Saturday, Jul 22: 8:00 AM - 12:00 PM
3612
Educational Course - Half Day (4 hours)
Palais
Room: 516AB
Abstract:
Open science tools and practices have been developed to advance reproducibility, as well as accessibility and transparency at all stages of the research cycle and across all levels of society. Together, they remove barriers to sharing and facilitate collaboration, with the goal of improving reproducibility and, ultimately, accelerating scientific discoveries. Empirical observations of how labs conduct research indicate that the adoption rate of open practices for transparent, reproducible, and collaborative science remains in its infancy. This is at odds with the overwhelming evidence for the necessity of these practices and their benefits for individual researchers, scientific progress, and society in general. However, information required for implementing open science practices throughout the different steps of a research project is scattered among many different sources. Even experienced researchers in the topic find it hard to navigate the ecosystem of tools and to make sustainable choices. In our recent publication (Niso, Botvinik-Nezer et al. 2022) we provide an integrated overview of community-developed resources that can support collaborative, open, and reproducible neuroimaging throughout the entire research cycle from inception to publication and across different neuroimaging modalities. With this course we want to lower the bar for the adoption of open science tools and practices across the research cycle from study inception to publication, by presenting a selection in an accessible and practice oriented way with hands-on experience.
Expected learning outcomes:
Open science is an important topic that requires adaptation of practices. Numerous tools and practices exist supporting open science along the workflow from study inception to publication. This educational course is oriented along this workflow and will introduce tools, in a clearly-structured manner, to support open neuroimaging at the different stages of research. Participants of this course will be able to pre-register their studies, organize and annotate their data and respective metadata into agreed-upon standards (e.g. BIDS, HED) and will be taught how to manage their data using datalad and what software to use for creating reproducible analysis workflows. Moreover, participants will learn about several tools for quality control of imaging data (covering several neuroimaging modalities), be provided some practical guidance with respect to data sharing and learn about open-source and vendor-neutral solutions for data acquisition. All the course materials (slides, hands-on content) will be openly available on a GitHub repository (https://oreoni.github.io/). This github repository is an interactive live book containing the (Niso, Botvinik-Nezer et al. 2022) manuscript content, that will be updated as new tools are developed.
- Participants are provided with an overview of the available tools and practices for open and reproducible neuroimaging along each research stage and how these tools and practices can be integrated together.
- Additionally, the audience will get hands-on experience with some of the tools to facilitate adoption into their everyday scientific practice.
Early and late career neuroimaging researchers who want to start or ramp-up the adoption of open science practices in their daily research routines.
Presentations
Decisions made at the planning stage of a study could substantially facilitate or hamper reproducibility and openness of the entire study. This talk will begin with a brief introduction to the entire course, followed by an overview of the tools and practices that are available for planning and initiating open and reproducible neuroimaging studies. First, we will cover tools and practices to ensure the quality of data and design, including Standard Operating Procedures and piloting. Second, we will focus on pre-registrations and Registered Reports, explaining what they are, why and when they are important, how they are done, and what are the main challenges researchers usually face when conducting them. This part will include hands-on experience with preparing a simple (mock) pre-registration. Third, we will describe important considerations and resources for writing informed consent and data sharing plans that allow subsequent ethical data sharing. In this part we will also elaborate on potential obstacles for data sharing and highlight practical solutions aiming at facilitating the lawful and effective sharing of neuroimaging data for the individual researcher. Finally, we will focus on the importance of monitoring the quality of research data, especially with regard to (human) neuroimaging data and provide an overview of quality control tools and analyses.
Presenter
Rotem Botvinik-Nezer, Dartmouth College
Psychological and Brain sciences
Hanover, NH
United States
Data acquisition in neuroimaging is typically performed using proprietary systems from vendors. This lack of transparency in the acquisition details and downstream proprietary processing hinders the development of end-to-end reproducible neuroimaging workflows. This is further compounded by heterogeneity in data formats, inconsistent definition of critical experimental parameters, and variations in technology, which may introduce spurious, non-biological differences between acquisition devices.
Fortunately, in recent years, the scientific community has developed a range of open-source solutions to address the challenges of standardizing data acquisition in neuroimaging research, specifically MRI and PET modalities. These solutions provide vendor-neutral ways of deploying acquisition software, enabling transparent access to all the necessary details while allowing for interoperable and open-source reconstruction frameworks. Additionally, the use of open-source software for stimulus presentation and response logging enables easy replication of neuroimaging experiments and helps establish a transparent and open research ecosystem.
In this talk, our goal is to provide an overview of the open-source solutions that are currently available for neuroimaging research. We will delve into various aspects of data acquisition, including the use of consensus protocols to standardize data acquisition, the availability of vendor-neutral data acquisition pulse sequences and reconstruction frameworks, and the use of open-source software for stimulus presentation. Finally, we will also discuss how these developments can help with data standardization and synchronization efforts in neuroimaging research and help researchers achieve greater reproducibility.
To involve the audience in a vendor-neutral acquisition process, the presentation will be accompanied by an interactive tutorial on reconstructing images from raw MRI data and performing downstream processing to calculate quantitative parameters. The tutorial will be hosted on https://neurolibre.org.
Presenter
Agah Karakuzu, École Polytechnique de Montréal Montréal, Québec
Canada
Research data management (RDM) relates to how data are organized, maintained, annotated, tracked, stored, and accessed throughout a research project. RDM forms the basic foundations of result reproducibility, data reusability, and research efficiency. And thus, it is good practice to develop, review and execute data management plans for every experiment. They are increasingly becoming widely required by funders even at the application phase. RDM includes standardization of data, with rich metadata and data annotation, together with efficient data management and tracking. Here we will focus on data organization and standards to increase FAIRness in neuroimaging. We will present the Brain Imaging Data Structure (BIDS), a community-led standard for organizing, describing, and sharing neuroimaging data. We will review multiple tools, converters and apps to facilitate BIDS adoption, contributing to maximize reproducibility, enable effective data sharing, and support good data management practices. We will also include practical hands-on aspects, showcasing BIDS for the harmonization of a neuroimaging dataset and providing pointers to similar pipelines for data obtained with different modalities, such as MRI, MEG, EEG and PET.
Presenter
Guiomar Niso, Instituto Cajal, CSIC Madrid, Madrid
Spain
Research Data Management (RDM) aspect can be presented as the “deciding on the data language of your study”. As with a language for communication (words, syntax, grammar) and medium for recording your thoughts and findings (notebooks, documents, papers) in that language for sharing and redistribution, RDM relies on agreeing on data standards (language) and tools (medium) to share and collaborate on your project.
In this section we will guide trainees through popular in neuroimaging experiments data standards (e.g., BIDS, HED), harmonization of data into those standards and enrichment with extra metadata, and tools to manage such data (DataLad) to streamline your data management for local compute or interaction with online platforms (such as Brainlife).
Presenter
Yaroslav Halchenko, Dartmouth College
Psychological and Brain Sciences
Hanover, NH
United States
Data processing and analysis comprises one of the core aspects of every scientific investigation. Throughout the last two decades a broad set of respective tools were developed for and from the neuroimaging community. However, the resulting diversity, interacting with a lack of standardization, created a hard-to-navigate landscape with various areas of conflict. This refers to the development and application of closed vs. open source software, software development practices, reproducibility of analysis, uncertain terminology and thus ultimately prominent outcome variability.
In this presentation, we will evaluate the neuroimaging software ecosystem concerning the FAIR-ness of the respective tools, as well as their quality control via development practices and provide guidelines on how to make data processing and analysis more open and reproducible. Besides a comprehensive overview of various neuroimaging software tools, this will entail an introduction to standardization of processing/analysis workflows and the included terminology. Finally, it will be addressed how the above mentioned aspects can be integrated to evaluate and reduce outcome variability.
After discussing each of these aspects, participants will gather the respective resources needed for their own workflows to generate an individual roadmap that they utilize after the educational course.
Sharing the complete workflow (data, preprocessing apps and analysis notebooks) with Brainlife. We will describe an approach to disseminate all the scientific results generated all the way from uploading your own data or by reusing data from OpenNeuro (via DataLad), as well as the preprocessing pipelines and the code used for the statistical analyses in a paper using the open cloud platform Brainlife. We will show it is possible to easily map DICOMs to BIDS structures, how to use pipelines such as FreeSurfer, fMRIPrep or QSIPrep while at the same time keeping track of provenance metadata so as to allow transparency in data processing. Finally, we will demonstrate how to create a transparent record of all the operations performed when implementing a full-stack of data operations such as those necessary when publishing a modern scientific article in neuroimaging research.
Presenter
Franco Pestilli, University Texas Austin
Psychology
AUSTIN, TX
United States