Dear Raju,
> We are planning 10-subject studies to assess transfer of finger
> movement sequence learning from Normal (N) to two Rotated settings (V
> and M). We are planning to use a control condition (F) where subjects
> follow random stimuli in the Normal Setting. Scanning to be conducted
> at 4T, TR=3secs, whole brain imaging (each scan is of 6mm thickness
> with 17 axial slices of 64x64). Because of the complexity of the
> settings V and M, we don't want to put them in the same study.
>
> Study 1: 6 runs adding to a total of 32.4 minutes
>
> Run1: F N V N F V N V F
> (each epoch lasts 36 seconds and run is 5.4 minutes.
> FNV are counterbalanced but with N preceding V to
> allow transfer of learning effects).
>
> Run2, Run3, Run4, Run5, Run6 are the same as Run1
>
>
> Study 2: 6 runs adding to a total of 32.4 minutes
>
> Same as Study 1 but replacing V with M.
>
> Subjects are scanned on the same day for Study 1 and Study 2 with a
> break in between. We will ask the subjects to come again on another day
> to repeat the two studies (counterbalancing the order across
> subjects). We hope this repetition would make sure the transfer
> effects are robust. Different sequences are generated randomly for each
> study and for each subject.
>
> Questions:
>
> 1) Do we need a control condition after each sequence condition? For
> example, instead of FNV, we need to use FNFV. Our feeling is we don't,
> because each condition puts the subject in a different mode even though
> N and V; N amd M use the same sequence.
The rationale for inserting a F between each condition (ar pair of
conditions) is to minimize the duty cycle of the design (to enable more
hi-pass filtering). In this case it might be important. if you wnat to
use long epochs.
> 2) Is there any upper limit on the length of the activation/control
> conditions. Most of the studies use 30s/30s type of designs. I looked
> at the discussion on State-related fMRI in SPM notes [Chapter 9, pg
> 11]. There is an upper limit of ~1 cycle/few minutes. How long can
> this be? In a motor sequencing task it is ideal to have sufficient
> practice in each epoch, so 40 or 50 seconds is better but then the
> cycle becomes 2 to 2.5 minutes long. Can the control condition be
> shorter duration?
Yes they can. The issue here is low freuquency confounds and keeping
the experimental variance as orthogonal as possible. I would have
thought that 36secs was a good compromise if a short (e.g.16 second
reference condition was intercalated between motor epoch (or pair of
epochs). Longer epochs or omission of the frequent reference condition
will render the differential effects among epochs of interest collinear
with low frequency artefacts. This will reduce the efficiency of their
estimation (if Hi-pass filtering is used) or sensitivity (if hi-pass
filtering is not and the undmodeled low frequency components end up in
the residuals).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|