Just a quick series of thoughts:
when you really intend to regress out BOLD activation caused by NEURAL
activity driving the movement of the head (which could be due to example
flexion/extension of nech/shoulder/torso muscles, to mention a few)
rather than MRI acquisition related signals, you might want to include
the 2nd derivitive of the parameter describing the relevent movement
component, and convolve that with the HRF. Many motor areas react
proportional to the applied force when I remember well, and as most of
you probably know Newtonion mechanics says that F=m * a, where a is the
acceleration of a body with mass m.
However, finding the relevant movent component from the 6 realignment
parameters would not be that straightforward, since biomechanically the
muscle force is applied along an angular moment with an axis of rotation
resembling a line through the back of the head touching the plane
supporting the head. This axis is also changing with head rotation, to
complicate things further.
To make a long story short: I wouldn't bother about that, unless you are
really interested in areas where neck/shoulder/etc mucsle motor activity
As said before, the effects of movement on B0 distortion and
spin-history artifacts is probably much larger.
Stephen J. Fromm schreef:
> On Tue, 18 Sep 2007 15:55:15 +0100, alistair galley
> <[log in to unmask]> wrote:
>> dear list,
>> apologies if this has been covered (i couldnt see it in the 100 or so
>> movement parameter related messages i surveyed)
> >from what i gather, including movement movement parameters in the design
>> matrix is a good idea even if you have already
>> realigned because it can remove additional artifacts not removed by rigid
>> body transformations, assuming that your movement params
>> are not correlated with your regressors of interest.
>> but this only regresses out artifacts, it does not regress out the bold
>> activity which may be underlying some of the movement.
>> so i was thinking, wouldnt it make sense to convolve the movement
>> with the hrf to model some of the underlying neuronal
>> causes of the movement. obviously the same caveat would apply with regard
>> these regressors not being correlated with your
>> regressors of interest.
>> does this make sense as something to do and has anyone done this?
> People do not typically do this, as far as I know. My overall take is
> that I wouldn't do this, myself.
> I can't see any reason why it would be a _really bad_ idea, but it's not
> clear to me that it would help much.
> As you probably know, the main concern is "regressing out" the effect of
> motion itself, because the changes in signal it can lead to are relatively
> Finally, were you to attempt to do what you suggest, you'd have to
> consider whether you'd really want to convolve the position parameters
> themselves with the HRF, or something like their first order differences
> (which would be velocity).
>> with very best wishes
Dr. S.F.W. Neggers
Division of Brain Research
Rudolf Magnus Institute for Neuroscience
Utrecht University Medical Center
Visiting : Heidelberglaan 100, 3584 CX Utrecht
Mail : Huispost A.01.126, P.O. Box 85500
3508 GA Utrecht, the Netherlands
Tel : +31 (0)30 2509609 Fax : +31 (0)30 2505443
E-mail : [log in to unmask]
Web : http://www.fmri.nl/people/bas.html