The OpenNeuro resource for sharing of neuroscience data

Abstract
Introduction
Results
Discussion
Materials and methods
Data availability
References
Article and author information
Metrics

Abstract

The sharing of research data is essential to ensure reproducibility and maximize the impact of public investments in scientific research. Here, we describe OpenNeuro, a BRAIN Initiative data archive that provides the ability to openly share data from a broad range of brain imaging data types following the FAIR principles for data sharing. We highlight the importance of the Brain Imaging Data Structure standard for enabling effective curation, sharing, and reuse of data. The archive presently shares more than 600 datasets including data from more than 20,000 participants, comprising multiple species and measurement modalities and a broad range of phenotypes. The impact of the shared data is evident in a growing number of published reuses, currently totalling more than 150 publications. We conclude by describing plans for future development and integration with other ongoing open science efforts.

Introduction

There is growing recognition of the importance of data sharing for scientific progress (National Academies of Sciences, Engineering, and Medicine, Policy and Global Affairs, Board on Research Data and Information, Committee on Toward an Open Science Enterprise, 2018). However, not all shared data are equally useful. The FAIR principles (Wilkinson et al., 2016) have formalized the notion that in order for shared data to be maximally useful, they need to be findable, accessible, interoperable, and reusable. An essential necessity for achieving these goals is that the data and associated metadata follow a common standard for organization, so that data users can easily understand and reuse the shared data. Here, we describe the OpenNeuro data archive (RRID:SCR_005031), accessible at https://openneuro.org, which enables FAIR-compliant data sharing for a growing range of neuroscience data types (currently including magnetic resonance imaging [MRI], electroencephalography [EEG], magnetoencephalography [MEG], and positron emission tomography [PET]) through the use of a common community standard, the Brain Imaging Data Structure (BIDS) (RRID:SCR_016124; Gorgolewski et al., 2016).

Starting with early pioneering efforts by Gazzaniga and Van Horn to establish an fMRI Data Center in 1999 (Van Horn and Gazzaniga, 2013), data sharing has become well established in the domain of neuroimaging (Milham et al., 2018; Poldrack and Gorgolewski, 2014; Poline et al., 2012). A major impetus for the growth of data sharing was the International Neuroimaging Data Sharing Initiative (Mennes et al., 2013), which published a landmark paper in 2010 (Biswal et al., 2010) demonstrating the scientific utility of a large shared resting fMRI dataset. The most prominent recent examples have been large-scale prospective data sharing projects, including the Human Connectome Project (HCP) (Van Essen et al., 2013), the NKI-Rockland sample (Nooner et al., 2012), Adolescent Brain Cognitive Development study (Casey et al., 2018), and the UK Biobank (Littlejohns et al., 2020). These datasets have provided immense value to the field and have strongly demonstrated the utility of shared data. However, their scientific scope is necessarily limited, given that each dataset includes only a limited number of imaging tasks and measurement types. Beyond these large focused data sharing projects, there is a ‘long tail’ of smaller neuroimaging datasets that have been collected in service of specific research questions. Making these available is essential to ensure reproducibility as well as to allow aggregation across many different types of measurements in service of novel scientific questions. The OpenNeuro archive addresses this challenge by providing researchers with the ability to easily share a broad range of neuroimaging data types in a way that adheres to the FAIR principles.

Goals and principles

The OpenNeuro archive evolved from the OpenfMRI archive (Poldrack et al., 2013), which was focused solely on the sharing of task-based human fMRI data. Some of the principles behind OpenNeuro were inherited from OpenfMRI, whereas others grew out of our experiences in that project as well as from new developments in the domain of open science.

Minimal restrictions on sharing

There is a range of restrictiveness across data archives with regard to their data use agreements (Jwa and Poldrack, 2021). At one end of the spectrum are highly restricted databases such as the Alzheimer’s Disease Neuroimaging Initiative, which requires researchers to submit their scientific question for review and requires the consortium to be included as a corporate author on any publications. OpenNeuro represents the other pole of restrictiveness, by releasing data (by default) under a Creative Commons Zero (CC0) Public Domain Dedication which places no restrictions on who can use the data or what can be done with them. While not legally required, researchers using the data are expected to abide by community norms and cite the data following the guidelines included within each dataset. The primary motivation for this policy is that it makes the data maximally accessible to the largest possible number of researchers and citizen-scientists.

Standards-focused data sharing

To ensure the utility of shared data for the purposes of efficient discovery, reuse, and reproducibility, standards are required for data and metadata organization. These standards make the structure of the data clear to users and thus reduce the need for support by data owners and curation by repository owners, as well as enabling automated QA, preprocessing, and analytics. Unfortunately, most prior data sharing projects have relied upon custom organizational schemes, which can lead to misunderstanding and can also require substantial reorganization to adapt to common analysis workflows. The need for a clearly defined standard for neuroimaging data emerged from our experiences in the OpenfMRI project; while the repository had developed a custom scheme for data organization and file naming, this scheme was ad hoc and limited in its coverage, and datasets often required substantial manual curation (involving laborious interaction with data owners). In addition, there was no way to directly validate whether a particular dataset met the standard.

For these reasons, we focused at the outset of the OpenNeuro project on developing a more robust data organization standard that could be implemented in an automated validator. We engaged a large group of researchers from the neuroimaging community to establish a standard that ultimately became the BIDS (Gorgolewski et al., 2016), which is now a highly successful community standard for a broad and growing range of neuroimaging data types. BIDS defines a set of schemas for file and folder organization and naming, along with a schema for metadata organization. The framework was inspired by the existing data organization frameworks used in many research laboratories, so that transitioning to the standard is relatively easy for most researchers. One of the important features of BIDS is its extensibility; using a scheme inspired by open-source software projects, community members can propose extensions to BIDS that encompass new data types. To date, modality extensions include MEG (Niso et al., 2018), scalp EEG (Pernet et al., 2019), intracranial EEG (Holdgraf et al., 2019), PET (Norgaard et al., 2021), and arterial spin labeling MRI. In addition to standards for raw data, the BIDS community has also developed a standard for the organization of the outputs of processing operations (known as ‘BIDS Derivatives’), providing a framework for sharing processed as well as raw data.

While BIDS and OpenNeuro are now independent projects, there is a strongly synergistic relationship. All data uploaded to OpenNeuro must first pass a BIDS validation step, such that all data in OpenNeuro are compliant with the BIDS specifications at upload time. Conversely, the OpenNeuro team has made substantial contributions to the BIDS standard and validator. The BIDS standard has been remarkably successful, with tens of thousands of datasets now available in the format, including but not limited to those contained in the OpenNeuro database. As a consequence, this model maximizes compatibility with processing and analysis tools (Gorgolewski et al., 2017), but more importantly, it effectively minimizes the potential for data misinterpretation (e.g., when owner and reuser have slightly different definitions of a critical acquisition parameter). Through the adoption of BIDS, OpenNeuro has moved away from project- or database-specific data structures designed by the owner or the distributor (as used in earlier projects such as OpenfMRI and HCP) and toward a uniform and unambiguous representation model agreed upon by the research community prior to sharing and reuse.

FAIR sharing

The FAIR principles (Wilkinson et al., 2016) have provided an important framework to guide the development and assessment of open data resources. OpenNeuro implements each of these principles.

Findable: Each dataset within OpenNeuro is associated with metadata, both directly from the BIDS dataset along with additional dataset-level metadata provided by the submitter at time of submission. Both data and metadata are assigned a persistent unique identifier (Digital Object Identifier [DOI]). Within the repository, a machine-readable summary of BIDS metadata is collected by the BIDS validator and indexed with an ElasticSearch mapping. In addition, dataset-level metadata are exposed according to the schema.org standard, which allows indexing by external resources such as Google Dataset Search.

Accessible: Data and metadata can be retrieved using a number of access methods (directly from Amazon S3, using the OpenNeuro command line tool, or using DataLad) via standard protocols (http/https). Metadata are also accessible programmatically via a web API. Metadata remain available even in the case that data must be removed (e.g., in cases of human subjects concerns). No authentication is necessary to access the data.

Interoperable: The data and metadata use the BIDS standard to ensure accessible representation and interoperation with analysis workflows, such as BIDS Apps (Gorgolewski et al., 2017). Ongoing work is extending the metadata representation to use richer formats and to link to relevant FAIR ontologies or vocabularies.

Reusable: The data are released with a clear data use agreement (currently defaulting to a CC0 public domain dedication). Through use of the BIDS standard, the data and metadata are consistent with community standards in the field.

Data versioning and preservation

OpenNeuro keeps track of all changes in stored datasets and allows researchers to unambiguously report the exact version of the data used for any analysis. OpenNeuro preserves all versions of the data through the creation of ‘snapshots’ that unequivocally point to one specific point in the lifetime of a dataset. Data management and snapshots are supported by DataLad (RRID:SCR_003931; Halchenko et al., 2021), a free and open-source distributed data management system (Hanke et al., 2021).

Protecting privacy and confidentiality of data

There is a direct relationship in data sharing between the openness of the data and their reuse potential; all else being equal, data that are more easily or openly available will be more easily and readily reused. However, all else is not equal, as openness raises concern regarding risks to subject privacy and confidentiality of data in human subjects research. Researchers are ethically bound to both minimize the risks to their research participants (including risks to confidentiality) and to maximize the benefits of their participation (United States. National Commission for the States, 1978). Because sharing of data will necessarily increase the potential utility of the data, researchers are ethically bound to share human subject data unless the benefits of sharing are outweighed by risks to the participant (Brakewood and Poldrack, 2013).

In general, risks to data privacy and confidentiality are addressed through deidentification of the data to be shared. For example, under the Health Insurance Portability and Accountability Act of 1996 (HIPAA) in the United States, deidentification can be achieved through the removal of any of 18 personal identifiers, unless the researcher has knowledge that the remaining data could be re-identified (known as the ‘safe harbor’ method). With regard to neuroimaging data, a particularly challenging feature is the facial structure that is present in some forms of imaging data, such as structural MRI images. It is often possible to reconstruct facial structures from these images, and there are proofs of concept that such data could be used to re-identify individuals from photographic databases (Schwarz et al., 2019). It is thus essential to remove any image features that could be used to reconstruct facial structure (Bischoff-Grethe et al., 2007). For this reason, all MRI data shared through OpenNeuro must have facial features removed prior to upload, in addition to the 18 personal identifiers outlined by HIPAA. An exception is provided in cases where an investigator has explicit permission to openly share the data without defacing, usually when the data are collected by the investigator themself. At present, data are examined by a human curator to ensure that this requirement has been met. In the future, we plan to deploy an automated face detection tool (Bansal et al., 2020) to detect any uploads that inadvertently contain facial features.

Truly informed consent requires that subjects be made aware that their data may be shared publicly, and that confidentiality cannot be absolutely guaranteed in the future. For this reason, we recommend that researchers planning to share their data via OpenNeuro use a consent form based on the Open Brain Consent (Bannier et al., 2021), which includes language that ensures subject awareness of the intent to share and its potential impact on the risk of participating. Of note, the Open Brain Consent has recently been adapted to include a data usage agreement that accommodates the European Union’s General Data Protection Regulation (GDPR 2016/679); however, data collected in countries covered by GDPR cannot be shared through OpenNeuro at present due to the requirement for restrictive data use agreements that are not currently supported by OpenNeuro.

Open source

The entirety of the code for OpenNeuro is available under a permissive open-source software license (MIT License) at https://github.com/OpenNeuroOrg/openneuro. This enables any researcher who wishes to reuse part or all of the code or to run their own instance of the platform.

Data submission and access

Figure 1 outlines the steps required for sharing a dataset using OpenNeuro. Once shared, data can be accessed by several available mechanisms.

Figure 1 with 2 supplements see all

Download asset Open asset

A schematic overview of the data upload process.

Web download: Each snapshot is associated with a link that provides immediate downloading of the dataset.

DataLad. DataLad (Halchenko et al., 2016) is a decentralized data management system built on top of git and git-annex. Through DataLad, researchers may install a complete copy of a dataset, while deferring the retrieval of file contents until needed, permitting lightweight views of large datasets. OpenNeuro’s versioned snapshots are implemented as git tags, which allows specific versions to be easily retrieved or compared. The decentralized protocol also allows mirrors of the datasets to be hosted on GitHub and https://datasets.datalad.org, ensuring access during service interruptions of the OpenNeuro website.

OpenNeuro command line tool: The OpenNeuro command line tool provides access to the latest snapshot of all datasets, and is generally more stable than browser downloads for large datasets.

Amazon S3: The latest snapshot as well as all previous versions of a dataset may be fetched using the Amazon Web Services (AWS) clients or directly via https.

User support

Support for individual datasets: Data users sometimes have questions regarding particular datasets. In order to facilitate discussion of these issues and to make those discussions available to the entire community, a discussion forum is provided on each dataset page. The dataset owner is automatically notified by email of any questions that are posted. In addition, users can ‘follow’ a dataset of interest and receive notifications of any comments posted to the dataset.

Site support: Two mechanisms are provided for users of the OpenNeuro site to obtain help with site issues. First, a helpdesk is available directly from the site, through which users can submit specific help questions. Second, users are recommended to post general questions to the Neurostars.org question and answer forum, so that the answers will be available to the entire community.

Data processing

Data processing was initially envisioned as an incentive for researchers to share their data, and the OpenNeuro site was launched in 2017 with the ability to perform cloud-based data processing using a limited set of analysis workflows. This feature was disabled in 2018, after an overhaul of the site’s initial storage infrastructure. At that time, we determined that it would be preferable to collaborate with an existing platform dedicated to cloud processing rather than rebuilding our own execution platform. At present, OpenNeuro has partnered with the Brainlife.io platform (RRID:SCR_020940), which provides a large set of cloud-based neuroimaging workflows for data analysis and visualization. Data hosted on OpenNeuro can be easily imported into Brainlife for analysis, and more than 400 OpenNeuro datasets are cached for quick access; in the first 6 months of 2021, more than 700 analyses were performed on these datasets. In the future we plan to partner with additional platforms, including the NEMAR platform for EEG/MEG analysis; the availability of the data via DataLad and Amazon S3 also enables any platform to make the data available to their users without requiring any agreement or effort from OpenNeuro.

Results

Usage and impact

The OpenNeuro site was launched in June 2017, and was originally seeded with all of the datasets previously shared through OpenfMRI, after converting them to the BIDS standard. All data presented below are current as of October 9, 2021. The database contains 604 datasets comprising data from 20,989 individual participants. Figure 2 shows cumulative figures for numbers of datasets and subjects since 2018, demonstrating sustained and continual growth in the archive since its inception.

Figure 2

Download asset Open asset

The volume of data available on OpenNeuro has shown a steady growth since its opening started operations in 2017.

Shown are figures from July 2018, when all data were migrated to a new DataLad storage backend, through the present date. The green line illustrates the cumulative growth in total number of datasets, and the red line shows the aggregate of subjects (in thousands).

The overwhelming majority of datasets are from humans (574 datasets, 95%), with a small but growing number of nonhuman species including mouse (17 datasets), rat (6 datasets), nonhuman primates (2 datasets), dogs (1 dataset), and juvenile pigs (1 dataset). Table 1 presents data for the prevalence of different modalities; while the majority of datasets include some form of MRI data, other supported modalities are present including electrophysiological measures and PET.

Table 1

Number of datasets by imaging modality; additional modalities present in fewer than three datasets are not included here.

Modality	Number of datasets
Anatomical MRI	501
Functional MRI	445
Electroencephalography	81
Diffusion-weighted MRI	53
Magnetoencephalography	23
Positron emission tomography	10
Intracranial EEG	8
Arterial spin labeling MRI	3

OpenNeuro is a recommended data repository for a number of publishers and journals, including Nature Scientific Data, PLOS, eLife, F1000 Research, Gigascience, BioMed Central, American Heart Association, and Wellcome Open Research. The database contains 407 DOIs for publications associated with datasets (including both primary scientific publications and data descriptors).

Multiple dimensions of ‘big data’

Discussions of ‘big data’ in neuroimaging (Poldrack and Gorgolewski, 2014; Smith and Nichols, 2018) have largely focused on datasets including large numbers of individuals. While these analyses are essential for robust population inference, it is also important to recognize that large numbers of subjects are only one dimension over which a neuroimaging dataset can be ‘big’. Here, we will define the number of subjects as the ‘width’ of the dataset, the number of different phenotypes measured for each individual as the ‘breadth’ of the dataset, and the number of measurements per individual as the ‘depth’ of the dataset.

The OpenNeuro database is distinguished by sharing datasets that are extensive along each of these dimensions (see Figure 3). With regard to width, the median dataset size is 23 subjects, with 31 studies having sample sizes larger than 100, and a maximum sample size of 928. With regard to breadth, notable datasets include the BOLD5000 dataset (Chang et al., 2019), which includes data from subjects viewing a total of 5000 natural images; the Individual Brain Charting dataset (Pinho et al., 2020; Pinho et al., 2018), which includes data from individuals each completing 24 different tasks, and the Multidomain Task Battery dataset (King et al., 2019), which includes data from individuals each completing 26 tasks. With regard to depth, the database currently includes the MyConnectome dataset (Poldrack et al., 2015), which includes extensive task, resting, and diffusion MRI data from more than 100 sessions for a single individual; the Midnight Scan Club dataset (Gordon et al., 2017), which includes extensive task and resting fMRI data from 10 individuals; and a number of other dense scanning datasets (Gonzalez-Castillo et al., 2015; Newbold et al., 2020; Salehi et al., 2020).

Figure 3

Download asset Open asset

OpenNeuro datasets vary substantially in number of participants (X axis), number of sessions per participant (Y axis), and number of tasks per participant (size/color of datapoints); axes are log-scaled for easier visualization.

Results are based on metadata derived directly from the 502 OpenNeuro datasets available via DataLad as of 10/9/2021.

Another unique feature of OpenNeuro is the breadth of phenotypes across datasets. To further characterize this, we searched the text associated with OpenNeuro datasets to identify terms related to psychological concepts and tasks as defined in the Cognitive Atlas ontology (Poldrack et al., 2011). Word clouds showing the top terms identified in this analysis are shown in Figure 1—figure supplement 2. This analysis shows a broad range of tasks and concepts associated with these datasets, highlighting the substantial conceptual and methodological breadth of the archive.

Data reuse

OpenNeuro has distributed a substantial amount of data; from May 2020 through April 2021, a total of 406 terabytes of data were distributed. Because data reuse is not directly measurable, we utilize published reuse of the shared data as a proxy. To identify published reuses of OpenNeuro data, we used Google Scholar and CrossRef to identify potential reuses, and then manually examined them to confirm that they were a legitimate reuse (as opposed to a primary publication of the data or data descriptor); note that this is an underestimate since many papers during this period reported analyses of data downloaded from OpenfMRI, which would not have been identified in our searches. We identified 165 publications that reused OpenNeuro datasets; this showed a sharp increase over time (see Figure 4). Of these publications, 112 were journal or conference papers, 42 were preprints, and 11 were other types of publications (such as theses or project reports). A total of 111 OpenNeuro datasets were reused at least once, with the most popular dataset (Poldrack et al., 2016) appearing in 28 published reuses. A significant number of publications reused multiple datasets; 31 of the 165 papers reused at least two datasets, with a maximum of 40 datasets reused (Esteban et al., 2019). Collecting these data from scratch would have required more than 21,000 individual subject visits; at an estimated scanning cost of $1000/session (based on the conservative cost estimate from Milham et al., 2018), this represents a total data reuse value of nearly 21 million US dollars. These reuses have a total of 1329 citations (according to Google Scholar as of June 15, 2021); the most highly cited reuse (Esteban et al., 2019) has more than 500 citations.

Figure 4

Download asset Open asset

Published reuses of OpenNeuro datasets, split by the type of reuse.

Note that the final bar includes only reuses identified through June 2021.

The published reuses of OpenNeuro data span from basic neuroscience to methodological studies and software development. In particular, several studies demonstrate how OpenNeuro data have enabled new insights into brain function. For example, Martins et al., 2021, used structural MRI data from several OpenNeuro datasets along with other shared data to examine different patient groups suffering from physical pain or depression. Their analyses demonstrated a specific pattern of anatomical change common to patients with pain syndromes but distinct from depression. This kind of analysis highlights the way in which OpenNeuro enables researchers to combine smaller datasets in order to test hypotheses using convergent data, which can help overcome the confounds and biases present in any particular study as well as increasing statistical power. Other basic neuroscience studies have used OpenNeuro data to model the role of temporal context in forgetting (Chien and Honey, 2020), characterize the role of edge communities in brain networks (Faskowitz et al., 2020), understand the relationship between functional connectivity and sustained attention (Rosenberg et al., 2020), and to demonstrate that functional parcellation changes as a function of task (Salehi et al., 2020).

Data from OpenNeuro have been particularly useful for the development of new software tools. Esteban et al., 2019, used the breadth and variety of datasets in the archive to assess the robustness of the fMRIPrep preprocessing workflow to many different fMRI datasets, incorporating a total of 40 datasets from OpenNeuro. Importantly, these datasets were used in an iterative manner to improve the robustness of the tool; thus, the breadth of the data were essential both for assessment and for improvement of the tool. Without OpenNeuro (and BIDS), amassing such a large and diverse group of datasets would have required immense efforts to reach out to many different research groups, request their data, and then format the data for common usage, whereas with OpenNeuro the entirety of these datasets can be downloaded automatically within a number of hours, immediately ready for analysis. Other software development projects have taken advantage of some of the particular unique datasets in OpenNeuro; for example, Takeda et al., 2019, took advantage of a unique dataset that combines EEG, MEG, and MRI data on the same individuals (Wakeman and Henson, 2015) to demonstrate the broad range of functions of their VBMEG toolbox. Other software publications using OpenNeuro data include FastSurfer (Henschel et al., 2020) for structural MRI analysis, and Brainstorm (Tadel et al., 2019) for MEG/EEG analysis.

The data in OpenNeuro have been particularly useful for methodological researchers. One prominent example was published by Bowring et al., 2019, who examined how the use of different analysis software impacted statistical results from fMRI activation analyses. Their study included an in-depth analysis of the publications associated with each of 55 datasets, in order to identify studies with analysis pipelines and activation results that could be easily compared with their multi-platform results. Based on this process, they selected three datasets and processed each using several different analysis pipelines; their results highlighted substantial similarity in unthresholded maps but substantial discordance in thresholded maps, highlighting the need for better understanding of the impact of software packages on statistical results. Another example that would have been challenging to perform without OpenNeuro was published by Dadi et al., 2020, who developed a set of functional atlases using 27 datasets. This breadth allowed them to ensure that the specific features of the atlas were not driven by any particular dataset or task. Other examples include studies that used OpenNeuro data to assess the impact of confound regression on fMRI signals and develop new methods for confound modeling (Aquino et al., 2020), and to develop and benchmark new methods for multiple comparison correction (Spisák et al., 2019).

Discussion

The OpenNeuro data archive plays an important role in advancing neuroscience research and ensuring its reproducibility by enabling the sharing of a broad range of neuroscience data types according to the FAIR principles. Its tight integration with the community-driven BIDS standard enhances the ease of sharing, the reusability of the shared data, and the extensibility of the archive in the future. The shared data have enabled a growing number of publications that provide novel neuroscientific insights, as well as supporting novel methodological advances and software development.

Lessons learned

The experiences of our group in developing the OpenNeuro project have provided a number of lessons that may be useful more generally for researchers interested in establishing a culture of data sharing within their scientific subdomain.

Foremost, we have found that the use of a common community-driven format for data organization is essential to effective sharing. In our case, the BIDS standard has enabled data owners to easily share a growing range of data types (through the use of client-side validator), and has enabled researchers to easily reuse the data. Because any dataset that passes the validator can be shared, the community’s efforts on extending the standard (which are implemented in the validator) has provided a steady stream of additions to the types of data that OpenNeuro can share. Another important point is that data sharing does not only include sharing with other researchers, but also with one’s own research group in the future; thus, the use of a well-structured data standard can help researchers ensure that data collected by current lab members can be effectively utilized by other lab members in future, as well as making it easy to share the data beyond one’s own lab. On the flipside, we continue to see that conversion of data into the BIDS standard remains a stumbling block for many researchers; the continued development of conversion tools is necessary to support these researchers.

Second, we have found that ‘it takes an ecosystem’ to make data sharing successful. OpenNeuro is only one of the data sharing projects within the field of neuroimaging, and each of the projects has its own particular features and advantages, but together these projects have increasingly led the field to view data sharing as a net positive for our field. In addition, the availability of these data resources has allowed others to build projects that support new mechanisms for data representation and distribution (such as the DataLad project) and new platforms for analysis (such as Brainlife.io). Together, these tools have provided researchers with additional incentives to share their data via OpenNeuro through its deep integration with those projects. While we believe that sharing is most effective when it is most open, we also realize that some researchers will be unable to share their data on OpenNeuro for ethical or regulatory reasons; for this reason, we believe that a variety of data sharing resources that vary in their sharing policies (Jwa and Poldrack, 2021) will remain essential to support the broadest possible degree of data sharing.

Finally, we would highlight the importance of domain-specific data repositories that support a particular research community. All of the sharing activities accomplished using OpenNeuro could in principle have been accomplished using more general data sharing repositories (such as Figshare or Dryad). A unique benefit of OpenNeuro has been in making a large number of datasets easily findable by researchers, rather than requiring a trawl through a much larger body of datasets to find ones that are relevant. By developing upload and download systems that are tailored for imaging data, OpenNeuro has also greatly lowered the barrier to sharing and reusing data. These benefits argue for the continued need for domain-specific data sharing projects designed in close consultation with researchers in the area. Domain specificity has also allowed OpenNeuro to nurture a community around the resource. Through our social media presence, we have engaged the community with regular blog posts that highlight the most open and sharing labs over the previous 6 months to promote more social incentives to sharing.

Long-term sustainability

A continual challenge for any investigator-initiated data repository is the long-term sustainability of the archive, in order to ensure researchers’ trust in the platform (Lin et al., 2020). The ongoing costs of running a repository are substantial, primarily due to the continuing cost of technological upkeep of a web platform with regard to security and stability, as well as the ongoing costs of storage and bandwidth on cloud platforms or hardware maintenance when using on-premise computing systems. Performant web applications require the use of cutting-edge software tools, which can often become deprecated or unstable over time, leading to substantial technical debt that must be continually addressed to maintain stable and secure operation.

One major challenge for repositories that are reliant upon federal grants is the usual 3-year funding period, in addition to the preference of standard grant mechanisms for funding novel projects rather than ongoing maintenance and operations. One welcome development has been the instigation of longer-term funding for data archives through the US BRAIN Initiative (Koroshetz et al., 2018), which has explicitly dedicated funding to the development and long-term sustainability of data archives for neuroscience data. These renewable 5-year grants (of which OpenNeuro is one of the recipients) provide a much-needed longer-term funding source for data repositories.

Another resource for longer-term sustainability is institutional data repositories, which are increasingly available at many universities. OpenNeuro is working with the Stanford Digital Repository to develop a plan to deposit all raw datasets within the university’s archive, which would provide a digital backstop to the archive’s cloud storage.

OpenNeuro has also been fortunate to be part of the Amazon Public Datasets project (https://registry.opendata.aws/openneuro/), which has provided free data storage and bandwidth for the openly available datasets in the OpenNeuro archive.

Current limitations and future directions

There are a number of additional features planned for future development. These include:

Enhanced metadata: At present, a limited amount of dataset-level metadata is collected beyond that present within the BIDS metadata. Working with the CEDAR Metadata Center (Musen et al., 2015), we plan to add the ability for researchers to enter additional metadata that is linked to standard ontologies, including those being developed for BIDS data in the context of the Neuroimaging Data Model (Maumet et al., 2016). These annotations will provide the basis for more powerful queries of the archive.

Sharing of derivatives: At present, OpenNeuro only shares raw data. However, the availability of a BIDS standard for the outputs of data processing (i.e. ‘derivative’ data) now provides the ability to include derivative data within a BIDS dataset. We plan to enable researchers to share derivatives, for example, allowing the sharing of preprocessed MRI data in addition to raw data. This will greatly enhance the reuse of data by researchers who do not have the resources or expertise to preprocess these complex datasets as well as provide a standard baseline for downstream analyses, reducing the potential effects of analytic flexibility (Botvinik-Nezer et al., 2020; Bowring et al., 2019).

Bringing computing to data: The availability of the OpenNeuro data on the AWS allows researchers direct access to computing on the data, but doing so requires a substantial degree of cloud computing expertise. To ease the application of computing to the data, we plan to adapt the DANDI Hub infrastructure developed by the Distributed Archives for Neurophysiology Data Integration (DANDI: https://www.dandiarchive.org/), which will allow direct access to the data via a Jupyter notebook.

Beyond MRI data: Driven by the initial seeding of data from OpenfMRI, and reflecting the fact that BIDS was originally MRI-centric, the data currently available from OpenNeuro are heavily skewed toward MRI, and fMRI in particular (Table 1). However, BIDS is quickly expanding to other modalities that can readily be uploaded to OpenNeuro, and there has been a rapid increase in sharing of other modalities; for example, more than 60 EEG datasets have been deposited since the publication of the BIDS-EEG standard in 2019 (Pernet et al., 2019). This organic expansion beyond MRI will be supported with the necessary adaptations (e.g., online visualization of new modalities) of OpenNeuro’s user interface.

Conclusion

Data sharing ensures the transparency and reproducibility of scientific research, and allows aggregation across datasets that improves statistical power and enables new research questions. The OpenNeuro repository plays a central role in the data sharing ecosystem by promoting maximally open sharing of data, and by enhancing open availability of data from a wide range of datasets spanning. The growth and impact of the repository demonstrate the viability of minimally restrictive sharing, and the importance of common standards such as BIDS for the effective sharing and reuse of data.

Materials and methods

OpenNeuro infrastructure

Request a detailed protocol

Code for the OpenNeuro platform is available at (https://github.com/OpenNeuroOrg/openneuro). The application utilizes a cloud-based containerized architecture and is built in JavaScript and Python with a MongoDB database for application data storage. OpenNeuro is hosted on AWS using the Kubernetes container orchestration platform. Services are deployed as containers and integrated via a JavaScript GraphQL API gateway and the AWS Application Load Balancer. Several clients access this API, the React website, OpenNeuro command line interface, and an ElasticSearch indexer. Datasets are stored as DataLad repositories and managed by a Python backend service container. Each DataLad repository is assigned to a ZFS pool backed by AWS Elastic Block Store. This allows DataLad versioning and filesystem level access to datasets with existing processing and validation tools. Persistent metadata such as user accounts and permissions are maintained in a MongoDB database. Ephemeral caching is provided by Redis. Search indexes, performance monitoring, and logging are implemented with ElasticSearch. CloudFront is used as a global cache and network to provide global presence.

Content analysis

Request a detailed protocol

Data regarding OpenNeuro contents and usage were current as of October 9, 2021. Code and data needed to execute all analyses and generate all figures are available from https://doi.org/105281/zenodo5559041.

Reuse analyses: Potential reuses were identified by first searching Google Scholar for the term “OpenNeuro”; note that this will exclude any paper that mention ‘OpenfMRI’ instead of OpenNeuro, thus the reported results are underestimates of the true impact of the data, given that many of the datasets in OpenNeuro came from OpenfMRI. Papers matching this search were examined manually to confirm that they had reused data; data descriptor papers were excluded from further analysis. Citation counts were obtained from Google Scholar using the Python package ‘scholarly’.

Dataset size analyses: Dataset size analyses were performed using DataLad to obtain the full BIDS metadata for the 502 datasets available as of October 9, 2021, and then using PyBIDS (Yarkoni et al., 2019) to load the metadata for each dataset.

Data availability

The OpenNeuro data repository is accessible at http://openneuro.org. The derived data used to generate the analyses and figures reported here are available at https://doi.org/10.5281/zenodo.5559041.

The following data sets were generated

1. Poldrack R
2. Esteban O
(2021) Zenodo
poldrack/OpenNeuro_analyses: Revision and updates.

https://doi.org/10.5281/zenodo.5559041

References

(2020) Identifying and removing widespread signal deflections from fMRI data: Rethinking the global signal regression problem
NeuroImage 212:116614.

https://doi.org/10.1016/j.neuroimage.2020.116614
- PubMed
- Google Scholar
1. Bannier E
2. Barker G
3. Borghesani V
4. Broeckx N
5. Clement P
6. Emblem KE
7. Ghosh S
8. Glerean E
9. Gorgolewski KJ
10. Havu M
11. Halchenko YO
12. Herholz P
13. Hespel A
14. Heunis S
15. Hu Y
16. Hu C-P
17. Huijser D
18. de la Iglesia Vayá M
19. Jancalek R
20. Katsaros VK
21. Kieseler M-L
22. Maumet C
23. Moreau CA
24. Mutsaerts H-J
25. Oostenveld R
26. Ozturk-Isik E
27. Pascual Leone Espinosa N
28. Pellman J
29. Pernet CR
30. Pizzini FB
31. Trbalić AŠ
32. Toussaint P-J
33. Visconti di Oleggio Castello M
34. Wang F
35. Wang C
36. Zhu H
(2021) The Open Brain Consent: Informing research participants and obtaining consent to share brain imaging data
Human Brain Mapping 42:1945–1951.

https://doi.org/10.1002/hbm.25351
- PubMed
- Google Scholar
Preprint
1. Bansal S
2. Kori A
3. Zulfikar W
4. Wexler J
5. Markiewicz C
6. Feingold F
7. Poldrack R
(2020) High-Sensitivity Detection of Facial Features on MRI Brain Scans with a Convolutional Network
bioRxiv.

https://doi.org/10.1101/2021.04.25.441373
- Google Scholar
1. Bischoff-Grethe A
2. Ozyurt IB
3. Busa E
4. Quinn BT
5. Fennema-Notestine C
6. Clark CP
7. Morris S
8. Bondi MW
9. Jernigan TL
10. Dale AM
11. Brown GG
12. Fischl B
(2007) A technique for the deidentification of structural brain MR images
Human Brain Mapping 28:892–903.

https://doi.org/10.1002/hbm.20312
- PubMed
- Google Scholar
1. Biswal BB
2. Mennes M
3. Zuo XN
4. Gohel S
5. Kelly C
6. Smith SM
7. Beckmann CF
8. Adelstein JS
9. Buckner RL
10. Colcombe S
11. Dogonowski AM
12. Ernst M
13. Fair D
14. Hampson M
15. Hoptman MJ
16. Hyde JS
17. Kiviniemi VJ
18. Kötter R
19. Li SJ
20. Lin CP
21. Lowe MJ
22. Mackay C
23. Madden DJ
24. Madsen KH
25. Margulies DS
26. Mayberg HS
27. McMahon K
28. Monk CS
29. Mostofsky SH
30. Nagel BJ
31. Pekar JJ
32. Peltier SJ
33. Petersen SE
34. Riedl V
35. Rombouts S
36. Rypma B
37. Schlaggar BL
38. Schmidt S
39. Seidler RD
40. Siegle GJ
41. Sorg C
42. Teng GJ
43. Veijola J
44. Villringer A
45. Walter M
46. Wang L
47. Weng XC
48. Whitfield-Gabrieli S
49. Williamson P
50. Windischberger C
51. Zang YF
52. Zhang HY
53. Castellanos FX
54. Milham MP
(2010) Toward discovery science of human brain function
PNAS 107:4734–4739.

https://doi.org/10.1073/pnas.0911855107
- PubMed
- Google Scholar
1. Botvinik-Nezer R
2. Holzmeister F
3. Camerer CF
4. Dreber A
5. Huber J
6. Johannesson M
7. Kirchler M
8. Iwanir R
9. Mumford JA
10. Adcock RA
11. Avesani P
12. Baczkowski BM
13. Bajracharya A
14. Bakst L
15. Ball S
16. Barilari M
17. Bault N
18. Beaton D
19. Beitner J
20. Benoit RG
21. Berkers RMWJ
22. Bhanji JP
23. Biswal BB
24. Bobadilla-Suarez S
25. Bortolini T
26. Bottenhorn KL
27. Bowring A
28. Braem S
29. Brooks HR
30. Brudner EG
31. Calderon CB
32. Camilleri JA
33. Castrellon JJ
34. Cecchetti L
35. Cieslik EC
36. Cole ZJ
37. Collignon O
38. Cox RW
39. Cunningham WA
40. Czoschke S
41. Dadi K
42. Davis CP
43. Luca AD
44. Delgado MR
45. Demetriou L
46. Dennison JB
47. Di X
48. Dickie EW
49. Dobryakova E
50. Donnat CL
51. Dukart J
52. Duncan NW
53. Durnez J
54. Eed A
55. Eickhoff SB
56. Erhart A
57. Fontanesi L
58. Fricke GM
59. Fu S
60. Galván A
61. Gau R
62. Genon S
63. Glatard T
64. Glerean E
65. Goeman JJ
66. Golowin SAE
67. González-García C
68. Gorgolewski KJ
69. Grady CL
70. Green MA
71. Guassi Moreira JF
72. Guest O
73. Hakimi S
74. Hamilton JP
75. Hancock R
76. Handjaras G
77. Harry BB
78. Hawco C
79. Herholz P
80. Herman G
81. Heunis S
82. Hoffstaedter F
83. Hogeveen J
84. Holmes S
85. Hu C-P
86. Huettel SA
87. Hughes ME
88. Iacovella V
89. Iordan AD
90. Isager PM
91. Isik AI
92. Jahn A
93. Johnson MR
94. Johnstone T
95. Joseph MJE
96. Juliano AC
97. Kable JW
98. Kassinopoulos M
99. Koba C
100. Kong X-Z
101. Koscik TR
102. Kucukboyaci NE
103. Kuhl BA
104. Kupek S
105. Laird AR
106. Lamm C
107. Langner R
108. Lauharatanahirun N
109. Lee H
110. Lee S
111. Leemans A
112. Leo A
113. Lesage E
114. Li F
115. Li MYC
116. Lim PC
117. Lintz EN
118. Liphardt SW
119. Losecaat Vermeer AB
120. Love BC
121. Mack ML
122. Malpica N
123. Marins T
124. Maumet C
125. McDonald K
126. McGuire JT
127. Melero H
128. Méndez Leal AS
129. Meyer B
130. Meyer KN
131. Mihai G
132. Mitsis GD
133. Moll J
134. Nielson DM
135. Nilsonne G
136. Notter MP
137. Olivetti E
138. Onicas AI
139. Papale P
140. Patil KR
141. Peelle JE
142. Pérez A
143. Pischedda D
144. Poline J-B
145. Prystauka Y
146. Ray S
147. Reuter-Lorenz PA
148. Reynolds RC
149. Ricciardi E
150. Rieck JR
151. Rodriguez-Thompson AM
152. Romyn A
153. Salo T
154. Samanez-Larkin GR
155. Sanz-Morales E
156. Schlichting ML
157. Schultz DH
158. Shen Q
159. Sheridan MA
160. Silvers JA
161. Skagerlund K
162. Smith A
163. Smith DV
164. Sokol-Hessner P
165. Steinkamp SR
166. Tashjian SM
167. Thirion B
168. Thorp JN
169. Tinghög G
170. Tisdall L
171. Tompson SH
172. Toro-Serey C
173. Torre Tresols JJ
174. Tozzi L
175. Truong V
176. Turella L
177. van ’t Veer AE
178. Verguts T
179. Vettel JM
180. Vijayarajah S
181. Vo K
182. Wall MB
183. Weeda WD
184. Weis S
185. White DJ
186. Wisniewski D
187. Xifra-Porxas A
188. Yearling EA
189. Yoon S
190. Yuan R
191. Yuen KSL
192. Zhang L
193. Zhang X
194. Zosky JE
195. Nichols TE
196. Poldrack RA
197. Schonberg T
(2020) Variability in the analysis of a single neuroimaging dataset by many teams
Nature 582:84–88.

https://doi.org/10.1038/s41586-020-2314-9
- PubMed
- Google Scholar
(2019) Exploring the impact of analysis software on task fMRI results
Human Brain Mapping 40:3362–3384.

https://doi.org/10.1002/hbm.24603
- PubMed
- Google Scholar
1. Brakewood B
2. Poldrack RA
(2013) The ethics of secondary data analysis: considering the application of Belmont principles to the sharing of neuroimaging data
NeuroImage 82:671–676.

https://doi.org/10.1016/j.neuroimage.2013.02.040
- PubMed
- Google Scholar
1. Casey BJ
2. Cannonier T
3. Conley MI
4. Cohen AO
5. Barch DM
6. Heitzeg MM
7. Soules ME
8. Teslovich T
9. Dellarco DV
10. Garavan H
11. Orr CA
12. Wager TD
13. Banich MT
14. Speer NK
15. Sutherland MT
16. Riedel MC
17. Dick AS
18. Bjork JM
19. Thomas KM
20. Chaarani B
21. Mejia MH
22. Hagler DJ
23. Daniela Cornejo M
24. Sicat CS
25. Harms MP
26. Dosenbach NUF
27. Rosenberg M
28. Earl E
29. Bartsch H
30. Watts R
31. Polimeni JR
32. Kuperman JM
33. Fair DA
34. Dale AM
35. ABCD Imaging Acquisition Workgroup
(2018) The Adolescent Brain Cognitive Development (ABCD) study: Imaging acquisition across 21 sites
Developmental Cognitive Neuroscience 32:43–54.

https://doi.org/10.1016/j.dcn.2018.03.001
- PubMed
- Google Scholar
1. Chang N
2. Pyles JA
3. Marcus A
4. Gupta A
5. Tarr MJ
6. Aminoff EM
(2019) BOLD5000, a public fMRI dataset while viewing 5000 visual images
Scientific Data 6:49.

https://doi.org/10.1038/s41597-019-0052-3
- PubMed
- Google Scholar
1. Chien HYS
2. Honey CJ
(2020) Constructing and Forgetting Temporal Context in the Human Cerebral Cortex
Neuron 106:675–686.

https://doi.org/10.1016/j.neuron.2020.02.013
- PubMed
- Google Scholar
(2020) Fine-grain atlases of functional modes for fMRI analysis
NeuroImage 221:117126.

https://doi.org/10.1016/j.neuroimage.2020.117126
- PubMed
- Google Scholar
1. Esteban O
2. Markiewicz CJ
3. Blair RW
4. Moodie CA
5. Isik AI
6. Erramuzpe A
7. Kent JD
8. Goncalves M
9. DuPre E
10. Snyder M
11. Oya H
12. Ghosh SS
13. Wright J
14. Durnez J
15. Poldrack RA
16. Gorgolewski KJ
(2019) fMRIPrep: a robust preprocessing pipeline for functional MRI
Nature Methods 16:111–116.

https://doi.org/10.1038/s41592-018-0235-4
- PubMed
- Google Scholar
1. Faskowitz J
2. Esfahlani FZ
3. Jo Y
4. Sporns O
5. Betzel RF
(2020) Edge-centric functional network representations of human cerebral cortex reveal overlapping system-level architecture
Nature Neuroscience 23:1644–1654.

https://doi.org/10.1038/s41593-020-00719-y
- Google Scholar
(2015) Task Dependence, Tissue Specificity, and Spatial Distribution of Widespread Activations in Large Single-Subject Functional MRI Datasets at 7T
Cerebral Cortex 25:4667–4677.

https://doi.org/10.1093/cercor/bhu148
- PubMed
- Google Scholar
1. Gordon EM
2. Laumann TO
3. Gilmore AW
4. Newbold DJ
5. Greene DJ
6. Berg JJ
7. Ortega M
8. Hoyt-Drazen C
9. Gratton C
10. Sun H
11. Hampton JM
12. Coalson RS
13. Nguyen AL
14. McDermott KB
15. Shimony JS
16. Snyder AZ
17. Schlaggar BL
18. Petersen SE
19. Nelson SM
20. Dosenbach NUF
(2017) Precision Functional Mapping of Individual Human Brains
Neuron 95:791–807.

https://doi.org/10.1016/j.neuron.2017.07.011
- PubMed
- Google Scholar
1. Gorgolewski KJ
2. Auer T
3. Calhoun VD
4. Craddock RC
5. Das S
6. Duff EP
7. Flandin G
8. Ghosh SS
9. Glatard T
10. Halchenko YO
11. Handwerker DA
12. Hanke M
13. Keator D
14. Li X
15. Michael Z
16. Maumet C
17. Nichols BN
18. Nichols TE
19. Pellman J
20. Poline JB
21. Rokem A
22. Schaefer G
23. Sochat V
24. Triplett W
25. Turner JA
26. Varoquaux G
27. Poldrack RA
(2016) The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments
Scientific Data 3:160044.

https://doi.org/10.1038/sdata.2016.44
- PubMed
- Google Scholar
1. Gorgolewski KJ
2. Alfaro-Almagro F
3. Auer T
4. Bellec P
5. Capotă M
6. Chakravarty MM
7. Churchill NW
8. Cohen AL
9. Craddock RC
10. Devenyi GA
11. Eklund A
12. Esteban O
13. Flandin G
14. Ghosh SS
15. Guntupalli JS
16. Jenkinson M
17. Keshavan A
18. Kiar G
19. Liem F
20. Raamana PR
21. Raffelt D
22. Steele CJ
23. Quirion PO
24. Smith RE
25. Strother SC
26. Varoquaux G
27. Wang Y
28. Yarkoni T
29. Poldrack RA
(2017) BIDS apps: Improving ease of use, accessibility, and reproducibility of neuroimaging data analysis methods
PLOS Computational Biology 13:e1005209.

https://doi.org/10.1371/journal.pcbi.1005209
- PubMed
- Google Scholar
Book
(2016)
DataLad--Decentralized Data Distribution for Consumption and Sharing of Scientific DatasetsOrganization of Human Brain Mapping Poster

Geneva, Switzerland: Organization of Human Brain Mapping Annual Meeting.
- Google Scholar
1. Halchenko Y
2. Meyer K
3. Poldrack B
4. Solanky D
5. Wagner A
6. Gors J
7. MacFarlane D
8. Pustina D
9. Sochat V
10. Ghosh S
11. Mönch C
12. Markiewicz C
13. Waite L
14. Shlyakhter I
15. de la Vega A
16. Hayashi S
17. Häusler C
18. Poline JB
19. Kadelka T
20. Skytén K
21. Jarecka D
22. Kennedy D
23. Strauss T
24. Cieslak M
25. Vavra P
26. Ioanas HI
27. Schneider R
28. Pflüger M
29. Haxby J
30. Eickhoff S
31. Hanke M
(2021) DataLad: distributed system for joint management of code, data, and their relationship
Journal of Open Source Software 6:3262.

https://doi.org/10.21105/joss.03262
- Google Scholar
(2021) In defense of decentralized research data management
Neuroforum 10:37.

https://doi.org/10.1515/nf-2020-0037
- Google Scholar
1. Henschel L
2. Conjeti S
3. Estrada S
4. Diers K
5. Fischl B
6. Reuter M
(2020) FastSurfer - A fast and accurate deep learning based neuroimaging pipeline
NeuroImage 219:117012.

https://doi.org/10.1016/j.neuroimage.2020.117012
- PubMed
- Google Scholar
1. Holdgraf C
2. Appelhoff S
3. Bickel S
4. Bouchard K
5. D’Ambrosio S
6. David O
7. Devinsky O
8. Dichter B
9. Flinker A
10. Foster BL
11. Gorgolewski KJ
12. Groen I
13. Groppe D
14. Gunduz A
15. Hamilton L
16. Honey CJ
17. Jas M
18. Knight R
19. Lachaux J-P
20. Lau JC
21. Lee-Messer C
22. Lundstrom BN
23. Miller KJ
24. Ojemann JG
25. Oostenveld R
26. Petridou N
27. Piantoni G
28. Pigorini A
29. Pouratian N
30. Ramsey NF
31. Stolk A
32. Swann NC
33. Tadel F
34. Voytek B
35. Wandell BA
36. Winawer J
37. Whitaker K
38. Zehl L
39. Hermes D
(2019) iEEG-BIDS, extending the Brain Imaging Data Structure specification to human intracranial electrophysiology
Scientific Data 6:102.

https://doi.org/10.1038/s41597-019-0105-7
- PubMed
- Google Scholar
Preprint
1. Jwa A
2. Poldrack R
(2021) The Spectrum of Data sharing Policies in Neuroimaging Data Repositories
PsyArXiv.

https://doi.org/10.31234/osf.io/cnuy7
- Google Scholar
(2019) Functional boundaries in the human cerebellum revealed by a multi-domain task battery
Nature Neuroscience 22:1371–1378.

https://doi.org/10.1038/s41593-019-0436-x
- PubMed
- Google Scholar
1. Koroshetz W
2. Gordon J
3. Adams A
4. Beckel-Mitchener A
5. Churchill J
6. Farber G
7. Freund M
8. Gnadt J
9. Hsu NS
10. Langhals N
11. Lisanby S
12. Liu G
13. Peng GCY
14. Ramos K
15. Steinmetz M
16. Talley E
17. White S
(2018) The State of the NIH BRAIN Initiative
The Journal of Neuroscience 38:6427–6438.

https://doi.org/10.1523/JNEUROSCI.3174-17.2018
- PubMed
- Google Scholar
1. Lin D
2. Crabtree J
3. Dillo I
4. Downs RR
5. Edmunds R
6. Giaretta D
7. De Giusti M
8. L’Hours H
9. Hugo W
10. Jenkyns R
11. Khodiyar V
12. Martone ME
13. Mokrane M
14. Navale V
15. Petters J
16. Sierman B
17. Sokolova DV
18. Stockhause M
19. Westbrook J
(2020) The TRUST Principles for digital repositories
Scientific Data 7:144.

https://doi.org/10.1038/s41597-020-0486-7
- PubMed
- Google Scholar
1. Littlejohns TJ
2. Holliday J
3. Gibson LM
4. Garratt S
5. Oesingmann N
6. Alfaro-Almagro F
7. Bell JD
8. Boultwood C
9. Collins R
10. Conroy MC
11. Crabtree N
12. Doherty N
13. Frangi AF
14. Harvey NC
15. Leeson P
16. Miller KL
17. Neubauer S
18. Petersen SE
19. Sellors J
20. Sheard S
21. Smith SM
22. Sudlow CLM
23. Matthews PM
24. Allen NE
(2020) The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions
Nature Communications 11:2624.

https://doi.org/10.1038/s41467-020-15948-9
- PubMed
- Google Scholar
Preprint
(2021) Transcriptional and Cellular Signatures of Cortical Morphometric Similarity Remodelling in Chronic Pain
bioRxiv.

https://doi.org/10.1101/2021.03.24.436777
- Google Scholar
1. Maumet C
2. Auer T
3. Bowring A
4. Chen G
5. Das S
6. Flandin G
7. Ghosh S
8. Glatard T
9. Gorgolewski KJ
10. Helmer KG
11. Jenkinson M
12. Keator DB
13. Nichols BN
14. Poline JB
15. Reynolds R
16. Sochat V
17. Turner J
18. Nichols TE
(2016) Sharing brain mapping statistical results with the neuroimaging data model
Scientific Data 3:160102.

https://doi.org/10.1038/sdata.2016.102
- PubMed
- Google Scholar
(2013) Making data sharing work: the FCP/INDI experience
NeuroImage 82:683–691.

https://doi.org/10.1016/j.neuroimage.2012.10.064
- PubMed
- Google Scholar
1. Milham MP
2. Craddock RC
3. Son JJ
4. Fleischmann M
5. Clucas J
6. Xu H
7. Koo B
8. Krishnakumar A
9. Biswal BB
10. Castellanos FX
11. Colcombe S
12. Di Martino A
13. Zuo XN
14. Klein A
(2018) Assessment of the impact of shared brain imaging data on the scientific literature
Nature Communications 9:2818.

https://doi.org/10.1038/s41467-018-04976-1
- PubMed
- Google Scholar
1. Musen MA
2. Bean CA
3. Cheung KH
4. Dumontier M
5. Durante KA
6. Gevaert O
7. Gonzalez-Beltran A
8. Khatri P
9. Kleinstein SH
10. O’Connor MJ
11. Pouliot Y
12. Rocca-Serra P
13. Sansone SA
14. Wiser JA
15. CEDAR team
(2015) The center for expanded data annotation and retrieval
Journal of the American Medical Informatics Association 22:1148–1152.

https://doi.org/10.1093/jamia/ocv048
- PubMed
- Google Scholar
Book
1. National Academies of Sciences, Engineering, and Medicine, Policy and Global Affairs, Board on Research Data and Information, Committee on Toward an Open Science Enterprise
(2018)
Open Science by Design: Realizing a Vision for 21st Century Research

National Academies Press.
- Google Scholar
1. Newbold DJ
2. Laumann TO
3. Hoyt CR
4. Hampton JM
5. Montez DF
6. Raut RV
7. Ortega M
8. Mitra A
9. Nielsen AN
10. Miller DB
11. Adeyemo B
12. Nguyen AL
13. Scheidter KM
14. Tanenbaum AB
15. Van AN
16. Marek S
17. Schlaggar BL
18. Carter AR
19. Greene DJ
20. Gordon EM
21. Raichle ME
22. Petersen SE
23. Snyder AZ
24. Dosenbach NUF
(2020) Plasticity and Spontaneous Activity Pulses in Disused Human Brain Circuits
Neuron 107:580–589.

https://doi.org/10.1016/j.neuron.2020.05.007
- PubMed
- Google Scholar
1. Niso G
2. Gorgolewski KJ
3. Bock E
4. Brooks TL
5. Flandin G
6. Gramfort A
7. Henson RN
8. Jas M
9. Litvak V
10. T Moreau J
11. Oostenveld R
12. Schoffelen J-M
13. Tadel F
14. Wexler J
15. Baillet S
(2018) MEG-BIDS, the brain imaging data structure extended to magnetoencephalography
Scientific Data 5:180110.

https://doi.org/10.1038/sdata.2018.110
- PubMed
- Google Scholar
1. Nooner KB
2. Colcombe SJ
3. Tobe RH
4. Mennes M
5. Benedict MM
6. Moreno AL
7. Panek LJ
8. Brown S
9. Zavitz ST
10. Li Q
11. Sikka S
12. Gutman D
13. Bangaru S
14. Schlachter RT
15. Kamiel SM
16. Anwar AR
17. Hinz CM
18. Kaplan MS
19. Rachlin AB
20. Adelsberg S
21. Cheung B
22. Khanuja R
23. Yan C
24. Craddock CC
25. Calhoun V
26. Courtney W
27. King M
28. Wood D
29. Cox CL
30. Kelly AMC
31. Di Martino A
32. Petkova E
33. Reiss PT
34. Duan N
35. Thomsen D
36. Biswal B
37. Coffey B
38. Hoptman MJ
39. Javitt DC
40. Pomara N
41. Sidtis JJ
42. Koplewicz HS
43. Castellanos FX
44. Leventhal BL
45. Milham MP
(2012) The NKI-Rockland Sample: A Model for Accelerating the Pace of Discovery Science in Psychiatry
Frontiers in Neuroscience 6:152.

https://doi.org/10.3389/fnins.2012.00152
- PubMed
- Google Scholar
Preprint
(2021) PET-BIDS, an extension to the brain imaging data structure for positron emission tomography
bioRxiv.

https://doi.org/10.1101/2021.06.16.448390
- Google Scholar
(2019) EEG-BIDS, an extension to the brain imaging data structure for electroencephalography
Scientific Data 6:103.

https://doi.org/10.1038/s41597-019-0104-8
- PubMed
- Google Scholar
1. Pinho A.L
2. Amadon A
3. Ruest T
4. Fabre M
5. Dohmatob E
6. Denghien I
7. Ginisty C
8. Becuwe-Desmidt S
9. Roger S
10. Laurier L
11. Joly-Testault V
12. Médiouni-Cloarec G
13. Doublé C
14. Martins B
15. Pinel P
16. Eger E
17. Varoquaux G
18. Pallier C
19. Dehaene S
20. Hertz-Pannier L
21. Thirion B
(2018) Individual Brain Charting, a high-resolution fMRI dataset for cognitive mapping
Scientific Data 5:180105.

https://doi.org/10.1038/sdata.2018.105
- PubMed
- Google Scholar
1. Pinho A
2. Amadon A
3. Gauthier B
4. Clairis N
5. Knops A
6. Genon S
7. Dohmatob E
8. Torre JJ
9. Ginisty C
10. Becuwe-Desmidt S
11. Roger S
12. Lecomte Y
13. Berland V
14. Laurier L
15. Joly-Testault V
16. Médiouni-Cloarec G
17. Doublé C
18. Martins B
19. Salmon E
20. Piazza M
21. Melcher D
22. Pessiglione M
23. van Wassenhove V
24. Eger E
25. Varoquaux G
26. Dehaene S
27. Hertz-Pannier L
28. Thirion B
(2020) Individual Brain Charting dataset extension, second release of high-resolution fMRI data for cognitive mapping
Scientific Data 7:353.

https://doi.org/10.1038/s41597-020-00670-4
- PubMed
- Google Scholar
1. Poldrack RA
2. Kittur A
3. Kalar D
4. Miller E
5. Seppa C
6. Gil Y
7. Parker DS
8. Sabb FW
9. Bilder RM
(2011) The cognitive atlas: toward a knowledge foundation for cognitive neuroscience
Frontiers in Neuroinformatics 5:17.

https://doi.org/10.3389/fninf.2011.00017
- PubMed
- Google Scholar
1. Poldrack RA
2. Barch DM
3. Mitchell JP
4. Wager TD
5. Wagner AD
6. Devlin JT
7. Cumba C
8. Koyejo O
9. Milham MP
(2013) Toward open sharing of task-based fMRI data: the OpenfMRI project
Frontiers in Neuroinformatics 7:12.

https://doi.org/10.3389/fninf.2013.00012
- PubMed
- Google Scholar
1. Poldrack RA
2. Gorgolewski KJ
(2014) Making big data open: data sharing in neuroimaging
Nature Neuroscience 17:1510–1517.

https://doi.org/10.1038/nn.3818
- PubMed
- Google Scholar
1. Poldrack RA
2. Laumann TO
3. Koyejo O
4. Gregory B
5. Hover A
6. Chen MY
7. Gorgolewski KJ
8. Luci J
9. Joo SJ
10. Boyd RL
11. Hunicke-Smith S
12. Simpson ZB
13. Caven T
14. Sochat V
15. Shine JM
16. Gordon E
17. Snyder AZ
18. Adeyemo B
19. Petersen SE
20. Glahn DC
21. Reese Mckay D
22. Curran JE
23. Göring HHH
24. Carless MA
25. Blangero J
26. Dougherty R
27. Leemans A
28. Handwerker DA
29. Frick L
30. Marcotte EM
31. Mumford JA
(2015) Long-term neural and physiological phenotyping of a single human
Nature Communications 6:8885.

https://doi.org/10.1038/ncomms9885
- PubMed
- Google Scholar
1. Poldrack RA
2. Congdon E
3. Triplett W
4. Gorgolewski KJ
5. Karlsgodt KH
6. Mumford JA
7. Sabb FW
8. Freimer NB
9. London ED
10. Cannon TD
11. Bilder RM
(2016) A phenome-wide examination of neural and cognitive function
Scientific Data 3:160110.

https://doi.org/10.1038/sdata.2016.110
- PubMed
- Google Scholar
1. Poline JB
2. Breeze JL
3. Ghosh S
4. Gorgolewski K
5. Halchenko YO
6. Hanke M
7. Haselgrove C
8. Helmer KG
9. Keator DB
10. Marcus DS
11. Poldrack RA
12. Schwartz Y
13. Ashburner J
14. Kennedy DN
(2012) Data sharing in neuroimaging research
Frontiers in Neuroinformatics 6:.

https://doi.org/10.3389/fninf.2012.00009
- Google Scholar
1. Rosenberg MD
2. Scheinost D
3. Greene AS
4. Avery EW
5. Kwon YH
6. Finn ES
7. Ramani R
8. Qiu M
9. Constable RT
10. Chun MM
(2020) Functional connectivity predicts changes in attention observed across minutes, days, and months
PNAS 117:3797–3807.

https://doi.org/10.1073/pnas.1912226117
- PubMed
- Google Scholar
1. Salehi M
2. Greene AS
3. Karbasi A
4. Shen X
5. Scheinost D
6. Constable RT
(2020) There is no single functional atlas even for a single individual: Functional parcel definitions change with task
NeuroImage 208:116366.

https://doi.org/10.1016/j.neuroimage.2019.116366
- PubMed
- Google Scholar
1. Schwarz CG
2. Kremers WK
3. Therneau TM
4. Sharp RR
5. Gunter JL
6. Vemuri P
7. Arani A
8. Spychalla AJ
9. Kantarci K
10. Knopman DS
11. Petersen RC
(2019) Identification of Anonymous MRI Research Participants with Face-Recognition Software
The New England Journal of Medicine 381:1684–1686.

https://doi.org/10.1056/NEJMc1908881
- PubMed
- Google Scholar
1. Smith SM
2. Nichols TE
(2018) Statistical Challenges in “Big Data” Human Neuroimaging
Neuron 97:263–268.

https://doi.org/10.1016/j.neuron.2017.12.018
- PubMed
- Google Scholar
1. Spisák T
2. Spisák Z
3. Zunhammer M
4. Bingel U
5. Smith S
6. Nichols T
7. Kincses T
(2019) Probabilistic TFCE: A generalized combination of cluster size and voxel intensity to increase statistical power
NeuroImage 185:12–26.

https://doi.org/10.1016/j.neuroimage.2018.09.078
- PubMed
- Google Scholar
1. States U
(1978)
The Belmont Report: Ethical Principles and Guidelines for the Protection of Human Subjects of Research

National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, The Belmont Report: Ethical Principles and Guidelines for the Protection of Human Subjects of Research, The Commission.
- Google Scholar
1. Tadel F
2. Bock E
3. Niso G
4. Mosher JC
5. Cousineau M
6. Pantazis D
7. Leahy RM
8. Baillet S
(2019) MEG/EEG Group Analysis With Brainstorm
Frontiers in Neuroscience 13:76.

https://doi.org/10.3389/fnins.2019.00076
- PubMed
- Google Scholar
(2019) MEG Source Imaging and Group Analysis Using VBMEG
Frontiers in Neuroscience 13:241.

https://doi.org/10.3389/fnins.2019.00241
- PubMed
- Google Scholar
(2013) The WU-Minn Human Connectome Project: an overview
NeuroImage 80:62–79.

https://doi.org/10.1016/j.neuroimage.2013.05.041
- PubMed
- Google Scholar
1. Van Horn JD
2. Gazzaniga MS
(2013) Why share data? Lessons learned from the fMRIDC
NeuroImage 82:677–682.

https://doi.org/10.1016/j.neuroimage.2012.11.010
- PubMed
- Google Scholar
1. Wakeman DG
2. Henson RN
(2015) A multi-subject, multi-modal human neuroimaging dataset
Scientific Data 2:150001.

https://doi.org/10.1038/sdata.2015.1
- PubMed
- Google Scholar
1. Wilkinson MD
2. Dumontier M
3. Aalbersberg IJJ
4. Appleton G
5. Axton M
6. Baak A
7. Blomberg N
8. Boiten J-W
9. da Silva Santos LB
10. Bourne PE
11. Bouwman J
12. Brookes AJ
13. Clark T
14. Crosas M
15. Dillo I
16. Dumon O
17. Edmunds S
18. Evelo CT
19. Finkers R
20. Gonzalez-Beltran A
21. Gray AJG
22. Groth P
23. Goble C
24. Grethe JS
25. Heringa J
26. ’t Hoen PAC
27. Hooft R
28. Kuhn T
29. Kok R
30. Kok J
31. Lusher SJ
32. Martone ME
33. Mons A
34. Packer AL
35. Persson B
36. Rocca-Serra P
37. Roos M
38. van Schaik R
39. Sansone S-A
40. Schultes E
41. Sengstag T
42. Slater T
43. Strawn G
44. Swertz MA
45. Thompson M
46. van der Lei J
47. van Mulligen E
48. Velterop J
49. Waagmeester A
50. Wittenburg P
51. Wolstencroft K
52. Zhao J
53. Mons B
(2016) The FAIR Guiding Principles for scientific data management and stewardship
Scientific Data 3:160018.

https://doi.org/10.1038/sdata.2016.18
- PubMed
- Google Scholar
1. Yarkoni T
2. Markiewicz CJ
3. de la Vega A
4. Gorgolewski KJ
5. Salo T
6. Halchenko YO
7. McNamara Q
8. DeStasio K
9. Poline JB
10. Petrov D
11. Hayot-Sasson V
12. Nielson DM
13. Carlin J
14. Kiar G
15. Whitaker K
16. DuPre E
17. Wagner A
18. Tirrell LS
19. Jas M
20. Hanke M
21. Poldrack RA
22. Esteban O
23. Appelhoff S
24. Holdgraf C
25. Staden I
26. Thirion B
27. Kleinschmidt DF
28. Lee JA
29. Notter MP
30. Blair R
(2019) PyBIDS: Python tools for BIDS datasets
Journal of Open Source Software 4:1294.

https://doi.org/10.21105/joss.01294
- PubMed
- Google Scholar

Article and author information

Author details

Christopher J Markiewicz

Department of Psychology, Stanford University, Stanford, United States

Contribution
Conceptualization, Project administration, Writing - original draft

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-6533-164X
Krzysztof J Gorgolewski

Department of Psychology, Stanford University, Stanford, United States

Contribution
Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Software

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0003-3321-7583
Franklin Feingold

Department of Psychology, Stanford University, Stanford, United States

Contribution
Project administration, Supervision, Software

Competing interests
No competing interests declared
Ross Blair

Department of Psychology, Stanford University, Stanford, United States

Contribution
Data curation, Software

Competing interests
No competing interests declared
Yaroslav O Halchenko

Department of Psychological & Brain Sciences, Dartmouth College, Hanover, United States

Contribution
Conceptualization, Data curation, Software

Competing interests
No competing interests declared
Eric Miller

Squishymedia, Portland, United States

Contribution
Conceptualization, Project administration, Data curation, Software

Competing interests
is owner of Squishymedia which is funded to perform software development work on OpenNeuro.
Nell Hardcastle

Squishymedia, Portland, United States

Contribution
Data curation, Software

Competing interests
is an employee of Squishymedia which is funded to perform software development work on OpenNeuro.

"This ORCID iD identifies the author of this article:" 0000-0002-3837-0707
Joe Wexler

Department of Psychology, Stanford University, Stanford, United States

Contribution
Investigation, Software

Competing interests
No competing interests declared
Oscar Esteban
1. Department of Psychology, Stanford University, Stanford, United States
2. Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland
Contribution
Data curation, Software

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-8435-6191
Mathias Goncavles

Department of Psychology, Stanford University, Stanford, United States

Contribution
Data curation, Software

Competing interests
No competing interests declared
Anita Jwa

Department of Psychology, Stanford University, Stanford, United States

Contribution
Formal analysis, Software

Competing interests
No competing interests declared
Russell Poldrack

Department of Psychology, Stanford University, Stanford, United States

Contribution
Conceptualization, Writing – review and editing, Funding acquisition, Methodology, Project administration, Data curation, Supervision, Writing - original draft, Software

For correspondence
russpold@stanford.edu

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-6755-0259

Funding

National Institute of Mental Health (R24MH117179)

Russell Poldrack

National Institute of Mental Health (R24MH114705)

Russell Poldrack

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The work described here has been supported by the National Institute of Mental Health of the National Institutes of Health under Award Numbers R24MH117179 and R24MH114705. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Development of OpenNeuro and OpenfMRI was also supported by a grant from the Laura and John Arnold Foundation, and the National Science Foundation (OAC-1131441). Sharing of OpenNeuro datasets has been enabled by support from AWS. We would like to thank all of the users who have uploaded data to OpenNeuro. Thanks to Franco Pestili for providing usage data on Brainlife.io, and Nico Dosenbach, Michael Hawrylycz, Karel Svoboda, and Armin Thomas for helpful comments on an earlier draft.

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.