Hi,
I've used Giedrius Buracas's m-sequence toolbox and also the Wager
genetic optimization toolbox, and know that they both support
multiple trial type experiments. We've also put together a toolbox that
you can check out at http://cfmri.ucsd.edu/ttliu/mttfmri_toolbox.html
The key with multiple trial type experiments
is that random search doesn't work so well once you get above
3 or 4 trial types -- that's why things like m-sequences are so useful.
That said, the experiment described is not that easily designed using
the above programs since there is a constraint on the ordering
of trials. e.g. cue-response-feedback.
Most of the programs above optimize assuming no constraints
on ordering of the trial types. In designing a similar experiment
in the past, I just used brute force calculations to come up with
an optimal design -- here you can just use whatever portion of
optseq, AFNI, etc. that computes your metric of interest until someone
comes up with a better method.
Good luck,
Tom
***********************************************************
Thomas Liu
Center for Functional Magnetic Resonance Imaging
University of California, San Diego
9500 Gilman Drive, Mail code 0677
La Jolla, CA 92093
Phone: (858) 822-0542
Fax: (858) 822-0605
http://fmriserver.ucsd.edu/ttliu
***********************************************************
On Sep 27, 2004, at 6:51 AM, Stephen Smith wrote:
> Hi - I'm not sure about the other software options, but Optseq does
> afaik
> allow multiple event types - see
> http://surfer.nmr.mgh.harvard.edu/optseq/
>
> Any thoughts on this Doug?
>
> Cheers, Steve.
>
>
>
> On Wed, 22 Sep 2004, X Liu wrote:
>
>> I am about to design an ER-fMRI experiment with multi-event trials. I
>> am
>> aware of a few optimization programs for single-event trials (e.g.,
>> Doug's
>> optseq, AFNI's RSFgen/waver/3dDeconvolve, UCSD's m-sequence, and
>> Wager's
>> genetic algorithm). But none of them can be readily applied to design
>> with
>> multi-event trials.
>> Say I have a trial with four events -- "cue" (5 levels), "response" (2
>> levels), "feedback" (2 levels), and "blank". I would like to estimate
>> the
>> brain activation of each level (except for "cue", just overall
>> effect) of
>> each event (except "blank") within a trial independently, but don't
>> want to
>> resort to long resting period after each trial. Does any one have
>> experience on this type of design and optimization?
>> I am thinking about optimizing just the "feedback" event since
>> "response"
>> event is subject specific and "cue" event partially depends on the
>> previous "response" event. Also, maybe after every 6 trials
>> (pseudo-block),
>> add a resting period of about 16 seconds, instead of long resting
>> period
>> after each trial.
>> A related issue is that the two levels of the "feedback" event do not
>> necessarily have equal probability. Does this affect the optimization
>> in
>> terms of looking at the contrast of these two levels?
>> Thanks very much for any suggestion.
>>
>
> Stephen M. Smith DPhil
> Associate Director, FMRIB and Analysis Research Coordinator
>
> Oxford University Centre for Functional MRI of the Brain
> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
>
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>
