Open Source Image Registration: The elastix Toolbox
Marius Staring (Leiden University Medical Center)
To improve accessibility, interoperability, efficiency, and sustainability of the biomedical image registration software elastix, by making it a library-first package, allowing integration with other software and improving its performance.
Image registration aims to find the spatial relationship between two or more images, and is abundantly used in biomedical image processing. elastix is an open source tool for performing image registration, and is among the most popular image registration toolboxes. But efforts are needed to sustain the software, better connect to users and the other software they use, and improve runtime performance. In a currently running project, elastix has been made available in Python and other scripting languages, and has improved performance by targeting the biology community. The Python interface is already a marked success, averaging 9,000 downloads monthly, emphasizing the need for seamless integration in user workflows. To facilitate the interoperability of elastix with other software, the library interface has been improved, with support for standard file formats (HDF5, TFM) for spatial transformations underway. This software development project will build on previous successes and prioritize better connection to elastix users, their software and deep learning ecosystems, and their urgent needs in terms of performance and functionality. These efforts will be underpinned by a substantial foundational improvement making elastix a library-first package.