This year marks the fourth edition of the workshop since 2021 with a focus on “Staying together: Long-term support and development of open MRI projects”. We want to discuss how to make open MRI efforts sustainable and long-lasting, overcome challenges with respect to funding, community building and democratizing contributions, as well as passing the torch between developer generations, and managing the overall challenges of mostly volunteer work. Of course, we will also highlight the latest open source tools for acquisition, reconstruction and analysis developed by the MR community.
This session is about how we can share knowledge about MRI scanners openly, including how to build an MRI scanner. We also want to take a look at how a network of open science collaborators and specialists can be successfully created.
In this talk I will discuss applications of AI techniques to preclinical and low-field MRI, and discuss open-source software packages and publishing methods. I will also review my own experiences in building open-source low field scanners.
The Scan With Me (SWiM) program is a train-the-trainer initiative developed by the Consortium for Advancement of MRI Education and Research in Africa (CAMERA) to upskill MRI personnel in low- and middle-income countries (LMICs). By combining free-access educational resources, expert lectures, hands-on training, and team-based learning, SWiM equips radiographers with the skills to perform high-quality MRI exams and train others in their facilities.
CONNExIN (COmprehensive Neuroimaging aNalysis Experience In resource constraiNed settings) is a hybrid initiative led by the Montreal Neurological Institute, McGill University, and the Consortium for Advancement of MRI Education and Research in Africa (CAMERA). CONNExIN (event.fourwaves.com/connexin) uses RAD-AID’s Teach-Try-Use strategy to advance MRI and PET neuroimaging skills in low- and middle-income countries through seminars and hands-on dementia data analysis.
Much of the work toward open and reproducible science has been spearheaded by researchers working in the brain space. Newer MRI applications, such as body and hyperpolarized MRI are less well represented in the open MRI community. In this session, we will discuss how lessons learned from open and reproducible brain MRI can be adopted in other areas.
In this talk I will share my personal experience in developing open-source software tools in the areas of hyperpolarized carbon-13 MRI and for lung MRI. These are emerging areas being developed by relatively small communities of researchers, which is a unique opportunity to use open science practices to better leverage limited resources.
Hyperpolarized Xenon MRI is a rapidly growing technique for assessing pulmonary gas exchange. This talk will describe efforts by the Xenon MRI community to establish standards for raw data sharing and analysis in order to directly compare imaging results across sites and to conduct multi-site clinical trials.
In recent years, the clinical interest in renal MRI has accelerated research and development and led to a growing excitement within an emerging multidisciplinary community. The RENALMRI.org community has developed multiple open access resources, that are made available through the renalmri.org website and are expected to be expanded over the next years. Among international renal MRI projects, RESPECT is devoted to the harmonisation and technical validation of a standardised multiparametric renal MRI protocol for personalised CKD management.
The first part of the talk will give a brief introduction to CMRsim as well as the specific requirements of cardiac MR simulation that motivated the development. The remainder will focus on the lessons learned along the way pertaining to open-source development as well as the current and future obstacles for open-source MR simulation.
How can I find funds to support my open-source project? To answer this and other related questions, in this session, we will discuss the funding landscape for scientific software development, which is crucial for software maintenance, and invite developers to share their experience with commercial software production. Moreover, we will cover topics on data sharing initiatives.
An interdisciplinary education system is being to train computational psychiatry experts who can apply quantitative methods to clinical problems. We envision an integrated computational psychiatry system that offers AI-aided, evidence-based assessment and intervention tools through digital health platforms, connecting researchers, clinicians, and patients to enhance online mental health services. The combination of big data from wearable devices and advanced AI technologies will advance precision psychiatry, enabling early risk screening and the development of targeted individualized treatments.
Open data plays a crucial role in enhancing research transparency, improving research reproducibility, and advancing scientific progress. This report uses the case study of Science Data Bank, a public data repository, to illustrate how data repositories serve as a bridge connecting scientists with data policies, data standards, and data dissemination in promoting open data.
A development journey of the Neurodesk platform and software project. The talk what the project is about, how it came to be, funding roadmap, challenges, breakthroughs and solutions along the way and possible future plans.
Open source tool development is crucial for reproducible science, but may not always align with existing funding mechanisms and vendor incentives. I’ll discuss our experience obtaining funding to develop and disseminate a vendor-agnostic functional MRI protocol based on Pulseq, with an emphasis on building connections with both the MRI research community and the vendors.
This session will feature distinguished speakers who have successfully transitioned between academia and industry. They will share their insights on career decisions and highlight how their expertise in MRI research has been pivotal in their professional journeys.
This talk will cover considerations that arise when transitioning from developing ML models for narrow research applications to models that scale and perform well across geographies and populations. The talk will also cover why industry researchers should engage with academics to ensure reproducibility and build trust in new technologies.
While the transition to “industry” may appear daunting, the startup world can offer a sense of familiarity to those wanting to branch away from academia. In this talk, I will share my journey and some pleasant surprises along the way.
Han will share his journey of navigating through the field of physics and AI, and his personal experience in work and life in academia and industrial positions as an international scholar. Fun fact: Han started his first industrial job during his 21-day stay in a quarantine hotel in Beijing in 2021.
In this talk I’ll share my academia to industry adventure, the realities of start-up and scale-up life, and reflect on which skills perspectives from academia I could leverage and which I had to transform or adapt.
In the realm of MRI, particularly in neuroimaging, the reproducibility crisis has emerged as a pivotal challenge, casting significant doubts on the reliability of many scientific findings. This session aims to delve deep into the critical issues surrounding replication failures in brain imaging studies and explore strategic pathways towards enhancing the robustness and validity of research outcomes.
Though the immense potentials and unprecedented increases in the AI-empowered neuroimaging-based studies, particularly in diagnostics, global population diversity in these computational models are highly underrepresented. Embracing AI techniques in the neuroimaging studies should be greatly welcomed, but as a result, sampling inequalities from AI model training and testing biases from AI model deployments substantially limited their real-world practices and ensuing benefits.
Understanding what range of effect sizes are “typical” for fMRI studies is essential for better study planning and interpretation of results, especially given recent concerns about reproducibility and statistical power in fMRI research. Here, we leverage large, publicly available datasets to benchmark expected effect sizes for “typical” studies and introduce a web app for exploring these effects. We emphasize the need to plan for small effect sizes and discuss paths forward, including our next steps developing an empirical power calculator web app.
Visual and semantic processing has been studied using both hypothesis-driven and data-driven approaches. However, the generalizability of research findings always depends on the selection of stimulus materials, and with inconsistent use of materials, it has been challenging to directly relate research findings across studies, research methods (e.g., behaviour, fMRI, electrophysiology), and domains (e.g., humans, animal models, artificial intelligence). To overcome these challenges, we have introduced the THINGS initiative, which is a global initiative of researchers to collect and share datasets all based on the THINGS database of object concepts and images. In this talk, I will highlight the opportunities afforded by this initiative, introduce available datasets, and highlight research findings that have underscored the value of this approach.
In this session, we will showcase innovative MRI data processing tools in a dynamic and interactive environment. Whether you’ve are interested in the developement of a novel algorithm, software package, or workflow, this session is your platform to learn more about it!
Machine Learning (ML) methods have evolved tremendously during the last decade with a number of backends supporting the development. However, support for high-dimensional and complex-valued data processing is often limited. Therefore, we developed a Machine Enhanced Reconstruction Learning and Interpretation Networks (MERLIN) framework that seamlessly integrates with existing ML solutions (Tensorflow/Keras and Pytorch) and complements them by high-dimensional, complex-valued and MR-specific operators, layers, and data pipelines.
Hyperpolarized 129Xe MRI is an emerging tool for assessing regional lung microstructure and function. This talk will introduce XIPline, an open-source, user-friendly software developed for standardizing the analysis of 129Xe MRI data, covering ventilation, diffusion, and gas exchange imaging. Ongoing efforts to streamline processing and increase research adoption will be highlighted.
This demo will feature the Shimming Toolbox tool used to perform shimming for the brain. Some of the features available will be shown such as volume-wise shimming and slice-wise shimming using spherical harmonics and custom coils.
MIRTorch, developed at the University of Michigan, focuses on reimplementing basic model-based MRI reconstruction in PyTorch. The project enables researchers to run fast reconstruction algorithms on GPUs, easily prototype new reconstruction methods, and explore innovative directions through differentiable programming.
These days, it’s common practice to share code and data via open repositories. These approaches could provide some difficulties, though. Patient data sharing repositories may differ due to variations in country ethical legislation. Furthermore, some research labs don’t have a code repository culture. In this session, we will review some different possibilities for code and data sharing and the importance of a culture of code sharing.
In this presentation, we will explore how the CAMH BrainHealth Databank (BHDB) is revolutionizing mental health research through its commitment to open science and data sharing. By securely collecting and making diverse data available for researchers, the BHDB accelerates the discovery of biological markers, treatments, and prevention strategies for mental illness. This initiative exemplifies how open data governance and ethical practices can foster collaboration and innovation in the global mental health research community.
Neurodesk is a platform for making reproducible Neuroimaging accessible to everyone. In this talk, he will give a quick overview of what Neurodesk is, how it works, how it supports open data analyses and I will briefly show our latest developments on running Neurodesk containers on the scanner through Open Recon.
With a strong impetus towards increasing the reproducibility and transparency of research findings there is growing interest in sharing more research artefacts including: preprints, open data but also open code. While a lot has been said on how to open data (and the so-called FAIR principles), code is a very special research artefact and as such requires specific solutions for sharing. Is sharing my code on Github and adding a link in my paper enough? Should I use the Gitlab instead? If I follow the best practices in code sharing will my results be reproducible?
This talk highlights the experiences of researchers from the African Brain Data Network whose goal is to make African brain data FAIR. It also includes their critical observations as they collected, curated, and shared the first African brain imaging dataset. These observations include the understanding of the importance of FAIR African brain data, the challenges FAIR brain data faces in Africa and the opportunities for growth.
This session will feature hands-on tutorials of essential open science software!
This talk will cover best practices for open science in MR research, highlighting the Brain Imaging Data Structure (BIDS) as a standard for organizing neuroimaging data. We’ll demonstrate how to connect data to Neurodesk, an open platform that improves tool accessibility and reproducibility, and convert it to BIDS format using various tools within Neurodesk.
This talk will provide an overview of open-source tools for MR image segmentation and registration with examples for applications to partial coverage brain data (typical at > 7 T). It will highlight widely used tools like ITK-SNAP, ANTs, FSL, AFNI, and Freesurfer whilst also emphasising the value of general-purpose image processing with Python.
In this talk, I will share practical tips for effectively using Visual Studio Code, a versatile code editor that can greatly enhance productivity. I will highlight several useful extensions (git, AI autocomplete, image viewer, formatting, …). Attendees will learn how to customize the environment to suit their specific needs and might find some new favourite extensions.
In this short introduction to git, we will begin by discussing “what is git and why is it useful?” and then cover some essential basic commands. Next, we will briefly go over the concept of branches and how they facilitate team collaboration. Finally, we will give a live demonstration of putting all these elements together in a small project.
Despite us giving all our heart, the MR community might not meet our beautifully crafted open science contribution with love at first sight. In this session, we want to hear from developers how they persevered, for example after a slow initial uptake of their contribution, and how they built a user community around it that enabled long-term success and continued development and support.
Cortical folding has been of great interest in recent years, but there is a missing link to connect theoretical models to evidence. In this talk, we show what we’ve done so far and how Open Science approaches are important to accomplish more.
Introduction to Brain Imaging Data Structure (BIDS) standard, how it came about, who is developing it, and what principles and procedures propell it forward.
Nonhuman primates (NHPs) serve as essential models in translational neuroscience. However, the small sample sizes in many NHP imaging studies limit cross-species comparisons with large-scale human brain data. In this presentation, I will introduce the Macaque Charts project, which uses over 1,000 MRI datasets from the PRIMatE Data Exchange (PRIME-DE) consortium to map lifespan brain trajectories in macaques.
In this joint talk, we will talk about our experience maintaining scientific software. We will discuss our experience with the BART project and the larger DebianMed initiative.
In this session, we aim to have a hands-on guide to working collaboratively on projects using Git and how to do code review in practice. Finally, we want to introduce concepts common in software development, such as the minimum viable product, that may help interested researchers improve their productivity.
Visualization plays a crucial role in neuroimaging acquisition, processing, inference and dissemination. The AFNI, FSL and FreeSurfer have created competing desktop tools, making it hard for users to transfer skills and data between these frameworks. We have harnessed the collective wisdom of these teams to develop NiiVue using web technologies. This allows a zero-footprint tool that works regardless of device (computer, phone, tablet), operating system, hardware or framework. In addition to traditional desktop use, this software enables innovative edge and cloud-based applications.
Essential Tremor (ET) is the most prevalent movement disorder with poorly understood etiology. We pre-registered our study, tested the hypothesis of cerebellar involvement through three types of neuroimaging biomarkers and further assessed the impacts of statistical models and cerebellar segmentation pipelines on the outcomes. Results showed a lack of replicability of literature findings. This study shows that current estimation of cerebellar involvement in ET is largely inflated and emphasizes the importance of sensitivity analysis in biological inferences.
In this talk, I will discuss a range of strategies that have helped new contributors to advance Physiopy’s code base, including the organization of codesprints, participation in Google’s Summer of Code, and the ongoing compilation of contribution standards and other community coding knowledge.
Transparency and reproducibility in scientific software development are critical to open science. This talk will discuss strategies for automatically testing and evaluating workflows using GitHub Actions, software containers, and open standards such as the Brain Imaging Data Structure (BIDS). The continuously integrated reconstruction challenge for Quantitative Susceptibility Mapping (QSM-CI) will be presented as an example of how a community-driven project can use these standards to unlock new possibilities in neuroimaging investigations that drive innovation.
This session will highlight the latest inspiring work of our community in the open MRI domain: tools that enable open science, reproducibility, and sustainability in MR research, as well as best practices for a completely open pipeline (from data acquisition to publication). We will hear from the Open Science Abstract Awardees at the recent ESMRMB conference 2024 in Barcelona, and give the stage to the finalists of our MRITogether24 Poster Awards, concluding with the award ceremony.
The importance of quantitative MRI is grounded in its capability to have objective measures of a specific parameter. This requires the enforcement of standardization of the measurement conditions. In this context, an open-source framework allows to homogenize the methodology applied in different scanners and centers making it more robust to use in the clinic.
A user project on our MR facility required precise MR spectroscopy voxel placements on unusual anatomical targets, as well as a good reproducibility in voxel placement. My colleagues and I developed two small tools to help alleviate these difficulties. We tried to make them low-dependency, well-documented, and user-friendly.
MRITogether24 Poster Awards
We often commit our time to Open MRI projects with an altruistic streak: the work is voluntary and based on our self-motivation. However, people in open science projects have a day job with other and higher priorities, and because the incentive structure in their career domain doesn’t align with sustainability and collaboration, progress in open science projects is often slow. Here, we want to learn from open science experts as well as other fields of volunteering how to navigate these sometimes conflicting demands and stay happy, healthy and productive in the process.
In the past seven years, several Chinese early-career researchers and I have been promoting open science among the Chinese-speaking community by initiating a grassroots network — the Chinese Open Science Network (COSN, Jin et al., 2023). In this talk, I will share our stories and the lessons we have learned.
Sina and Johanna bring together years of experience as researchers and volunteers in the MRI field and beyond, offering a candid perspective on the intersection of open science, academia, and sustainable volunteering. Drawing from their diverse engagements in open science initiatives through volunteering, mentoring, and teaching, they will share insights on both the short-term and long-term benefits of scientific volunteering. This talk aims to equip attendees with practical strategies for navigating the rewards and challenges of volunteering while highlighting the lasting impact of contributing to open science and academic communities.
Throughout my professional life, volunteering side projects have been a part of my work portfolio. As a former neuroscientist studying self-control, stress and resilience, my insights from my research together with my growing project management practice helped me develop my own take on how to run myself and my work in a sustainable fashion so I could bring meaningful complex projects to fruition. With a nod to the realities and insights of cognitive neuroscience on self-control and stress, I am happy to share some of my takeaways and steps on how to practically approach and steer your course while handling a demanding project and life portfolio that meets not only your ambition, but also your needs when it comes to nurturing your soul and health.