BrainLiner: A Platform for Sharing and Searching Time-aligned Neural and Behavioral Data

Neuroscientists have used machine learning techniques in recent years to explore the representation of information in the brain to much success. Neural decoding models, for instance, have been developed that can predict behavioral parameters, stimuli, and mental states from measured brain activity. Unraveling more sophisticated neural representation or training generic statistical models will require large amounts of data across multiple subjects.

To remedy the need for easy access to large data sets, we created BrainLiner.jp as part of the Japanese Strategic Research Program for Brain Science (SRPBS), as a web portal for searching for and sharing neurophysiological data. Real-time text-based search allows visitors to the site to find what they are looking for within seconds. We are also implementing a prototype data-driven search that quantizes time-series data for fast performance.

Data files contain aligned neural and behavioral signal time series, and can be shared on the portal in the Neuroshare file format. Neuroshare files are then converted to a standard Matlab format on our servers and users can then choose to download either Neuroshare or Matlab files. The use of a standard file structure for all data sets enables efficient data sharing, since once a user learns the data format for one project, they can easily use data from another project, since they are all the same. This also enables automated analyses of data files across projects.

Uploaded files are by default licenses under the Creative Commons CC-BY license, though uploading users can choose between using a CC-0 license or writing a custom license. This allows users to freely share their data in a readily comprehensible way.

Additionally, as part of our project we have released open-source versions of our Neuroshare conversion tools [1], which include Matlab and Java-based reader and writers for Neuroshare files.

Since neuroscientists do not perform their work alone in a vacuum, we added integration with Google+, Facebook, and Twitter, so users can express interest in and share data sets that are useful. Users can also log in to our site using their Google, Facebook, or Twitter accounts, in order to share data; no registration required.

[1] http://www.cns.atr.jp/[…]/