Presets

See also

Quickstart: from a raw recording to clean epochs runs the eeg-erp preset from raw data to epochs.

Presets are ready-made pipelines. They are ordinary Pipeline objects: change any parameter with set_params, edit steps, or export them to YAML.

import meeg_utils as meu

meu.presets.available()
pipe = meu.Pipeline.preset("had-meeg", datatype="meg")

Recommended presets

eeg-erp, eeg-rest, meg-erp and meg-rest are the pipelines this library recommends. Every parameter comes from a published default (MNE-BIDS-Pipeline, the requirements of ICLabel and MEGnet, PREP, ZapLine-plus, autoreject, Tanner et al. 2015 for the high-pass); parameters without an accepted default must be given. The sources are listed step by step in docs/presets/recommended.md.

Each has two stages, applied per run:

pre = meu.Pipeline.preset("meg-erp", system="neuromag")   # or "ctf", "kit"
clean = pre.fit_transform(run)

ep = meu.Pipeline.preset("meg-erp", stage="epochs", event_id=["face", "house"],
                         head_destination=meu.io.average_dev_head_t(runs))
epochs = ep.fit_transform(clean)
# several runs: meu.epochs.combine([...])

• preprocessing – EEG: bridged electrodes, 0.1 Hz high-pass, ZapLine-plus, PREP, interpolation, average reference, ICA with ICLabel. MEG: system-specific noise reduction (Neuromag: SSS; CTF: grade-3 gradiometers; KIT: reference regression), 0.1 Hz high-pass, ZapLine-plus, 250 Hz, 20-component ICA with MEGnet.

• epochs – ERP: 40 Hz low-pass, -0.2 to 0.5 s around event_id, pre-stimulus baseline. Rest: consecutive windows of epoch_duration. Both end with local autoreject.

Dataset presets

Dataset presets reproduce the preprocessing of a published dataset with its known problems fixed. Every difference from the original code is documented with its reason.

had-meeg

The per-run preprocessing of HAD-MEEG for CTF MEG (datatype="meg") or EEG (datatype="eeg"): 0.1-100 Hz, 250 Hz, bad channels (Maxwell / PREP), interpolation, ZapLine-plus, CTF grade-3 compensation or average reference, and ICA labelled with MEGnet / ICLabel.

The list of corrections, including those for the original epoching stage (head alignment, baseline, mastoid reference), is in docs/presets/had-meeg.md.