“Building the Jupyter Community in Musculoskeletal Imaging Research”

Authors: Serena Bonaretti, Donnie Cameron, Michael Kuczynski, and Gianluca Iori, on behalf of all the participants to the workshop

On June 9–11, 2022, 25 researchers in Quantitative Musculoskeletal (MSK) Imaging met in Maastricht (The Netherlands) for a three-day Jupyter Community Workshop (JCW). Our aim was twofold: to learn tools for open and reproducible research in medical imaging — such as Python, Jupyter Notebook, and ITK—and to create open source software for MSK imaging research. During the three days, we had parallel sessions of tutorials and coding, we hosted speakers to give us inspiration and teach us how to do open and reproducible research, and we discussed the next steps for our community. You can find our workshop page here. In this blog post, we’ll give a detailed breakdown of what happened at the workshop.

Tutorials

As some of us did not have experience with the Jupyter/Python ecosystem, we organized tutorials at various levels of difficulty:

• Introduction to the Jupyter ecosystem and Python, and to Pandas and Matplotlib (tutorial material here);
• Introduction to the biomedical image analysis and visualization toolkit ITK (tutorial material here).

Computational projects

We split into three groups based on our research interest, and we worked on three computational projects, all of which are now close to publication:

• Ciclope: a Python package that processes micro computed tomography images to generate micro finite element models. The software allows the user to create reproducible and fully open-source pipelines for simulating the mechanical behavior of trabecular bone using the finite element method. The Jupyter notebook examples created within the workshop illustrate a complete pipeline from 3D image preprocessing to finite element model generation, solution, and postprocessing.
• ORMIR_XCT: a Python package for computing joint space parameters and trabecular bone microarchitecture in second generation high resolution peripheral quantitative computed tomography (HR-pQCT, XtremeCT2, Scanco Medical) images. This package contains all necessary scripts to convert Scanco image data (AIM/ISQ) and perform standard bone and joint analyses that are currently performed using the scanner manufacturer’s built-in software. A Jupyter notebook was developed during this workshop to demonstrate this package’s functionality by perform joint space analysis of a sample finger joint.
• MuscleBIDS: a Python package for reading and writing a standardized data format for muscle MR imaging that is based on BIDS. Given the diversity of imaging data formats that exists internationally, such standardized formats are crucial for obtaining reproducible and comparable results. In addition to continuing the development of the tool to support more image contrasts and scanner vendors, a Jupyter notebook was produced, showcasing the current functionality of the package.

Invited speakers

We were excited to have two extraordinary speakers in person at our workshop:

• Lorena Barba, who demonstrated the crucial role of open science in solving global problems (e.g. during the COVID pandemic) and showed current and future infrastructure for conducting open science (find her presentation here);
• Chris Holdgraf, who talked about current efforts in developing and deploying Jupyter in research and education with different tools and adaptations depending on field requirements, e.g. JupyterLab, Jupyter Book (with MyST), and JupyterHub (find his presentation here).

Community discussion (i.e. what we learned and what’s next)

We dedicated the last session of the workshop to a free exchange of ideas, focusing on what we learned from this experience and defining our next steps:

• Workshop organization. The choice of the venue was successful: staying in a modern, tech-style venue with space for work and play got us into the right mindset for coding and learning. On the other hand, we should have organized tutorials and coding sessions sequentially, as the parallel organization did not favor those who were eager to both learn and contribute to the computational projects. In addition, some new members of the community had difficulties finding a way to contribute to the computational projects, which had already been started by some labs;
• Community organization. We understood that we needed to structure our community in a more formal way. Thus, we established a Technical Advisory Board, who are going to create guidelines for software homogenization and standardization, and a Community Advisory Board, whose task is to organize and share information about the community, and structure material for newcomers so that they can easily integrate into the community. Finally, we also decided to formally name our community the ‘Open and Reproducible Musculoskeletal Imaging Research’ (ORMIR) community;
• Next steps. The Community and Technical Advisory Boards are already working to improve the community website, and create learning material and coding guidelines. We applied for a grant to continue developing our community, and we are looking for more funding opportunities to expand and maintain our projects.

In conclusion, the Jupyter Community Workshop was a powerful collective learning experience that inspired us to start, develop, and continue our journey towards open and reproducible research in quantitative musculoskeletal imaging.

This event would not have been possible without the generous support provided by Bloomberg and Amazon Web Services. If your organization would like to support programs like the Jupyter Community Workshops please contact NumFOCUS at [email protected]