> The attached image may help in giving you an idea of what I am trying to do.
> It depicts the time course of a stimulus and the theoretical evoked
> hemodynamic response. Below these curves are the non-overlapping FIR
> regressors. The reason I am looking at only 1 bin is because I want to get
> information from the dynamic response to the stimulus. I want to know which
> areas show activation first, compared to others. I thought the FIR method is
> a good approach, because it helps me model the HRF sequentially in time
> rather well.
The timing information from an FIR model is great to have, but can
also be difficult to interpret. This is particularly true with respect
to making inferences about timing across different areas - it's hard
to know whether these differences are due to differences in
vasculature or hemodynamic response, or actual differences in the
timecourse of the neural activity (which is probably what is of more
interest). Two relevant papers (one on timing responses, and one as an
example of using an FIR model to look at timing) are:
Miezin FM, Maccotta L, Ollinger J, Petersen SE, Buckner RL (2000)
Characterizing the hemodynamic response: Effects of presentation rate,
sampling procedure, and the possibility of ordering brain activity
based on relative timing. NeuroImage 11:735-759.
Davis MH, Ford MA, Kherif F, Johnsrude IS (2011) Does semantic context
benefit speech understanding through "top-down" processes? Evidence
from time-resolved sparse fMRI. J Cogn Neurosci 23:3914-3932.
The other point is that the dynamic response to the stimulus occurs
over several seconds, and so looking at a single bin will probably not
capture the full effect - although I understand the desire to come up
with a simple metric! :)
> There is also the other approach to use time shifted stimulus functions
> convolved with the HRF. This approach might work, but it makes a big
> assumption about the HRF across different areas of the brain.
True - analyses all have tradeoffs. Nevertheless, it may be worth
trying this if the FIR approach proves too challenging!
> Regarding the window length, I have an ISI of 10 s (for a total of 20
> presentations), which is half of what you suggested. I have a bin for every
> scan (TR = 0.514 s) and I am still deciding on how many bins to use. I
> should probably stick to 19 bins, otherwise I get into the next stimulus
> response. I do understand that I have a lot less power by using only 1 bin,
> but I cannot use an F-test over all bins, since this will give me only one
> image over all time points.
The best window length doesn't depend on your ISI, but on the duration
of the presumed hemodynamic response. There is no problem with
regressors overlapping the onset of another event. I would still use
20-30 seconds for the window length.
With such a short TR, you can of course use a short bin width, but
given how smooth the hrf is, you would probably be ok using a longer
one (even a bin width of 1 second would still give you ample timing
Hope this helps!
Dr. Jonathan Peelle
Center for Cognitive Neuroscience and
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104