MEF3 tools#

The Multiscale Electrophysiology Format (MEF 3.0) is a compressed, optionally encrypted file format designed for large-scale, long-term electrophysiology recordings. BNEL develops and maintains an open toolchain for working with MEF 3.0 data, from low-level reading and writing to high-throughput streaming for viewers and data-processing applications, and a desktop app for working with recordings without programming.

mef3io

The current library for reading and writing MEF 3.0 data from Python, MATLAB and C++, and for de-identifying it from Python and C++. Replaces mef_tools.

https://bnelair.github.io/mef3io/
mef3io-server

A gRPC server and client for efficient, concurrent streaming of MEF3 data, with caching and prefetch for viewers and data-processing applications.

https://bnelair.github.io/mef3io-server/
mef3io-studio (pre-release)

Desktop app for exploring and anonymizing MEF 3.0 recordings, built on mef3io (Windows, macOS, Linux).

https://github.com/bnelair/mef3io-studio
mef3io-coregistration

Read two MEF3 recordings that were coregistered (aligned onto one clock) as if they were one recording. Built on mef3io.

https://github.com/bnelair/mef3io-coregistration

mef3io#

mef3io is the recommended library for MEF 3.0 data. It reads and writes MEF 3.0 recordings (EEG, iEEG, ECG and other regularly sampled signals) from Python, MATLAB and C++, with one C++ core shared by all three, and makes de-identified copies from Python and C++ (from MATLAB, through python -m mef3io anonymize in a terminal). Sessions written by mef3io open in the established MEF 3.0 tools (meflib, pymef, mef_tools) and vice versa. It is the successor to mef_tools and the starting point for new projects that work with MEF 3.0 files.

mef3io-server#

mef3io-server is a gRPC server and client for efficient, concurrent access to MEF3 electrophysiology data. Every data call is addressed by channels and time. It is backed by a shared per-channel cache, parallel decoding across worker processes (built on mef3io), and configurable prefetch, and is aimed at scenarios that need fast, repeated access to large recordings, in particular:

  • pre-caching in viewers, so that scrolling and paging through long recordings stays responsive, and

  • data-processing applications, where many workers stream overlapping segments concurrently.

It can also run in Docker.

mef3io-studio#

Note

Pre-release. mef3io-studio is in early development. Only test versions are available so far, and the builds are not code-signed yet.

mef3io-studio is a desktop app for exploring and anonymizing MEF 3.0 recordings, with no programming needed. It runs on Windows, macOS and Linux and is built on mef3io, so every change it makes to a file is done by the same code that the Python and MATLAB libraries use. There is no pip package: download the installer for your system from the GitHub Releases page.

mef3io-coregistration#

mef3io-coregistration reads two MEF3 recordings that were coregistered (aligned onto one clock) as if they were one recording. It is only the reader: it opens a saved coregistration result (*.coreg.json) and reads the recordings through it. It is built on mef3io.

Deprecated and legacy#

Warning

mef_tools is deprecated and kept as an archive for legacy support only. It is no longer actively developed. New projects should use mef3io (and mef3io-server where streaming is needed) instead.

mef_tools was the original high-level BNEL library for reading and writing MEF3 files. It builds on the upstream pymef/meflib stack. Existing code that depends on it will continue to work, but it will not receive new features. mef3io provides drop-in classes for it (from mef3io import MefReader, MefWriter). The components below are listed for legacy context: