> Would you recommend that one warps the native space c1 images, rather than
> the imported images in the warp step? The idea would be to reduce the
> number of resamplings by 1.
I think I would, although I don't know how much difference it makes in
practice. Note also that the new "modulated" warping to MNI space does
things a bit differently. Rather than resampling the original image and
multiplying by the Jacobian determinants, instead it scans through the
original and adds the values to the appropriate location in the warped image.
This exactly preserves the tissue volumes. A side effect is that it can
introduce aliasing effects, but these are reduced if the original image is at
a higher resolution.
>
> Would this be done with the create warped step or through the deformation
> toolbox?
In SPM8(b), I would suggest using the new function that warps data to MNI
space.
Best regards,
-John
> On Wed, Apr 1, 2009 at 4:42 PM, John Ashburner
<[log in to unmask]>wrote:
> > This kind of works, but when applying a GLM to the imported data, the
> > information in the second matrix will not be written to the NIfTI headers
> > of
> > the beta and contrast images. This means that DARTEL won't apply the
> > right warps.
> >
> > DARTEL "knows" the relationship between the orientations in the imported
> > images, and the originals via this matrix. The idea of this is that
> > DARTEL can be used to warp original (non-imported) scans. If it is used
> > to try to warp an imported scan, which does not have the second matrix
> > (eg beta or contrast images), then it all goes horribly wrong.
> >
> > Best regards,
> > -John
> >
> > On Wednesday 01 April 2009 21:13, MCLAREN, Donald wrote:
> > > While it was originally written written to import c1, c2, and c3 files;
> > > there is no requirement that the images be segments.
> > >
> > > We've used the function for DTI data as well.
> > >
> > > It should be noted that DARTEL can't use fMRI data to create the
> > > deformation field, but that by importing another modality into DARTEL
> > > the same way as segments, then the deformations can be applied.
> > >
> > > This is true, at least, for DTI data
> > >
> > > On Wed, Apr 1, 2009 at 3:04 PM, John Ashburner
> >
> > <[log in to unmask]>wrote:
> > > > If you have the imported GM and WM (rc1 and rc2) files, then you
> > > > would run DARTEL using these. If I remember correctly the
> >
> > "import_for_dartel.m
> >
> > > > function was to take c1 and c2 images, and do the necessary tweaks to
> > > > obtain
> > > > the rc1 and rc2 images that DARTEL can work with. You would not
> > > > apply
> >
> > it
> >
> > > > to
> > > > the fMRI data.
> > > >
> > > > Best regards,
> > > > -John
> > > >
> > > > On Wednesday 01 April 2009 19:19, Amir Tahmasebi wrote:
> > > > > Hi John,
> > > > >
> > > > > I used the "import_for_dartel.m" file from another posting dated on
> > > > > Wed. 6 Feb. 2008.
> > > > >
> > > > > Regards,
> > > > > Amir
> > > > >
> > > > > John Ashburner wrote:
> > > > > > I'm not quite sure what you mean by " use the code provided by
> > > > > > John to transfer both fMRI and structural data to DARTEL space".
> > > > > > If this means using the create DARTEL warped images, then then
> > > > > > doing this again to
> > > >
> > > > the
> > > >
> > > > > > contrast images (step 4) is going to move the data out of
> >
> > alignment.
> >
> > > > > > If you have SPM8b (soon to be SPM8), then maybe try skipping the
> >
> > "use
> >
> > > > the
> > > >
> > > > > > code provided by John" step. Generate the contrast images in
> >
> > native
> >
> > > > > > space (without smoothing), and simply warp them using the new
> > > >
> > > > "Normalise
> > > >
> > > > > > to MNI space" option of DARTEL.
> > > > > >
> > > > > > Best regards,
> > > > > > -John
> > > > > >
> > > > > > On Wednesday 01 April 2009 18:24, you wrote:
> > > > > >> Dear John and SPMers,
> > > > > >>
> > > > > >> I am facing a problem while performing DARTEL-based
> > > > > >> normalization
> >
> > of
> >
> > > > my
> > > >
> > > > > >> fMRI data followed by group analysis.
> > > > > >>
> > > > > >> I followed the below routine to conduct first level and second
> >
> > level
> >
> > > > > >> analysis on my fMRI data:
> > > > > >>
> > > > > >> (1) coregister fMRI and structural for each subject
> > > > > >> (2) perform DARTEL among segmented structural data
> > > > > >> (2) use the code provided by John to transfer both fMRI and
> > > > > >> structural data to DARTEL space
> > > > > >> (3) perform first-level analysis of the fMRI data --> output:
> > > > > >> con
> > > >
> > > > files
> > > >
> > > > > >> (4) in order to perform second level analysis, I apply DARTEL
> > > >
> > > > parameters
> > > >
> > > > > >> to warp my con files (using "create warp").
> > > > > >>
> > > > > >> The problem is at step 4: the warped con images seems not to
> > > > > >> have correct orientation. However, if I apply the same
> > > > > >> parameters to
> >
> > the
> >
> > > > fMRI
> > > >
> > > > > >> data that are transformed to DARTEL space (from step 2), the
> > > > > >> data look OK.
> > > > > >>
> > > > > >> Is there anything that I am missing?
> > > > > >>
> > > > > >> Thank you,
> > > > > >>
> > > > > >> Regards,
> > > > > >> Amir
> > > > > >>
> > > > > >> John Ashburner wrote:
> > > > > >>> This will be straightforward in SPM8 - and it will combine the
> > > > > >>> transforms with an affine transform so that the spatially
> > > > > >>> normalised images are in MNI space, rather than the population
> > > > > >>> average. SPM8
> > > >
> > > > will
> > > >
> > > > > >>> be released very soon, but I think the necessary routines are
> > > > > >>> already available in the recent updates for SPM8b.
> > > > > >>>
> > > > > >>> An alternative strategy could be to spatially normalise
> > > > > >>> contrast
> > > >
> > > > images
> > > >
> > > > > >>> from a first level analysis done in native space, prior to
> > > > > >>> doing the 2nd level analysis. This could be done one scan per
> > > > > >>> subject.
> > > > > >>>
> > > > > >>> Best regards,
> > > > > >>> -John
> > > > > >>>
> > > > > >>> On Monday 30 March 2009 17:24, Amir Tahmasebi wrote:
> > > > > >>>> Dear John,
> > > > > >>>>
> > > > > >>>> I am using DARTEL to warp fMRI data. Using DARTEL's "create
> > > > > >>>> warped" tool, for every image you have to provide one flow
> >
> > field.
> >
> > > > > >>>> Is there
> > > >
> > > > any
> > > >
> > > > > >>>> way that I can apply one flow field to a bunch of images at
> >
> > once?
> >
> > > > > >>>> Thank you,
> > > > > >>>>
> > > > > >>>> Regards,
> > > > > >>>> Amir
> > > > > >>>>
> > > > > >>>> John Ashburner wrote:
> > > > > >>>>> DARTEL is run as described in the user guide, but there is an
> > > > > >>>>> extra option that allows the resulting deformations to be
> > > > > >>>>> automatically combined with an additional affine transform.
> >
> > The
> >
> > > > > >>>>> DARTEL warps encode the inter-subject variability, and the
> >
> > affine
> >
> > > > > >>>>> transform
> > > >
> > > > warps
> > > >
> > > > > >>>>> the average of the population to MNI space.
> > > > > >>>>>
> > > > > >>>>> Best regards,
> > > > > >>>>> -John
> > > > > >>>>>
> > > > > >>>>> On Tuesday 24 March 2009 17:49, Amir Tahmasebi wrote:
> > > > > >>>>>> Hi John,
> > > > > >>>>>>
> > > > > >>>>>> Thanks for your reply.
> > > > > >>>>>>
> > > > > >>>>>> Does this mean that the fMRI data are only affinely
> > > > > >>>>>> registered between subjects?
> > > > > >>>>>>
> > > > > >>>>>> Regards,
> > > > > >>>>>> Amir
> > > > > >>>>>>
> > > > > >>>>>> John Ashburner wrote:
> > > > > >>>>>>> If you have SPM8b (latest updates), then you should find
> > > > > >>>>>>> this
> > > >
> > > > much
> > > >
> > > > > >>>>>>> easier, as there is now an option to spatially normalise
> > > > > >>>>>>> (and spatially blur) fMRI to MNI space (as it incorporates
> > > > > >>>>>>> an
> > > >
> > > > additional
> > > >
> > > > > >>>>>>> affine mapping between DARTEL template and MNI template).
> > > > > >>>>>>> I haven't had a chance to update the user guide for DARTEL
> > > > > >>>>>>> yet, but it should be reasonably straightforward to use.
> > > > > >>>>>>>
> > > > > >>>>>>> I expect that the t stats should be increased, and look
> >
> > forward
> >
> > > > to
> > > >
> > > > > >>>>>>> any feedback on whether this is the case.
> > > > > >>>>>>>
> > > > > >>>>>>> Best regards,
> > > > > >>>>>>> -John
> > > > > >>>>>>>
> > > > > >>>>>>> On Tuesday 24 March 2009 16:22, Amir Tahmasebi wrote:
> > > > > >>>>>>>> Dear John and SPMers,
> > > > > >>>>>>>>
> > > > > >>>>>>>> I am trying to use DARTEL for normalization of fMRI data.
> > > > > >>>>>>>> I have looked at "how to use DARTEL for normalizing fMRI
> > > > > >>>>>>>> con maps?" thread but it is not exactly what I was looking
> > > > > >>>>>>>> for.
> > > > > >>>>>>>>
> > > > > >>>>>>>> I was wondering how should I use the outputs of DARTEL
> > > > > >>>>>>>> ("u_rc1*.nii" and "Template_6.nii") to normalize my fMRI
> >
> > data
> >
> > > > > >>>>>>>> which is already coregistered to the corresponding
> >
> > structural
> >
> > > > > >>>>>>>> data.
> > > > > >>>>>>>>
> > > > > >>>>>>>> Thank you,
> > > > > >>>>>>>> Amir
|