Example usage of BIDS conversion for MEG data using the data2bids function of Fieldtrip
Background: The Brain Imaging Data Structure is a standard for organising and describing neuroimaging datasets.(BIDS; https://bids.neuroimaging.io/) MEG-BIDS is the magnetoencephalography extension of BIDS: https://bids-specification.readthedocs.io/en/stable/04-modality-specific-files/02-magnetoencephalography.html
A number of options exist to assist in converting a raw MEG dataset to a BIDS compliant structure. For example: 1. MNE has MNE-BIDS https://mne.tools/mne-bids/ 2. 'Biscuit' - a python GUI from Macquarie University, using MNE-BIDS and associated functions. https://macquarie-meg-research.github.io/Biscuit/ 3. Fieldtrip has 'data2bids'. http://www.fieldtriptoolbox.org/reference/data2bids/
If converting relatively small datasets, or incrementally as project is running, Biscuit is a very useful tool. Where large datasets are to be converted, a scripting solution (eg. using MNE-BIDS in python or data2bids in Fieldtrip/MATLAB seems the best approach (in the interests of consistency and time)
Below is an example usage of data2bids to generate a BIDS compliant dataset from MEG data collected on the Elekta Neuromag system at Swinburne University, in Melbourne, Australia. Author: David White, davidjwhite33@gmail.com
The dataset is from a multi-site clinical trial exploring a 12-month nutrional intervention in older adults with subjective memory impairments or mild cognitive impairment (The 'CANN' trial; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6222033/). MEG data is collected at the Melbourne (Swinburne ) site only, with 2 runs of a virtual water maze task in addition to 2 resting recordings (6 mins each, eyes open, fixation cross). Data was acquired at baseline and after 12 months treatment.
Data has been pulled from the Swinburne MEG data storage (neuro-mgmt1.cc.swin.edu.au:projects/sinuhe_MEG_data), in this case using rsync to manage a local copy of the acquired data over the course of the project.
Directory stucture on this storage uses acquisition date to store data in directories within a project folder, with raw MEG data (.fif) are as named when acquired. Manual intervention has placed all raw files in a single directory, with files named according to ID|task|session, eg. 001_rest1_V1.fif. Unfortunately, files have upper or lower case..
The script will generate a bids-compliant structure. Some additional notes to consider:
- README file in bids_dir is not generated by this script: "a free form text file (README) describing the dataset in more details SHOULD be provided. The README file MUST be either in ASCII or UTF-8 encoding".
- coordsystem.json sidecars (optional): head coil coordinates not currently written by data2bids, so this recommended info is is not in coordsystem.json sidecars ('Biscuit' will read the from the fif)
- channels.tsv files do not contain filter settings, nor info of the EOG/ECG channels (these just retain raw labels (BIO001))
- events.json file (optional) could be written ("As with all other tabular data, _events files may be accompanied by a JSON file describing the columns in detail")