Developing formats for sharing electrophysiology data using as a testbed

Jeff Teeters (UC Berkeley), Friedrich Sommer (UC Berkeley)

The website hosts a repository of electrophysiology data useful for analysis in computational neuroscience.  Multiple laboratories have contributed data.  Each laboratory contributed data in the format used in their lab, and they all used different formats.

These differences in data format present a barrier to efficient sharing of data because it requires data users to learn a new storage scheme for every data set.  Different formats also preclude the development of tools that would work with multiple data sets, for example, to do meta-analysis or online visualization.

A number of projects are developing techniques to store electrophysiology data in a way to facilitate data sharing, including: BrainML, SignalML, NDF, Neuroshare and MIEN. Of these, Neuroshare, NDF and MIEN most closely fulfill the requirements of  These three projects are similar in that they follow a paradigm in which recorded data is stored using predefined formats for different types of data. The types of data include: time series (continuously recorded analog signals), events, neural events (spike times), segments (short sections of continuously recorded signals) and histograms (include in MIEN).

We are investigating to what extent the data contributed to can be converted into a common storage format that follows this paradigm.  So far, we have concluded that:

  1. The paradigm is sufficient to allow encoding most of the data, but for at least one data set, ad hoc supplementary information will be required.
  2. Stimulus information, which is essential to allow interpreting data, can be incorporated using the event data type.  But the details of how that is done needs to be developed for each type of stimulus.
  3. To make using the data easier, an API to allow accessing the data should be developed.  Neuroshare and the NEO may serve as a guide for this.
  4. Using HDF5 as the storage container allows storing metadata in the same file as data and online visualization of data using the OPeNDAP protocol.

Web resources:  – Collaborative Research in Computational Neuroscience - Data Sharing
BrainML -
SignalML -
Neuroshare -
HDF5 -

Preferred presentation format: Poster
Topic: Infrastructural and portal services

Document Actions