Hi Richard,
Looks like you're not actually using VBM but working with diffusion data. I am aware of statistical issues surrounding single subject analyses in VBM but if you've got essentially parametric maps IMHO you should be fine.
You've created a template - good. Obviously this is one datapoint now so as you've correctly identified you can't do stats on it. I would suggest the following:
1. check that your template corresponds to MNI space with CheckReg; set origins etc.
2. normalise all patients _and_ all controls to this template (-> w* images)
3. smooth all w* images (search the list for tips on level of smoothing depending on smoothness of your images, voxel sizes, expected changes etc.) (-> sw* images)
4. under PET models, for each control (and then for each patient) run single-subject: conditions and covariates, without covariates, pretending all sw* scans come from one subject under different conditions: 0 1 1 1 1 1 1 1 1 1 1 1 1 (one "1" less for the controls, as you'll be testing each one against the rest). This should be the same for SPM5, 2 or 99.
5. evaluate for increases and decreases in the single subject compared to the control group with the contrasts 1 -1 and -1 1.
Some people do a similar thing in one big design matrix but that doesn't seem right to me as you would be deriving your normal variance from controls but also from patients (for groups, we've got unpublished data showing this is a problem) - if you run your analysis as many times as you've got controls and patients, you'll get around that.
HTH,
Alexander
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]On
Behalf Of Richard Bauld
Sent: 15 December 2005 18:01
To: [log in to unmask]
Subject: [SPM] VBM Statistics
Hey SPMers,
I'm a medical student currently working on a project looking at
differences in FA in different populations compared to a normal brain for
an intercalated degree.
We've created a normal FA map from 12 controls and have been using this as
the template to normalise our patient data to. Currently I'm working on a
population suffering from MS. The factor we're investigating is simply the
templates ability to identify lesions, so that we can then move on to
other populations.
I've normalised all the patient FA maps to the template and now need to
compare each patient to the template so I can decide whether the template
is of any use. However, I cannot for the life of me work out how to do
this.
I've looked through Basic Models, etc, but the tests don't give me any
output that is of any use. I was wondering whether anyone could point me
in the right direction, as I am at a complete loss!
Thank you all already, as many of your posts have already helped me get to
where I am at the moment
Richard
|