Hello Richard and Geraint,
Thanks for your excellent suggestions. As I'm
relatively new to SPM I wonder if you would mind clearing up
a few things up for me?
The data were acquired with a block design, 15 seconds OFF,
15 seconds on, etc, etc. Did the method Richard proposed require
some re-ordering of the data? Basically I'm not too sure about
how to specify the required model.
Also, considering modification of BOLD signal due to scanner
'variation' - does SPM detrend the time course data? If so
wouldn't a time dependent reduction in BOLD signal be physiological
rather than artifactual?
Thanks again for your help.
Jonathan Brooks Ph.D. (Research Fellow)
Magnetic Resonance and Image Analysis Research Centre
University of Liverpool, Pembroke Place, L69 3BX, UK
tel: +44 151 794 5629 fax: +44 151 794 5635
On Wed, 26 Jul 2000, Richard Perry wrote:
>Dear Geraint (and Jon),
>>Richard also said:
>>> One could imagine a condition by time
>>> interaction occurring non-specifically, perhaps even because of the
>>> physics of the scanner.
>>Although I agree with Richard's general points, I'm not sure about this
>>specific point. Overall fMRI signal in all conditions may wander over time
>>for a variety of experimentally uninteresting reasons. For example, low
>>frequency drifts in signal are common to all conditions and usually removed.
>>But relative differences in signal between two conditions that change over
>>time ('condition-by-time interactions') have by their nature
>>condition-specific causes. I agree these may be trivial or uninteresting,
>>but in Richard's example I think would have to be some condition-specific
>>scanner physics that affected the active condition but not the rest
>>condition (or vice-versa) in a time-dependent way.
>I guess that, to give an over-simplistic example, I was imagining a
>case in which the scanner drifts into a state in which it is much
>less sensitive to BOLD contrast. During the first few scans there is
>a large differential response. As the session continues, the
>sensitivity to BOLD decreases, and therefore the differential
>response decreases. This would appear as a condition-by-time
>interaction (possibly even without any 'main effect' of time), but
>one could not necessarily infer that the physiological response to
>the 'active task' compared to the 'baseline' changes over time, which
>is what one is usually interested in.
>If I wanted to provide evidence for a genuine time-dependent change,
>my own preference would be to look for a three way interaction. If
>one could demonstrate, for example, that in context A there is a
>differential response (task - baseline) which changes over time,
>whilst in context B there is a differential response which does not
>change over time, then I would feel happier that the change is at the
>neural level. However, an alternative would perhaps be to look for
>some way of 're-setting the physiology' so that the decline in
>differential signal can be repeated several times during a single
>session, making a scanner-dependent explanation unlikely.
>from: Dr Richard Perry,
>Clinical Research Fellow, Wellcome Department of Cognitive Neurology,
>Darwin Building, University College London, Gower Street, London WC1E
>Tel: 0171 504 2187; e mail: [log in to unmask]