MNI Non-parametric Non-uniformity Normalization (N3) Package


DESCRIPTION
-----------

An artifact often seen in MRI is for the signal intensity to vary smoothly
across an image.  Variously referred to as RF inhomogeneity, shading
artifact, or intensity non-uniformity, it is usually attributed to such
factors as poor radio frequency (RF) field uniformity, eddy currents driven
by the switching of field gradients, and patient anatomy both inside and
outside the field of view.

The perl script nu_correct implements a novel approach to correcting for
intensity non-uniformity in MR data that achieves high performance without
requiring supervision.  By making relatively few assumptions about the
data, the method can be applied at an early stage in an automated data
analysis, before a tissue intensity or geometric model is available.
Described as Non-parametric Non-uniform intensity Normalization (N3), the
method is independent of pulse sequence and insensitive to pathological
data that might otherwise violate model assumptions.  To eliminate the
dependence of the field estimate on anatomy, an iterative approach is
employed to estimate both the multiplicative bias field and the
distribution of the true tissue intensities.  Preprocessing of MR data
using N3 has been shown [2] to substantially improve the accuracy of
anatomical analysis techniques such as tissue classification and cortical
surface extraction.

References:

   [1] J.G. Sled, A.P. Zijdenbos and A.C. Evans, "A non-parametric method
       for automatic correction of intensity non-uniformity in MRI data",
       in "IEEE Transactions on Medical Imaging", vol. 17,  pp. 87--97,
       February 1998. 

   [2] J.G. Sled, "A Non-parametric Method for Automatic Correction of
       Intensity Non-uniformity in MRI Data", Master's Thesis, McGill 
       University, Montreal, QC, May 1997
       (available at http://www.bic.mni.mcgill.ca/~jgsled/thesis/)

   [3] J.G. Sled, A.P. Zijdenbos and A.C. Evans, "A Comparison of
       Retrospective Intensity Non-uniformity Correction Methods for MRI",
       Information Processing in Medical Imaging, pp. 459-464, 1997 


GETTING N3
----------

You can obtain a copy of the latest version of the MNI_N3
software from:

    http://packages.bic.mni.mcgill.ca/tgz

Look for a file named N3-x.yy.tar.gz (where x.yy is the release
number).
 
MNI_N3 requires that both the MINC and netCDF packages be built and
installed first.  The MINC file format is a highly flexible medical
image file format built on top of the netCDF generalized data format.

You will also need Perl 5, version 5.003 or later.
You can always find the latest version of Perl at at any
mirror of the Comprehensive Perl Archive Network (CPAN), in src/5.0/.  
   
Lastly, you will need the perl library mni_perllib available at

    http://packages.bic.mni.mcgill.ca/tgz


INSTALLATION
------------

In order to make the MNI N3 utilities:
 
   0a) You will need a recent version (egcs-1.0 or later) of the GNU C++
       compiler (g++) unless you are compiling under IRIX (see below).  
       GNU C++ can be obtained at

        http://gcc.gnu.org/

      In addition, you will also need a GNU version of make (gmake on
      some systems), available at any GNU ftp site.  

        ftp://prep.ai.mit.edu/pub/gnu/

   0b) SGI users do not need to obtain g++ as N3 will also compile using
       the DCC compiler packaged with IRIX.  SGI users still, however, need
       to obtain a version of GNU make.

   1) You may have to build and install the netCDF library.

   2) You may have to build and install the MINC library 
      (check that you have version 0.6 or later).

   3) You may have to build and install Perl 5.  (check using perl -V that
      it is version 5.003 or later)

   4) You may have to build and install the mni_perllib package.  If you
      decide not to install the library in its default location
      (/usr/local/lib/perl5/site_perl on many systems) then you will
      need to set your PERLLIB environment variable.  For example, if you
      were to install the libraries to /usr/local/mni/lib/perl5 then you
      would set the PERLLIB variable with 

        setenv PERLLIB /usr/local/mni/lib/perl5

      (assuming you use a csh-variant) in your ~/.cshrc or the global
       /etc/cshrc file.  (Or, if you're a Bourne-shell, ksh, zsh, or bash
       type of person, put

         PERLLIB=/usr/local/mni/lib/perl5/site_perl ; export PERLLIB

      in ~/.profile or /etc/profile.)
      
      Also, note that during installation of the mni_perllib package you
      will be queried for the default location where model data included
      with the MNI_N3 package will be stored.  In the future, most MNI
      software packages will install their model data in a subdirectory of
      a common data directory, which by default is called /usr/local/mni/data.
      While the default location of model data can be overridden on a
      package by package basis, doing so will require each user to specify
      the model data's location using the environment variable
      MNI_DATAPATH.
  
   5) We have used the GNU autoconf program to create a `configure'
      script that will automatically determine various system-
      dependent values and generate Makefiles needed to
      build the package.

      In the directory where you unpacked the N3 tar file,
      type

         ./configure

      `configure' works by poking around your system to determine all
      the system-specific values needed to build and install
      MNI_N3.  Once it has done that, it creates Makefile.include,
      which is the granddaddy Makefile used by all the programs in the
      package.  If `configure' runs successfully, skip to step 7 below.

      If `configure' complains about not being able to find certain
      things, though, you may have to give it some hints.  In
      particular, this be necessary if you have installed netCDF, MINC,
      Volume_IO, or Perl 5 in weird places.  The easiest way to do that
      is with the --prefix option; for instance, at the MNI, we keep
      home-grown software under /usr/local/bic, so here we run
      `configure' like this:

         ./configure --prefix=/usr/local/bic

      and the MINC and Volume_IO files will be found in various
      directories under /usr/local/bic.  (netCDF is *not* home-grown, so
      we keep it under /usr/local.  Luckily, `configure' is smart enough
      to look there if all else fails.)

      If you simply can't get `configure' to find some files, you can set
      various environment variables, e.g. NETCDF_INCLUDE to specify the
      directory where netcdf.h lives. For instance, if you installed netCDF
      to /tmp, its include file would be in /tmp/include, and you could
      say:
 
         env NETCDF_INCLUDE=/tmp/include ./configure

      Here are all the environment variables you might want to set to
      help configure find things:

         NETCDF_INCLUDE    where to find netCDF header (netcdf.h)
         NETCDF_LIB        where to find netCDF library (libnetcdf.a
                           on most Unix-like systems)
         MINC_INCLUDE      where to find MINC header (minc.h)
         MINC_LIB          where to find MINC library (libminc.a)
         VOLUMEIO_INCLUDE  where to find Volume_io headers (volume_io.h,
                           volumes.h, etc.)
         VOLUMEIO_LIB      where to find Volume_io library (libvolume_io.a)
         PERL              where to find Perl binary (must be Perl 5.003)

      Note that these are all *inputs* required to build N3.
      There are also a number of output directories that determine where
      various files are installed to; these are explained below.

      If you have built shared-library versions of netCDF, MINC, or
      Volume_io, configure should be able to find these.  It looks for
      libraries first with a ".a" extension, then with ".so"; this
      appears to be a quite common convention under various Unices, and
      is known to work under IRIX and Linux.  If your system has a
      different scheme for naming shared libraries, you might need to
      set the LIBEXT variable; e.g. to only look for ".so" files, 

         env LIBEXT=".so" ./configure

      (LIBEXT is just a space-separated list of extensions, including
      the dot on each one.)

   6) You might also want to set various site-specific installation
      directories when running `configure'.  MNI_N3 installs many
      files in various directories, and expects to find other files in
      certain directories, so it's important to know where you want
      stuff to go.  By default, it all gets put under /usr/local, unless
      you specify a different base directory with `--prefix').  In
      particular:

         binary executables:      /usr/local/bin
         scripts:                 /usr/local/bin
         average brain mask:      /usr/local/share/N3

     (Binaries and scripts are distinguished in case you have a
      multi-platform environment and want to share scripts between
      machines, e.g. using /usr/local/share.)  All of these
      directories can be customized, either by setting an environment
      variable before running `configure' or by editing
      Makefile.include after `configure' has run.  If you wish to use
      environment variables, set them as follows:

         BINDIR         directory for binary executables
         SCRIPTDIR      directory for scripts (machine-independent executables)

      Note that if you feel like setting ALL of these (e.g., to install
      MNI_N3 to /tmp or your home directory), you're probably
      better off to use `--prefix'.    

      If you prefer to edit Makefile.include, just read the instructions
      there -- but be warned that any changes you make will be lost if
      you re-reun `configure'!

   7) In the directory where you have unpacked both the MNI_N3 package and
      the test data, type:

        make build
      
        make test

        make install

      Note that you may need to set the environment variable TMPDIR to some
      more appropriate place than /usr/tmp on systems where /usr/tmp has 
      little available space.

   8) Now, you're ready to test nu_correct on real data.  Start by typing

        nu_correct -help
 
      nu_correct makes use of a number of other binaries and scripts that 
      are also useful on their own.  Their functionality is summarized below:

      nu_correct    estimates and removes intensity non-uniformity from
                    a volume.
      nu_estimate   estimates the intensity non-uniformity in a volume, storing
                    the result in a compact format referred to as an .imp file.
      nu_evaluate   corrects the intensity non-uniformity in a volume using
                    the given .imp file.
      imp2field     expands an .imp file into a MINC volume representation
                    of the correction field.
      field2imp     converts a MINC volume representation of a correction 
                    field into an .imp file.
      volume_stats  prints a variety of summary statistic for a volume 
                    or region of interest.
      volume_hist   produces a histogram of a volume.
      spline_smooth smoothes a volume using tensor cubic B splines or thin
                    plate splines.
    
Further information

For further information you may want to checkout the N3 home page at
http://www.bic.mni.mcgill.ca/software/N3 where you can find the FAQ for
N3.  Also, you can write to me, John Sled, at: 

John Sled
McConnell Brain Imaging Centre            Tel: (514) 398-8547
Montreal Neurological Institute           Fax: (514) 398-8948
3801 University Street, Room WB208        Email: [email protected]
Montreal, Quebec  H3A 2B4 Canada