NeuroVault.org: a web-based repository for collecting and sharing unthresholded statistical maps of the human brain.
Gorgolewski KJ, Varoquaux G, Rivera G, Schwarz Y, Ghosh SS, Maumet C, Sochat VV, Nichols TE, Poldrack RA, Poline JB, Yarkoni T, Margulies DS. · 2015
Read NeuroVault.org: a web-based repository for collecting and sharing unthresholded statistical maps of the human brain. by Gorgolewski KJ, Varoquaux G, Rivera G, Schwarz Y, Ghosh SS, Maumet C, Sochat VV, Nichols TE, Poldrack RA, Poline JB, Yarkoni T, Margulies DS. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4392315.txt.
Introduction
Krzysztof Gorgolewski and a team of co-authors launched NeuroVault, a web home for brain maps, in Frontiers in Neuroinformatics in 2015. Brain imaging studies produce statistical maps showing where activity or structure differs, yet papers usually print only thresholded pictures, and the full maps are rarely shared. NeuroVault lets researchers store, share, view, and decode unthresholded maps in a browser, with no extra software to install.
Deposited maps are linked to the Neurosynth database, which can decode a map into the mental terms it most resembles. A public programming interface exposes the collection so other tools can build on it. The resource is aimed at meta-analysis and co-activation studies, where many maps must be compared at once. Sharing the full map, not just its peaks, makes later synthesis far more faithful.
Catalog record
| CREATOR | Gorgolewski KJ, Varoquaux G, Rivera G, Schwarz Y, Ghosh SS, Maumet C, Sochat VV, Nichols TE, Poldrack RA, Poline JB, Yarkoni T, Margulies DS. |
|---|---|
| YEAR | 2015 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
NeuroVault.org: a web-based repository for collecting and sharing unthresholded statistical maps of the human brain By Krzysztof J Gorgolewski, Gael Varoquaux, Gabriel Rivera, Yannick Schwarz, Satrajit S Ghosh, Camille Maumet, Vanessa V Sochat, Thomas E Nichols, Russell A Poldrack, Jean-Baptiste Poline, Tal Yarkoni, Daniel S Margulies Published in: Frontiers in Neuroinformatics DOI: 10.3389/fninf.2015.00008 ABSTRACT Abstract Here we present NeuroVault—a web based repository that allows researchers to store, share, visualize, and decode statistical maps of the human brain. NeuroVault is easy to use and employs modern web technologies to provide informative visualization of data without the need to install additional software. In addition, it leverages the power of the Neurosynth database to provide cognitive decoding of deposited maps. The data are exposed through a public REST API enabling other services and tools to take advantage of it. NeuroVault is a new resource for researchers interested in conducting meta- and coactivation analyses. Keywords: data sharing, statistical parameter mapping (SPM), meta-analysis, repository, database INTRODUCTION Non-invasive neuroimaging techniques such as MRI and PET have enabled unprecedented insight into the localization of various functions in the human brain. As the number of studies using such techniques continues to grow exponentially, the challenge of assessing, summarizing, and condensing their findings poses ever-greater difficulty. Even though a single study can take years to conduct, cost hundreds of thousands of dollars, and require the effort of dozens of highly trained scientists and volunteers, the output is usually reduced to an academic article, and the original data are rarely shared (Poline et al., 2012). Unfortunately, due to the historical legacy of reporting knowledge in written form (of an academic paper), the final documented results consist mostly of subjective interpretation of data with very little machine-readable information. While the introduction of common stereotaxic spaces (e.g., Talairach and MNI305) has provided an initial framework for a standard of reporting activation locations to subsequently enable meta-analyses, there are several issues with this coordinate-based strategies. First, peak coordinates are not able to fully describe the 3D shape and extent of a suprathreshold volume on a statistical map. Many papers use figures (2D or 3D) to present these statistical maps, but authors must decide which aspects of the 3D data cube to show. To fully explore all layers of the data one would need to be able to interrogate it in an interactive fashion. Furthermore, published figures are not machine-readable, and researchers that are interested in comparing their own results with published literature are forced to manually reconstruct regions of interest (ROIs) using spheres placed at the limited reported activation locations. A second issue is the difficulty of putting one's results in the context of other studies. The overwhelming number of brain imaging results published each year makes manual comparison both unfeasible and prone to bias. There are attempts to automatically aggregate knowledge across large sets of neuroimaging studies. For example, Neurosynth (Yarkoni et al., 2011) is a meta-analysis database that collects coordinates of activation foci from published papers and generates topic maps based on the spatial distribution of those coordinates. Such maps can aid in interpretation of new results. However, comparing a new result to a set of topic maps has so far not been implemented in a user-friendly fashion. Finally, and most importantly, making meta-analytic inferences using only peak coordinates (or statistically thresholded maps) is problematic. It is easy to imagine a subthreshold effect that is consistent across many studies. Such an effect would not be picked up by existing meta-analysis methods (Laird et al., 2005; Yarkoni et al., 2011) because it would never be reported in the tables of peak coordinates. Considering how underpowered most human neuroscience studies are, this situation is not that unlikely. Discarding information that is below threshold in this fashion is akin to not publishing null results (Rosenthal, 1979), a dangerous practice that creates a publication bias skewing our perception of accumulated knowledge. Using fully unthresholded statistical maps instead of solely peak coordinates would provide a significant advance in meta-analytic power. Coordinate-based meta-analysis (CBMA) methods show only modest overlap with image-based meta-analysis (IBMA; meta-analysis based on unthresholded statistical maps) methods and are less powerful (Salimi-Khorshidi et al., 2009). However, IBMA methods struggle with access to the data. Peak coordinates are easier to obtain and share because coordinate tables are an integral component of traditional neuroimaging papers, whereas very few papers provide links to unthresholded statistical maps (usually by an ad hoc means such as the author's web site). NeuroVault.org is an attempt to solve these problems. It is a web-based repository that makes it easy to deposit and share statistical maps. It provides attractive visualization and cognitive decoding of the maps that can improve collaboration efforts and readability of the results. At the same time, it also provides an API for methods researchers to download the data, perform powerful analyses, or build new tools. RESULTS In the following section we describe the architecture and features of NeuroVault and present two example analyses. Platform One of the key features of NeuroVault is the ease of uploading and sharing statistical brain maps. Figure 1 presents a schematic overview of the platform. After logging in, users can upload a broad range of neuroimaging images and associated metadata. These data are then immediately accessible (subject to user-controlled privacy settings) via both an interactive HTML-based interface, and a comprehensive RESTful web API that facilitates programmatic interoperability with other resources. In the following sections, we discuss different aspects of the platform. [Figure: Schematic overview of the NeuroVault platform. To begin working with NeuroVault, users are asked to create an account or log in using their Facebook or Google account. After login, the user creates a collection (representing a paper or a study). At this stage, users can provide a DOI pointing to a paper associated with the collection and/or fill in a number of fields describing the study (see Supplementary Table 1 for details). This additional information is, however, optional. After the collection is created, users can add images. This can be done one-by-one or in bulk by uploading whole folders. Again, there is an option to add more metadata describing the images. The process of creating a collection and uploading statistical maps to NeuroVault takes only 5–10 min. When the maps are uploaded, users can start benefiting from permanent link to their results, interactive web-based visualization, and real-time image decoding.] Image upload The NeuroVault upload process emphasizes speed and ease of use. Users can rely on existing social media accounts (Google or Facebook) to log in, and can upload individual images, or entire folders (see Figure 1). Users can arrange their maps into collections or to group them with tags. Each collection and statistical image in NeuroVault gets a permanent link (URL) that can be shared with other researchers or included in papers or other forms of publication (blogs, tweets, etc…). Users can specify whether each collection is public or private. The latter have a unique obfuscated URL that is not discoverable on the NeuroVault website, and thus are accessible only by whomever the owner decides to share the URL with. The option of creating private collections gives users freedom to decide who can access their data, and can facilitate a scenario