Hi,
On 1 Aug 2009, at 17:56, Mengfei H wrote:
> Dear all,
>
> From speaking with others it sounds like most people using Melodic
> to remove
> unwanted noise components from motion artefacts, whereas the second
> function of Melodic in revealing unexpected activation is much less
> commonly
> used.
>
> Since I am conducting an exploratory study that is less heavily
> hypothesis-
> driven, I'm assuming that in this case, it may be useful to look at
> Melodic for
> task-related activation patterns.
>
> 1) I would like to first verify my Melodic setup:
>
> I have 10 subjects, same stimulus timing across all subjects.
> However 5
> subjects see the sequence of images in one order, and the other 5
> see the
> images in the reverse sequence. Within each group of 5, 2 subjects
> see the
> image 1 under condition A (1-A), image 2 under condition B (2-B),
> image 3
> under condition B (3-B), etc. The 3 other subjects are given the
> opposite
> conditions: 1-B, 2-A, 3-A, etc. Should I run 4 Multi-session Tensor-
> ICAs?
You're right that this is slightly tricky as they don't all have the
same paradigm timing. If you want one simultaneous analysis of all
datasets then you could use the "Concat-ICA" option in the GUI, though
post-ICA interpretation may be tricky. Alternatively you could run ICA
separately for each run, and start to build up your analysis that way,
comparing and combining different components if you can associate them
with particular conditions.
>
> To reveal any task-related activations, do I just go to Post-stats
> tab of the
> Melodic Gui and load an individual subject's design.mat file for the
> time-series
> model field and a design.con file for the time-series contrast
> field? Should I load
> anything for the session/subjects model/contrasts fields?
>
> 2) Can you explain how Melodic reveals task-related activation that
> a Feat
> analysis would not have?
ICA doesn't follow a prespecified design, and so is flexible in
showing you what is actually in the data, which may be different in
various ways from what you would predict with the GLM.
Cheers.
>
> Thank you very much for your help!
> Mengfei
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|