>
> On Mon, 2012-06-04 at 14:38 +0000, Christine Zakrzewski wrote:
> > Just be careful, some information in the nifti header (sform, qform)
> > created with afni programs, like to3d and dimon, will not be read in
> > correctly by fsl--which could cause right/left flipping. I would
> > reccomend using 3dclustsim to correct for multiple comparisons and
> > clusterize in afni.
> > Be careful.
> > Christine
> >
> >
> > > Date: Mon, 4 Jun 2012 15:49:05 +0200
> > > From:
[log in to unmask]> > > Subject: [FSL] using TFCE on tstat maps
> > > To:
[log in to unmask]> > >
> > > Hi all,
> > >
> > > I received some tstat maps that were created in a higher-order
> > analysis using AFNI. I would like to apply cluster-based thresholding,
> > ideally using TFCE. Could anyone comment on the best way to go about
> > this? I guess I should first convert to Zstat maps?
> > >
> > > Cheers,
> > >
> > > Diederick
> >