Getting started

Getting help

Don’t panic!

If you have problems downloading or installing plastimatch, please send an email to our mailing list. We’re friendly people.

Downloading the code

The recommended method for downloading plastimatch is to use subversion to download the source code, and then compile the source. To download using subversion, use the following command:

git clone

If you are using Windows, you will need a git client. We recommend cygwin ( for command-line users, and TortoiseGit ( for graphical users.

If you have already downloaded a previous version, you can update to the latest version by executing the following command from within the plastimatch source directory:

git pull

Build dependencies

Debian installs

On debian, all the needed dependencies are already included. The following command will install all the needed prerequisites.:

sudo apt-get install g++ make git cmake-curses-gui \
  libblas-dev liblapack-dev libsqlite-dev \
  libdcmtk-dev libdlib-dev libfftw3-dev \
  libgdcm2-dev libinsighttoolkit4-dev \
  libpng-dev libtiff-dev uuid-dev zlib1g-dev

Cmake (required)

Plastimatch uses cmake, so you must download and install cmake before you can build plastimatch. Download from here:

Cmake 2.8.12 or higher is required.

C/C++ Compiler (required)

You will need a C/C++ compiler. If you are running Windows, we recommend Microsoft Visual Studio (Express or Full). Versions 2008 and higher are supported. You can download the Microsoft Visual Studio Express compiler from here:

On windows, you may use the MinGW compiler, however, ITK does not support cygwin.

On unix systems, g++ and clang are supported. If you use ITK 3.20, however, only g++ is supported. The clang compiler will work for newer versions of ITK.

On OSX, you need the Xcode package, and you must also install the command line tools. If you wish to use g++ instead of clang, do something like the following:

CC=/usr/bin/gcc CXX=/usr/bin/g++ ccmake /path/to/plastimatch/

ITK (required)

ITK is required for the main plastimatch program. But if you only want the DRR and FDK programs, you don’t need it. Get ITK from here:

We currently support version ITK 3.20.1 and greater, and ITK 4.1 and greater. For ITK 4, you will need to install DCMTK if you want DICOM support.

ITK < 3.20.1          Not supported
ITK 3.20.1            Supported (with caveats)
ITK 3.20.2            Recommended
ITK < 4.1             Not supported
ITK >= 4.1            Recommended (install DCMTK)

ITK 3.20.1 is a decent version, but it has a few bugs that cause problems on recent versions of gcc. These bugs are fixed in the ITK 3.20.2 maintenance release. To get ITK 3.20.2, do the following:

git clone git://
cd ITK
git checkout -b release-3.20 origin/release-3.20

When you build ITK, the following settings are recommended or required:

CMAKE_BUILD_TYPE                          Release
BUILD_EXAMPLES                            OFF
BUILD_SHARED_LIBS                         (EITHER)
BUILD_TESTING                             OFF
ITK_USE_REVIEW                            ON         # Below ITK 4.5
Module_ITKReview                          ON         # ITK 4.5 and greater

DCMTK (optional)

DCMTK is needed for DICOM-RT support with ITK 4. The supported version is 3.6. On linux, feel free to use the dcmtk that comes from your package manager (that’s what I do).

There are special considerations to building dcmtk:

  1. PNG, TIFF, and ZLIB are not required
  2. Set CMAKE_INSTALL_PREFIX to an install directory of your choice; I use $HOME/build/dcmtk-3.6.0-install
  3. On linux x86_64 platforms, you need to add -fPIC to CMAKE_CXX_FLAGS and CMAKE_C_FLAGS
  4. On windows, you need to set DCMTK_OVERWRITE_WIN32_COMPILER_FLAGS to OFF
  5. After building, you need to install; on linux do “make install”, or on Visual Studio build the INSTALL target
  6. When you run cmake on plastimatch, set DCMTK_DIR to the install directory

VTK (optional)

VTK is required for compiling reg-2-3, for 2D-3D image registration. You don’t need VTK if you only need plastimatch. Get VTK from here:

Only VTK version 5.6.1 is supported. On linux x86_64 platforms, you will need to adjust the compile flags, and add “-fPIC” to CMAKE_CXX_FLAGS and CMAKE_C_FLAGS.

In addition, VTK 5.6.1 has a small bug which prevents it from compiling on gcc version 4.6. You will need to edit the VTK source code. Specifically, you need to edit the file Utilities/vtkmetaio/metaUtils.cxx, and add the following line somewhere near the top of the file (for example after line 20):

#include <cstddef>

CUDA (optional)

CUDA is needed if you want GPU acceleration of the DRR, FDK, and B-Spline registration codes. You need to install the driver and toolkit, but the SDK is not needed.

Please note that CUDA is constantly evolving in order to provide new high performance computing features. The following table will help you with selecting the correct CUDA version to install/upgrade:

CUDA 2.X              Not supported
CUDA 3.X              Supported
CUDA 4.X              Supported
CUDA 5.0              Supported, Required for Kepler

Download CUDA from here:

FFTW (optional)

The FFTW library is used to implement the ramp filter for FDK cone-beam reconstruction. So if you are not using the FDK code, you don’t need this. We recommend the most current version of FFTW 3.

On windows, the precompiled DLLs work fine. However, you do need to create the import libraries. See this page for details:

WxWidgets (optional)

WxWidgets is needed if you want to build “Mondoshot”, the dicom screen capture program. Download WxWidgets from here:

Compiling plastimatch (Windows)

Before compiling plastimatch, compile or install the desired prerequisites. At a minimum, you must compile required packages such as ITK. Be sure to build ITK and plastimatch using the same build type (e.g. both as Debug, or both as Release).

Run CMake as follows:

  1. Select source directory and binary directory
  2. Click configure
  3. Select makefile format (e.g. MS VC 2005)
  4. <CMake configures>
  5. Set the ITK directory (sometimes it might be found automatically)
  6. Set directories for optional components (such as slicer)
  7. Click configure
  8. <CMake configures>
  9. Click OK
  10. <CMake generates>

Then build in Visual Studio as follows:

  1. Navigate to your binary directory
  2. Open the project file plastimatch.sln into MSVC.
  3. Change the build type (e.g. release, debug) to match ITK (and other dependencies. You probably want release.
  4. Click “Build Solution”. Let the project build.

Special instructions for running cmake with MSYS/gcc on Windows

There is a trick to building with MSYS/gcc. The trick is that you need to run the win32 cmake from the MSYS command line instead of the GUI. For example, here is the command that I use:

$ mkdir /c/gcs6/build/plastimatch-mingw
$ cd /c/gcs6/build/plastimatch-mingw
$ /c/Program\ Files/CMake\ 2.8/bin/cmake \
    -DITK_DIR=/c/gcs6/build/itk-mingw \
    -G"MSYS Makefiles" \

Then, edit CMakeCache.txt to set your options. Re-run cmake to create the MSYS Makefile, and then run make to build.

Special instructions for Visual Studio 2010

The CUDA compiler nvcc is not compatible with Visual Studio 2010. That is why we use Visual Studo 2008. But, if you insist on using VS 2010, there are some workarounds (Google is your friend).

Compiling plastimatch (Unix)

Build plastimatch as follows:

  1. mkdir /path/to/build/files; cd /path/to/build/files
  2. ccmake /path/to/source/files
  3. Type “c” to configure
  4. <CMake configures>
  5. Set the ITK directory (it may be found automatically)
  6. Set directories for other optional components (if necessary)
  7. Type “c” to configure
  8. <CMake configures>
  9. Type “g” to generate
  10. <CMake generates>
  11. Type “make”

Users with multicore systems can speed up the process of compiling plastimatch considerably by invoking make with the -j option. For example, a user with a dual-core system would type:

make -j 2

whereas a user with a quad-core system would type:

make -j 4

You can probably get even better performance by increasing the the number of processes (specified by the -j option) beyond the number of cores. One rule of thumb is to use approximately 1.5 times the number of available CPUs (see [1], [2]).

Compiling the 3D Slicer extensions

The 3D Slicer extension is now included in SlicerRT. Please see the developer instructions on the SlicerRT assembla page for detailed instructions.