NeuroVault Public Data Archive

This repository contains archived data from NeuroVault.org.

The data has been dumped from a subset production database concerned with the "statsmaps" application, and censoring has been performed to remove user related information.

Tables

Within the statsmaps subset, the following tables are included:

Collections

• statmaps_collection This is the main table that indexes NeuroVault Collections. Critically the id column is the primary key for the table, and is used to link to other tables (refered to in other tables as collection_id)
• statmaps_collection_contributors Association of collections with user_id

Collection Items / Images

• statmaps_basecollectionitem This is the main table that indexes NeuroVault Collection Items. Critically the id column is the primary key for the table, and is used to link to other tables as basecollectionitem_ptr_id
• statmaps_image Images are a type of Collection Item. This table contains the image specific information. The column basecollectionitem_ptr_id in this table corresponds to statmaps_basecollectionitem.id
• statmaps_statisticmap StatisticMaps are a type of Images, with additional meta-data, such as smoothness_fwhm, cognitive_paradigm_cogatlas_id. The column image_ptr_id' in this table corresponds to statmaps_basecollectionitem.id & statmaps_image.basecollectionitem_ptr_id
• statmaps_atlas Atlases are a type of Images, with an additional reference to label_description_file. the column image_ptr_id refers to statmaps_basecollectionitem.id & statmaps_image.basecollectionitem_ptr_id.

Cogntive Atlas

• statmaps_cognitiveatlastask Mapping of cog_atlas_id (referenced in statmaps_statisticmap) to task names
• statmaps_cognitiveatlascontrast Mapping of cog_atlas_id referenced in statmaps_statisticmap) to contrast names

Other tables

• statmaps_collection_communities Grouping of collections into "communities".
• statmaps_communities Listing of communities. This is specific to two communities, "Developmental Neuroscience" and "Nutritional Neuroscience"

Importing into a SQLite database

In order to quickly explore the contents of these CSV files, it may be convenient to import them into an SQLite database. The 'create_db.sh' script creates such a database (with the primary and foreign key constraints mentioned above) and fills it. It can be used like this:

tar xzf november_2022.tar.gz
cd november_2022
../create_db.sh

sqlite3 neurovault.sqlite3

Gotchas when loading these CSVs

Two properties of the data that will silently distort an analysis:

• The literal string "None" is a real primary key. It belongs to the statmaps_cognitiveatlastask row named "None / Other". Every cognitive_paradigm_cogatlas_id in the dump joins to that table, with no orphans, so "None" records a genuine user selection rather than a missing value. Pandas' default na_values converts it to NaN, merging "the user chose None/Other" with "the user answered nothing" — in february_2024 that is 301,879 rows versus 75,452. Read with keep_default_na=False and test for blank strings explicitly.

• Owner 5761 is the Neuroscout mass-upload account, a bulk push from a separate platform: 3,889 collections and 289,248 statistic maps in february_2024, or 51.7% of all maps. Most analyses of user-contributed content should exclude it.

Also note that statmaps_collection.description contains newlines, so line counts do not equal row counts — use a real CSV parser.

Analyses built on these dumps live in the companion nv-audit repository.