At Tuesday 29.08.2006 11:37, Jan Gläscher wrote:
>I found this entry in the SPM archives very helpful:
>http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind04&L=SPM&P=R392137&I=-3&X=5B59A1759A28595368
>
>Briefly (citing Will Penny), if your images are
>scaled to global mean of 100 then your betas
>will reflect the "percent *global* signal
>change". This is the usual case with fMRI
>time-series. You can access the beta by using
>the 'plot' button (parameter estimates)
>
>If you want the "percent *local* signal change"
>you'd have to compute it yourself using this formula:
>beta(task)*100 / beta(constant term)
>
>You'll find much more on this if you search
>archives for either "% signal change" or
>"percent signal change". It's a very common
>question and there is lots of information around.

Some days ago I wondered if it is meaningful to
scale the signal change as obtained by the beta estimate by 'constant term'.

In the MARSBAR FAQ for example signal change is
somewhat related to the 'constant term' which is
obviously a kind of baseline. in the given
example a calculated BOLD gain induced by a
stimulation is 0.2 SPM scaled 'units'. to get the
% signal change the 0.2 units are now related to
the SPM scaled 'mean' of the voxel - in the
example 192. so we end up with a % signal change of 'roughly 0.1'.

IMHO this approach is assuming that the specified
'mean' is meaningful in a sense that a BOLD gain
can only be related to this baseline. while the
induced gain is regarded as a kind of secondary
measure of neuronal activity (deoxyHB or oxyHB in
B0) I think that the 'baseline' can not easily be
regarded as related to neuronal activity. so
relating the BOLD gain to any kind of 'baseline'
would certainly include parts of the BOLD signal
that are _not_ related to the deoxyHB or oxyHB effects in B0.

In the EEG (EP) world this problem is faced as
well but solved by a baseline correction, i.e. by
intentionally ignoring the baseline. Assuming
that there is a kind of baseline in the
experiment (I am aware that this is not easy to
realize) wouldnt it make sense to proceed this
way in fMRI, too? The 'mean' isn't homogenous in
the grey matter so 'scaled' and 'baseline
corrected' signal change values tend to show
significant differences in different cortex
regions that induce different results, e.g. in
the 'raw' signal change and in a subsequent
second level analysis (at least in my data).

However, SPM is transforming the data during the
preprocessing (to a mean of 100). Does anyone has
an idea what would then be the impact of the
magical scaling of SPM on a comparison of two conditions?

Best regards

Christoph




--
Christoph Christmann [log in to unmask]
Psychologist, MEE Fon/Fax +49 621 1703 63 18/05
Central Institute of Mental Health http://www.zi-mannheim.de
Department of Clinical D-68159 Mannheim J5
and Cognitive Psychology
**********************************************************************************************
IMPORTANT: The contents of this email and any attachments are confidential. They are intended for the
named recipient(s) only.
If you have received this email in error, please notify the system manager or the sender immediately and do
not disclose the contents to anyone or make copies thereof.
*** eSafe scanned this email for viruses, vandals, and malicious content. ***
**********************************************************************************************