Skip to main content

Generate displacement fields with known volume changes

Project description

disptools

Generate displacement fields with known volume changes

GitHub release PyPI Wheels License Appveyor

This library provides utilities to generate and manipulate displacement fields with known volume changes. It implements three search-based algorithms for the generation of deformation fields, along with a small collection of utility functions, and provides optional GPU acceleration through a CUDA implementation.

The three algorithms implemented are referred as:

  • gradient: a gradient descent method (default).
  • greedy: a greedy search method proposed in [1].
  • matching: a volume matching method proposed in [2] and [3]. The implementation comes from the PREDICT atrophysim tool.

The library is built on top of SimpleITK, in order to provide a simple yet powerful set of functionalities for image analysis. Images stored as numpy arrays can be easily converted from and to SimpleITK and ITK image objects.

Documentation

The complete documentation for this package is available on https://martinopilia.com/disptools.

Quick example

import SimpleITK as sitk
import disptools.displacements as dsp
import disptools.drawing as drw

# Create an example Jacobian map
# A spherical ROI with a Jacobian of 1.1 (expansion)
jacobian = drw.create_sphere(10, 40, fg_val=1.1, bg_val=1.0)

# Create a binary mask for the ROI
mask = drw.create_sphere(10, 40) > 0

# Generate the displacement
displacement = dsp.displacement(jacobian, mask=mask)

# Check the correctness of the result within the ROI
error = jacobian - dsp.jacobian(displacement)
error = sitk.Mask(error, mask)

A 3D rendering of the resulting displacement field with ParaView, and a visualisation of the the error on the Jacobian within the ROI:

displacement error

Architecture

The project is structured in three layers:

  • a pure standard C99 library (whose headers are in src/headers), with no external dependencies, implementing the core algorithms for the generation of deformation fields. It is a standalone library that can be directly included in a C or C++ project. It is paired with an optional CUDA library, whose headers are in cuda/headers, that depends on src/headers and provides a GPGPU implementation of the key routines.
  • a Python C extension package _disptools (whose source is in the file python_c_extension/_disptools.c), providing a bare Python wrapper to the aforementioned library, using the NumPy C API to pass arrays. This can be directly included in a Python project with no dependencies other than NumPy.
  • a Python package (disptools), that wraps the _disptools package providing file IO (through SimpleITK) and implementing high-level features (such as the multi-resolution framework) and auxiliary utilities and functions.
    • disptools.displacements: module providing the main functions for the generation and manipulation of displacement fields.
    • disptools.drawing: collection of utilities to create test images.
    • disptools.io: collection of utilities to read and write to file.
    • disptools.measure: collection of utilities to measure some image features.
    • disptools.simulatrophy: routines to interface with the Simul@atrophy tool.
    • disptools.predict: routines to interface with the PREDICT tool.

Install

This package is available on PyPI both as source distribution and as a Windows pre-compiled binary wheel. You can install it with pip:

 python3 -m pip install disptools

Build from source

Requirements

The library is a cross-platform Python 3.5+ package, with a compiled C extension. The Python dependencies are:

Build dependencies are a standard C compiler (tested with gcc 8.2 on Linux, mingw-w64 7.2 and MSVC 19 on Windows 10), CMake, the numpy and the setuptools packages. scikit-build may be required to build the other Python dependencies.

Some optional dependencies are required only for a limited set of features, and the package should build and run without them:

Build options

The following build flags are recognised by setup.py:

  • --opt: enable non-portable optimisations.
  • --debug: disable optimisations, compile with debug symbols.
  • --cuda: enable CUDA support.

Additional flags starting with -D are also accepted and passed directly to CMake.

Windows (Visual Studio) and Linux

Install the dependencies with your favourite package manager. For example, with pip:

python3 -m pip install scikit-build numpy scipy SimpleITK

The package provides a setuptools based install script. To install the library, run from the project root folder

python3 setup.py install

which should build the C extension and install the Python package.

Windows (MinGW)

  1. First, be sure that mingw, CMake and Python are installed and their executables are in your PATH.

  2. Ensure that gcc is working correctly:

> gcc --version
gcc (x86_64-posix-seh-rev1, Built by MinGW-W64 project) 7.2.0
Copyright (C) 2017 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  1. Ensure that distutils correctly recognises your version of Visual Studio (even when using mingw). Open the file C:\Users\yourname\AppData\Local\Programs\Python\Python3x\Lib\distutils\cygwinccompiler.py (the exact location may vary according to your setup) and check that your version of Visual Studio is present in the function get_msvcr(); if not, adjust it according to the following:
def get_msvcr():
    """Include the appropriate MSVC runtime library if Python was built
    with MSVC 7.0 or later.
    """
    msc_pos = sys.version.find('MSC v.')
    if msc_pos != -1:
        msc_ver = sys.version[msc_pos+6:msc_pos+10]
        if msc_ver == '1300':
            # MSVC 7.0
            return ['msvcr70']
        elif msc_ver == '1310':
            # MSVC 7.1
            return ['msvcr71']
        elif msc_ver == '1400':
            # VS2005 / MSVC 8.0
            return ['msvcr80']
        elif msc_ver == '1500':
            # VS2008 / MSVC 9.0
            return ['msvcr90']
        elif msc_ver == '1600':
            # VS2010 / MSVC 10.0
            return ['msvcr100']
        elif msc_ver == '1700':
            # Visual Studio 2012 / Visual C++ 11.0
            return ['msvcr110']
        elif msc_ver == '1800':
            # Visual Studio 2013 / Visual C++ 12.0
            return ['msvcr120']
        elif msc_ver == '1900':
            # Visual Studio 2015 / Visual C++ 14.0
            # "msvcr140.dll no longer exists" http://blogs.msdn.com/b/vcblog/archive/2014/06/03/visual-studio-14-ctp.aspx
            return ['vcruntime140']
        else:
            raise ValueError("Unknown MS Compiler version %s " % msc_ver)
  1. Ensure that the library vcruntime140.dll is present in your library path. Otherwise, download it and place it in C:\Users\yourname\AppData\Local\Programs\Python\Python3x\libs (the exact path may vary according to your setup).

  2. Install the dependencies:

> python -m pip install scikit-build numpy scipy SimpleITK
  1. Clone the sources of this package with git, or download and extract them as a zip archive. Move to the root folder of the sources (C:\Users\yourname\disptools in this example), specify the right compiler, and launch the setup script to build and install the package.
> cd C:\Users\yourname\disptools
> python setup.py setopt --command=build --option=compiler --set-value=mingw32
> python setup.py install

References

  • [1] van Eede, M. C., Scholz, J., Chakravarty, M. M., Henkelman, R. M., and Lerch, J. P. Mapping registration sensitivity in MR mouse brain images. Neuroimage 82 (2013), 226–236.
  • [2] Karaçali, B., and Davatzikos, C. Estimating topology preserving and smooth displacement fields. IEEE Transactions on Medical Imaging 23, 7 (2004), 868–880.
  • [3] Karaçali, B., and Davatzikos, C. Simulation of tissue atrophy using a topology preserving transformation model. IEEE transactions on medical imaging 25, 5 (2006), 649–652.

License

The software is distributed under the MIT license.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
disptools-0.4.0-cp35-cp35m-win32.whl (67.8 kB) Copy SHA256 hash SHA256 Wheel cp35
disptools-0.4.0-cp35-cp35m-win_amd64.whl (168.5 kB) Copy SHA256 hash SHA256 Wheel cp35
disptools-0.4.0-cp36-cp36m-win32.whl (67.8 kB) Copy SHA256 hash SHA256 Wheel cp36
disptools-0.4.0-cp36-cp36m-win_amd64.whl (168.5 kB) Copy SHA256 hash SHA256 Wheel cp36
disptools-0.4.0-cp37-cp37m-win32.whl (67.8 kB) Copy SHA256 hash SHA256 Wheel cp37
disptools-0.4.0-cp37-cp37m-win_amd64.whl (168.5 kB) Copy SHA256 hash SHA256 Wheel cp37
disptools-0.4.0.tar.gz (73.4 kB) Copy SHA256 hash SHA256 Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page