Dear SPM Mailing List,
I'm currently trying examining the "true" hemodynamic response upon
spontaneous slow calcium waves in anesthetized rats, derived from a fMRI
measurement while simultaneously detecting the calcium concentration by
an optic fiber. Therefore, I use optically detected onsets of these slow
waves and their durations in an event-related design. As basisfunction,
I choose the FIR set. With a window size of 25 seconds and a # bins = 25
(corresponding to the TR=1 s), I assume that I can capture the whole
response after each onset.
However, the results confuse me a little bit. I can see at bin 0
(representing the bin when the onset occurs) that the actual response
has already reached it's peak level, and decays as expected like the
canonical HRF.
We then wanted to figure out, if we really capture the whole
hemodynamic response upon spontaneous slow calcium waves and decided to
shift the onset times by hand by 5 seconds. This means, an onset that
occurred at t=15 seconds was defined as t=10 seconds.
According to that method, we interpreted the actual response starting
at bin #5 as the response upon a slow calcium wave. This time, the
extracted response started (more or less) at baseline level followed by
the expected rise and decay similar to the canonical HRF.
If we shift the onset times by hand by 20 seconds (and extend the
window size to capture the 20 seconds after the real onset), we found
out that the hemodynamic response at bin #20 is starting at baseline
level more accurate. The shape of the response is the same for all 3
methods.
Now I try to figure out what's the reason for this. As far as I know,
an FIR filter needs some measurement values until it reached
steady-state. According to literature in system theory, this is defined
by the order of the FIR filter.
An order of 20 needs at least 21 measurement values until it reached
steady-state. My question is:
Is the order of the FIR set defined by window size, # of bins or the
duration of the bins? Is there a transient-state and steady-state when
using the FIR set in SPM12?
And where can I find the matlab function that is used for deconvolve
the raw signal with the dirac impulses (= the boxcar matrix created by
selecting the FIR set)?
I hope someone can help me with these questions.
BR
Hendrik
|