Print

Print
Dear Marta

just for emphasising what Chris wrote:
While many people include the realignment parameter in the design matrix, I guess, almost the same number of people don't do that.
And, as Chris wrote earlier, including the realignment parameter in case of paradigm correlated movements "kills" your activation, since the realignment parameter are capturing all variance and they are in the worst case not orthogonal to your paradigm.
Having said that, your highest priority should be to prevent those paradigm correlated movements since you never can be sure whether you are looking at an activation or a movement artefact. And the suggestions that Chris made for preventing movements may be already enough in your case.

Good luck

Karsten

-- 
-------------------------------------------------------------
Karsten Specht, PhD

Department of Biological and Medical Psychology

Bergen fMRI group

University of Bergen
Jonas Lies vei 91
5009 Bergen
Norway
Tel.: +47-555-86279
Fax: +47-555-89872
[log in to unmask] 
http://fmri.uib.no/
Von: Gandolla Marta <[log in to unmask]>
Gesendet: 26.aug.2010 20:00:19
An: [log in to unmask]
Betreff: Re: [SPM] realignment parameters effect

well, sorry for bothering you all again...
my concern is that quite everyone in literature uses realignment parameters as covariats of no interest... do you think it is ok not to use them?
thanks for the help
marta

2010/8/26 Chris Watson <[log in to unmask]>
Don't include motion parameters as regressors.

I'm sure you already do this, but I've seen that it's good to have a practice session outside of the scanner to see how much they move when doing the task. Then you can coach them into trying to move their body less. Also, putting some pillows under their knees, so their feet are elevated, helps quite a bit. And of course vacuum bags, tape across the forehead, etc.

Gandolla Marta wrote:
Hi Chris and list,
   so, what do you suggest?

2010/8/26 Chris Watson <[log in to unmask] <mailto:[log in to unmask]>>

   I think that for something as simple as ankle dorsiflexion, if the
   results are in a plausible location, you can believe them. The
   motion really isn't that bad. It's hard to keep a patient still
   when doing foot movements, so there will probably always be
   stimulus-correlated motion, and if you include the motion
   parameters as regressors (as you have seen already), it'll get rid
   of any true activation.

   Gandolla Marta wrote:

       Hi Chris and everyone else!!
         indeed, we are looking at right ankle dorsiflexion and yes,
       I definitively agree that the activation seems to be quite
       posterior.
       the graph I posted has on x axis the number of scan (TR=3secs,
       5 minutes of acquisition). on the y axis there are mm for
       blue, green and d line, and degrees for yellow, cyan and
       magenta line. I superimposed all the realignemnt parameters
       along with the block design to search for the eventual
       correlation. I attached in any case te SPM output of realignment.

       all my best
       marta
       2010/8/26 Chris Watson
       <[log in to unmask]
       <mailto:[log in to unmask]>
       <mailto:[log in to unmask]
       <mailto:[log in to unmask]>>>


          Is it right ankle dorsiflexion that you're looking at?
          It looks pretty posterior to the motor strip; what does it look
          like overlaid on a structural?
          What's the scale (units) of the graph you posted? Can you
       post the
          output of SPM's realignment?


          Gandolla Marta wrote:

              Hi everyone,
                 I attached a pdf file instead of a docx of the previous
              e-mail. sorry for the inconvenience!!
                 I'd need some help about realignment parameters
       effect that
              seems to be huge in the acquisition I will now describe.
              We have a 30 secs block design, starting from rest. The
              patient is performing active ankle dorsiflexion during
       on blocks.
              We did the following preprocessing steps:
              - realignment
              - coreg
              - normalize
              - smooth (6mm)
               then we implemented the GLM with a 8 columns design matrix
              (1-condition with the 30 secs block design, 2-7- realign
              param, 8- baseline) and we found zero activation
       (p<0.01 FWE
              corrected).
              we then chacked with the 2 columns design matrix
       (1-condition
              with the 30 secs block design, 2- baseline) and we found a
              quite important activation (fig.1). should we trust this
              activation? the realignemnt param plot along with the block
              design protocol is shown in fig.2. is it possible that all
              this effect depends on the correlation between the
       realignment
              parameters and the condition column in the design matrix?
              should trust only the design with the realignement
       parameters
              as covariats of no interest? we are thinking in this
       case why
              we don't see any activation... do we miss something?
               thanks for your help
              best regards
              marta