Dear Andy,
Another perspective on this is afforded by work on nonlinear responses
using Volterra Series. This work suggests that nonlinear interactions
among stimuli (c.f. a hemodynamic 'refractoriness') became pronounced
at SOAs of about 1 second were not insubstantial at 2 seconds. This
deviation from the linear model speaks to a minimum SOA of 1 to 2
seconds. These results pertain to secondary and tertiary cortical
responses to single words.
I hope this helps - Karl
> We are working on methods for analysis of fMRI data with very short ISI
> and varying ISI. A rough draft of a paper is available on my web page
> (www.sph.umich.edu/~jonraz/papers.html). As shown in this paper, even with
> ISI=TR, if we are willing to assume additivity, then it is possible to
> estimate the differential response. With a randomly varying ISI, it is
> possible to estimate the hemodynamic responses to each stimulus type. I
> am working with Bruce Turetsky and Dan Ragland on designing an experiment
> with randomly varying ISI.
>
> I do not believe that there is a minimum mean ISI that is necessary.
> Clearly, if the maximum ISI is greater than the duration of the
> hemodynamic response, then the responses can be estimated. The mean ISI
> can be quite short, however, by including lots of short ISIs in the
> experiment. This assumes additivity and that the response does not
> vary according to ISI. For estimating differences, you do not need to vary
> the ISI. Just use the shortest ISI that is not not shorter than the TR and
> that you believe is reasonable from a biological and psychological
> perspective.
>
> Jonathan
>
>
> On Thu, 10 Jun 1999, Andrew J. Saykin wrote:
>
> > Dear SPM group,
> > In planning event related fMRI experiments with the expectation of
> > analysis by SPM99 I am wondering what the minimum average ISI can be. We
> > would like to optimize the number of stimuli we can present in a scan
> > session. Assume the ISI is randomized within a defined window about the
> > mean ISI and a TR of 3.3s for whole brain acquisition. I realize that with
> > ISI's less than 16-20s we will limit our ability to characterize
> > differential onset and temporal pattern of HRFs. So my question is, if we
> > primarily want to detect relative HRF amplitude differences among classes of
> > items, what is the smallest mean ISI that can be appropriately analyzed by
> > SPM99. Has anyone tested this or run simulations? I'd greatly appreciate any
> > thoughts.
> >
> > Best wishes,
> > Andy Saykin
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|