Hi,
I asked a variant of this question a while back, but nobody replied so I
thought I'd try again, this time with a more concrete example.
We are writing up a paper in which we used SPM99 to analyse some fMRI
data. We are using a threshold of P<0.05 (corrected) and gets lots of
activation in relevant areas. In the literature it seems almost
obligatory to put in a table of clusters (Zscores/Talairach
coordinate/Cluster Sizes etc.), and there are various mailbase questions
about extracting the table structure from SPM. I happened to use Robert
Welsh's script:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0104&L=spm&P=R27120
and then used the Talairach Daemon to help me identify the anatomical
locations.
So far so good.
However, some of my clusters are rather large and span quite different
anatomical regions. For example, I have a cluster which starts off just
anterior to SMA (on the midline) and weaves it way down to the medial
frontal pole (MFP). The MFP activation is very interesting, but in the
Table of Clusters it does not feature as a cluster or a sub-cluster. So
my beautifully constructed Excel tables only tell part of the story for
our paper.
If I make my statistical threshold more stringent (P<0.01 corrected)
then my MFP activation becomes distinct and appears in my Table of
Clusters.
Now, I realise there is nothing conceptually difficult here and nothing
is "wrong" with the way SPM reports clusters. I could amend my Tables to
say something like "Cluster extends to the Medial Frontal Pole", but it
seems a little inelegant. I could potentially have to do this for many
clusters, which would make the usefulness of a table questionable. I can
also, obviously, explain in the text where we found statistically
significant activation.
I just wonder if anyone has any ideas about what is the best way to
report these kind of activation maps. The nice thing about the "Table of
Clusters" is that it provides a common, well understood, method of
communicating the position and size of the activation clusters that were
found.
Some Question/Thoughts:
1. The reason I have got some quite large extended clusters is that I
have a relatively complex task with a fairly low-level baseline. Should
I really be designing my paradigms so that I only get relatively small
clusters? After all, large cluster sizes do interact with Global
Scaling/Normalisation. See:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0001&L=spm&P=R6648 and
http://www.mariarc.liv.ac.uk/scaletest.html
2. I wonder how much activation goes unreported in the neuroimaging
literature because authors simply publish the Cluster Table?
3. Perhaps Talairach (oops, I mean MNI) coordinates should only be
reported for relatively small discrete clusters and I should do away
with a Table altogether?
Perhaps some kind of generic flat-map technique for visualising all the
statistically significant activations in a single, standard, 2D space
would be the way to go in future.
Any thoughts/ideas welcome
Krish
--
Dr Krish D Singh
Department of Vision Sciences
Aston University
Aston triangle
Birmingham
B4 7ET
ENGLAND
tel: +44 (0)121 359 3611 ext 5176/5190
fax: +44 (0)121-333 4220
email: [log in to unmask]
|