SPM5 Tissue Segmentation - Gray matter segmentation
===================================================

To obtain the GM segmentation using SPM5, I used the
tissue segmentation algorithm that is included in the
software package. One item to note is that the positional
normalization of the IBSR 20 Normals T1 dataset does
not work right out of the box with SPM's segmentation
routine.

To make it work, I had to use SPM's display tool and
reorient the image approximately so that the Anterior 
Commissure was closer to mm coordinate [0,0,0]. Tutorials
regarding how to use the display tool are available 
on the Internet.

After reorienting the images and saving a copy, still
using the ANALYZE HDR/IMG file format, I proceeded to
use SPM5's segmentation algorithm with the default
parameters to segment the image into the various tissue
classes. The images in this archive are only of the
GM segmentation, and the names of the files have the c1
prefix.

To return the segmented images to the original positionally
normalized three dimensional space, I simply copied the
original HDR files for the 20 T1 images and renamed them
with the c1 prefix and placed them in the same directory
as the segmented IMG files, overwriting the modified HDR
files.

One will notice that I have thresholded the resulting
images as well, using a 50% threshold. This was performed
using the 3dcalc utility from AFNI. The parameter for the
step function will depend on the data format that is used.
For instance, if the data is represented by floating point
number, then 50% would correspond to 0.5. Otherwise, if the
data is represented by unsigned 8 bit integer, 50% would
correspond to an intensity value of 127. 

i.e. 3dcalc -a input.hdr -prefix output.hdr -expr 'step(a-.5)'

OR 

i.e. 3dcalc -a input.hdr -prefix output.hdr -expr 'step(a-127)'

To check that this procedure worked, I overlaid the
segmented image on top of the original image in AFNI
and the results displayed correctly.


==========================

If this data is to be made available publicly, please acknowledge
that it is data resulting from the following publication:

   O. Tsang, A. Gholipour, N. Kehtarnavaz, I. Panahi, K. Gopinath,
   and R. Briggs,  Comparison  of Tissue Segmentation Algorithms
   in Neuroimage Analysis Software Tools, Proceedings of the
   30th IEEE EMBS Annual International Conference, Vancouver,
   BC., Aug 2008.