Raj -
Firstly, I apologise that these issues are potentially confusing.
However, they are discussed at length on pages 14-15 of the
SPM99 manual:
http://www.fil.ion.ucl.ac.uk/spm/course/manual/man.htm
and also on the SPM mail-archive:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0009&L=spm&P=R4666
In any case, I reiterate some of them below.
Secondly, Russ has already answered your general question.
As to your specific questions:
> >1. I know that I'm meant to express stimulus onsets
> > in terms of TRs from the start of the session.
> >
> > Does SPM think that the start of the session is:
> > (a) The time of the *first* slice in the first acquisition ?
> > (b) The time of the *reference* slice in the first acquisition ?
With the default value of fMRI_T0 (1), SPM effectively assumes that
the session starts with the first slice of the first acquisition (a).
This will be appropriate if you slice-time correct to the first slice
acquired in time (which is what many people do).
If you slice-time correct to a later slice in time, you will need to
adjust fMRI_T0 in proportion to fMRI_T, as Russ described.
> >2. I'm also not quite clear on how the value of fMRI_T0 alters the answer
> > to the above question. The definitions of fMRI_T0 that have been
> > posted to the list express it in terms of "which time bin to put into
> > the regressor". But it's not clear to me how precisely that translates
> > into how to specify stimulus onset timings. If anyone could give
> > a definition of fMRI_T0 (and fMRI_T) in terms of how they affect
> > how I should express stimulus onsets in terms of TRs from the start
> > of the session, rather than in terms of bins and regressors, then that
> > would help to make things a lot clearer for me.
SPM divides each TR into a finite number of timebins (that number
being fMRI_T). This creates a "high-resolution" timespace (a vector
with fMRI_T * N elements, where N is the number of scans). So
when the user enters onsets in fractions of scans, SPM creates
a delta function (in an event-related experiment) in the bin nearest
to each onset (ie, the onsets are rounded to the nearest bin).
The convolution with the basis functions is then performed this
highresolution space, before being downsampled again to create
a value for each scan (row) in the design matrix. The specific bin
that is chosen as the sample point is that given by fMRI_T0.
Attached is a WinZipped PDF of a Figure that might help explain.
> >Clustered volume acquisition
> >TR = 4.5s
> >25 slices
> >Acquisition duration = 2100ms
> >Time-per-slice = 84ms
> >TA = (nslices-1) * time_per_slice = 2.016
> >
> >batch_slicetiming=[25 23 21 19 17 15 13 11 9 7 5 3 1 ...
> > 24 22 20 18 16 14 12 10 8 6 4 2];
> >refslice = 1;
> >
> >The stimuli are speech sounds, 1.24 secs long, timed
> >to onset w.r.t. the start of the first acq as follows:
> >
> >Onsets = 2.68 + TR*[ 0 1 2 3 4 5 ... etc ]
> >
> >This means that each stimulus starts and ends with 580ms
> >of silent time at either end of it, to avoid masking by
> >the scanner beeps.
> >
> >It seems to me that I should adjust these onset times
> >in some way in order to take accout of when my refslice
> >happens and what I set fMRI_T0 to be, but it is somewhat
> >opaque to me how exactly I should do this
This case is a bit more trickier than usual, because your TA
is much less than your TR. SPM works in TRs, so each TR
will be divided into 4.5s/fMRI_T=4.5s/16=0.28s timebins.
You have slice-time corrected to the middle of the TA (spatial
slice 1, ie the bottom slice in Analyze format). This is not
the same as the middle of the TR. Thus you need to change
the value of fMRI_T0 (in the defaults-fmri-stats menu) to:
round ( ( (TA/2)/TR ) * fMRI_T) = 4
I hope the logic for the above will become clear.
Note that if you slice-time corrected to the first slice in time
(spatial slice 25), then you would not need to change any
defaults. The logic for shifting to the middle slice in space
only really applies for sequential (ascending/descending)
acquisitions, where any interpolation error is pushed to
the extreme (top/bottom) slices. Your interleaving means
things are not so simple, so the first slice in time might be
easier all round.
Hope this helps
Rik
--
---------------------------------------------------------
DR RICHARD HENSON
Wellcome Department of Imaging Neuroscience
& Institute of Cognitive Neuroscience
Institute of Neurology
University College London
17 Queen Square
London, WC1N 3AR
England
EMAIL: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~rhenson
TEL1 +44 (0)20 7679 1131
TEL2 +44 (0)20 7833 7472
FAX1 +44 (0)20 7813 1420
FAX2 +44 (0)20 7813 2835
MOB +44 (0)794 1377 345
---------------------------------------------------------
--
|