def task_ica():
"""Step 3: ICA"""
for subject in subjects:
yield dict(
name=str(subject),
file_dep=[fname.raw_detrend(subject=subject), '03_ica.py'],
targets=[fname.ica(subject=subject)],
actions=[f'ipython 03_ica.py {subject:d}'],
)
def task_ica():
"""Step 03: Use ICA to clean up ECG and EOG artifacts"""
for subject in subjects:
filt_fnames = [fname.filt(subject=subject, run=run,
fmin=bandpass_fmin, fmax=bandpass_fmax)
for run in range(1, 7)]
ica_fname = fname.ica(subject=subject)
yield dict(
name=subject,
task_dep=['filter'],
file_dep=filt_fnames + ['03_ica.py'],
targets=[ica_fname],
actions=['python 03_ica.py %s' % subject],
)
def task_epochs():
"""Step 4: Construct epochs"""
for subject in subjects:
yield dict(
name=str(subject),
file_dep=[
fname.raw_filt(subject=subject),
fname.ica(subject=subject), '04_epochs.py'
],
targets=[
fname.epochs(subject=subject),
fname.epochs_long(subject=subject)
],
actions=[f'ipython 04_epochs.py {subject:d}'],
)
def task_epochs():
"""Step 04: Cut the data into epochs"""
for subject in subjects:
filt_fnames = [fname.filt(subject=subject, run=run,
fmin=bandpass_fmin, fmax=bandpass_fmax)
for run in range(1, 7)]
ica_fname = fname.ica(subject=subject)
epo_fname = fname.epo(subject=subject)
yield dict(
name=subject,
task_dep=['ica'],
file_dep=filt_fnames + [ica_fname, '04_epochs.py'],
targets=[epo_fname],
actions=['python 04_epochs.py %s' % subject],
)
import mne
import argparse
import numpy as np
from config import fname, events_id, subjects_with_extra_stim_artifacts, stim_artifact_sensor, n_jobs
# Handle command line arguments
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument('subject', metavar='sub###', type=int, help='The subject to process')
args = parser.parse_args()
subject = args.subject
print('Processing subject:', subject)
report = mne.open_report(fname.report(subject=subject))
raw = mne.io.read_raw_fif(fname.raw_filt(subject=subject))
ica = mne.preprocessing.read_ica(fname.ica(subject=subject))
# Create short epochs for evoked analysis
events = mne.find_events(raw, shortest_event=0.01)
epochs = mne.Epochs(raw, events, events_id, tmin=-0.2, tmax=0.5, reject=None, baseline=(-0.2, 0), preload=True)
report.add_figs_to_section(epochs.average().plot_joint(times=[0.035, 0.1]), ['Evokeds without cleaning (grads)', 'Evokeds without cleaning (mags)'], 'Sensor level', replace=True)
# Apply ICA to remove EOG and ECG artifacts
epochs = ica.apply(epochs)
# Do a first pass for fixing the stim artifact
mne.preprocessing.fix_stim_artifact(epochs)
epochs.save(fname.epochs(subject=subject), overwrite=True)
# For some subjects, there are more stim artifacts that we need to remove using ICA
if subject in subjects_with_extra_stim_artifacts:
# Find onsets of heart beats and blinks. Create epochs around them
ecg_epochs = create_ecg_epochs(raw, tmin=-.3, tmax=.3, preload=False)
eog_epochs = create_eog_epochs(raw, tmin=-.5, tmax=.5, preload=False)
# Find ICA components that correlate with heart beats.
ecg_epochs.decimate(5)
ecg_epochs.load_data()
ecg_epochs.apply_baseline((None, None))
ecg_inds, ecg_scores = ica.find_bads_ecg(ecg_epochs, method='ctps')
ecg_scores = np.abs(ecg_scores)
rank = np.argsort(ecg_scores)[::-1]
rank = [r for r in rank if ecg_scores[r] > 0.05]
ica.exclude = rank[:n_ecg_components]
print(' Found %d ECG indices' % (len(ecg_inds),))
# Find ICA components that correlate with eye blinks
eog_epochs.decimate(5)
eog_epochs.load_data()
eog_epochs.apply_baseline((None, None))
eog_inds, eog_scores = ica.find_bads_eog(eog_epochs)
eog_scores = np.max(np.abs(eog_scores), axis=0)
# Remove all components with a correlation > 0.1 to the EOG channels and that
# have not already been flagged as ECG components
rank = np.argsort(eog_scores)[::-1]
rank = [r for r in rank if eog_scores[r] > 0.1 and r not in ecg_inds]
ica.exclude += rank[:n_eog_components]
print(' Found %d EOG indices' % (len(eog_inds),))
# Save the ICA decomposition
ica.save(fname.ica(subject=subject))
# Read events from the stim channel
mask = 4096 + 256 # mask for excluding high order bits
events = mne.find_events(raw,
stim_channel='STI101',
consecutive='increasing',
mask=mask,
mask_type='not_and',
min_duration=0.003)
# Compensate for projector delay
events[:, 0] += int(round(0.0345 * raw.info['sfreq']))
# Load the ICA object
print(' Using ICA')
ica = read_ica(fname.ica(subject=subject))
# Make epochs. Because the original 1000Hz sampling rate is a bit excessive
# for what we're going for, we only read every 5th sample. This gives us a
# sampling rate of ~200Hz.
epochs = mne.Epochs(raw,
events,
events_id,
epoch_tmin,
epoch_tmax,
baseline=baseline,
decim=5,
preload=True)
# Save evoked plot to the report
with mne.open_report(fname.report(subject=subject)) as report: