import os.path as op
import mne
import numpy as np
from config import bem_ico, excludes, map_subjects, meg_dir, spacing
conditions = {
'aud_left_minus_right', 'vis_left_minus_right', 'aud_left_eq',
'aud_right_eq', 'vis_left_eq', 'vis_right_eq'
}
subjects = [s for s in map_subjects.values() if s not in excludes]
for condition in conditions:
stcs = list()
for subject in subjects:
stc_fname = op.join(
meg_dir, subject, '-'.join([
f'{subject}_audvis', 'dSPM', spacing, 'inverse', 'morph',
'filt', 'sss', condition
]))
stc = mne.read_source_estimate(stc_fname)
stcs.append(stc)
data = np.average([s.data for s in stcs], axis=0)
stc_ave = mne.VectorSourceEstimate(data, stcs[0].vertices, stcs[0].tmin,
stcs[0].tstep, 'fsaverage')
ave_fname = op.join(
meg_dir, '-'.join(
[spacing, f'ico{bem_ico}', 'dSPM', 'morph', 'condition', 'ave']))
stc_ave.save(ave_fname)
trans_fname = op.join(meg_dir, subject, f'{subject}_audvis_raw-trans.fif')
# If you follow the step in 2_setup_source_space.py
src_fname = op.join(subjects_dir, subject, 'bem',
f'{subject}-{spacing}-src.fif')
bem_fname = op.join(subjects_dir, subject, 'bem',
f'{subject}-ico{bem_ico}-bem-sol.fif')
# If you are practicing with the sample dataset
# src_fname = op.join(subjects_dir, subject, 'bem',
# f'{subject}-oct-6-src.fif')
# bem_fname = op.join(subjects_dir, subject, 'bem',
# f'{subject}-5120-5120-5120-bem-sol.fif')
info = mne.io.read_info(evoked_fname)
fwd = mne.make_forward_solution(info,
trans_fname,
src_fname,
bem_fname,
meg=True,
eeg=False)
mne.write_forward_solution(fwd_fname, fwd, overwrite=True)
print(f'Computed forward solution for {subject}')
parallel, run_func, _ = parallel_func(run_forward, n_jobs=n_jobs)
subjects = [
subject for subject in map_subjects.values() if subject not in excludes
]
parallel(run_func(subject) for subject in subjects)
picks=picks,
event_id=event_id,
tmin=tmin,
tmax=tmax,
baseline=baseline,
decim=4, # raw sampling rate is 600 Hz, subsample to 150 Hz
preload=True, # for autoreject
reject_tmax=reject_tmax,
reject_by_annotation=True)
del raw, annot
# autoreject (local)
if autoreject:
# local reject
# keep the bad sensors/channels because autoreject can repair it via
# interpolation
picks = mne.pick_types(epochs.info, meg=True, exclude=[])
ar = AutoReject(picks=picks, n_jobs=n_jobs, verbose=False)
print(f'Run autoreject (local) for {subject} (it takes a long time)')
ar.fit(epochs)
print(f'Drop bad epochs and interpolate bad sensors for {subject}')
epochs = ar.transform(epochs)
print(f'Dropped {round(epochs.drop_log_stats(), 2)}% epochs for {subject}')
epochs.save(epo_fname, overwrite=True)
for subject in map_subjects.values():
if subject not in excludes:
run_epochs(subject)