There's no need for the standard normalize. You would use the Segment button.
When you do this, there is no point in generating any normalised, modulated
or native space images - unless they are to be used at some later point. You
just need the seg_sn.mat file, which allows you to import into DARTEL.
I didn't understand what you meant by "TEMPLATE DONE". Here, you just need to
run DARTEL, such that it creates its own template from an average of all the
individuals.
When you create warped, you would specify that modulation is to be included.
There is no need for an additional modulation step.
Getting results in MNI space can be achieved by using the Normalise button to
obtain an affine spatial normalisation of the Template_6.nii file, by
matching it with the tpm/grey.nii . Then the resulting Template_6_sn.mat can
be used with the following, which can be used to modify the headers of the
smoothed DARTEL warped and modulated images so that they are brought into MNI
space....
% Select files
PN = spm_select(1,'.*_sn.mat','Select sn.mat file');
PI = spm_select(inf,'nifti','Select images');
% Determine affine transform from header
sn = load(deblank(PN));
M = sn.VG(1).mat/(sn.VF(1).mat*sn.Affine);
% Scaling by inverse of Jacobian determinant, so that
% total tissue volumes are preserved.
scale = 1/abs(det(M(1:3,1:3)));
% Pre-multiply existing headers by affine transform
for i=1:size(PI,1),
% Read header
Ni = nifti(deblank(PI(i,:)));
% Pre-multiply existing header by affine transform
Ni.mat = M*Ni.mat;
Ni.mat_intent='MNI152';
% Change the scalefactor. This is like doing a "modulation"
Ni.dat.scl_slope = Ni.dat.scl_slope*scale;
% Write the header
create(Ni);
end
Whether or not you use parametric or non-parametric statistics is up to you.
There is evidence that parametric is fine for large numbers of subjects, and
balanced designs (Ashburner & Friston, NeuroImage). If the designs are not
balanced (eg comparing one subject against a population), then you may need
to smooth the data by about 12mm in order to ensure that the parametric stats
are reliable (Salmond et al, NeuroImage). That's all I can say.
Best regards,
-John
On Tuesday 05 August 2008 21:36, you wrote:
> Dr. Asburner:
> Please check the slide is our understandig of DARTEL, please writte if we
> are wrong and please correct the slide. Thank you Luis Almeida.
>
>
> De : "John Ashburner" [log in to unmask]
> Para : "LUIS GUILLERMO ALMEIDA" [log in to unmask]
> Copia :
> Fecha : Tue, 05 Aug 2008 18:50:19 +0100
> Asunto : Re: Ref:Re: Ref:Re: Question about SPM5
>
> > If you download and install the latest updates for SPM5, then it is in
> > there. This document (
> > http://www.fil.ion.ucl.ac.uk/~john/misc/dartel_guide.pdf ) should explain
> > how to use it.
> >
> > -John
> >
> > On Tuesday 05 August 2008 18:17, you wrote:
> > > Dear JOHN:
> > > How I use DARTEL and where I obtain?
> > >
> > >
> > > De : "John Ashburner" [log in to unmask]
> > > Para : "LUIS GUILLERMO ALMEIDA" [log in to unmask]
> > > Copia :
> > > Fecha : Tue, 05 Aug 2008 17:31:57 +0100
> > > Asunto : Re: Ref:Re: Question about SPM5
> > >
> > > > Dear Luis,
> > > > I would much rather use the Unified Segmentation approach, rather
> > > > than the "optimised VBM". For best results though, I would suggest
> > > > also trying the DARTEL toolbox (found in the latest updates of SPM5).
> > > >
> > > > best regards,
> > > > -John
> > > >
> > > > On Tuesday 05 August 2008 15:42, you wrote:
> > > > > Thank you. What do you think obut using optimized VBM, now with
> > > > > SPM5 in better or not to use it, (Good et al 2001)
> > > > >
> > > > >
> > > > > De : "John Ashburner" [log in to unmask]
> > > > > Para : "LUIS GUILLERMO ALMEIDA" [log in to unmask]
> > > > > Copia :
> > > > > Fecha : Tue, 05 Aug 2008 11:34:50 +0100
> > > > > Asunto : Re: Question about SPM5
> > > > >
> > > > > > Dear Luis,
> > > > > > This is a result of the default bounding box not covering the
> > > > > > entire brain. If you want to extend the bounding box to cover the
> > > > > > whole cerebellum, then set the third element to some more
> > > > > > negative value (maybe subtract about 30mm from it). See the
> > > > > > manual for more information on this.
> > > > > >
> > > > > > Best regards,
> > > > > > -John
> > > > > >
> > > > > > On Tuesday 05 August 2008 04:47, you wrote:
> > > > > > > Dear Jhon:
> > > > > > > I have a question about the image produced by SPM 5 after
> > > > > > > spatial normalization. I am using the T1 templete which comes
> > > > > > > with the SPM5 Software, and I am Trying to get quality images,
> > > > > > > but I dont understand why the brain stem and cerebellum apears
> > > > > > > cutted as you can see in the slide in the attechment (please
> > > > > > > see the slide attached). My question is why it is cutted?, It
> > > > > > > is a problem of the visor of MRICro? or is a Failure in
> > > > > > > precessing data in SPM5. Can you help me to fix it? Again, and
> > > > > > > again thank you for your assistence, is not for you, I never
> > > > > > > ended my PhD. Lis Almeida Querétaro México.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > De : "John Ashburner" [log in to unmask]
> > > > > > > Para : "LUIS GUILLERMO ALMEIDA" [log in to unmask]
> > > > > > > Copia :
> > > > > > > Fecha : Thu, 31 Jul 2008 12:07:52 +0100
> > > > > > > Asunto : Re: Ref:Re: Information about SPM5
> > > > > > >
> > > > > > > > Dear Luis,
> > > > > > > > Figure 8 of the following paper shows the largest t value for
> > > > > > > > different levels of smoothing when a FWHM of 13mm is used.
> > > > > > > >
> > > > > > > > "Voxel-Based Morphometry Using the RAVENS Maps: Methods and
> > > > > > > > Validation Using Simulated Longitudinal Atrophy"
> > > > > > > > Christos Davatzikos, Ahmet Genc, Dongrong Xu and Susan M.
> > > > > > > > Resnick NeuroImage (2001) - not sure of volume, pages etc.
> > > > > > > >
> > > > > > > > This was with an old spatial normalisation approach, which
> > > > > > > > was a bit less accurate. The unified segmentation approach
> > > > > > > > should be slightly more accurate, so 10-12mm would seem to be
> > > > > > > > reasonable. Based on this work (which used only simulated
> > > > > > > > atrophy) 8mm smoothing for more precise registration would
> > > > > > > > also seem to be about right.
> > > > > > > >
> > > > > > > > Best regards,
> > > > > > > > -John
> > > > > > > >
> > > > > > > > > So you recomend to use between 10 and 12mm of FWHM if I am
> > > > > > > > > doing VBM comparing differences between two groups in
> > > > > > > > > smooting step before segmentation?. On the other hand, can
> > > > > > > > > you give to me the reference of the study of Davatzikos'
> > > > > > > > > group?. Thank you very much. Luis Almeida
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > De : "John Ashburner" [log in to unmask]
> > > > > > > > > Para : "LUIS GUILLERMO ALMEIDA" [log in to unmask]
> > > > > > > > > Copia :
> > > > > > > > > Fecha : Wed, 30 Jul 2008 19:32:29 +0100
> > > > > > > > > Asunto : Re: Information about SPM5
> > > > > > > > >
> > > > > > > > > > Dear Luis,
> > > > > > > > > >
> > > > > > > > > > > Hi again, my name is Luis Almeida from UNAM in México
> > > > > > > > > > > and I am using SPM5 in order to make my PhD degree. I
> > > > > > > > > > > am working with VBM but I have some questions. First:
> > > > > > > > > > > Hoy do yuo choose te size of FWHM in the smoothing
> > > > > > > > > > > step?
> > > > > > > > > >
> > > > > > > > > > Ideally, once you have finished a VBM analysis, and
> > > > > > > > > > submitted your results for publication, then you can
> > > > > > > > > > experiment with the data. You could try different amounts
> > > > > > > > > > of smoothing to see what gives the most significant
> > > > > > > > > > corrected p-values - and then use these settings for your
> > > > > > > > > > next analysis. Alternatively, you could use similar
> > > > > > > > > > values to everyone else. For the spatial normalisation
> > > > > > > > > > with the Segment button, I would suggest 10-12mm. If you
> > > > > > > > > > try DARTEL for your spatial normalisation, then you could
> > > > > > > > > > try about 8mm. Note that these values are just educated
> > > > > > > > > > guesses (although the 10-12mm is from an evaluation paper
> > > > > > > > > > by Davatzikos' group).
> > > > > > > > > >
> > > > > > > > > > Note that the results may be easier to interpret if you
> > > > > > > > > > use more smoothing, as this is likely to down-weight some
> > > > > > > > > > of the differences due to systematic mis-registration.
> > > > > > > > > >
> > > > > > > > > > > Second. Once the program has computed the differences
> > > > > > > > > > > in the botom apears some information, can you explain
> > > > > > > > > > > me what it means? (see the attached power point slide.
> > > > > > > > > > > Third..Why the acivation area doesn´t represent the
> > > > > > > > > > > real volume of the difference?, so we must use another
> > > > > > > > > > > toolbox as MRICRO or IBASPM in order to calculate the
> > > > > > > > > > > real volume of the difference?. Thank you very much.
> > > > > > > > > >
> > > > > > > > > > I would suggest looking at some of the work by Tom
> > > > > > > > > > Nichols for the FDR information. Much of the other
> > > > > > > > > > information may be explained by chapters 14 and 15 of
> > > > > > > > > > http://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/
> > > > > > > > > >
> > > > > > > > > > best regards,
> > > > > > > > > > -John
