def save_beat_times(beats, stimulus_id, cue=False, data_root=None, offset=None, overwrite=False, version=None):
if version is None:
version = DEFAULT_VERSION
if data_root is None:
data_root = os.path.join(deepthought.DATA_PATH, 'OpenMIIR')
if cue:
assert offset is None # no offset in cue files
beats_filepath = os.path.join(data_root, 'meta',
'beats.v{}'.format(version),
'{}_cue_beats.txt'.format(stimulus_id))
else:
beats_filepath = os.path.join(data_root, 'meta',
'beats.v{}'.format(version),
'{}_beats.txt'.format(stimulus_id))
if os.path.exists(beats_filepath) and not overwrite:
log.info('Skipping existing {}'.format(beats_filepath))
else:
ensure_parent_dir_exists(beats_filepath)
with open(beats_filepath, 'w') as f:
if offset is not None:
f.write('# offset of cue: {}\n'.format(offset))
for beat in beats:
f.write('{}\n'.format(beat))
log.info('Saved beat times in {}'.format(beats_filepath))
def clean_data(mne_data_root, subject, verbose=True, overwrite=False):
## check whether output already exists
output_filepath = os.path.join(mne_data_root,
'{}_filtered-raw.fif'.format(subject))
if os.path.exists(output_filepath):
if not overwrite:
log.info('Skipping existing {}'.format(output_filepath))
return
input_filepath = os.path.join(mne_data_root, '{}-raw.fif'.format(subject))
raw = mne.io.Raw(input_filepath, preload=True, verbose=verbose)
## apply bandpass filter
raw.filter(0.5,
30,
filter_length='10s',
l_trans_bandwidth=0.1,
h_trans_bandwidth=0.5,
method='fft',
iir_params=None,
picks=None,
n_jobs=1,
verbose=verbose)
ensure_parent_dir_exists(output_filepath)
raw.save(output_filepath, overwrite=overwrite, verbose=False)
def save_beat_times(beats, stimulus_id, cue=False, data_root=None, offset=None, overwrite=False, version=None):
if version is None:
version = DEFAULT_VERSION
if data_root is None:
data_root = os.path.join(deepthought.DATA_PATH, 'OpenMIIR')
if cue:
assert offset is None # no offset in cue files
beats_filepath = os.path.join(data_root, 'meta',
'beats.v{}'.format(version),
'{}_cue_beats.txt'.format(stimulus_id))
else:
beats_filepath = os.path.join(data_root, 'meta',
'beats.v{}'.format(version),
'{}_beats.txt'.format(stimulus_id))
if os.path.exists(beats_filepath) and not overwrite:
log.info('Skipping existing {}'.format(beats_filepath))
else:
ensure_parent_dir_exists(beats_filepath)
with open(beats_filepath, 'w') as f:
if offset is not None:
f.write('# offset of cue: {}\n'.format(offset))
for beat in beats:
f.write('{}\n'.format(beat))
log.info('Saved beat times in {}'.format(beats_filepath))
def extract_events_from_file(data_root, subject, verbose=True):
bdf_filepath = os.path.join(data_root, 'eeg', 'biosemi', '{}.bdf'.format(subject))
markers_filepath = os.path.join(data_root, 'eeg', 'biosemi', '{}_EEG_Data.mat'.format(subject))
raw = read_raw_edf(bdf_filepath, preload=True, verbose=verbose)
log.debug(raw)
log.debug(raw.info)
events = extract_events_from_raw(raw, markers_filepath, subject, verbose)
# Writing events
output_filepath = os.path.join(data_root, 'eeg', 'imported', subject, '{}-eve.fif.gz'.format(subject))
ensure_parent_dir_exists(output_filepath)
mne.write_events(output_filepath, events)
def generate_plots(y_real, y_pred, output, dataset_name, output_path):
font = {
# 'family' : 'normal',
'weight': 'normal',
'size': 24
}
mpl.rc('font', **font)
# Compute confusion matrix
cm = confusion_matrix(y_real, y_pred)
shuffle_classes = LabelConverter().shuffle_classes
cmt = np.zeros([24, 24])
for i in xrange(24):
for j in xrange(24):
cmt[shuffle_classes[i], shuffle_classes[j]] = cm[i, j]
labels = swapped_meta_labels
print cmt
print classification_report(y_real, y_pred)
misclass = (y_real != y_pred).mean()
print 'misclassification rate: {:.4f}'.format(misclass)
print 'accuracy: {:.4f}'.format(100. * (1 - misclass))
fig = plt.figure(1, figsize=(10, 10), dpi=600)
axes = fig.add_subplot(111)
# Show confusion matrix in a separate
axes.matshow(cmt)
# plt.title('Confusion matrix')
# axes.colorbar()
plt.ylabel('True label')
plt.xlabel('Predicted label')
axes.set_xticks(xrange(24))
axes.set_xticklabels(labels, rotation=90, fontsize=15)
axes.set_yticks(xrange(24))
axes.set_yticklabels(labels, fontsize=15)
plot_file = os.path.join(experiment_root, 'confusion',
'{}_confusion.pdf'.format(dataset_name))
ensure_parent_dir_exists(plot_file)
plt.savefig(plot_file, bbox_inches='tight')
def extract_events_from_file(data_root, subject, verbose=True):
bdf_filepath = os.path.join(data_root, 'eeg', 'biosemi',
'{}.bdf'.format(subject))
markers_filepath = os.path.join(data_root, 'eeg', 'biosemi',
'{}_EEG_Data.mat'.format(subject))
raw = read_raw_edf(bdf_filepath, preload=True, verbose=verbose)
log.debug(raw)
log.debug(raw.info)
events = extract_events_from_raw(raw, markers_filepath, subject, verbose)
# Writing events
output_filepath = os.path.join(data_root, 'eeg', 'imported', subject,
'{}-eve.fif.gz'.format(subject))
ensure_parent_dir_exists(output_filepath)
mne.write_events(output_filepath, events)
def generate_plots(y_real, y_pred, output, dataset_name, output_path):
font = {
# 'family' : 'normal',
'weight' : 'normal',
'size' : 24}
mpl.rc('font', **font)
# Compute confusion matrix
cm = confusion_matrix(y_real, y_pred);
shuffle_classes = LabelConverter().shuffle_classes;
cmt = np.zeros([24,24]);
for i in xrange(24):
for j in xrange(24):
cmt[shuffle_classes[i],shuffle_classes[j]] = cm[i,j];
labels = swapped_meta_labels;
print cmt
print classification_report(y_real, y_pred)
misclass = (y_real != y_pred).mean();
print 'misclassification rate: {:.4f}'.format(misclass);
print 'accuracy: {:.4f}'.format(100. * (1 - misclass));
fig = plt.figure(1, figsize=(10, 10), dpi=600)
axes = fig.add_subplot(111)
# Show confusion matrix in a separate
axes.matshow(cmt)
# plt.title('Confusion matrix')
# axes.colorbar()
plt.ylabel('True label')
plt.xlabel('Predicted label')
axes.set_xticks(xrange(24));
axes.set_xticklabels(labels, rotation=90, fontsize=15)
axes.set_yticks(xrange(24));
axes.set_yticklabels(labels, fontsize=15)
plot_file = os.path.join(experiment_root, 'confusion', '{}_confusion.pdf'.format(dataset_name));
ensure_parent_dir_exists(plot_file);
plt.savefig(plot_file, bbox_inches='tight')
def clean_data(mne_data_root, subject, verbose=True, overwrite=False):
## check whether output already exists
output_filepath = os.path.join(mne_data_root,
'{}_filtered-raw.fif'.format(subject))
if os.path.exists(output_filepath):
if not overwrite:
log.info('Skipping existing {}'.format(output_filepath))
return
input_filepath = os.path.join(mne_data_root,
'{}-raw.fif'.format(subject))
raw = mne.io.Raw(input_filepath, preload=True, verbose=verbose)
## apply bandpass filter
raw.filter(0.5, 30, filter_length='10s',
l_trans_bandwidth=0.1, h_trans_bandwidth=0.5,
method='fft', iir_params=None,
picks=None, n_jobs=1, verbose=verbose)
ensure_parent_dir_exists(output_filepath)
raw.save(output_filepath, overwrite=overwrite, verbose=False)
def import_and_process_metadata(biosemi_data_root,
mne_data_root,
subject,
verbose=True,
overwrite=False):
## check whether output already exists
output_filepath = os.path.join(mne_data_root, '{}-raw.fif'.format(subject))
if os.path.exists(output_filepath):
if not overwrite:
log.info('Skipping existing {}'.format(output_filepath))
return
## import raw BDF file from biosemi
bdf_filepath = os.path.join(biosemi_data_root, '{}.bdf'.format(subject))
## NOTE: marks EXT1-4 channels as EOG channels during import
log.info('Importing raw BDF data from: {}'.format(bdf_filepath))
raw = read_raw_edf(bdf_filepath,
eog=RAW_EOG_CHANNELS,
preload=True,
verbose=verbose)
log.info('Imported raw data: {}'.format(raw))
sfreq = raw.info['sfreq']
if sfreq != 512:
log.warn('Unexpected sample rate: {} Hz'.format(sfreq))
log.warn('Re-sampling to 512 Hz')
fast_resample_mne(raw,
512,
res_type='sinc_best',
preserve_events=True,
verbose=True)
## mark all unused channels as bad
raw.info['bads'] += [
u'C1', u'C2', u'C3', u'C4', u'C5', u'C6', u'C7', u'C8', u'C9', u'C10',
u'C11', u'C12', u'C13', u'C14', u'C15', u'C16', u'C17', u'C18', u'C19',
u'C20', u'C21', u'C22', u'C23', u'C24', u'C25', u'C26', u'C27', u'C28',
u'C29', u'C30', u'C31', u'C32', u'D1', u'D2', u'D3', u'D4', u'D5',
u'D6', u'D7', u'D8', u'D9', u'D10', u'D11', u'D12', u'D13', u'D14',
u'D15', u'D16', u'D17', u'D18', u'D19', u'D20', u'D21', u'D22', u'D23',
u'D24', u'D25', u'D26', u'D27', u'D28', u'D29', u'D30', u'D31', u'D32',
u'E1', u'E2', u'E3', u'E4', u'E5', u'E6', u'E7', u'E8', u'E9', u'E10',
u'E11', u'E12', u'E13', u'E14', u'E15', u'E16', u'E17', u'E18', u'E19',
u'E20', u'E21', u'E22', u'E23', u'E24', u'E25', u'E26', u'E27', u'E28',
u'E29', u'E30', u'E31', u'E32', u'F1', u'F2', u'F3', u'F4', u'F5',
u'F6', u'F7', u'F8', u'F9', u'F10', u'F11', u'F12', u'F13', u'F14',
u'F15', u'F16', u'F17', u'F18', u'F19', u'F20', u'F21', u'F22', u'F23',
u'F24', u'F25', u'F26', u'F27', u'F28', u'F29', u'F30', u'F31', u'F32',
u'G1', u'G2', u'G3', u'G4', u'G5', u'G6', u'G7', u'G8', u'G9', u'G10',
u'G11', u'G12', u'G13', u'G14', u'G15', u'G16', u'G17', u'G18', u'G19',
u'G20', u'G21', u'G22', u'G23', u'G24', u'G25', u'G26', u'G27', u'G28',
u'G29', u'G30', u'G31', u'G32', u'H1', u'H2', u'H3', u'H4', u'H5',
u'H6', u'H7', u'H8', u'H9', u'H10', u'H11', u'H12', u'H13', u'H14',
u'H15', u'H16', u'H17', u'H18', u'H19', u'H20', u'H21', u'H22', u'H23',
u'H24', u'H25', u'H26', u'H27', u'H28', u'H29', u'H30', u'H31', u'H32',
u'EXG7', u'EXG8', u'GSR1', u'GSR2', u'Erg1', u'Erg2', u'Resp', u'Plet',
u'Temp'
]
log.info('Marked unused channels as bad: {}'.format(raw.info['bads']))
if not recording_has_mastoid_channels(subject):
raw.info['bads'] += [u'EXG5', u'EXG6']
picks = mne.pick_types(raw.info,
meg=False,
eeg=True,
eog=True,
stim=True,
exclude='bads')
## process events
markers_filepath = os.path.join(biosemi_data_root,
'{}_EEG_Data.mat'.format(subject))
log.info('Processing events, external source: {}'.format(markers_filepath))
events = extract_events_from_raw(raw, markers_filepath, subject, verbose)
raw._data[-1, :].fill(0) # delete data in stim channel
raw.add_events(events)
# crop to first event - 1s ... last event + 20s (longer than longest trial)
onesec = raw.info['sfreq']
tmin, tmax = raw.times[[
events[0, 0] - onesec, events[-1, 0] + 20 * onesec
]]
log.info('Cropping raw inplace to {:.3f}s - {:.3f}s'.format(tmin, tmax))
raw.crop(tmin=tmin, tmax=tmax, copy=False)
# fix sample offser -> 0
raw.last_samp -= raw.first_samp
raw.first_samp = 0
ensure_parent_dir_exists(output_filepath)
log.info('Saving raw fif data to: {}'.format(output_filepath))
raw.save(output_filepath, picks=picks, overwrite=overwrite, verbose=False)
del raw
raw = fix_channel_infos(output_filepath, verbose=verbose)
log.info('Imported {}'.format(raw))
log.info('Metadata: {}'.format(raw.info))
def save_ica(self, description):
ica_data_root = self.get_ica_data_root()
ica_filepath = os.path.join(
ica_data_root, '{}-{}-ica.fif'.format(self.subject, description))
ensure_parent_dir_exists(ica_filepath)
self.ica.save(ica_filepath)
def import_and_process_metadata(biosemi_data_root, mne_data_root, subject, verbose=True, overwrite=False):
## check whether output already exists
output_filepath = os.path.join(mne_data_root,
'{}-raw.fif'.format(subject))
if os.path.exists(output_filepath):
if not overwrite:
log.info('Skipping existing {}'.format(output_filepath))
return
## import raw BDF file from biosemi
bdf_filepath = os.path.join(biosemi_data_root, '{}.bdf'.format(subject))
## NOTE: marks EXT1-4 channels as EOG channels during import
log.info('Importing raw BDF data from: {}'.format(bdf_filepath))
raw = read_raw_edf(bdf_filepath, eog=RAW_EOG_CHANNELS, preload=True, verbose=verbose)
log.info('Imported raw data: {}'.format(raw))
sfreq = raw.info['sfreq']
if sfreq != 512:
log.warn('Unexpected sample rate: {} Hz'.format(sfreq))
log.warn('Re-sampling to 512 Hz')
fast_resample_mne(raw, 512, res_type='sinc_best', preserve_events=True, verbose=True)
## mark all unused channels as bad
raw.info['bads'] += [u'C1', u'C2', u'C3', u'C4', u'C5', u'C6', u'C7', u'C8', u'C9', u'C10',
u'C11', u'C12', u'C13', u'C14', u'C15', u'C16', u'C17', u'C18', u'C19', u'C20',
u'C21', u'C22', u'C23', u'C24', u'C25', u'C26', u'C27', u'C28', u'C29', u'C30',
u'C31', u'C32', u'D1', u'D2', u'D3', u'D4', u'D5', u'D6', u'D7', u'D8',
u'D9', u'D10', u'D11', u'D12', u'D13', u'D14', u'D15', u'D16', u'D17', u'D18',
u'D19', u'D20', u'D21', u'D22', u'D23', u'D24', u'D25', u'D26', u'D27', u'D28',
u'D29', u'D30', u'D31', u'D32', u'E1', u'E2', u'E3', u'E4', u'E5', u'E6',
u'E7', u'E8', u'E9', u'E10', u'E11', u'E12', u'E13', u'E14', u'E15',
u'E16', u'E17', u'E18', u'E19', u'E20', u'E21', u'E22', u'E23', u'E24',
u'E25', u'E26', u'E27', u'E28', u'E29', u'E30', u'E31', u'E32', u'F1',
u'F2', u'F3', u'F4', u'F5', u'F6', u'F7', u'F8', u'F9', u'F10', u'F11',
u'F12', u'F13', u'F14', u'F15', u'F16', u'F17', u'F18', u'F19', u'F20',
u'F21', u'F22', u'F23', u'F24', u'F25', u'F26', u'F27', u'F28', u'F29',
u'F30', u'F31', u'F32', u'G1', u'G2', u'G3', u'G4', u'G5', u'G6', u'G7',
u'G8', u'G9', u'G10', u'G11', u'G12', u'G13', u'G14', u'G15', u'G16', u'G17',
u'G18', u'G19', u'G20', u'G21', u'G22', u'G23', u'G24', u'G25', u'G26', u'G27',
u'G28', u'G29', u'G30', u'G31', u'G32', u'H1', u'H2', u'H3', u'H4', u'H5',
u'H6', u'H7', u'H8', u'H9', u'H10', u'H11', u'H12', u'H13', u'H14', u'H15',
u'H16', u'H17', u'H18', u'H19', u'H20', u'H21', u'H22', u'H23', u'H24', u'H25',
u'H26', u'H27', u'H28', u'H29', u'H30', u'H31', u'H32',
u'EXG7', u'EXG8',
u'GSR1', u'GSR2', u'Erg1', u'Erg2', u'Resp', u'Plet', u'Temp']
log.info('Marked unused channels as bad: {}'.format(raw.info['bads']))
if not recording_has_mastoid_channels(subject):
raw.info['bads'] += [u'EXG5', u'EXG6']
picks = mne.pick_types(raw.info, meg=False, eeg=True, eog=True, stim=True, exclude='bads')
## process events
markers_filepath = os.path.join(biosemi_data_root, '{}_EEG_Data.mat'.format(subject))
log.info('Processing events, external source: {}'.format(markers_filepath))
events = extract_events_from_raw(raw, markers_filepath, subject, verbose)
raw._data[-1,:].fill(0) # delete data in stim channel
raw.add_events(events)
# crop to first event - 1s ... last event + 20s (longer than longest trial)
onesec = raw.info['sfreq']
tmin, tmax = raw.index_as_time([events[0,0]-onesec, events[-1,0]+20*onesec])
log.info('Cropping raw inplace to {:.3f}s - {:.3f}s'.format(tmin, tmax))
raw.crop(tmin=tmin, tmax=tmax, copy=False)
# fix sample offser -> 0
raw.last_samp -= raw.first_samp
raw.first_samp = 0
ensure_parent_dir_exists(output_filepath)
log.info('Saving raw fif data to: {}'.format(output_filepath))
raw.save(output_filepath, picks=picks, overwrite=overwrite, verbose=False)
del raw
raw = fix_channel_infos(output_filepath, verbose=verbose)
log.info('Imported {}'.format(raw))
log.info('Metadata: {}'.format(raw.info))
def save_ica(self, description):
ica_data_root = self.get_ica_data_root()
ica_filepath = os.path.join(ica_data_root,
'{}-{}-ica.fif'.format(self.subject, description))
ensure_parent_dir_exists(ica_filepath)
self.ica.save(ica_filepath)
