Marina,

This tar contains four MRI datasets in DICOM format.  Each data set
consists of 124 images of 256x256 unsigned shorts. The filenames for each MRI
slice have the pattern CT1.*.#, CT1.*.##, CT1.*.###( i.e. ignore all
files that have .hist or .hds or other non-numeric extensions). The
easiest way to read these images is to skip N bytes at the start of
each image where N is computed as (filesize - 256*256*2).  I have
matlab code to do that if you need it.

The four directories are:

	AL_AZMA__GHAZI_DAHAS_AXIAL____11_apr_2001__MRI32/
	AL_AZMA__GHAZI_DAHAS_AXIAL____11_apr_2001__SEG36/

	KEOMURJIAN__LUCINE___AXIAL____29_apr_2001__MRI16/
	KEOMURJIAN__LUCINE___AXIAL____29_apr_2001__SEG31/

where the MRI in the directory name indicates that the data is the raw
MRI scan i.e. each voxel is a 16-bit value that represents the MR
signal at that spatial location. The directories with SEG in the name
are segmentations of these MR scans.

In both scans 3 structures have been segmented and assigned the
following labels: 

	Brain 1 
	Vessels 3
	Tumor Bright 5
	(Background 0)


In addition, I've included a solaris 5.2 executable m3list that can be
used to dump the header of a DICOM image.  I doubt you will be able to
use it, but figured I'd take a chance.

-Tina

