def preprocess(loadedraw, events,\
APPLY_CAR,\
l_freq,\
h_freq,\
filter_method,\
tmin,\
tmax,\
tlow,\
thigh,\
n_jobs,\
picks_feat,\
baselineRange,\
verbose=False):
# Load raw, apply bandpass (if applicable), epoch
raw = loadedraw.copy()
# Di
# %% Spatial filter - Common Average Reference (CAR)
if APPLY_CAR:
raw._data[1:] = raw._data[1:] - np.mean(raw._data[1:], axis=0)
# print('Preprocess: CAR done')
# %% Properties initialization
tdef = trigger_def('triggerdef_errp.ini')
sfreq = raw.info['sfreq']
event_id = dict(correct=tdef.by_key['FEEDBACK_CORRECT'], wrong=tdef.by_key['FEEDBACK_WRONG'])
# %% Bandpass temporal filtering
b, a, zi = pu.butter_bandpass(h_freq, l_freq, sfreq,
raw._data.shape[0] - 1) # raw._data.shape[0]- 1 because channel 0 is trigger
if filter_method is 'NC' and cv_container is None:
raw.filter(l_freq=2, filter_length='10s', h_freq=h_freq, n_jobs=n_jobs, picks=picks_feat, method='fft',
iir_params=None) # method='iir'and irr_params=None -> filter with a 4th order Butterworth
# print('Preprocess: NC_bandpass filtering done')
if filter_method is 'LFILT':
# print('Preprocess: LFILT filtering done')
for x in range(1, raw._data.shape[0]): # range starting from 1 because channel 0 is trigger
# raw._data[x,:] = lfilter(b, a, raw._data[x,:])
raw._data[x, :], zi[:, x - 1] = lfilter(b, a, raw._data[x, :], -1, zi[:, x - 1])
# self.eeg[:,x], self.zi[:,x] = lfilter(b, a, self.eeg[:,x], -1,zi[:,x])
# %% Epoching and baselining
# = tmin-paddingLength
# t_upper = tmax+paddingLength
t_lower = 0
t_upper = thigh
# t_lower = 0
# t_upper = tmax+paddingLength
epochs = mne.Epochs(raw, events=events, event_id=event_id, tmin=t_lower, tmax=t_upper, baseline=baselineRange,
picks=picks_feat, preload=True, proj=False, verbose=verbose)
total_wframes = epochs.get_data().shape[2]
print('Preprocess: Epoching done')
# if tmin != tmin_bkp:
# # if the baseline range was before the initial tmin, epochs was tricked to
# # to select this range (it expects that baseline is witin [tmin,tmax])
# # this part restore the initial tmin and prune the data
# epochs.tmin = tmin_bkp
# epochs._data = epochs._data[:,:,int((tmin_bkp-tmin)*sfreq):]
return tdef, sfreq, event_id, b, a, zi, t_lower, t_upper, epochs, total_wframes
# Spatial filter - Common Average Reference (CAR)
if APPLY_CAR:
raw._data[1:] = raw._data[1:] - np.mean(raw._data[1:], axis=0)
processing_steps.append('Car')
tdef = trigger_def('triggerdef_errp.ini')
sfreq = raw.info['sfreq']
paddingIdx = int(paddingLength * sfreq)
event_id = dict(correct=tdef.by_name['FEEDBACK_CORRECT'],
wrong=tdef.by_name['FEEDBACK_WRONG'])
# %% Dataset wide processing
# Bandpass temporal filtering
b, a, zi = pu.butter_bandpass(
h_freq, l_freq, sfreq, raw._data.shape[0] -
1) # raw._data.shape[0]- 1 because channel 0 is trigger
if FILTER_METHOD is 'NC':
raw.filter(
l_freq=l_freq,
h_freq=h_freq,
n_jobs=n_jobs,
picks=picks_feat,
method='iir',
iir_params=None
) # method='iir'and irr_params=None -> filter with a 4th order Butterworth
if FILTER_METHOD is 'LFILT':
for x in range(1, raw._data.shape[0]
): # range starting from 1 because channel 0 is trigger
# raw._data[x,:] = lfilter(b, a, raw._data[x,:])
raw._data[x, :], zi[:, x - 1] = lfilter(b, a, raw._data[x, :], -1,