Hi
I am having some trouble with highpass filtering (using commandline 'ip'
under Cygwin). It is stated that ip expects to get input for the '-t' option
in volumes ("... with sigmas in volumes, not seconds ..."). However, in the
Feat GUI the high-pass cutoff should be given in seconds. Inputing values of
TR=5s, high-pass cutoff=55s (i.e. 11 volumes) in the GUI executes ip with
the following settings: ip -t 5.5 -1 (110s --> -t 11.0 -1)
Why is the highpass only half the number of volumes? What exactely is Sigma
in the case of temporal filtering? It could not be the same as Sigma=FWHM /
2.35 for spatial filtering?
A second question: Consider the following design R-A1-f-R-A2-f-R-A3-f ....
repeated three times with R being rest (4scans), Ax being the active
conditions (6scans) and f being feedback (1scan). The order of the three
different active tasks is randomized over repeated runs.
What would be the 'best' cutoff in this case? The way I understand your
suggestions it should be '(R+A+f) = 11scans/55s' not 'R+A+f)*3 =
33scans/165s' (taking into account the three different tasks)?
Does it make sense to visualize the frequency spectrum of the (filtered)
regressors in comparision to the real data (e.g. FFT over a masked 4D
containing grey matter only) to get an idea of the 'quality' of the
highpass? How could I do a FFT without using Matlab?
Thanks and best regards
Thomas
|