John,
Thanks for your comments.
> I hope you don't mind me shifting your words around a little, but you could
> also say that the slice-timing correction disrupts the known absolute
> location at which a given data point was acquired. It is the fMRI slice in
> absolute time that needs to be shifted in space.
>
> Consider an fMRI dataset containing ascending slices where there was a
> jump of 1 slice. What you do depends on what you consider to be the
> greatest source of error: a timing error of 1 TR, or a spatial error
> of one slice.
Actually, I still believe that it is generally appropriate to apply the
slice timing correction first. The argument is based upon the observation
that the motion correction routine assumes that all motion takes place in
the periods between TRs (i.e., the entire volume is treated as a rigid body,
acquired at the same point in time). If this assumption is embraced, I think
you'll agree that applying the slice-timing correction first must be
correct. The SPM archives have the extended form of this argument:
http://www.mailbase.ac.uk/lists/spm/1998-07/0065.html
I'm now less certain about the situation given violations of this assumption
(in contrast to my pronouncement in that archived message!). Specifically,
intra-TR movement technically invalidates both the motion correction AND
slice-timing routines. I'm going to think about it a bit and I'd appreciate
any insight you have into the problem.
> I don't think that there are any theoretical reasons why there should not
> be any power above the Nyquist frequency.
Well, sinc-interpolation will produce invalid results if there is power
above the Nyquist frequency.
> A typical HRF to a delta function
> will contain frequencies higher than the frequency of the acquisition of
> sucessive scans.
It will? I think a few groups now have shown that there is not really
much (if any) power in the HRF above a frequency of about 0.25 Hz. In
particular, Oliver Josephs' event-related fMRI paper demonstrated quite
nicely that even when sampled with an effective TR of (I think) around 160
msecs, the HRF is rather smooth.
Thanks again for your thoughts on these issues.
Geoff
--
Geoffrey Karl Aguirre, Ph.D.
Univ. of Pennsylvania
(215) 614-1976
mailto:[log in to unmask]
http://cortex.med.upenn.edu/~aguirre
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|