def apply_ICA(subject, runlist, badEEG, badICA):
from mne.io.pick import _picks_by_type as picks_by_type
from mne.preprocessing import read_ica
from mne.io import Raw
data_path = '/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/'
for run in runlist:
raw_fname = (data_path + subject + '/' + run + '_trans_sss.fif')
# read raw data
raw = []
raw = Raw(raw_fname, preload=True)
raw.info['bads'] = badEEG
# Load ICA file for each data type (MEG and EEG)
for ch_type, picks in picks_by_type(raw.info, meg_combined=True):
# Read ICA file
ica = []
ica = read_ica(data_path + subject + '/mne_python/ICA/ICA_' +
ch_type + '_allRuns')
# Apply ICA projection
ica.exclude += badICA[ch_type]
raw = ica.apply(raw, copy=True)
# Interpolate bad EEG channels
raw = raw.copy().interpolate_bads()
# Save artifacts-corrected raw data
raw.save((data_path + subject + '/' + run + 'ICAcorr_trans_sss.fif'),
overwrite='True')
def browse_components(subject, runlist, badEEG):
import numpy as np
import mne
import os.path as op
import os
from mne.report import Report
from mne.io.pick import _picks_by_type as picks_by_type
from mne.preprocessing import read_ica
from mne.io import Raw
data_path = '/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/'
report = Report(subject)
raw_fnames = []
for run in runlist:
raw_fnames.append(data_path + subject + '/' + run + '_trans_sss.fif')
# read raw data
raw = Raw(raw_fnames, preload=True)
raw.info['bads'] = badEEG
# Highpass filter 1Hz on EOG/ECG channels
#picks = mne.pick_types(raw.info, meg=False,eeg=False, eog=True, ecg=True)
#raw.filter(l_freq = 1, h_freq=30, picks = picks)
#picks = mne.pick_types(raw.info, meg=True,eeg=True, eog=True, ecg=True)
#raw.filter(l_freq = None, h_freq=30, picks = picks,n_jobs = 4)
picks = mne.pick_types(raw.info, meg=False, eeg=False, eog=False, ecg=True)
raw.filter(l_freq=1, h_freq=30, picks=picks)
picks = mne.pick_types(raw.info, meg=False, eeg=False, eog=True, ecg=False)
raw.filter(l_freq=1, h_freq=5, picks=picks)
picks = mne.pick_types(raw.info, meg=True, eeg=True, eog=True, ecg=True)
raw.filter(l_freq=None, h_freq=30, picks=picks, n_jobs=4)
results_dir = op.join(data_path, subject, 'artefactICA')
if not op.exists(results_dir):
os.makedirs(results_dir)
# Plot all ICA component
for ch_type, picks in picks_by_type(
raw.info, meg_combined=True): # bad EEG channels are excluded
# Read ICA file for each data type (MEG and EEG) - created in preprocessing_ica
ica = read_ica(data_path + subject + '/mne_python/ICA/ICA_' + ch_type +
'_allRuns')
print subject + ' ' + ch_type
#fig = ica.plot_sources(raw, block = True) # the viewer will open and ICA component can be screened
#report.add_figs_to_section(fig, 'All ICA sources (%s)' % ch_type,'ICA decomposition')
fig = ica.plot_components(colorbar=True)
for i in range(np.shape(fig)[0]):
report.add_figs_to_section(fig[i],
'All ICA components (%s)' % ch_type,
'ICA decomposition')
report.save((results_dir + '/AllRuns_checkcomponents.html'),
open_browser=False,
overwrite=True)
del raw
# contrast['name']))
# continue
# Prepare plot delta (subtraction, or regression)
fig1, ax1 = plt.subplots(1, len(chan_types))
if type(ax1) not in [list, np.ndarray]:
ax1 = [ax1]
# Prepare plot all conditions at top level of contrast
fig2, ax2 = plt.subplots(len(evokeds), len(chan_types))
ax2 = np.reshape(ax2, len(evokeds) * len(chan_types))
# Loop across channels
for ch, chan_type in enumerate(chan_types):
# Select specific types of sensor
info = delta.info
picks = [i for k, p in picks_by_type(info)
for i in p if k in chan_type['name']]
# --------------------------------------------------------------
# Plot delta (subtraction, or regression)
# adjust color scale
mM = np.percentile(np.abs(delta.data[picks,:]), 99.)
# plot mean sensors x time
ax1[ch].imshow(delta.data[picks,:], vmin=-mM, vmax=mM,
interpolation='none', aspect='auto',
cmap='RdBu_r', extent=[min(delta.times),
max(delta.times), 0, len(picks)])
# add t0
ax1[ch].plot([0, 0], [0, len(picks)], color='black')
ax1[ch].set_title(chan_type['name'] + ': ' + delta.comment)
ax1[ch].set_xlabel('Time')
def ICA_denoising(subject, run, badEEG):
import mne
import os.path as op
import os
import matplotlib
matplotlib.use('Agg')
from mne.report import Report
from mne.io import Raw
from meeg_preprocessing import compute_ica
from mne.io.pick import _picks_by_type as picks_by_type
###########################################################################
# jobs and runtime performance
n_components = 0.99
decim = 5 # decimation
n_max_ecg, n_max_eog = 5, 5 # limit components detected due to ECG / EOG
ica_reject = {'mag': 5e-12, 'grad': 5000e-13, 'eeg': 200e-6}
###########################################################################
data_path = '/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/'
raw_fname = data_path + subject + '/' + run + '_trans_sss.fif'
report = Report(subject)
# read raw data
raw = Raw(raw_fname, preload=True)
raw.info['bads'] = badEEG
# Highpass filter 1Hz on EOG/ECG channels
picks = mne.pick_types(raw.info, meg=False, eeg=False, eog=True, ecg=True)
raw.filter(l_freq=1, h_freq=30, picks=picks)
picks = mne.pick_types(raw.info, meg=True, eeg=True, eog=True, ecg=True)
raw.filter(l_freq=None, h_freq=30, picks=picks)
###########################################################################
# Create the report directory if it doesn't exist
results_dir = op.join(data_path, subject, 'artefactICA')
if not op.exists(results_dir):
os.makedirs(results_dir)
# Create the results directory if it doesn't exist
ica_dir = op.join(data_path, subject, 'mne_python/ICA')
if not op.exists(ica_dir):
os.makedirs(ica_dir)
report = mne.report.Report(subject)
###########################################################################
for ch_type, picks in picks_by_type(
raw.info, meg_combined=True): # bad EEG channels are excluded
ica, _ = compute_ica(raw,
picks=picks,
subject=subject,
n_components=n_components,
n_max_ecg=n_max_ecg,
n_max_eog=n_max_eog,
reject=ica_reject,
random_state=666,
decim=decim,
report=report)
ica.save((ica_dir + '/ICA_' + ch_type + '_' + run))
report.save((results_dir + '/' + run + '.html'),
open_browser=False,
overwrite=True)
del raw
for filePath in glob.glob(nip_path + '*_auto_ica_sss_raw.fif'):
temp = mne.io.Raw(filePath, preload=True, add_eeg_ref=False)
if 'MISC005' in temp.ch_names and nip in ['vV100048', 'Pb160320']:
temp.drop_channels(['MISC005'])
if not 'raw' in locals():
raw = temp
else:
raw.append(temp)
# Exclude bad EEG channels
raw.info['bads'] = bad_EEG[nip]
###########################################################################
####################### Cycle across channel type #########################
###########################################################################
for ch_type, picks in picks_by_type(raw.info, meg_combined=True):
#######################################################################
##################### Reject head-speaker artifacts ###################
#######################################################################
# Choose good channels (ch_types + misc)
picks_misc = np.concatenate([picks, mne.pick_types(raw.info, misc=True, meg=False)])
# Plot before transformation
misc_epochs = create_misc_epochs(raw, reject=reject, picks=picks_misc,
baseline=(None, 0), ch_name='MISC004',
tmax=0.5)
misc_average = misc_epochs.average()
report.add_figs_to_section(misc_average.plot(),
def check_ICA_Cleaning(subject, runlist, badEEG, badICA):
import mne
import os.path as op
import os
from mne.report import Report
from mne.io.pick import _picks_by_type as picks_by_type
from mne.preprocessing import read_ica
from mne.io import Raw
from mne.preprocessing import create_ecg_epochs, create_eog_epochs
data_path = '/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/'
report = Report(subject)
raw_fnames = []
for run in runlist:
raw_fnames.append(data_path + subject + '/' + run + '_trans_sss.fif')
# read raw data
raw = Raw(raw_fnames, preload=True)
raw.info['bads'] = badEEG
# Highpass filter 1Hz on EOG/ECG channels
picks = mne.pick_types(raw.info, meg=False, eeg=False, eog=True, ecg=False)
raw.filter(l_freq=1, h_freq=2, picks=picks)
picks = mne.pick_types(raw.info, meg=True, eeg=True, eog=True, ecg=True)
raw.filter(l_freq=None, h_freq=30, picks=picks, n_jobs=4)
results_dir = op.join(data_path, subject, 'artefactICA')
if not op.exists(results_dir):
os.makedirs(results_dir)
# Create ICA file for each data type (MEG and EEG)
for ch_type, picks in picks_by_type(
raw.info, meg_combined=True): # bad EEG channels are excluded
# Read ICA file
ica = []
ica = read_ica(data_path + subject + '/mne_python/ICA/ICA_' + ch_type +
'_allRuns')
for badICAsel in badICA[ch_type]:
# Select bad component and assess artefact rejection quality
ica_exclude = badICAsel
ecg_evoked = create_ecg_epochs(raw, tmin=-.5, tmax=.5,
picks=picks).average()
fig = ica.plot_overlay(ecg_evoked,
exclude=ica_exclude) # plot ECG cleaning
report.add_figs_to_section(
fig,
str(badICAsel) + ' Rejection overlay (%s)' % ch_type, 'ECG')
eog_evoked = create_eog_epochs(raw, tmin=-.5, tmax=.5,
picks=picks).average()
fig = ica.plot_overlay(eog_evoked,
exclude=ica_exclude) # plot EOG cleaning
report.add_figs_to_section(
fig,
str(badICAsel) + ' Rejection overlay (%s)' % ch_type, 'EOG')
fig = ica.plot_overlay(raw)
report.add_figs_to_section(
fig,
str(badICAsel) + ' Rejection overlay (%s)' % ch_type,
'Workroad artefact')
report.save((results_dir + '/AllRunsChecked.html'),
open_browser=False,
overwrite=True)