elastix 5.3.0 released!

[Niels_Dekker]

I’m very pleased to announce that we have just released version 5.3.0 of our image registration toolbox elastix, at https://github.com/SuperElastix/elastix/releases

Binaries of elastix executables are provided for Windows, MacOS (ARM) and Ubuntu. But of course, you can also build them yourself, using ITK >= 5.4.1.

A new ITKElastix Python package based on elastix 5.3.0 is intended to be released next week, hopefully.

elastix 5.3.0 includes the following improvements:

• Added a new metric component: IMPACT (Image Metric with Pretrained model-Agnostic Comparison for Transmodality registration), by Valentin Boussot et al.
  • Reference: “IMPACT : A Generic Semantic Loss for Multimodal Image Registration”, Valentin Boussot et al, arXiv:2503.24121
  • In order to use this metric, download LibTorch and enable CMake option USE_ImpactMetric. More information: GitHub - vboussot/ImpactLoss: IMPACT Reg: A Generic Semantic Loss for Multimodal Image Registration
• Supported the TOML format for registration parameter files and transform parameter files
  • As alternative to the “legacy” elastix specific text file format
• Supported fixed width notation for the elastix parameter ResultImagePixelType
  • Supported parameter values: “int8”, “int16”, “int32”, “int64”, “uint8”, “uint16”, “uint32”, “uint64”, “float32”, “float64”
• Supported registering a 3D slice on a 3D volume
  • The default pyramid (“FixedSmoothingImagePyramid”, “MovingSmoothingImagePyramid”) now accepts 3D images that have a thickness of 1 pixel
  • Inspired by forum.image.sc topic Elastix questions related to 2D => 3D registration, started by Christian Tischer
• Improved transformix command-line option “-def” for VTK files
  • When the input file is a binary VTK file, the output file is now also a binary VTK file (rather than an ASCII file)
  • When the input file is a VTK file of 32-bit floating points, the output file is now also a VTK file of 32-bit floating points (rather than 64-bit floating points)
  • Inspired by pull request ENH: Support for binary point file (.bin) reading and writing by ChristophKirst · Pull Request #943 · SuperElastix/elastix · GitHub
• Library enhancements:
  • Added SetFixedPoints and SetMovingPoints member functions to ElastixRegistrationMethod
    • Supported specifying the input points for “CorrespondingPointsEuclideanDistanceMetric”
  • Made random number generation used by samplers deterministic, even when running registrations multithreaded
    • Aims to address Deterministic multithreading usage of itkMersenneTwisterRandomVariate… by michalmeszaroswork · Pull Request #5287 · InsightSoftwareConsortium/ITK · GitHub

For all the details, please check https://github.com/SuperElastix/elastix/releases