Mark Bad Channels and Segments in Raw MEG/EEG Data

Description

Marks bad channels and time segments in MNE raw MEG/EEG data using raw.info['bads'] and mne.Annotations. Channels marked as bad are excluded from analysis, and annotations describe segments of data to be discarded (e.g., movement artifacts, signal dropout).

The app generates:

• The raw data with updated info['bads'] and annotations
• A power spectral density (PSD) plot (bad channels excluded)
• A product.json summarizing the marked channels and annotations

Inputs

• raw (neuro/meeg/mne/raw): continuous raw MEG/EEG data to mark bad channels and segments in (required)
• channels (neuro/meg/fif-override, channels.tsv file): optional BIDS channels.tsv file; rows whose status column is bad are added to raw.info['bads'] (optional)

Outputs

• out_dir/raw.fif (neuro/meeg/mne/raw, tag bads_marked): raw data with marked bad channels and annotations
• out_figs/psd.png: power spectral density plot computed with bad channels excluded
• product.json: summary of marked channels and annotations, with the PSD plot embedded

Configuration Parameters

key	type	default	description
bads	string	""	Comma-separated list of channel names to mark as bad (e.g., "MEG2423,MEG2422,EEG001"). Names not present in the data are ignored.
annotations	string (multiline)	""	Bad data segments, one per line: onset, duration, description[, channels]. If channels is omitted the annotation applies to all channels.
reset_bads	boolean	false	If true, clears any bad channels already marked in the input file before adding the new ones from bads/channels.

Example annotations value:

2, 2, bad_segment
5, 2, movement_artifact, MEG2121, MEG2122

Usage

Running on Brainlife.io

1. Select your raw MEG/EEG dataset as the raw input.
2. Optionally provide a channels.tsv file as the channels input to mark channels whose status column is bad.
3. Optionally set bads, annotations, and/or reset_bads in the configuration form.
4. Submit the process.

Local Testing

# Edit config.json to point "raw" (and optionally "channels") at real data, then:
python main.py

Technical Details

• Bad Channels: Marked in raw.info['bads'] and excluded from analysis in downstream apps
• Annotations: Stored in raw.annotations and can be used for segment-level quality control
• Channel Validation: Invalid channel names in both bads and annotations are validated against the input file
• Preload: Data is loaded and saved efficiently without full preload unless needed

Authors

• Maximilien Chaumon, Paris Brain Institute

Citations

We kindly ask that you cite the following articles when publishing papers and code using this app:

Hayashi, S., Caron, B.A., Heinsfeld, A.S. et al. brainlife.io: a decentralized and open-source cloud platform to support neuroscience research. Nat Methods 21, 809–813 (2024). https://doi.org/10.1038/s41592-024-02237-2

Gramfort A, et al. & Hämäläinen MS. MEG and EEG data analysis with MNE-Python. Frontiers in Neuroscience, 7(267):1–13 (2013). https://doi.org/10.3389/fnins.2013.00267

Funding Acknowledgement

brainlife.io is publicly funded and for the sustainability of the project we kindly ask that you acknowledge the following funding sources:

License

Copyright (c) 2026 MEEG Brainlife team. Licensed under AGPL-3.0, see license.txt.