Hi Michael
> I am wondering if the canonical HRF is appropriate for a all brain regions and all tasks.
> Probably not or maybe.
I think that the general rule of thumb is that, generally, the
canonical HRF is a reasonable approximation, for most subjects, for
most brain areas, for most tasks. For example, if you look at:
Aguirre GK, Zarahn E, D'Esposito M (1998) The variability of human
BOLD hemodynamic responses. NeuroImage 8, 360-369.
D'Esposito M, Zarahn E, Aguirre GK, Rypma B (1999) The effect of
normal aging on the coupling of neural activity to the bold
hemodynamic response. NeuroImage 10, 6-14.
you can generally see an increase in response 3-7 seconds after an event.
As you suggest, using other basis sets would almost certainly better
capture the range of responses seen, even in these more 'canonical'
regions. However, then you have the task of interpreting the more
complicated set of basis functions. For example, if you do an FIR
analysis with 5 bins, this lets you not make a priori assumptions
about when the peak occurs. But then to test for a response, you
might do an F test to see whether any bin shows a response; but then,
being an F test, you don't know (a) which bin it was or (b) whether it
was a positive or negative response; thus, I think in general there is
more work to be done before you can interpret the results of the more
flexible basis sets.
If there is a particular brain region you are interested in and you
have reason to think it might not be well approximated by a canonical
HRF, one option is to run a pilot study to determine what kind of
function might best characterize it, and then use that to inform your
analysis of real experimental data. Although this is a sensible
approach I don't think it's often done due to (real or perceived) time
and expense constraints.
> The time and dispersion derivatives only account for small
> differences in timing which, as I understand, do not matter much in a 30 sec block design.
You're right, for a block design, the response should saturate at some
point, which makes differences in shape of the basis functions much
less relevant.
Hope this helps,
Jonathan
|