Just another couple of cents for you: Your problem is a long standing one in
that SPM doesn't handle the temporal discontinuity of sparse designs very
well (probably should be called "clustered acquisitions" to avoid
confusion). Essentially this means it is difficult to derive any temporal
resolution with sparse designs and so "sparse-efMRI" seems a bit
contradictory to me as the analysis of sparse designs often seems equivalent
to an epoch-based not event-related analysis. In any case, its largely
pointless to convolve the scan time series with the HRF as the time series
itself is so poorly resolved in the SPM design. This is a limitation for
many auditory studies using SPM analysis. I think the approach Grieg
describes is the most favoured work around (use a time bin of 1 for FIR)
although I'm not sure there is a perfect solution to this problem. In your
orginal post I wondered whether your TR (as Bas advised SPM "TR" is
different from an MR TR ) is 16.5 seconds not 4 seconds. When this setting
is incorrect it will impact upon things like temporal filtering etc in SPM.
I've pasted some links to other spm-mails that deal with FIR and sparse
analysis but to answer some of your original questions:
The SPM-TR refers to the time between onsets of succesive volume
acquisitions. This includes the silent period in sparse designs and any gap
for continuous acquistions.
I think the SOTs in the example data/tutorial you refer to below might use
scans not seconds as the metric.
It is possible to define "negative" TRs in SPM2 but not SPM99, not sure
about SPM5. I suspect you may not need these though(see above).
Hope this helps.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask]] On Behalf Of Neggers, S.F.W.
> Sent: Friday, 15 February 2008 7:41 PM
> To: [log in to unmask]
> Subject: Re: [SPM] Sparse efMRI, TR, Interscan Interval & SOTs
> Dear Christopher,
> TR is a well defined term from NMR and indicates the time
> between subsequent excitations of, in the case of (f)MRI, a
> certain proton. It is sometimes misinterpreted by fMRI users
> as being the duration of a scan.
> For most EPI sequences, including Sparse EPI, TR should equal
> the time between scans including the time you do not read nor
> excite any signal in your design. Please note that TR can be
> (much) shorter than scan duration, for example for 3D EPI it
> is a few 10s of ms (so TR does not per se equal scan
> duration!!!). Beware when choosing your flip angle, it should
> incorporate that you have a very long TR.
> Only for the settings of your slice timing correction, I
> think, you should notice the fact that the acquisition of you
> slices is terminated long before a proton feels an RF
> excitation again. Otherwise, I think you can enter TR (in
> model specification) as being the time between subsequent
> exciations of a certain slice, eg the time you acquire a
> volume plus the 'silent time'. Enter your onsets in seconds,
> or in scans (in the latter case, in units of TR including
> silent time).
> I have no experience with sparse designs though, just my 2
> cts from a theoretical point of view.
> Good luck,
> Dr. S.F.W. Neggers
> Division of Brain Research
> Rudolf Magnus Institute for Neuroscience Utrecht University
> Medical Center
> Visiting : Heidelberglaan 100, 3584 CX Utrecht
> Room B.01.1.03
> Mail : Huispost B.01.206, P.O. Box 85500
> 3508 GA Utrecht, the Netherlands
> Tel : +31 (0)88 7559609
> Fax : +31 (0)88 7555443
> E-mail : [log in to unmask]
> Web : http://www.fmri.nl/people/bas.html
> -----Oorspronkelijk bericht-----
> Van: SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask]] Namens Christopher Benjamin
> Verzonden: vrijdag 15 februari 2008 3:30
> Aan: [log in to unmask]
> Onderwerp: [SPM] Sparse efMRI, TR, Interscan Interval & SOTs
> Dear SPMers
> I'm still unable to resolve the issue below and am hoping to
> clarify, as I cannot see the solution. Specifically, when
> running an efMRI paradigm with a large interval between
> acquisitions, does 'Interscan Interval' = TR in the level 1
> model (as it notes in the manual, and in the chapter 27 -
> face data - example), or does it include the silent period?
> If it is the TR, I assume to proceed you adjust your stimulus
> onset times relative to each scan. It looks like this is
> what was done in the face data example, as, the TR is listed
> as 2 secs, indicating the experiment ran for at least 702
> seconds (351 images), but the SOTS do not go past 351 secs.
> If so, is it the case that SPM cannot accommodate a design
> where stimulus onsets occur more than a TR's length before
> scan acquisition? Here, when you make the first SOT 0secs,
> the second SOT occurs before the first.
> A related query: If 'Interscan Interval' = TR, I am also
> wondering if specifying the SOTS relative to the scan onset
> times and ignoring the silent period between scans could
> create problems. Doesn't SPM need the full time course of
> the experiment (inc. silent periods) to correctly convolve
> the SOTs with the HRF? Without the full time course wouldn't
> it treat each stimulus as activating a region from resting
> state, and ignore the effects of repeated activation, thus
> modeling the activation incorrectly?
> Any advice much appreciated.
> Christopher Benjamin
> > Dear list,
> > I am relatively new to SPM and have a quick query about model
> > specification for a sparse event- related fMRI design.
> > Specifically, in my design I have a TR of 4 seconds and the delay
> > between one TR commencing and the next is 16.5 seconds (to
> allow for
> > stimulus presentation and the HRF). In level one model
> > would I
> > - specify 16.5 for the Interscan Interval (not the 4s TR);
> > - specify the SOTs in one long train (seconds) relative to
> the first
> > SOT (specified as time = 0; Nb.
> > have set Units of design = seconds, time bins = number of slices).
> > Many thanks
> > Christopher