DISCLAIMER

This software has been developed for research purposes only, and hence should not be used as a diagnostic tool. In no event shall the authors or distributors be liable to any direct, indirect, special, incidental, or consequential damages arising of the use of this software, its documentation, or any derivatives thereof, even if the authors have been advised of the possibility of such damage.

CODES AND EXECUTABLES

C++ codes of the multi-scale LDDMM algorithm of [Risser et al. TMI 2011], the geodesic shooting algorithm of [Vialard et al. IJCV 2011], and tools for the treatment of 3D images are freely available. The codes can be downloaded using svn:

svn co https://utilzreg.svn.sourceforge.net/svnroot/utilzreg utilzreg

Recent executables for mac OSX, Linux 64 (compiled on Debian/i686) and Windows (compiled on Windows 7 - AMD E1) of utilzreg are also available here. Toy example 2D/3D images are also given.

HOW TO COMPILE THE CODE?

To compile the C++ code, simply use CMake and then your usual compiler. You have to set nothing at all with CMake, the program is self-contained. Until now, we have compiled it under Mac OS 10.6.7 and recent versions of Linux (Ubuntu/Debian) with gcc 4.2.1 and higher. We have also compiled the code under Windows using MS Visual C++ 2010 Express after revision 46 (19.09.2012).

HOW TO USE AN EXECUTABLE?

All programs are used in command line. To use an executable, say [toto], open a terminal, then go to the directory where [toto] is saved, then write the command line './[toto]'. It will explain you how to use [toto]. If you have a message like "Permission denied", enter the command 'chmod u+x [toto]' and try again './[toto]'.

HOW TO VISUALISE RESULTS?

Several of softwares allow to visualise 2D/3D/3D+t nifti images. Among them, we definitively like paraView, itk-SNAP, and 3D Slicer. Most of the time, we however use rview, the visualisation program of irtk.

HOW TO CONVERT A 2D OR 3D IMAGE INTO THE NIFTI FORMAT?

We use Convert3D. It converted all the 3D image formats we tried and also png images (the only 2D format we tried).

REMARKS

* All the codes work on non-compressed nifty images (.nii).
* Kernel sizes are in millimetres and not voxels.
* Some nifti images may not be opened for a mysterious reason. If utilzreg can't open a nifti image, load it using itksnap or rview and then save it. The saved nifti image will be OK for utilzreg.
* To perform multi-resolution registration -- different to multi-kernel registration --, register first subsampled images -> the subsampled velocity field will be saved in the files "[prefixOutput]_VelocityField_*.nii" -> you can then use this velocity field as an input when registering the images at a finer resolution by using the option "-PrefixOutputs".

DEVELOPERS

Laurent Risser, PhD - Imperial College London (2009-2010), University of Oxford (2011), CNRS - Institut de Mathématiques de Toulouse (2012-.)
Francois-Xavier Vialard, PhD - Imperial College London (2009-2011), Université Paris Dauphine (2011-.)

Acknowledgements

We warmly thank Daniel Rueckert (Imperial College London), Julia Schnabel (University of Oxford) and Darryl D. Holm (Imperial College London) who supported us in the development of uTIlzreg. We are also grateful to the Imperial College Strategic Initiative Fund, the European Research Council, the EPSRC EP/H050892/1, the Cancer Research UK and the EPSRC Oxford Cancer Imaging Centre (OCIC) for fundings.