Dear Ignazio,
I presume this is FDG data?
The problem is that a t-test isn't appropriate to compare one patient against five controls - try again with 12 or 15 controls, ideally more. You're simply seeing some artefacts outside the head - with crazy t scores.
The rest of your preprocessing looks essentially fine. Once you've got enough data to do stats, you may wish to spend some time on the global normalisation issue. For FDG you should probably either normalise by dose and body weight (i.e. SUVs) or use a group-specific ANCOVA (which is what I normally do) or, in the context of Alzheimer's, probably scale by an unaffected area which you can get from a first-pass SPM (see Yakushev I et al. Neuroimage recently for details).
Good luck,
Best wishes,
Alexander
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of Vilardi Ignazio [[log in to unmask]]
Sent: 13 May 2009 17:45
To: [log in to unmask]
Subject: [SPM] Alzheimer analysis
Dear all,
Sorry for this email, but there is nobody in my department I could ask to. Please, may you tell me if I’m carrying out correctly the first analysis?
I have one AD patient and 5 normal patients for the control.
I just click on the button “Normalise: Estimate and Write”;
Under “Data” I create 6 subjects;
Under “source image” and “images to write” I insert the same scan for each patient (the hdr/img file);
In “estimation options”, in the “template image” I insert the template provided by SPM programs\spm\templates\PET.nii
“Source Image Smoothing” = 8;
The other option left as defaults.
After “Normalise”, click on “Smooth”;
In “Image to Smooth”, I insert the 6 images with the prefix w, images created with the Normalise function;
“FWHM”=10;
“Data type” = same;
After Smoothing, I click on “basic models”;
Under “design”, I choose “two sample t-test”;
“Group 1 scans” I choose the AD patient (sw file)
“Group 2 scans” I choose the 5 normal patients;
“Indipendence”=Yes;
“Variance”=unequal;
“Grand mean scaling”=No;
“ANCOVA”=No
No covariates created and other options left as defaults.
After creating SPM.mat, I click on “estimate”, choose the SPM.mat.
Then, I click on “results”, select “t-contrasts” defining one;
“Name”=nc>ad and “contrast”=[1 -1], “submit” and “ok”.
“Mask with other” = no;
“Title for comparison” nc>ad;
“P value” = FWE and 0.05;
Threshold = 0;
Attached to this email there is the file for the final results.
Why I have these warnings?
Warning: Returning NaN for out of range arguments
> In c:\programmi\spm5\spm_Pcdf.m at line 88
In c:\programmi\spm5\spm_P_RF.m at line 106
In c:\programmi\spm5\spm_P.m at line 48
In c:\programmi\spm5\spm_list.m at line 487
I much appreciate if somebody can help me.
Many thanks
Best regards
Ignazio
|