Hi,

I see your scans from Insight are 2-byte signed integer voxels encoded (within the NRRD file) with GZIP.  I'm guessing you just want to get to the raster datacube, and don't care about all the interpolation and sampling management stuff that the NRRD Library provides (http://teem.sourceforge.net/nrrd/). 

One possible option, that I didn't try because I don't feel like building the "unu" command line utility (http://teem.sourceforge.net/unrrdu/index.html) from source code, is to use unu to convert your NRRD files to VTK format  (http://teem.sourceforge.net/nrrd/otherformat.html) which Mathematica does import (see Help on Import).  But "unu" conversion support is incomplete, and I don't know how completely M supports VTK. 

The other is to take an NRRD file from Insight, then use a hex editor to chop out its header, (chop right before the "1f 8b" gzip hex lead-in) then save-as .gzip.  Then use 7-Zip to extract  the .gzip file into a binary file.  From here, I can read and display in M, as so:
   Since Mathematica itself can decode GZIP, the entire process should be able to be able to be automated within M, i.e. no need to externally use a hex editor and GZIP.   If you can't take it from here, let me know, and if I'm not too busy, I'll finish a set of Mathematica commands to automate everything (that is, ignoring generality; only specifically handling these 5 files). 

Also, it's possible to write custom import and export converters that work directly with Import and Export commands. If you want to go to this depth, see http://reference.wolfram.com/mathematica/tutorial/DevelopingAnImportConverter.html and http://reference.wolfram.com/mathematica/tutorial/DevelopingAnExportConverter.html .