Proportional scaling will divide the intensities in the preprocessed
data by the global value that you specify prior to fitting the GLM
through it. The threshold you specify will analyse those parts of the
images that have greater intensity than this. I would suggest
specifying a lower value in order to include the interesting parts of
the data in the analysis.
Best regards,
-John
Hello,
I have a question regarding global normalization for VBM. I've
been following John Ashburner's SPM Course 2010 VBM Tutorial,
with absolute thresholding at 0.1. I've been using GM volume
(from the native space segmentation volumes) calculated using
get_totals.m. When using Proportional Normalization to the
globals I get an error saying "no in-mask voxels". To my
understanding the proportional normalization divides the voxels
by the global prior to masking, so with modulated data (voxels
running from 0 to 1) and GM volumes (running around 800 mm^3),
the data clearly does not survive the absolute masking. As
John's manual suggests both proportional normalization to either
TIV or GMV, and Absolute thresholding, for modulated data, I'm
wondering if I'm simply using the wrong values for the globals.
Any help would be kindly appreciated!
With thanks,
Jared
--
John Ashburner <[log in to unmask]>
|