Dear Alina,
It would be useful if you could send me the SPM.mat corresponding to
your initial analysis (to look at the residuals' covariance matrix
SPM.xVi.Cy and the globals' estimate SPM.xGX.rg).
There seems to be a lot of spatial variability in the signal intensity
so it might not be too surprising if the rescaling of the data based on
the mean signal in space and time is not entirely satisfactory. Which
value do you actually want to compute? For the effect size, you might
want to rescale the time series with a local estimate of the mean
signal. It might be what MarsBar does or some code from Cyril:
http://journal.frontiersin.org/article/10.3389/fnins.2014.00001/full#h6
Best regards,
Guillaume.
On 05/04/17 11:21, Alina D wrote:
> Hello,
>
> Thank you very much for your help.
> I tried with no correlations and the design matrix indeed looks ok
> Inline image 4Inline image 6
>
> New mask file:
> Inline image 5
> And this is how the original images look like(after smoothing):
> Inline image 7
>
> Is there a way to fix this?
>
> Thank you very much!
>
> Kind regards,
> Alina
>
> On Tue, Apr 4, 2017 at 6:14 PM, Guillaume Flandin <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Dear Alina,
>
> The change in the design matrix between before and after estimation is
> due to non-sphericity whitening (adjusting for serial correlations). In
> your case, it seems it went wrong - to understand better what is going
> on, could you rerun your entire analysis while setting the "serial
> correlations" option in fMRI model specification to "none"?
> Also, what does your difference images look like? Before entering the
> GLM, the data are rescaled according the estimated global signal, and
> this estimation is based on a crude intensity thresholding approach to
> define the brain mask. What does the mask.nii image look like?
>
> Best regards,
> Guillaume.
>
>
>
> On 04/04/17 16:18, Alina D wrote:
> > Hello,
> >
> > These are actually arterial spin labelling difference images, so the
> > only processing was smoothing them by 6mm. Then I specified everything
> > by clicking on the specify first level command. What I noticed is that
> > after I am doing this step, indeed the " clean" design matrix (that I
> > just send before) comes up. Nevertheless, after model estimation and in
> > the results file the design matrix looks again like the initial one,
> > with the odd looking constant term. Could this be the source of error?
> > And if yes, why is this happening?
> >
> > Best wishes,
> > Alina
> >
> > On Apr 4, 2017 5:04 PM, "Guillaume Flandin" <[log in to unmask] <mailto:[log in to unmask]>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>>> wrote:
> >
> > Dear Alina,
> >
> > Your design matrix looks better this time: it contains two regressors
> > (task and constant) so you have two regression coefficients beta_1 and
> > beta_2. You are right to find the values rather large.
> > Where are the data coming from? How did you preprocess them? Did you use
> > SPM's batch interface to specify and estimate this first level GLM?
> >
> > Best regards,
> > Guillaume.
> >
> >
> > On 04/04/17 12:54, Alina D wrote:
> > > I apologize, I selected the wrong file accidentally.
> > > These are the betas. Could you please tell me what is the first
> > one and
> > > what is the second one?
> > > beta =
> > >
> > > 134.8669
> > > 196.3979
> > >
> > > And this is the correct design matrix :
> > >
> > > Inline image 2
> > >
> > > Best wishes,
> > >
> > > Alina
> > >
> > > On Mon, Apr 3, 2017 at 5:45 PM, Guillaume Flandin
> > <[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> > > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>> wrote:
> > >
> > > Dear Alina,
> > >
> > > Your second regressor, the constant term, looks odd. How did
> > you specify
> > > your GLM?
> > >
> > > Best regards,
> > > Guillaume.
> > >
> > > On 03/04/17 16:38, Alina D wrote:
> > > > Hello,
> > > >
> > > > Thank you again. This is how the design matrix looks like. I
> > only have
> > > > one stimulation condition.
> > > > Best wishes,
> > > > Alina
> > > > Inline image 2
> > > >
> > > > On Mon, Apr 3, 2017 at 5:29 PM, Guillaume Flandin
> > <[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>
> > > > <mailto:[log in to unmask]
> <mailto:[log in to unmask]> <mailto:[log in to unmask]
> <mailto:[log in to unmask]>>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>>> wrote:
> > > >
> > > > Dear Alina,
> > > >
> > > >
> > > > OK, you probably read the data correctly. Could
> you show us
> > > what your
> > > > design matrix look like? To explore a bit further,
> you can
> > > also use the
> > > > "plot" button with the "contrast estimates" and
> "fitted
> > responses"
> > > > options.
> > > >
> > > > Best regards,
> > > > Guillaume.
> > > >
> > > >
> > > > On 03/04/17 16:20, Alina D wrote:
> > > > > Hello,
> > > > >
> > > > > Thank you very much for replying. I did Extract
> data >
> > betas
> > > > this
> > > > > voxel and the only thing that happens is that matlab
> > returns
> > > this :
> > > > >
> > > > > beta =
> > > > >
> > > > > 87.2718
> > > > > 265.9859
> > > > >
> > > > > Before I tried to load the beta_0001.nii file and to
> > get the
> > > intensitied
> > > > > of some voxels from there.
> > > > >
> > > > > Qse = 'beta_0001.nii';
> > > > > Wse = spm_vol(Qse);
> > > > > Bse = spm_read_vols(Wse);
> > > > >
> > > > > I stored the coordinates of the voxels in a
> matrix of
> > the form:
> > > > > myindex3d = zeros(length(myindex),3)
> > > > >
> > > > > And then looped through the total number and
> ploted it :
> > > > > scatter_data(ii) =
> > > Bse(myindex3d(ii,1),myindex3d(ii,2),myindex3d(ii,3))
> > > > >
> > > > > What would you suggest?
> > > > >
> > > > > Best wishes,
> > > > > Alina
> > > > >
> > > > > On Mon, Apr 3, 2017 at 5:05 PM, Guillaume Flandin
> > > <[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>
> > > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>>
> > > > > <mailto:[log in to unmask]
> <mailto:[log in to unmask]>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>>>
> > > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> > <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>>>> wrote:
> > > > >
> > > > > Dear Alina,
> > > > >
> > > > > How did you extract the data for a given voxel?
> > What is
> > > > displayed if,
> > > > > when looking at results, you move the 'current
> > voxel' (red
> > > > arrow) to
> > > > > your voxel of interest and do a right click
> while
> > on the
> > > glass
> > > > brain and
> > > > > select Extract data > betas > this voxel in the
> > contextual
> > > > menu that
> > > > > opens?
> > > > >
> > > > > Best regards,
> > > > > Guillaume.
> > > > >
> > > > >
> > > > > On 03/04/17 12:51, Alina Dinu wrote:
> > > > > > Dear all,
> > > > > >
> > > > > > I am interested in the beta values for certain
> > activated
> > > > voxels.I
> > > > > extracted them from the beta_0001 file. I have a
> > couple of
> > > > questions:
> > > > > > 1. Could someone please tell me if this is
> this the
> > > correct
> > > > file?
> > > > > (given that I have a block design with only one
> > condition)
> > > > > > 2. After I plot them I notice that the values
> > are too high
> > > > (in the
> > > > > order of 80-100) which is nonsensical for beta
> > values. Could
> > > > someone
> > > > > please tell me if there is any kind of scalling
> > done in
> > > this file?
> > > > > How are these values coded?
> > > > > > 3. Could someone please tell me if this
> file also
> > > includes the
> > > > > confound mean predictor?
> > > > > >
> > > > > > Kind regards,
> > > > > > Alina
> > > > > >
> > > > >
> > > > > --
> > > > > Guillaume Flandin, PhD
> > > > > Wellcome Trust Centre for Neuroimaging
> > > > > University College London
> > > > > 12 Queen Square
> > > > > London WC1N 3BG
> > > > >
> > > > >
> > > >
> > > > --
> > > > Guillaume Flandin, PhD
> > > > Wellcome Trust Centre for Neuroimaging
> > > > University College London
> > > > 12 Queen Square
> > > > London WC1N 3BG
> > > >
> > > >
> > >
> > > --
> > > Guillaume Flandin, PhD
> > > Wellcome Trust Centre for Neuroimaging
> > > University College London
> > > 12 Queen Square
> > > London WC1N 3BG
> > >
> > >
> >
> > --
> > Guillaume Flandin, PhD
> > Wellcome Trust Centre for Neuroimaging
> > University College London
> > 12 Queen Square
> > London WC1N 3BG
> >
>
> --
> Guillaume Flandin, PhD
> Wellcome Trust Centre for Neuroimaging
> University College London
> 12 Queen Square
> London WC1N 3BG
>
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|