Hi Gavinda

I have a few comments based on our experience of using fMRI to track
cardio-respiratory related changes, which are more or less continuous in
nature.


On Thu, 20 Jul 2006 09:31:17 +1200, Govinda Poudel
<[log in to unmask]> wrote:

>Hi there,
>
>
>
>I am designing a continuous visuomotor tracking experiment. Being a beginner
>to fMRI, I am investigating on possibilities of using various
>well-established paradigms for the experiment. I have some basic questions,
>which I need to clarify.
>
>
>
>1) How can we fit a continuous tracking experiment into more
>conventional and well established discrete "blocked", "event-related" or
>mixed designs? Are there different paradigm approaches for continuous
>experiments?

You need to start by deciding what you are looking for. fMRI will give you
activation at one time relative to activity at another time, i.e., relative
to a baseline period. So you need to decide what is your baseline and what
is the task during which you want to see activation. From there you want to
design an experiment that will hopefully give you repeated changes between
baseline and activation periods. For your case that may be "tracking" and
"not-tracking" conditions to start with, or it may be a correlation with
error, eye movement etc., in which case your '"baseline" is no error, or
perhaps no movement.

>2) Any important issues that I need to take into account while running
>a continuous experiment?

I'm not familiar with the timing of visuomotor tracking task. You don't want
to have very long time periods during your activation phases if you are
simply looking for an "on/off" response. There are technical issues (scanner
drift) and physiological effects (habituation/learning/adaptation) that make
it to hard to measure the "same" activation over a long time period. On the
flip side, you'll need to make sure that the timing of your performance
measure in the correlation is matched to the fMRI. For example, if you have
a change in performance that lasts only half a second, the down-sampling to
the fMRI collection rate may hide the change (assuming at TR of 2 s or
more). So, if you hypothesize a full regional BOLD response to a 0.5 s
performance change, you'll need to adapt the performance measure.

I'd also be careful to check for task-related motion (do people
unconsciously move their head while tracking?).

>3) Since it is not a discrete stimulus-response type experiment, we
>will base our analysis on behaviour (such as tracking performance, eye
>movement etc). How well established are these type of analysis? Any paper
>you are aware of?
>

This really comes back to question 1 - fMRI is always going to show you
relative activation, so the question is what activation relative to what
reference? If your performance measure is continuous in nature and are
looking for a linear correlation with fMRI changes, then you run into the
problems that a) your neural activity may not change linearly with the
performance changes, or b) the BOLD signal may not change linearly with
linear changes in neural activity; you then have to be careful in how you
interpret any findings.


>
>Regards,
>
>Govinda
>

Hope this helps; I'm sure others have more direct experience with
correlation measures, and hopefully they'll correct or add to the above!

Best regards,
Paul