Appropriate preprocessing?

[David Ruhl]

Hello,
I am starting to use your toolbox (specifically the mutualinf
function) to develop a connectivity matrix for graph theoretic
analysis. I am curious as to exactly what preprocessing parameters I
ought to be using (most of my neuroimaging experience is with meg/
eeg). I have images realigned and normalized in spm. I thought I
ought not do any spatial smoothing, since I'm interested in
connectivity between two directly adjacent regions. Other than the
temporal smoothing function you provide, need I do any other
preprocessing?
Thank you!

David Ruhl
Research Associate II
Mind Research Network

[Dongli Zhou]

Dear David,

Thanks a lot for your interest in our toolbox. Here is a message from another author of our toolbox regarding your question.

We do motion-correction, detrending, outlier-rescaling (via "NIScorrect"), minimal temporal smoothing (5 pt peaked kernal) and spatial smoothing (6mm FWHM). That said, for regions next to each other, taking out the temporal and spatial smoothing might be a good idea. Doing the outlier correction, in my experience, does help to decrease spurious connectivity estimates.

 - Greg

Hope it will help.

Best,
Dongli

--

You received this message because you are subscribed to the Google Groups "MATLAB Toolbox for Functional Connectivity" group.
To post to this group, send email to fc-to...@googlegroups.com.
To unsubscribe from this group, send email to fc-toolbox+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/fc-toolbox?hl=en.

[David Ruhl]

Dongli (& Greg),
Thank you for your reply - I will also detrend and remove outliers.
I'm afraid I don't understand, however, why *temporal* smoothing would
be a bad idea in my case (is it not performed on each voxel timeseries
independently?). Is there some alternate temporal smoothing scheme
you recommend (my impression from your paper is that failing to do so
can cause erroneously low connectivity estimates)?
Thank you!
-David

On Dec 6, 10:27 pm, Dongli Zhou <dongli.z...@gmail.com> wrote:
> Dear David,
>
> Thanks a lot for your interest in our toolbox. Here is a message from
> another author of our toolbox regarding your question.
>
> We do motion-correction, detrending, outlier-rescaling (via "NIScorrect"),
> minimal temporal smoothing (5 pt peaked kernal) and spatial smoothing (6mm
> FWHM). That said, for regions next to each other, taking out the temporal
> and spatial smoothing might be a good idea. Doing the outlier correction, in
> my experience, does help to decrease spurious connectivity estimates.
>
>  - Greg
>
> Hope it will help.
>
> Best,
> Dongli
>

> On Sun, Dec 6, 2009 at 3:04 PM, David Ruhl <d.a.r...@gmail.com> wrote:
> > Hello,
> > I am starting to use your toolbox (specifically the mutualinf
> > function) to develop a connectivity matrix for graph theoretic
> > analysis.  I am curious as to exactly what preprocessing parameters I
> > ought to be using (most of my neuroimaging experience is with meg/
> > eeg).  I have images realigned and normalized in spm.  I thought I
> > ought not do any spatial smoothing, since I'm interested in
> > connectivity between two directly adjacent regions.  Other than the
> > temporal smoothing function you provide, need I do any other
> > preprocessing?
> > Thank you!
>
> > David Ruhl
> > Research Associate II
> > Mind Research Network
>
> > --
>
> > You received this message because you are subscribed to the Google Groups
> > "MATLAB Toolbox for Functional Connectivity" group.
> > To post to this group, send email to fc-to...@googlegroups.com.
> > To unsubscribe from this group, send email to

> > fc-toolbox+...@googlegroups.com<fc-toolbox%2Bunsu...@googlegroups.com>
> > .

[Dongli Zhou]

Hi, David,

Here is the deal. If smooth the data with a priori kernal and then run the toolkit allowing more smoothing, it may yield too much smoothing. Also smoothing may differ between studies based on preproc stream. If do not smooth the data with a priori and do the smoothing within the toolkit, firstly, data is smoothed optionally; secondly, same smoothing rule are applied to all studies.

Hope this could help!

Best,
Dongli

To unsubscribe from this group, send email to fc-toolbox+...@googlegroups.com.