Dear Jesper,
Thanks for your reply. I apologize for not being clear enough in my
first message.
We are doing drug studies with O15-PET to study the acute effects of
CNS-depressant drugs in a single session. We have 3 conditions in the
baseline (2 scans each) and we repeat the same 3 conditions after giving
drug (again 2 scans each). In SPM96 this was a multi-subject, 6
conditions with replications, and sometimes we threw in some covariates as
well, just for fun. We also have a small group of subjects who received
placebo, so we could also choose to do a multi-group, conditions (by
covariates) analysis with replications.
Doing some preliminary analysis with SPM96, we got very strong
changes in rCBF over widespread regions in cortex and subcortical areas,
even with only 5 or 6 subjects.
Now with SPM99 we have doubled the number of subjects but we see
almost no change at all in rCBF, even though subjects are experiencing
strong sedative effects. Surely this cannot be correct! I suspect that the
difference is due to the way SPM99 handles scan order sequence. Instead of
specifyng the replicated scans by condition, we now put in the order of
scans in chronological order. In our design, unfortunately but
unavoidably, scan order is completely confounded with drug administration.
(Scans 1-6 = baseline, Scans 7-12 = Drug.) When SPM99 corrects for time and
order effects, it removes all our experimental effect at the same time.
(Ouch!)
The effect of the drug is long enough that we can't counterbalance the
design by giving the drug first to half the subjects, waiting for it to
wear off, and then administering the baseline condition. We chose to do
both conditions in the same session to avoid between-session changes in
state, and problems of repositioning the subjects in the scanner for their
2nd session.
Is there a way to either a) disable this automatic correction for time
effect in SPM99, or b) access the SPM96 design matrix specification from
inside SPM99 (as the Help file suggests one can) ?
I'd appreciate any comments or suggestions. Thank you in advance!!
Ruth Reinsel
************************************************
Ruth Reinsel Ph.D.
Dept. of Anesthesiology & Critical Care Medicine
Memorial Sloan-Kettering Cancer Center
1275 York Avenue, Box 24
New York, N. Y. 10021
Tel: (212) 639-8111
Fax: (212) 772-8646
e-mail: [log in to unmask]
************************************************
______________________________ Reply Separator _________________________________
Subject: Re: how to run SPM96 type analysis from within SPM99?
Author: jesper ([log in to unmask]) at Internet
Date: 07/12/2000 4:58 AM
Dear Ruth,
[log in to unmask] wrote:
> Hello SPM helpers,
> We did some preliminary analyses of PET O-15 data (5 subjects) some
> time back using SPM96 implemented under Unix and Matlab 4.2. Now we have
> collected more data (11 subjects) and we have installed SPM99 on Windows NT
> with Matlab 5.3.
> When re-analyzing the data (including realignment and normalization
> within SPM99) we seem to be getting quite different results. I believe the
> difference is at least partly due to SPM99 making the correction for time
> effects within session (scan number).
> I would like to compare the SPM96 and SPM99 analysis on the same
> datafiles, but we no longer have access to the Unix installation of SPM96.
> According to the SPM HELP files, it should be possible to access the SPM96
> format for setting up the design matrix. However I can't find the SPM96
> options on the menu for PET/SPECT analysis. Can you please tell me how to
> implement an SPM96 analysis for PET under SPM99?
>
As far as I know there is no "SPM96 mode" in SPM99. There are a number of
small changes between the analysis of PET data between SPM96 and SPM99 (even
though most changes affect mainly fMRI). Certainly the spatial preprocessing
has been changed (improved) and I would suspect that you will find different
results e.g. for realignment of data with large movements and for spatial
normalisation of noisy images.
In the stats section the main difference affecting you is that inferences are
now based directly on t-fields (rather than gaussianised t-fields).
However, the general design options available in SPM96 are all available in
SPM99, albeit with sometimes different names. There is certainly no "general"
time effects correction in SPM99. If you could just let us know what design
option you used in SPM96 (possibly a small tiff of the design matrix) I am
sure we could help you set up the same design with SPM99.
>
> Thank you in advance,
>
> Ruth Reinsel
Good luck Jesper
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|