def get_response_epoch(subject, session):
epochs, meta = preprocessing.get_epochs_for_subject(subject,
"stimulus",
sessions=session)
epochs = epochs.pick_channels(
[x for x in epochs.ch_names if x.startswith("M")])
response, meta = preprocessing.get_epochs_for_subject(subject,
"response",
sessions=session)
response = response.pick_channels(
[x for x in response.ch_names if x.startswith("M")])
# Find trials that are present in both time periods
overlap = list(
set(epochs.events[:, 2]).intersection(set(response.events[:, 2])))
epochs = epochs[[str(l) for l in overlap]]
response = response[[str(l) for l in overlap]]
id_time = (-0.25 <= epochs.times) & (epochs.times <= 0)
means = epochs._data[:, :, id_time].mean(-1)
epochs._data -= means[:, :, np.newaxis]
response._data = response._data - means[:, :, np.newaxis]
# Now also baseline stimulus period
epochs, meta = preprocessing.get_epochs_for_subject(subject,
"stimulus",
sessions=session)
epochs = epochs.pick_channels(
[x for x in epochs.ch_names if x.startswith("M")])
id_time = (-0.25 <= epochs.times) & (epochs.times <= 0)
means = epochs._data[:, :, id_time].mean(-1)
epochs._data -= means[:, :, np.newaxis]
data_cov = pymeglcmv.get_cov(epochs, tmin=0, tmax=1.35)
return data_cov, response
def get_stim_epoch(subject, session):
epochs, meta = preprocessing.get_epochs_for_subject(subject,
"stimulus",
sessions=session)
epochs = epochs.pick_channels(
[x for x in epochs.ch_names if x.startswith("M")])
id_time = (-0.25 <= epochs.times) & (epochs.times <= 0)
means = epochs._data[:, :, id_time].mean(-1)
epochs._data -= means[:, :, np.newaxis]
data_cov = pymeglcmv.get_cov(epochs, tmin=0, tmax=1.35)
return data_cov, epochs
def get_stim_epoch(subject, session, recording):
filenames = glob(get_filenames(subject, session, recording)[0])
epochs = preprocessing.load_epochs(filenames)
epochs = preprocessing.concatenate_epochs(epochs, None)
epochs = epochs.pick_channels(
[x for x in epochs.ch_names if x.startswith('M')])
id_time = (-0.25 <= epochs.times) & (epochs.times <= 0)
means = epochs._data[:, :, id_time].mean(-1)
epochs._data -= means[:, :, np.newaxis]
data_cov = lcmv.get_cov(epochs, tmin=0, tmax=3)
return data_cov, epochs, filenames
def get_response_epoch(subject, session, recording):
stimulus = glob(get_filenames(subject, session, recording)[0])
response = glob(get_filenames(subject, session, recording)[1])
stimulus = preprocessing.load_epochs(stimulus)
stimulus = preprocessing.concatenate_epochs(stimulus, None)
stimulus = stimulus.pick_channels(
[x for x in stimulus.ch_names if x.startswith('M')])
response = preprocessing.load_epochs(response)
response = preprocessing.concatenate_epochs(response, None)
response = stimulus.pick_channels(
[x for x in response.ch_names if x.startswith('M')])
id_time = (-0.25 <= stimulus.times) & (stimulus.times <= 0)
means = stimulus._data[:, :, id_time].mean(-1)
stimulus._data = stimulus._data - means[:, :, np.newaxis]
response._data = response._data - means[:, :, np.newaxis]
data_cov = lcmv.get_cov(stimulus, tmin=0, tmax=3)
return data_cov, response
def do_source_recon(subj, session, njobs=4):
epochs_filename_stim = os.path.join(data_folder, "epochs", subj, session,
'{}-epo.fif.gz'.format('stimlock'))
epochs_filename_resp = os.path.join(data_folder, "epochs", subj, session,
'{}-epo.fif.gz'.format('resplock'))
trans_filename = os.path.join(data_folder, "transformation_matrix",
'{}_{}-trans.fif'.format(subj, session))
if os.path.isfile(epochs_filename_stim):
runs = sorted([
run.split('/')[-1] for run in glob.glob(
os.path.join(data_folder, "raw", subj, session, "meg", "*.ds"))
])
center = int(np.floor(len(runs) / 2.0))
raw_filename = os.path.join(data_folder, "raw", subj, session, "meg",
runs[center])
# # make transformation matrix:
# sr.make_trans(subj, raw_filename, epochs_filename, trans_filename)
# load labels:
labels = sr.get_labels(subject=subj,
filters=['*wang*.label', '*JWG*.label'],
annotations=['HCPMMP1'])
labels = sr.labels_exclude(labels=labels,
exclude_filters=[
'wang2015atlas.IPS4',
'wang2015atlas.IPS5',
'wang2015atlas.SPL', 'JWG_lat_Unknown'
])
labels = sr.labels_remove_overlap(labels=labels,
priority_filters=['wang', 'JWG'])
print(labels)
# load epochs:
epochs_stim = mne.read_epochs(epochs_filename_stim)
epochs_stim = epochs_stim.pick_channels(
[x for x in epochs_stim.ch_names if x.startswith('M')])
epochs_resp = mne.read_epochs(epochs_filename_resp)
epochs_resp = epochs_resp.pick_channels(
[x for x in epochs_resp.ch_names if x.startswith('M')])
# baseline stuff:
overlap = list(
set(epochs_stim.events[:, 2]).intersection(
set(epochs_resp.events[:, 2])))
epochs_stim = epochs_stim[[str(l) for l in overlap]]
epochs_resp = epochs_resp[[str(l) for l in overlap]]
id_time = (-0.3 <= epochs_stim.times) & (epochs_stim.times <= -0.2)
means = epochs_stim._data[:, :, id_time].mean(-1)
epochs_stim._data = epochs_stim._data - means[:, :, np.newaxis]
epochs_resp._data = epochs_resp._data - means[:, :, np.newaxis]
# TFR settings:
fois_h = np.arange(42, 162, 4)
fois_l = np.arange(2, 42, 2)
tfr_params = {
'HF': {
'foi': fois_h,
'cycles': fois_h * 0.4,
'time_bandwidth': 5 + 1,
'n_jobs': njobs,
'est_val': fois_h,
'est_key': 'HF'
},
'LF': {
'foi': fois_l,
'cycles': fois_l * 0.4,
'time_bandwidth': 1 + 1,
'n_jobs': njobs,
'est_val': fois_l,
'est_key': 'LF'
}
}
# get cov:
data_cov = lcmv.get_cov(epochs_stim, tmin=0, tmax=1)
noise_cov = None
# get lead field:
forward, bem, source = sr.get_leadfield(
subject=subj,
raw_filename=raw_filename,
epochs_filename=epochs_filename_stim,
trans_filename=trans_filename,
conductivity=(0.3, 0.006, 0.3),
njobs=njobs)
# do source level analysis:
for tl, epochs in zip(['stimlock', 'resplock'],
[epochs_stim, epochs_resp]):
for signal_type in ['LF', 'HF']:
print(signal_type)
# events:
events = epochs.events[:, 2]
data = []
filters = lcmv.setup_filters(epochs.info,
forward,
data_cov,
None,
labels,
njobs=njobs)
# in chunks:
chunks = 100
for i in range(0, len(events), chunks):
filename = os.path.join(data_folder, "source_level", 'lcmv_{}_{}_{}_{}_{}-source.hdf'.\
format(subj, session, tl, signal_type, i))
# if os.path.isfile(filename):
# continue
M = lcmv.reconstruct_tfr(filters,
epochs.info,
epochs._data[i:i + chunks],
events[i:i + chunks],
epochs.times,
est_args=tfr_params[signal_type],
njobs=njobs)
M.to_hdf(filename, 'epochs')
del M