| title | tags | authors | affiliations | date | bibliography | |||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
neurone_loader: A Python module for loading EEG data recorded with Bittium NeurOne |
|
|
|
05 July 2019 |
paper.bib |
In the neuroimaging domain, most imaging equipment comes with its own binary format to store raw imaging data. It has always been a challenge for researchers to handle these formats and convert them into formats compatible with the processing tools used in their research. For example at this time the popular EEGLAB toolbox for MATLAB supports over 30 different data formats.
Bittium NeurOne is a research-grade EEG system that comes with a recording software which, like many others, has its own proprietary binary format for storing raw recordings. For a long time, MATLAB has been a popular processing environment in neuroscientific research, but lately especially with the rise of machine learning methods python has also become more present in neuroscientific workflows. Despite this, Bittium at this time only provides a MATLAB conversion script for their data format, so python-based workflows had to include a conversion step using the proprietary MATLAB software.
To enable pure python processing of data recorded with Bittium NeurOne we created
the neurone_loader python package. It provides the ability to load data in the
proprietary NeurOne binary format in pure python. The data can be used directly
as numpy arrays or imported into python-mne [@Gramfort2013] [@Gramfort2014],
a popular python package for processing neuroimaging data.
Since EEG data is often recorded at very high sampling rates and single recordings
can reach sizes of hundreds of gigabytes neurone_laoder was designed with
memory restrictions in mind. To make it easier to work with data this large even
in environments with (relatively) restricted memory resources, most data will be
loaded lazily and redundant data will be removed from memory as soon as it has been
copied.
We acknowledge the work of Andreas Henelius at the Finnish Institute of Occupational Health figuring out how to read most the NeurOne binary format in pure python as part of his export2hdf project.