My study uses 10 SPECT images of normal subjects on which each image was undergone to 2 different image processing procedure (A and B) resulting in 20 imagens (2 per subject) which were divided in two groups:
Group A (Images of subjects 1 to 10 with processing procedure A) and Group B (Images of subjects 1 to 10 with processing procedure B). In the visualization of the resulting images of the procedure B, we note that their voxel value was tripled if compared with the images of procedure A.
We want to measure the differences between this 2 procedures. We want to know what are regions have increase (-1 1) of signal and the regions where eventually there were decrease (1 -1) of the signal.
Our difficult is understand the function of a parameter in the Paired Design Model. If the parameter "Grand mean scaling" is turned on, I can see increase and decrease of the 2 conditions (fig 1), on other hand, if "Grand mean scaling" is turned off, the SPM result shows only increase, but not decreases(fig 2).
What mean the use of "Grand mean scaling" in this case?
The design parameters are:
Design
Paired t-test
Independence: Yes
Variance:Unequal
Grand mean scaling: Yes/No
ANCOVA:No
Covariates:0
Masking
Threshold Masking:
Relative
Threshold: 0.8
Implicit Mask: Yes
Global Calculation: mean
Global Normalization:
Overall grand mean scaling:Yes
Grand mean scaled value:50
Normalization: none
|