 |
 |
Hi,
I've spent some time on inter-subject global scaling of T1 and T2 weighted MR images using their (scalp-masked) histograms. The measures I've investigated to characterise global intensity were derived from linear fitting of the histogram upslope and downslope and included X crossings of these slopes and a 'hybrid' weighted mean (?=median for a histogram) using the linear fits below 50% of the histogram peak level and the actual histogram above.
In an SPM2 clinical study of 2 groups we found significant differences for weighted T2 using the 'hybrid' weighted mean and for weighted T1 using the X crossing of the upslope. (The global levels were entered as nuisance covariates in SPM). We've submitted this for publication.
The code for this is written in IDL and is downloadable as hgram.zip and lblib.zip from
http://www.anzsnm.org.au/servlet/NM?command=changePage&page=289
It will batch process your studies and write an SPM-ready matlab function, the name of which can be entered at the SPM Nuisance Covariates prompt.
Note that there is a freely distributable utility called the 'IDL Virtual Machine' that permits you to run such a program, if already compiled to a .sav file.
Contact me if you need help with this.
Leighton
Leighton Barnden PhD
Medical Scientist
Nuclear Medicine, The Queen Elizabeth Hospital
Woodville Road, Woodville
SA 5011, Australia
ph. +61 8 8222 6438