def selectTFData(self, laplacian, excl_factor): ''' Arguments - - - - - Returns - - - - ''' # load processed behavior and eeg beh = self.beh EEG = self.EEG # check whether trials need to be excluded if type(excl_factor) == dict: # remove unwanted trials from beh beh, EEG = trial_exclusion(beh, EEG, excl_factor) # select electrodes of interest picks = mne.pick_types(EEG.info, eeg=True, exclude='bads') eegs = EEG._data[:,picks,:] if laplacian: embed() x,y,z = np.vstack([EEG.info['chs'][i]['loc'][:3] for i in picks]).T leg_order = 10 if picks.size <=100 else 12 eegs = laplacian_filter(eegs, x, y, z, leg_order = leg_order, smoothing = 1e-5) return eegs, beh
def selectBDMData(self, EEG, beh, time, excl_factor=None):
'''
Arguments
- - - - -
EEG (object):
beh (dataFrame):
time (tuple | list): time samples (start to end) for decoding
excl_factor (dict): see Classify documentation
Returns
- - - -
eegs (array): eeg data (trials X electrodes X time)
beh (dict): contains variables of interest for decoding
'''
# check whether trials need to be excluded
if type(excl_factor) == dict: # remove unwanted trials from beh
beh, EEG = trial_exclusion(beh, EEG, excl_factor)
# apply filtering and downsampling (if specified)
if self.bdm_type != 'broad':
EEG = EEG.filter(h_freq=self.bdm_band[0],
l_freq=self.bdm_band[1],
method='iir',
iir_params=dict(ftype='butterworth', order=5))
if self.downsample != int(EEG.info['sfreq']):
print('downsampling data')
EEG.resample(self.downsample)
# select time window and EEG electrodes
s, e = [np.argmin(abs(EEG.times - t)) for t in time]
picks = mne.pick_types(EEG.info, eeg=True, exclude='bads')
picks = select_electrodes(np.array(EEG.ch_names)[picks], self.elec_oi)
eegs = EEG._data[:, picks, s:e]
times = EEG.times[s:e]
# if specified average over trials
#beh, eegs = self.averageTrials(beh, eegs, self.to_decode, cnd_header, 4)
# store dictionary with variables for plotting
plot_dict = {
'ch_names': EEG.ch_names,
'times': times,
'info': EEG.info
}
with open(
self.FolderTracker(['bdm', self.to_decode],
filename='plot_dict.pickle'),
'wb') as handle:
pickle.dump(plot_dict, handle)
return eegs, beh, times
def selectERPData(self, time, l_filter=False, excl_factor=None):
'''
THIS NEEDS TO BE ADJUSTED FOR TIMINGS
Arguments
- - - - -
Returns
- - - -
'''
beh = self.beh
EEG = self.eeg
# check whether trials need to be excluded
if type(excl_factor) == dict: # remove unwanted trials from beh
beh, EEG = trial_exclusion(beh, EEG, excl_factor)
# read in eeg data
if l_filter:
EEG.filter(l_freq=None, h_freq=l_filter)
# select time window and EEG electrodes
EEG = EEG.crop(tmin=time[0], tmax=time[1])
ch_names = EEG.ch_names
self.eeg = EEG
# store dictionary with variables for plotting
plot_dict = {
'ch_names': ch_names,
'times': EEG.times,
'info': EEG.info
}
with open(
self.FolderTracker(['erp', self.header],
filename='plot_dict.pickle'.format(
self.header)), 'wb') as handle:
pickle.dump(plot_dict, handle)
def select_erp_data(
self,
excl_factor: dict = None,
topo_flip: dict = None) -> Tuple[pd.DataFrame, mne.Epochs]:
"""
Selects the data of interest by excluding a subset of trials
Args:
excl_factor (dict, optional): If specified, a subset of trials that
matches the specified criteria is excluded from further analysis
(e.g., dict(target_color = ['red']) will exclude all trials
where the target color is red). Defaults to None.
topo_flip (dict, optional): If specified a subset of trials is
flipped such that it is as if all stimuli of interest are
presented right (see lateralized_erp).
Returns:
beh: pandas Dataframe
epochs: epochs data of interest
"""
beh = self.beh.copy()
epochs = self.epochs.copy()
# if not already done reset index (to properly align beh and epochs)
beh.reset_index(inplace=True, drop=True)
# if specified remove trials matching specified criteria
if excl_factor is not None:
beh, epochs = trial_exclusion(beh, epochs, excl_factor)
# check whether left stimuli should be
# artificially transferred to left hemifield
if topo_flip is not None:
(header, left), = topo_flip.items()
epochs = self.flip_topography(epochs, beh, left, header)
else:
print('No topography info specified. In case of a lateralized '
'design. It is assumed as if all stimuli of interest are '
'presented right (i.e., left hemifield')
return beh, epochs
def selectBDMData(self, epochs, beh, time, excl_factor = None):
'''
Arguments
- - - - -
EEG (object):
beh (dataFrame):
time (tuple | list): time samples (start to end) for decoding
excl_factor (dict): see Classify documentation
Returns
- - - -
eegs (array): eeg data (trials X electrodes X time)
beh (dict): contains variables of interest for decoding
'''
# check whether trials need to be excluded
if type(excl_factor) == dict: # remove unwanted trials from beh
beh, epochs = trial_exclusion(beh, epochs, excl_factor)
# apply filtering and downsampling (if specified)
if self.bdm_type != 'broad':
print('eeg data is filtered before downsampling and slicing the time window of interest')
epochs.filter(self.bdm_band[0],self.bdm_band[1], method = 'iir',
iir_params = dict(ftype = 'butterworth', order = 5))
epochs.apply_hilbert(envelope=True)
tf_data = abs(epochs._data)**2
# db convert
# base_slice = slice(*[np.argmin(abs(epochs.times - t)) for t in self.baseline])
# print('baseline correct TF data via db convert')
# tf_data = tf_data.reshape(-1,epochs.times.size)
# tf_data = np.array(np.matrix(tf_data) / np.matrix(tf_data[:,base_slice]).mean(axis = 1)
# ).reshape(-1,len(epochs.info['ch_names']),epochs.times.size)
#epochs._data = 10 * np.log10(tf_data)
# or 'standard' baseline correction
epochs._data = abs(epochs._data)**2
epochs.apply_baseline(baseline = self.baseline) # check whether this is correct???
if self.downsample < int(epochs.info['sfreq']):
if self.window_size[0] == 1:
print('downsampling data')
epochs.resample(self.downsample, npad='auto')
else:
self.down_factor = int(epochs.info['sfreq'])/self.downsample
print('downsampling will be done after data averaging')
# select time window and EEG electrodes
s, e = [np.argmin(abs(epochs.times - t)) for t in time]
picks = mne.pick_types(epochs.info, eeg=True, eog= True, misc = True, exclude='bads')
picks = select_electrodes(np.array(epochs.ch_names)[picks], self.elec_oi)
eegs = epochs._data[:,picks,s:e]
times = epochs.times[s:e]
# transform eeg data in case of sliding window approach
if self.window_size[0] > 1:
eegs = self.sliding_window(eegs, self.window_size[0], self.window_size[1], self.window_size[2])[:,:,self.window_size[0]-1:]
times = np.linspace(times[0], times[-self.window_size[0]], eegs.shape[-1])
s_freq =epochs.info['sfreq']
time_red = 1/s_freq * 1000 * self.window_size[0]
warnings.warn(f'Final timepoint in analysis is reduced by {time_red} ms as each timepoint in analysis now reflects {self.window_size[0]} data samples at a sampling rate of {s_freq} Hz', UserWarning)
return eegs, beh, times
