PREPRINT: Open and reproducible research in musculoskeletal imaging: Why it matters and how to implement it with the guidelines of the ORMIR community

Open and reproducible research in MSK imaging: Why it matters and how to implement it with the guidelines of the ORMIR community 1 JBMRPlus,2025, 1–9 doi: DOI HERE Advance Access Publication Date: Day Month Year Perspective Paper Open and reproducible research in musculoskeletal imaging: Why it matters and how to implement it with the guidelines of the ORMIR community Serena Bonaretti ,1,∗Mojtaba Barzegari ,2Melissa Bevers ,3,4,5 Steven Boyd ,6,7 Andrew J. Burghardt ,8Donnie Cameron ,9Francesco Chiumento ,10 Gianluigi Crimi11, Gerard Degenhart ,12 Pholpat Durongbhan ,13 Michelle Alejandra Espinosa Hernandez ,14,13 Giulia Fraterrigo 11, Ali Ghasem-Zadeh ,15 Lorenzo Grassi ,16 Jukka Hirvasniemi ,17 Mahdi Hosseinitabatabaei ,18 Gianluca Iori ,19 Joeri Kok ,5Michael Kuczynski ,20,7 YoungJun Lee ,21 Cecilia Liberati 22, Sarah Manske ,6,7 Matt McCormick ,23 Maria Monzon ,24 Martino Pani,25 Simone Poncioni ,26 Jilmen Quintiens 22, Sabine R¨auber ,27 Paul Ritsche ,28 Alfonso Dario Santamaria11, Francesco Santini ,27 Fabio Sarto ,29 Enrico Schileo11, Vincent Stadelmann ,30 Kathryn S. Stok ,13 Rachel Surowiec ,21 Fulvia Taddei11, Jared Vicory,31 Matthias Walle ,7Mariska Wesseling ,32 Danielle Whittier,20,7 Bettina Willie 33, Andy Kin On Wong 34,35 and Dˇzenan Zuki´c 31 1Independent Researcher, Switzerland, 2Department of Chemical Engineering and Chemistry, Eindhoven University of Technology (TU/e), The Netherlands, 3Department of Internal Medicine, VieCuri Medical Center, The Netherlands, 4NUTRIM Institute of Nutrition and Translational Research In Metabolism, Maastricht University, The Netherlands, 5Department of Biomedical Engineering, Eindhoven University of Technology, The Netherlands, 6Department of Radiology, Cumming School of Medicine, University of Calgary, Canada, 7McCaig Institute for Bone and Joint Health, University of Calgary, Alberta, Canada, 8Department of Radiology & Biomedical Imaging, University of California, San Francisco, CA, USA, 9Department of Medical Imaging, Radboud University Medical Center, The Netherlands, 10School of Electronic Engineering, Dublin City University (DCU), Ireland, 11Bioengineering and Computing Laboratory, IRCCS Istituto Ortopedico Rizzoli, Italy, 12Core Faicility MicroCT, University Clinic for Radiology, Medical University Innsbruck, Austria, 13Department of Biomedical Engineering, The University of Melbourne, Victoria, Australia, 14Rehabilitation Sciences Institute, University of Toronto, ON, Canada, 15Departments of Endocrinology and Medicine, Austin Health, The University of Melbourne, Victoria, Australia, 16Department of Biomedical Engineering, Lund University, Sweden, 17Department of Radiology & Nuclear Medicine, Erasmus MC University Medical Center Rotterdam, The Netherlands, 18Department of Bioengineering, Faculty of Engineering, McGill University, Quebec, Canada, 19Center for Photon Science, Paul Scherrer Institut PSI, Switzerland, 20Department of Cell Biology and Anatomy, Cumming School of Medicine, University of Calgary, Alberta, Canada, 21Weldon School of Biomedical Engineering, Purdue University, IN, USA, 22Biomechanics Section, Department of Mechanical Engineering, KU Leuven, Belgium, 23Fideus Labs, NC, USA, 24Department of Health Science and technology, ETH Zurich, Zurich, Switzerland, 25School of Electrical and Mechanical Engineering, University of Portsmouth, Hampshire, United Kingdom, 26ARTORG Center for Biomedical Engineering Research, University of Bern, Switzerland, 27Basel Muscle MRI, Department of Biomedical Engineering, University of Basel, Basel, Switzerland, 28Department of Sport, Exercise and Health, University of Basel, Switzerland, 29Department of Biomedical Sciences, University of Padova, Italy, 30Department of Research and Development, Schulthess Klinik, Switzerland, 31Medical Computing, Kitware, Inc., NC, USA, 32Department of Orthopedics and Sports Medicine, Erasmus Medical Center, The Netherlands, 33Faculty of Dental Medicine and Oral Health Sciences, McGill University and Shriners Hospital for Children-Canada, Quebec, Canada, 34Joint Department of Medical Imaging; Schroeder Arthritis Institute, University Health Network, ON, Canada and 35Dalla Lana School of Public Health, University of Toronto, ON, Canada ∗Corresponding author: serena.bonaretti.resea[email protected] Abstract The Open and Reproducible Musculoskeletal Imaging Research (ORMIR) community is a scientific community dedicated to promoting openness and reproducibility in musculoskeletal imaging research. In this perspective paper, we outline the motivations for conducting transparent research and provide practical guidelines to implement it. We start with defining open and reproducible research and describing the benefits and challenges of working transparently. Next, we redefine the outputs of a computational research study as a combination of data, code, and a publication, recommend a folder and file structure that reflects these three study outcomes, and describe how to maintain and update such structure during the study and at study publication. Finally, we emphasize that working in an open and reproducible manner is a learning process and the best way to acquire the necessary competencies is simply to start. Lay summary: The ORMIR community promotes openness and reproducibility in musculoskeletal imaging research. In this perspective paper, we explain why transparency matters and recommend how to conduct a computational study in an open and reproducible manner focusing on its three outputs: data, code, and publication. Finally, we highlight that the best way to learn these practices is simply to start. Keywords: Open research, reproducible research, data sharing, open-source code, computational narratives, guidelines, license, ORMIR Introduction In 2012, Panagiotopoulou et al. published an article titled “What makes an accurate and reliable subject-specific finite element model? A case study of an elephant femur” (1). In this work, they sought “to construct the first reliable finite element model of the femur of an adult Asian elephant” from computed tomography images. As part of the study, the authors conducted convergence analyses for the size, type, and number of elements used in the geometry discretization of the femur anatomy. In the following years, the paper received 33 citations1. Some of these publications referred to the methodology and findings of this study in their introduction or discussion, while other publications built upon the study methodology and findings. Among the latter, in 2018 Brassey et al. published a study in which they used the same size, type, and number of elements as in Panagiotopoulou’s publication, motivating that “previous research has found such meshes perform well” and “strain values have been found to converge when element numbers exceed 200,000” (2). While building upon previous work is typical in research, in this case there is an unexpected outcome. In 2014—four years before Brassey’s publication—Panagiotopoulou’s paper had been self-retracted because the authors “uncovered problems with the methods [...] and therefore some of the final data” (3). This story highlights three important considerations. First, Panagiotopoulou et al. deserve high praise for their scientific integrity. Second, what occurred to Brassey et al. could happen to any researcher who builds a study on previous findings. Third, we researchers can be inspired to ask fundamental questions about some practical needs that we face in our daily work: How can we and our collaborators create and publish verifiable methodologies and findings that others can build upon? How can we avoid building our studies on incorrect methods and results? And how can we check the correctness of methodologies and results when we review a paper and assess its scientific value? The answer to these questions is open and reproducible research. Open and reproducible research: Definitions, benefits, and challenges Openness refers to “an approach to research based on open cooperative work that emphasizes the sharing of knowledge, results and tools as early and widely as possible” (4). Reproducibility— intended here as computational reproducibility—is the ability of researchers to duplicate the results of a previous study using the “same input data, computational steps, methods, and conditions of analysis” (5). Together, openness and reproducibility constitute away of working whose ultimate goal is to improve the reliability and efficiency of scientific progress (6; 7). The scientific advantages of open and reproducible research are several and include evaluating the correctness of scientific claims (8), building on previous work with confidence and efficiency (9)—by reusing and extending existing computational pipelines—and comparing and validating new methods and algorithms against existing ones, without having to reimplement them (10). These scientific benefits are accompanied by academic advantages. Publishing open data can increase the number of citations by approximately 70%, regardless of journal impact factor, 1Retrieved from Google Scholar: https://scholar.google.com. Accessed: July 21, 2025. author country of origin, and date of publication (11). Similarly, making code available online can increase citations by a factor of three (12). Furthermore, research laboratories can greatly benefit from adopting open and reproducible practices. Such a transparent methodological approach fosters an intrinsic “discipline” that supports the continuity of maintenance and extension of computational tools created in the labs, which would otherwise risk a short lifespan as they are usually developed by successive and distinct generations of PhD students and postdocs (6). Despite the aforementioned advantages, working in an open and reproducible manner has been challenging. Until a few years ago, guidelines were scarce, computational tools were not fully developed, and permanent, large, and free repositories for data and code were unavailable (13). In addition, many scientists were concerned about not receiving proper recognition, having to allocate additional time and effort to prepare data and code for release, and providing advantages to competing datasets or software when releasing seminal work (14). However, the most critical limitation to sharing data and code may be related to the way we conceive the outcome of computational studies. What is the outcome of a computational research study? Traditionally, the main outcome of a computational research study is a scientific paper. As we know, papers are concise reports limited by a defined number of words (typically 3,000–6,000) and a restricted number of figures and tables (usually 3–6). To meet these constraints, only a selection of the methods and results developed and obtained during a study are presented in a paper. Crucial implementation details (e.g., computational initializations and parameters) are often omitted (12), as well as intermediate or individual findings, because results are usually presented as aggregated statistics, such as means and standard deviations. In this context, a scientific paper does not represent “the scholarship itself, it’s merely scholarship advertisement”, as Donoho famously wrote in 2009 (6). In other words, a scientific paper is just the tip of the iceberg representing research efforts (15). The actual work— that is, the data and code—remains hidden underwater (Figure 1, left). In open and reproducible research, data and code emerge to be an integrative part of the outcome of a computational study. Together with publications, they constitute the three essential, interlocked pieces of the puzzle necessary to understand and reproduce the entire study (Figure 1, right). From the raw data of the input, through the code, researchers can recreate the results presented in the publication, and subsequently reuse and extend the workflow for new studies. However, for this to be feasible, data and code should be clearly structured and documented (16). More recently, numerous guidelines have been available on how to publish computational studies in an open and reproducible manner while enabling attribution and facilitating sharing (8; 9; 17; 18; 19). Moreover, researchers in many fields have formed scientific communities to establish consensus-based standards and develop computational tools specific to their disciplines (20; 21; 22). 2 © The Author 2025. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.p[email protected] Open and reproducible research in MSK imaging: Why it matters and how to implement it with the guidelines of the ORMIR community 3 Publications Data and Code Publications Code Data Fig. 1: Representation of research outcomes. Traditionally, publications are the main research outcome. They constitute the shared, visible tip of an iceberg, whereas data and code remain hidden in the lab (left; modified from https://stodden.net/AMP2011/slides/ 2011-amp-reproducible-research.pptx.pdf). Conversely, in open and reproducible research data, code, and publication constitute the three interlocking pieces of the puzzle and are essential to each other to represent the entire study (right). ORMIR guidelines for open and reproducible computational studies in MSK imaging research In musculoskeletal (MSK) imaging, the Open and Reproducible Musculoskeletal Imaging Research (ORMIR) community was founded in 2020, with the aims of creating open, reproducible, well-tested, and well-documented code to analyze MSK images; standardizing data acquisition and management to favor data sharing and algorithm comparison; and promoting a culture of openness and reproducibility in musculoskeletal imaging research for a faster advancement of the field (23). The community is currently composed of approximately 70 international researchers from academia and industry at various career stages, who have been active in creating guidelines, templates, and software to manage and analyze MSK images—see www.ormir.org. In the next two sections of this article, we will present the ORMIR guidelines on how to create an open and reproducible computational study in MSK imaging. These guidelines are based on the principle that open and reproducible research constitutes a methodological approach that should be implemented throughout an entire study, rather than being considered as a timeand resource-intensive task at publication. Therefore, first, we will focus on how to conduct a computational MSK research study for an open and reproducible publication. Then, we will provide suggestions on how to integrate and share the three study outcomes at publication. All recommendations are summarized in Fig.2 and can be adapted to the specific needs of any computational study. Conducting a computational MSK research study for an open and reproducible publication Begin by organizing your research project into a folder structure similar to the one in Fig. 3. You can download the folder structure from Zenodo (www.doi.org/10.5281/zenodo.17206440) or implement it manually. Start with a main folder and name it after the study. Within the main folder, create three additional folders, each for one of the three study outcomes: data,code, and publication. The structure and content of each folder is described in more detail in the following sections. Data Within the data folder, add four folders: raw,derived, and results for data at different stages of the reproducible workflow, and data docs for documentation. In the folder raw, store the original data of the study, that is, the image files—usually in DICOM or proprietary format—and demographic or clinical information of the subjects—commonly in tabular format. These data constitute the input of your workflow. Do not modify them either via computational or manual intervention to ensure the reproducibility of the whole study (8; 24). In the folder derived, save the intermediate data generated during the computational workflow. For the internal structure of this folder, follow the specifications defined by the ORMIR Musculoskeletal Imaging Data Structure (ORMIR-MIDS) (25), which extends the Brain Imaging Data Structure (BIDS) principles (26) specifically for MSK imaging. Reorganize the image files in raw into a hierarchy of nested folders arranged by subject, session, and imaging modality using the ORMIR-MIDS Python package, which also converts the images from the Digital Imaging and Communications in Medicine (DICOM) format (27) to the Neuroimaging Informatics Technology Initiative (NIfTI) format (28) (Fig. 3, right). During the study, store the intermediate outputs of the computational workflow—such as segmentation masks, morphological measurements, and finite element analyses—in this folder structure. Saving these intermediate outputs will facilitate debugging and potential reuse of partial findings (9). 4Serena Bonaretti et al. Conducting a computational MSK research study for an open and reproducible publication Publishing a computational MSK research study as open and reproducible Data • Create 4 folders: raw, derived, results, and data_docs •Never modify the data in raw •Use the ORMIR-MIDS structure for the intermediate data •The data in results will appear in the publication •Document the who/what/when/where/why/how • Use open file formats and meaningful file names • Verify sharing feasibility: •Consider consent and anonymization •Consult the ORMIR Data Sharing Guidelines • Select a repository providing DOI, version control, and embargo if needed • Avoid links to cloud services or personal websites • Choose an institutional or a Creative Commons license Code • Create 4 folders: src, notebooks, batch_scripts, and code_docs •Add the Python modules to src •Use Jupyter Notebooks to analyze tabular data, generate graphs for publications, and create demos– use the ORMIR template •Add the bash scripts to batch_scripts •Document code using Sphynx, Read the Docs or Jupyter Book •Write code for humans: use variables with descriptive names and follow style conventions • Use version control (GitHub) • Change the GitHub repository visibility to public • Use Binder to share notebooks interactively • Host code documentation on GitHub or Read the Docs • Sync the code with a repository providing DOI for citations • Choose an open-source license Publication • Write the manuscript in Latex, Markdown, or Google Docs • Add links to code and data repositories to the paper • Add links to computational workflows to the graphs • Upload to a preprint server • Submit to an open access journal Takeaways • It is a learning process, and you will learn by doing • In case of doubts, ask the ORMIR community • It does not have to be perfect. Just start! • Others can now build on your research • You have contributed to a more transparent and faster advancement of MSK imaging research Fig. 2: Summary of best practices for open and reproducible computational studies in musculoskeletal imaging research. In the folder results, save the final output of the computational workflow, including tabular files reporting morphological, densitometric, mechanical, or statistical calculations across subjects. These are the data that you will present in the Results section of the publication. Finally, in the folder data docs, add the data documentation, such as acquisition protocol, ethical approval, and any metadata— that is, information about the data. As currently ORMIR does not provide specific guidelines for data documentation, simply collect all the reasonable information that specify the “who/what/when/where/why/how” of your dataset (13), so that other researchers can understand and reuse it. At every stage of the computational workflow, use open and widely supported file formats (24). For imaging data, prefer NIfTI or other open standards to scanner-specific proprietary formats. For tabular data, use comma-separated values (.csv) or tabseparated values (.tsv) over proprietary formats, such as Excel. Finally, give files meaningful names that clearly indicate their content (24). Code In the folder code, create the four folders src,notebooks, batch scripts, and code docs—in larger projects add the additional folder tests for automated unit tests (29). In the folder src (abbreviation for source), add the core algorithms of the study. Write code in the open-source programming language Python, which has gained popularity in medical imaging for the availability of general-purpose image processing packages— such as ITK (30) and SimpleITK (31)—as well as deep learning and machine learning packages—including PyTorch (32), TensorFlow (33), and Scikit-learn (34). Within src, organize your code into Python modules, that is, .py files. For larger projects, group the modules into a Python package and change the folder name to the package name, as required by the packaging tools (35). In the folder notebooks, add the study’s Jupyter Notebooks, which are the ideal format for integrating code with narrative, equations, and visualizations to effectively document reproducible workflows (9). Download the Jupyter Notebook template created by the ORMIR community from Zenodo (www.doi.org/10.5281/ zenodo.17206440) and adapt it to your study. The template has a clear structure composed of aims, computational steps, outputs, and dependencies—that is, a print of the hardware characteristics and Python package versions (9). Use notebooks to analyze tabular data, generate figures for the publication, collect the main function calls from the modules in src into a computational workflow with narrative, and create examples or demos for your project. In the folder batch scripts, save the bash files. These files are the scripts that call the algorithms from the folder src and that you use for large-scale computations on servers. In all cases, write code “for humans”, that is, make it readable rather than unnecessarily complex (36). To create clear code, use variables with descriptive names that reflect their role in the computation (36) and adhere to the Python coding convention PEP8 (37). Moreover, modularize code into functions—or classes—that represent units of operations (24; 9). Document each function following the NumPy Style (38) or Google Python Style (39)— which clearly describe functionality, inputs, and outputs—and add inline comments to explain non-obvious steps and clarify the underlying logic. Finally, synchronize the code to a repository like GitHub (https://github.com/). Such repositories provide version control to preserve the history of changes and revert to previous Open and reproducible research in MSK imaging: Why it matters and how to implement it with the guidelines of the ORMIR community 5 study_name code src notebooks preprocessing.py preprocess_images.ipynb segmentation.py segment_images.ipynb data raw derived 902145989_001.dcm sub-01 798346781_001.dcm sub-02 batch_scripts run_segmentation.sh run_preprocessing.sh results morphology.csv statistics.csv publication code_docs sphynx/readthedocs/jupyterbook paper.tex ses-01 mr-anat ct sub-01_ses-01_T1w.json sub-01_ses-01_ct.json sub-01_ses-01_T1w.nii.gz sub-01_ses-01_ct.nii.gz sub-01_ses-01_seg.nii.gz sub-01_ses-01_seg.nii.gz sub-01_ses-01_morph.tsv ses-02 ORMIR-MIDS data_docs acquisition.pdf Fig. 3: Folder and file structure of a computational research study for an open and reproducible publication. code in case of errors, and they are convenient for coordinating programming tasks in collaborative software development (17). Finally, in the folder code docs (or simply docs) create code documentation for developers and users, especially if you are building a large project (40). In the documentation for developers, carefully explain the technical details of the code. Create this documentation using Sphynx (41) or Read the Docs (42) as they are compatible with the NumPy and Google Python styles. In the documentation for users, explain in nontechnical language how to install and execute your code and provide examples or demos. Write this documentation using Jupyter Book (43). Publication In the last folder publication, store all the material relative to the scientific paper. Write the manuscript in a format that permits version control, such as LaTeX or Markdown, and opt for online tools that allow tracking changes and collaborative editing, such as Overleaf (www.overleaf.com/) or Google Docs (https://docs. google.com) (24). Publishing a computational MSK research study as open and reproducible At publication, the primary task is to decide how and where to share the three study outcomes. Here are the details for each of them. Data Sharing study data requires attention to two main aspects: assessing the feasibility of sharing and choosing a data repository (44; 45). The feasibility of sharing depends on legal and ethical factors, including whether the study participants signed an appropriate consent form, the level of data anonymization or deidentification required, and the privacy legislation of your institution and country. These factors can preclude sharing some or all the content of the folders raw,derived, and data docs. However, you can always share the final aggregated data of your study—that is, the content of the folder results. For further details, consult the ORMIR Data Sharing Guidelines (46). To date, only a few dedicated MSK repositories exist where you can share your data, such as the Universal Musculoskeletal Ultrasound Repository (UMUD) (47). If your data does not fit the score of such repositories, share the data in your institution 6Serena Bonaretti et al. repository, following their regulation. Otherwise, use general purpose repositories such as Zenodo (https://zenodo.org)—where ORMIR is present as a community label—or Figshare (https: //figshare.com). Both these repositories provide a digital object identifier (DOI) to ensure permanent availability and enable citation, versioning for data, and an embargo period if you cannot immediately share your data. Avoid links to personal websites or cloud storage, because these links tend to expire or be deleted (48; 49). Code To make your code open, simply change the GitHub repository visibility from private to public. Complete the README.md file of the repository using the template provided by the ORMIR community (www.doi.org/10.5281/zenodo.17206440). Whenever possible, share the notebooks in an interactive and reproducible cloud environment, such as Binder (https://mybinder.org) (50). Host the code documentation created with Sphynx or Jupyter Book on GitHub Pages (51; 52) or the hosting service provided by Read the Docs. Finally, synchronize the GitHub repository with Zenodo (53), Fighshare (54), or Software Heritage (55) to obtain a DOI that enables reproducibility and citations. Publication Add the links to data and code repositories to the manuscript. If possible, add the specific links to the data and code used to create the graphs in their caption to enable readers to “recalculate the figure from all its data, parameters, and programs” (56). Upload the paper to a preprint archive, such as arXiv (www.arxiv.org), bioRxiv (www.biorxiv.org), or medRxiv (www.medrxiv.org), and then submit it to a journal that supports open access. In the not too distant future, we may be able to write fully executable and interactive papers in which readers can rerun the described experiments (13). This vision is supported by existing pilot projects, where Jupyter Notebooks are submitted together with conference abstracts (57; 58). Licenses Finally, add a license to your data, code, and publication to specify how you want them to be reused and modified. Without a license, you would retain all the rights and other researchers would not be allowed to use your material (24). If available, use the licenses recommended by your institution. Otherwise, here are the ORMIR guidelines. For data and publication, use the Creative Commons (CC) license, whose features—attribution (BY), Non Commercial use (NC), Share-Alike (SA), and No Derivative (ND)—can be combined to create a customized license. Creative Commons also provides the CC0 license, which permits releasing material to the public domain without retaining any copyright (https:// creativecommons.org/share-your-work/cclicenses/). For code, two main types of open-source licenses exist: restrictive—which extend openness to derived work (e.g., GNU licenses)—and permissive, which do not require openness for derived work (e.g., MIT, BSD, or Apache) For a complete list of licenses with their permissions and limitations, see https: //choosealicense.com/appendix. Choose a license that is compatible with the licenses of the computational tools you used in your study and that you consider the most appropriate for the future of your project (18). Just start! Open and reproducible research is a way of working that supports our needs as researchers to assess scientific claims, build on previous reliable work, and contribute to faster advancement of our research field. In this perspective paper, we proposed concrete guidelines created by the ORMIR community on how to work in an open and reproducible manner. Adopting this methodological approach is a learning process that is most effective when learned by doing. If you have doubts about some guidelines or tools, contact the ORMIR community. Do not aspire to perfection from the very beginning. Just start! Your efforts will prove valuable because other researchers will be able to deeply understand and build on your research, and you will have contributed to a more transparent and faster advancement of MSK imaging research. Competing interests The authors have no competing interests to declared. Author contributions statement Authorship was offered to any individual who has contributed to the development of the ORMIR community and its guidelines. Authors were asked to denote their contributions to the project according to the Contributor Roles Taxonomy (CRediT). Role abbreviations. C: Conceptualization, DC: Data curation for initial use and later re-use, FA: Funding acquisition, M: Methodology, PA: Project administration, R: Resources, S: Software, SU: Supervision, V: Validation, W: Writing—original draft, and RE: Writing—review & editing. Serena Bonaretti: C, FA, M, PA, S, SU, W . Mojtaba Barzegari: C,S. Melissa Bevers: RE. Steven Boyd: S,RE. Andrew J. Burghardt: RE. Donnie Cameron: S, RE. Francesco Chiumento: RE. Gianluigi Crimi: S, RE. Gerard Degenhart: RE. Pholpat Durongbhan: RE. Michelle Alejandra Espinosa Hernandez: RE. Giulia Fraterrigo: S, V, RE. Ali Ghasem-Zadeh: RE. Lorenzo Grassi: RE. Jukka Hirvasniemi: S, RE. Mahdi Hosseinitabatabaei: RE. Gianluca Iori: DC, M, S, SU. Joeri Kok: RE. Michael Kuczynski: S, RE. YoungJun Lee: RE. Cecilia Liberati: RE. Sarah Manske: C, RE. Matt McCormick: C, S, RE. Maria Monzon: S, RE. Martino Pani: RE. Simone Poncioni: S, RE. Jilmen Quintiens: RE. Sabine R¨auber: RE. Paul Ritsche: RE. Alfonso Dario Santamaria: S, V, RE. Francesco Santini: C, FA, PA, S, RE. Fabio Sarto: RE. Enrico Schileo: C, SU, RE. Vincent Stadelmann: FA, C, RE. Kathryn S. Stok: C, SU, RE. Rachel Surowiec: RE. Fulvia Taddei: C, RE. Jared Vicory: S, RE. Matthias Walle: RE. Mariska Wesseling: S. Danielle Whittier: RE, SU. Bettina Willie: RE. Andy Kin On Wong: SU, RE. Dˇzenan Zuki´c: S, RE. Acknowledgments We thank all the individuals who have contributed to and supported the open-source ecosystem, including developers, maintainers, and advocates, whose efforts have made open and reproducible research possible. The development of the ORMIR community and its guidelines has been supported by a NumFOCUS-funded Jupyter Community Workshop and a Swiss National Science Foundation grant (n. 221844). Finally, we acknowledge the use of ChatGPT for editorial support, including grammar checking and flow enhancement. Open and reproducible research in MSK imaging: Why it matters and how to implement it with the guidelines of the ORMIR community 7 References 1. O Panagiotopoulou, SD Wilshin, EJ Rayfield, SJ Shefelbine, and JR Hutchinson. What makes an accurate and reliable subject-specific finite element model? a case study of an elephant femur. Journal of the Royal Society Interface, 9(67):351–361, 2012. doi:10.1098/rsif.2011.0323. 2. CA Brassey, JD Gardiner, and AC Kitchener. Testing hypotheses for the function of the carnivoran baculum using finite-element analysis. Proceedings of the Royal Society B: Biological Sciences, 285(1887):20181473, 2018. doi:10.1098/ rspb.2018.1473. 3. O Panagiotopoulou, SD Wilshin, EJ Rayfield, SJ Shefelbine, and JR Hutchinson. What makes an accurate and reliable subject-specific finite element model? a case study of an elephant femur. Journal of the Royal Society Interface, 11(20140854), 2014. doi:10.1098/rsif.2014.0854. 4. European Research Executive Agency. Open science. Accessed: 27 August 2025. URL: https://rea.ec.europa.eu/ open-science_en. 5. National Academies of Sciences, Medicine, Policy, Global Affairs, Board on Research Data, Division on Engineering, Physical Sciences, Committee on Applied, Theoretical Statistics, Board on Mathematical Sciences, et al. Reproducibility and replicability in science. National Academies Press, 2019. doi:10.17226/25303. 6. DL Donoho, A Maleki, IU Rahman, M Shahram, and V Stodden. Reproducible research in computational harmonic analysis. Computing in Science & Engineering, 11(1):8–18, 2008. doi:10.1109/MCSE.2009.15. 7. MR Munaf`o, BA Nosek, DVM Bishop, KS Button, CD Chambers, N Percie du Sert, U Simonsohn, E-Jan Wagenmakers, JJ Ware, and JPA Ioannidis. A manifesto for reproducible science. Nature human behaviour, 1(1):0021, 2017. doi:http: //doi.org/10.1038/s41562-016-0021. 8. GK Sandve, A Nekrutenko, J Taylor, and E Hovig. Ten simple rules for reproducible computational research. PLoS computational biology, 9(10):e1003285, 2013. doi:http://doi.org/ 10.1371/journal.pcbi.1003285. 9. A Rule, A Birmingham, C Zuniga, I Altintas, S-C Huang, R Knight, N Moshiri, MH Nguyen, SB Rosenthal, F P´erez, and PW Rose. Ten simple rules for writing and sharing computational analyses in jupyter notebooks, 2019. doi: 10.1371/journal.pcbi.1007007. 10. J Freire and F Chirigati. Provenance and the different flavors of computational reproducibility. IEEE Data Engineering Bulletin, 41(1):15, 2018. URL: http://sites.computer.org/ debull/A18mar/p15.pdf. 11. HA Piwowar, RS Day, and DB Fridsma. Sharing detailed research data is associated with increased citation rate. PloS one, 2(3):e308, 2007. doi:10.1371/journal.pone.0000308. 12. P Vandewalle. Code sharing is associated with research impact in image processing. Computing in Science & Engineering, 14(4):42–47, 2012. doi:10.1109/MCSE.2012.63. 13. Y Gil, CH David, I Demir, BT Essawy, RW Fulweiler, JL Goodall, L Karlstrom, H Lee, HJ Mills, J-H Oh, and SA Pierce. Toward the geoscience paper of the future: Best practices for documenting and sharing research from data to software to provenance. Earth and Space Science, 3(10):388–415, 2016. doi:http://dx.doi.org/10.1002/2015EA000136. 14. V Stodden. The scientific method in practice: Reproducibility in the computational sciences. MIT Sloan research paper, (4773-10), 2010. doi:10.2139/ssrn.1550193. 15. J Freire. A provenance-based infrastructure for creating reproducible papers. In Reproducible Research: Tools and Strategies for Scientific Computing. A workshop in association with Applied Mathematics Perspectives 2011, University of British Columbia, Canada. July 13–16, 2011. URL: https: //stodden.net/AMP2011/index.html. 16. X Chen, S Dallmeier-Tiessen, R Dasler, S Feger, P Fokianos, JB Gonzalez, H Hirvonsalo, D Kousidis, A Lavasa, S Mele, and DR Rodriguez. Open is not enough. Nature Physics, 15(2):113–119, 2019. doi:10.1038/s41567-018-0342-2. 17. Y Perez-Riverol, L Gatto, R Wang, T Sachsenberg, J Uszkoreit, Fd V Leprevost, C Fufezan, T Ternent, SJ Eglen, DS Katz, and TJ Pollard. Ten simple rules for taking advantage of git and github. PLoS computational biology, 12(7):e1004947, 2016. doi:10.1371/journal.pcbi.1004947. 18. A Morin, J Urban, and P Sliz. A quick guide to software licensing for the scientist-programmer. PLOS Computational Biology, 8(7):1–7, 07 2012. doi:10.1371/journal.pcbi. 1002598. 19. A Prli´c and JB Procter. Ten simple rules for the open development of scientific software. PLoS computational biology, 8(12):e1002802, 2012. doi:10.1371/journal.pcbi.1002802. 20. RA Poldrack, CJ Markiewicz, S Appelhoff, YK Ashar, T Auer, S Baillet, S Bansal, L Beltrachini, C G Benar, G Bertazzoli, et al. The past, present, and future of the brain imaging data structure (bids). Imaging Neuroscience, 2:1–19, 2024. doi:10.1162/imag_a_00103. 21. RC Gentleman, VJ Carey, DM Bates, B Bolstad, M Dettling, S Dudoit, B Ellis, L Gautier, Y Ge, J Gentry, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome biology, 5(10):R80, 2004. doi:10.1186/gb-2004-5-10-r80. 22. TE Odaka, A Banihirwe, G Eynard-Bontemps, A Ponte, G Maze, K Paul, J Baker, and R Abernathey. The Pangeo ecosystem: interactive computing tools for the geosciences: benchmarking on HPC. In Annual Workshop on HPC User Support Tools, pages 190–204. Springer, 2019. doi:10.1007/ 978-3-030-44728-1_12. 23. S Bonaretti, L Barzaghi, MM Barzegari, AJ Burghardt, D Cameron, J Carballido-Gamio, G Crimi, P Durongbhan, M Espinosa Hernandez, G Fraterrigo, L Grassi, H Greer, G Iori, M Kuczynski, S Manske, M McCormick, N Neeteson, M Niethammer, M Pani, J Quintiens, MM Sadeghi, F Santini, E Schileo, KS Stok, F Taddei, JJ Tse, J Vicory, M Wesseling, AKO Wong, and D Zuki´c. Introducing the open and reproducible musculoskeletal imaging research (ORMIR) community, July 2023. doi:10.5281/zenodo.8119243. 24. G Wilson, J Bryan, K Cranston, J Kitzes, L Nederbragt, and TK Teal. Good enough practices in scientific computing. PLoS computational biology, 13(6):e1005510, 2017. doi:10.1371/ journal.pcbi.1005510. 25. S Bonaretti, MA Espinosa Hernandez, F Chiumento, Y Founas, M Froeling, J Hirvasniemi, G Iori, Y Lee, S Matuschik, M Monzon, F Santini, and D Cameron. ORMIR-MIDS: An open standard for curating and sharing musculoskeletal imaging data. In 24th International Workshop on Quantitative Musculoskeletal Imaging (QMSKI), The Barossa Valley, South Australia, November 2024. doi:10.1093/jbmrpl/ziaf035. 8Serena Bonaretti et al. 26. Krzysztof J Gorgolewski, Tibor Auer, Vince D Calhoun, R Cameron Craddock, Samir Das, Eugene P Duff, Guillaume Flandin, Satrajit S Ghosh, Tristan Glatard, Yaroslav O Halchenko, et al. The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Scientific data, 3(1):1–9, 2016. doi:10.1038/sdata. 2016.44. 27. WD Bidgood Jr and SC Horii. Introduction to the ACRNEMA DICOM standard. Radiographics, 12(2):345–355, 1992. doi:10.1148/radiographics.12.2.1561424. 28. RW Cox, J Ashburner, H Breman, K Fissell, C Haselgrove, CJ Holmes, JL Lancaster, DE Rex, SM Smith, JB Woodward, et al. A (sort of) new image data format standard: NiFTI-1. In 10th annual meeting of the organization for human brain mapping, volume 22, page 01, 2004. 29. KJ Millman and F P´erez. Developing open-source scientific practice. In Implementing reproducible research, pages 149– 183. Chapman and Hall/CRC, 2018. URL: https://doi.org/ 10.1201/9781315373461. 30. M McCormick, X Liu, J Jomier, C Marion, and L Ibanez. Itk: enabling reproducible research and open science. Frontiers in neuroinformatics, 8:13, 2014. doi:10.3389/fninf.2014. 00013. 31. BC Lowekamp, DT Chen, L Ib´a˜nez, and D Blezek. The design of simpleitk. Frontiers in neuroinformatics, 7:45, 2013. doi: 10.3389/fninf.2013.00045. 32. A Paszke, S Gross, F Massa, A Lerer, J Bradbury, G Chanan, T Killeen, Z Lin, N Gimelshein, L Antiga, and A Desmaison. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019. URL: https://proceedings.neurips.cc/paper/2019/ hash/bdbca288fee7f92f2bfa9f7012727740-Abstract.html. 33. M Abadi, A Agarwal, P Barham, E Brevdo, Z Chen, C Citro, GS Corrado, A Davis, J Dean, M Devin, et al. Tensorflow: Large-scale machine learning on heterogeneous systems, 2015. URL: https://www.usenix.org/system/ files/conference/osdi16/osdi16-abadi.pdf. 34. F Pedregosa, G Varoquaux, A Gramfort, V Michel, B Thirion, O Grisel, M Blondel, P Prettenhofer, R Weiss, V Dubourg, J Vanderplas, A Passos, D Cournapeau, M Brucher, M Perrot, and E Duchesnay. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:2825–2830, 2011. 35. Python Packaging Authority. Python packaging user guide. Accessed: 28 August 2025. URL: https://packaging.python. org/en/latest/. 36. DE Knuth. Literate programming. The computer journal, 27(2):97–111, 1984. doi:10.1093/comjnl/27.2.97. 37. G van Rossum, B Warsaw, and A Coghlan. Pep8 - Style guide for Python code. Accessed: 28 August 2025. URL: https://peps.python.org/pep-0008/. 38. Numpy style guide. Accessed: 28 August 2025. URL: https: //numpydoc.readthedocs.io/en/latest/format.html. 39. Google Python style guide. Accessed: 28 August 2025. URL: https://google.github.io/styleguide/pyguide.html. 40. FdV Leprevost, VC Barbosa, EL Francisco, Y Perez-Riverol, and PC Carvalho. On best practices in the development of bioinformatics software. Frontiers in genetics, 5:199, 2014. doi:10.3389/fgene.2014.00199. 41. A Turner, B Tran, C Sewell, F Freitag, JL Andersen, JF B., S Finucane, T Shimizukawa, and T Komiya. Sphynx. Accessed: 28 August 2025. URL: https://www.sphinx-doc.org/en/ master/index.html. 42. Read the Docs. Accessed: 28 August 2025. URL: https: //about.readthedocs.com/. 43. Executable Books Community. Jupyter Book. Accessed: 28 August 2025. URL: https://jupyterbook.org/en/stable/ intro.html. 44. J Scheibner, M Ienca, S Kechagia, JR Troncoso-Pastoriza, JL Raisaro, J-P Hubaux, J Fellay, and E Vayena. Data protection and ethics requirements for multisite research with health data: a comparative examination of legislative governance frameworks and the role of data protection technologies  . Journal of Law and the Biosciences, 7(1):lsaa010, 05 2020. doi:10.1093/jlb/lsaa010. 45. JC Batlle, K Dreyer, B Allen, T Cook, CJ Roth, AB Kitts, R Geis, CC Wu, MP Lungren, J Patti, et al. Data sharing of imaging in an evolving health care world: report of the ACR data sharing workgroup, part 1: data ethics of privacy, consent, and anonymization. Journal of the American College of Radiology, 18(12):1646–1654, 2021. doi: 10.1016/j.jacr.2021.07.014. 46. A Burghardt, AKO Wong, F Taddei, K Stok, S Manske, S Bonaretti, and V Stadelmann. ORMIR data sharing guidelines - v1.1 (v1.1). Zenodo, 2025. doi:10.5281/zenodo. 17206129. 47. P Ritsche, F Sarto, F Santini, C Leitner, M Franchi, O Faude, and N Cronin. UMUD: A web application for easy access to musculoskeletal ultrasonography datasets, January 2025. Preprint. doi:10.31219/osf.io/syr4z. 48. RP Dellavalle, EJ Hester, LF Heilig, AL Drake, JW Kuntzman, M Graber, and LM Schilling. Going, going, gone: Lost internet references. Science, 302(5646):787–788, 2003. doi:10.1126/science.1088234. 49. M Klein, H Van de Sompel, R Sanderson, H Shankar, L Balakireva, K Zhou, and R Tobin. Scholarly context not found: one in five articles suffers from reference rot. PloS one, 9(12):e115253, 2014. doi:10.1371/journal.pone.0115253. 50. Jupyter Project, M Bussonnier, J Forde, J Freeman, B Granger, T Head, C Holdgraf, K Kelley, G Nalvarte, A Osheroff, M Pacer, Y Panda, F P´erez, B Ragan-Kelley, and C Willing. Binder 2.0—reproducible, interactive, sharable environments for science at scale. Scipy, 2018. URL: https://proceedings. scipy.org/articles/Majora-4af1f417-011.pdf. 51. GitHub Pages documentation. Accessed: 28 August 2025. URL: https://docs.github.com/en/pages. 52. Jupyter Book - Publish your book online. Accessed: 28 August 2025. URL: https://jupyterbook.org/en/stable/ start/publish.html. 53. GitHub Docs - Referencing and citing content. Accessed: 28 August 2025. URL: https://docs.github. com/en/repositories/archiving-a-github-repository/ referencing-and-citing-content. 54. How to connect Figshare with your GitHub account. Accessed: 28 August 2025. URL: https://info.figshare.com/user-guide/ how-to-connect-figshare-with-your-github-account/. 55. Welcome to software heritage documentation - get a permanent identifier for your source code. Accessed: 28 August 2025. URL: https://docs.softwareheritage.org/ Open and reproducible research in MSK imaging: Why it matters and how to implement it with the guidelines of the ORMIR community 9 #landing-basics. 56. JF Claerbout and M Karrenbach. Electronic documents give reproducible research a new meaning. In SEG technical program expanded abstracts 1992, pages 601–604. Society of Exploration Geophysicists, 1992. doi:10.1190/1.1822162. 57. G Caprarelli, B Sedora, M Ricci, S Stall, and M Giampoala. Notebooks now! the future of reproducible research. Earth and Space Science, 10(12):e2023EA003458, 2023. doi:10.1029/ 2023EA003458. 58. C Erdmann, S Stall, C Gentemann, C Holdgraf, FPA Fernandes, Gehlen KP-von, and M Corvellec. Guidance for agu authors - jupyter notebooks, May 2021. doi:10.5281/zenodo. 5651648.