83 — Neurodata Without Borders: Creating a Common Data Format for Neurophysiology

This paper announces the release of the NWB format specification, a data standardization proposed by the Neurodata Without Borders group. NWB files attempt to unify the disparate datatypes in the field of functional neurophysiology: The first page of the paper’s text lists just a small selection of more than ten competing formats in this space.

So why is this not a case of #927?

Based on the Allen Institute’s Orca data format, the NWB data model attempted to more closely model the required features of a neuroscience data format while still generalizing to accomodate disparate recording types (for example, multi-cell and whole-cell recording methods are both supported by the NWB data model.)

NWB files, openable with HDF browsers such as HDFView, provide a consistent API for different “extensions” of the specification, which allows the namespacing of content so that multiple projects can implement the spec as they deem appropriate.

One frustrating issue I find with the NWB spec is that it entirely ignores anatomical data as a first-class citizen; because so much of my work focuses on 3D or multichannel-3D imagery, the NWB format is mostly not useful to me. (The overlay of functional data and anatomical data is interesting and has a lot of promise, but in particlar, sharing data in this format isn’t as efficient as, say, storing the image volumes in a conventional TIFF stack, compressed numpy volume, or HDF5 file.)

Where I do see this format becoming very useful is in providing a strategy for unifying the neuroscience community; if the electrophys community adopts the NWB format entirely, I could see NWB becoming a data-standard authority, and the unification of other data-types such as anatomical datasets could be simplified by looking to NWB for insight.

The 2.0 of this specification was just released at SfN 2017; I’m looking forward to seeing what’s new in the spec!