Hi, ive just started to use spm and am not entirely sure what is the best
way to go about things, any help is greatly appreciated!
Using SPM2 to analyse mri data from mice which have been injected after
about 4 baseline images have been taken without injected material. Looking
at the brain would like to find activation sites connected to the
injection, would expect a steady rise after injection then a plateau. Only
using one subject per analysis.
As im not exactly sure what the timecourse should look like im finding it
a bit difficult to decide the covariate to put in. I have been looking at
the image in an imaging program to find the timecourse which there is in
the brain and then entering that as the covariate but I feel by doing that
spm is only bringing up the timecourse which I have decided, its not
actually telling me anything new! As I can't think of a way to obtain a
very good estimate of the timecourse before doing the experiment im
guessing there might not be a good solution to this.
As I am very new to spm there are some options which im not sure about. I
am using a scan with 66time points with 10slices per time point. Injection
occurs after the 4th timepoint. I have been using "Single
subject:covariates only" in the PET and SPECT interface.
Main decisions im not sure about!
Smoothing - Gaussian filter width to be used (have used twice the voxel
size), not sure of the size of activations, probably will be larger than
this.
Global normalisation method Proportional scaling or ANCOVA.
Other choices I am using, please let me know if anything looks really
wrong!
- No centering of covariate around mean
- No grand mean scaling
- Relative threshold masking of 0.9
- Global calculation is mean voxel value
- t-contrast of 1
- p-value is FWE 0.05
- extent threshold of 10 voxels
Thanks! Sorry about the long email. Im not very good with matlab by the
way and so would have trouble changing or adding in scripts.
Sean Donnellan
ICL
|