Dear Claus:
==============Original message text===============
On Mon, 30 Apr 2007 3:19:40 pm CDT Claus Lamm wrote:
Dear FIR aficionados,
Two questions:
1) I was wondering whether there are any recommendations ('hard' or 'soft')
on how to select bin size in doing a FIR analysis. More specifically, I wonder how to interpret
the results if bin size is shorter than the TR.
>>> I don't think finer units would be helpful. There is no need for this type of interpolation
using an FIR model. If you want finer resolution in the time course it would be best to use a
shorter TR. The bin size should really = TR.
2) I guess the way a FIR analysis is setup in SPM5 it does not take into
account the hemodynamic delay - i.e., the first bin after stimulus onset
would not actually reflect the response at stimulus onset, but 'activation'
that happened about 4-5 sec ago. So in order to see onset-related resposnes
I would have to assess a bin 4-5 s post stimulus.Am I correct with this?
>>> Actually it does take into account the delay because it allows a separate estimation of each bin. Thus if we assume a TR of 2 secs, you will find that bin 1 (relative to some trial) might show very little effect of a task, while bins 2-5 would show increasing task effects. There is a very nice demonstration of this if you want to work through the example given in the Laboratory book for the 2005 USA-SPM course (http://www.fil.ion.ucl.ac.uk/spm/course/usa/Lab-Book.pdf) Look at the chapter "Lab 5: Multi Subject fMRI"
An FIR model is equivalent to what is called selective averaging but because SPM instantiates it as
a basis set, it does not require that the task and MRI scan timing be explicitly aligned to one
another.
Importantly SPM5 allows the proper estimation of repeated measures so you really should include all
the time bins for each condition, forward them to the second level and use an F-test to test for
differences between conditions. i.e., a condition x time interaction. When using other programs
sometimes users will choose a particular set of time bins to compare. Deciding whether to do this
is left up to the user.
In general: are there any full papers out there describing the FIR approach in some more detail?
>>> From the SPM5 manual:
R.N.A. Henson. Analysis of fmri time series. In R.S.J. Frackowiak, K.J. Friston, C. Frith,
R. Dolan, K.J. Friston, C.J. Price, S. Zeki, J. Ashburner, and W.D. Penny, editors, Human
Brain Function. Academic Press, 2nd edition, 2003.
R.N.A. Henson, M.D. Rugg, and K.J. Friston. The choice of basis functions in event-related
fmri. NeuroImage, 13(6):127, June 2001. Supplement 1.
good luck
darren
THANKs for any feedback,
Claus
===========End of original message text===========
--------------------------------
Darren R. Gitelman, M.D.
Department of Neurology
710 N. Lake Shore Dr. #1122
Chicago, IL 60611
Voice: (312) 908-8614
Fax: (312) 908-5073
--------------------------------
|