imagedata-registration 0.2.4

NPreg

NPreg by Erlend Hodneland is implemented in Python, and available as a self-supported PyPi package. There are three implementations of NPreg:

• Pure Python/NumPy code. Source code will run on any Python platform.

• Cython code. Binary code compiled for supported platforms.

• CuPy/CUDA code. Source code which will run on platforms with a working CuPy and CUDA Toolkit.

FSL

FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSL) has several methods for image registration. Using FSL image registration from Python requires the FSL interface from nipype, as well as a the FSL executables. Each FSL method may have different requirements. The common factor is that FSL methods will read and write NIfTI image files.

A function register_fsl is provided here. This function will register a moving Series to a fixed Series. The default registration method is fsl.MCFLIRT. The function will accept other registration methods.

ITK Elastix

The popular Elastix GUI is based on the C++ ITK image registration routines. Like the FSL methods, there are numerous Elastix methods available, all with different requirements. The SimpleElastix (https://simpleelastix.readthedocs.io/index.html) Python library is one particular interface to the Elastix/ITK routines.

NPreg on CUDA GPU

imagedata-registration will benefit from a CUDA GPU. If this is available, install CuPy (https://docs.cupy.dev).

• First, install the CUDA Toolkit: see https://developer.nvidia.com/cuda-toolkit.

• There are different options for installing CuPy. See: https://docs.cupy.dev/en/stable/install.html

FSL

The imagedata-registration FSL module is a wrapper around the official FSL tools. A native FSL installation is required on the host computer.

SimpleElastix

SimpleElastix must be installed separately:

pip install SimpleITK-SimpleElastix