This tutorial shows how to clean MEG data with Maxwell filtering.
Maxwell filtering in MNE can be used to suppress sources of external intereference and compensate for subject head movements.
See Maxwell filtering for more details.
import mne
from mne.preprocessing import maxwell_filter
data_path = mne.datasets.sample.data_path()
Set parameters
raw_fname = data_path + '/MEG/sample/sample_audvis_raw.fif'
ctc_fname = data_path + '/SSS/ct_sparse_mgh.fif'
fine_cal_fname = data_path + '/SSS/sss_cal_mgh.dat'
Preprocess with Maxwell filtering
raw = mne.io.read_raw_fif(raw_fname)
raw.info['bads'] = ['MEG 2443', 'EEG 053', 'MEG 1032', 'MEG 2313'] # set bads
# Here we don't use tSSS (set st_duration) because MGH data is very clean
raw_sss = maxwell_filter(raw, cross_talk=ctc_fname, calibration=fine_cal_fname)
Select events to extract epochs from, pick M/EEG channels, and plot evoked
tmin, tmax = -0.2, 0.5
event_id = {'Auditory/Left': 1}
events = mne.find_events(raw, 'STI 014')
picks = mne.pick_types(raw.info, meg=True, eeg=False, stim=False, eog=True,
include=[], exclude='bads')
for r, kind in zip((raw, raw_sss), ('Raw data', 'Maxwell filtered data')):
epochs = mne.Epochs(r, events, event_id, tmin, tmax, picks=picks,
baseline=(None, 0), reject=dict(eog=150e-6))
evoked = epochs.average()
evoked.plot(window_title=kind, ylim=dict(grad=(-200, 250),
mag=(-600, 700)))
Total running time of the script: ( 0 minutes 6.252 seconds)