Paul -
I think Richard is correct.
1. I don't think there is a problem with SOAs < TRs; it is the effective
sampling rate (via the jitter Richard talked about) that matters.
2. Ironically, slice-timing correction is potentially more important the
longer the TR, yet also less accurate for longer TRs.
Remember, slice-timing correction is only really important for people
doing t-tests on a single canonical HRF regressor (in event-related designs).
You can also cater for different slice times by using F-tests on a more
comprehensive basis set (eg including temporal derivatives), in which
case you would not need to perform any slice-time correction.
Thus I would use slice-timing correction for any TR between 0-3s. For
longer TRs, I might not perform any slice-timing correction, but use a more
general basis set, particularly if I had short SOAs (ie when there is
signal energy at frequencies close to those that will be aliased by the
slice-timing correction, ie >1/(2TR)).
Someone please correct me if I'm wrong though!
Rik
Paul Fletcher wrote:
> Hello Richard,
> I'm well. Thanks for the reply.
>
> Re. the TR SOA. Your reply makes sense. I was under the impression though,
> from a previous talk by Rik, that, when TR exceeds SOA, and you regularly
> have two events per scan, power to detect to differences would be greatly
> diminished. I will copy this message to Rik in order to clariy and then
> broadcast the result.
>
> re slice timing correction, my answer was based upon some discussion with
> Christian Buchel who suggested that a TR of 4 or seconds or below means
> that the timing correction has little effect as the data are so temporally
> smooth. Once again, will copy to Rik who may have responded to this anyway.
>
> Thanks again for reply - will clraify asap.
> Best wishes
> Paul
>
> At 18:22 29/11/00 +0000, you wrote:
> >Dear Paul,
> >
> >Hope all is well with you.
> >
> >> >we are currently designing a rapid event-related fMRI study with
> >>>interests in differential and main effects:
> >>>3 conditions, 50 trials in each condition, 50 null-events, random
> >>>presentation, SOA 3.5 sek, TR 4.0 sec.
> >>
> >>>Can the TR be longer than the SOA, or should the SOA be longer than the
> >> >TR?
> >>I think that you should certainly keep your SOA greater than your TR -
> >>otherwise you will have a single scan covering more than one event.
> >
> >Hmm. I don't think that I quite agree with this. Surely in ANY
> >rapid event-related fMRI study a single scan covers the haemodynamic
> >response to more than one event. The fact that in this case it
> >covers the onsets of more than one event doesn't seem to be a
> >problem. With the 'jitter' between TR and SOA that they have
> >suggested, and plenty of repetitions, they will have an effective
> >sampling frequency of 0.5 seconds, which I would have thought is more
> >than adequate.
> >
> >>Re. The slice timing issues. Yes, I believe it does work with interleaved
> >>slice acquisition. One question for the experts, is it actually of any
> >>value at these shortish TRs?
> >
> >I seem to remember that Rik has suggested that a TR of 4 sec is a bit
> >on the LONG side for the sinc interpolation algorithm to work
> >correctly. So perhaps for this reason they should try to reduce the
> >TR (which may then make it smaller than the SOA anyway). But I won't
> >copy this to the list, as Rik usually keeps an eye out for these
> >sorts of questions, so he'll probably send an 'expert' answer,
> >
> >All the best,
> >
> >Richard.
> >--
> >from: Dr Richard Perry,
> >Clinical Lecturer, Wellcome Department of Cognitive Neurology,
> >Institute of Neurology, Darwin Building, University College London,
> >Gower Street, London WC1E 6BT.
> >Tel: 0207 679 2187; e mail: [log in to unmask]
> >
> -----------------------------------------------------------------
> Paul Fletcher,
> Research Department of Psychiatry,
> University of Cambridge,
> Addenbrooke's Hospital,
> Hills Road,
> Cambridge,
> UK
> CB2 2QQ
>
> Tel 01223 336 988
> Fax 01223 336 581
--
---------------------------8-{)}-------------------------
DR R HENSON
Institute of Cognitive Neuroscience &
Wellcome Department of Cognitive Neurology
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
FAX +44 (0)20 7813 1420
---------------------------------------------------------
--
|