111 lines
4.1 KiB
Python
111 lines
4.1 KiB
Python
import mne
|
|
from mne.preprocessing import create_eog_epochs
|
|
|
|
from mne_bids import BIDSPath, read_raw_bids
|
|
from utils.ccs_eeg_utils import read_annotations_core
|
|
from utils.file_utils import get_epochs
|
|
|
|
|
|
def load_unprocessed_subject(subject, dataset):
|
|
"""
|
|
Load the eeg data of a subject
|
|
:param subject: The subject of which the data will be loaded
|
|
:param dataset: The dataset which will be loaded
|
|
:return: the subject data
|
|
"""
|
|
bids_path = BIDSPath(subject=subject, task=dataset, session=dataset, datatype='eeg', suffix='eeg',
|
|
root='Dataset\\' + dataset)
|
|
raw = read_raw_bids(bids_path)
|
|
# Add annotations
|
|
read_annotations_core(bids_path, raw)
|
|
return raw
|
|
|
|
|
|
def filter_data(raw):
|
|
"""
|
|
Filter the data of a single subject with a bandpass filter.
|
|
The lower bound ist 0.5Hz to compensate the slow drifts.
|
|
The upper bound is 50Hz to compensate the high frequencies, including the power line spike at 60Hz
|
|
:param raw: The data to be filtered
|
|
:return: The filtered data
|
|
"""
|
|
raw.filter(0.5, 48, fir_design='firwin')
|
|
return raw
|
|
|
|
|
|
def plot_filter_data():
|
|
ds = 'N170'
|
|
for subj in ['014']:
|
|
data = load_unprocessed_subject(subj, ds)
|
|
data.load_data()
|
|
# data.plot(n_channels=len(data.ch_names), block=True, scalings=40e-6)
|
|
filter_data(data)
|
|
fig = mne.viz.plot_raw_psd(data, fmax=80, average=True, show=False)
|
|
fig.savefig("plots/frequency_filtered_subj_" + subj + "_48Hz.png")
|
|
# data.plot(n_channels=len(data.ch_names), block=True, scalings=40e-6)
|
|
|
|
|
|
def plot_filter_data_epoched(subj):
|
|
ds = 'N170'
|
|
data = load_unprocessed_subject(subj, ds)
|
|
data.load_data()
|
|
filter_data(data)
|
|
get_epochs(data)[0].average().plot()
|
|
|
|
|
|
def plot_cleaning():
|
|
ds = 'N170'
|
|
for subj in ['014']:
|
|
data = load_unprocessed_subject(subj, ds)
|
|
data.load_data()
|
|
filter_data(data)
|
|
folder = "Dataset\\" + ds + "\\sub-" + subj + "\\ses-" + ds + "\\eeg\\"
|
|
filepath = folder + "sub-" + subj + "_task-" + ds
|
|
print(filepath)
|
|
ann = mne.read_annotations(filepath + "_" + "badannotations.csv")
|
|
data.annotations.append(ann.onset, ann.duration, ann.description)
|
|
data.plot(n_channels=len(data.ch_names), block=True, scalings=40e-6)
|
|
|
|
|
|
def plot_ica():
|
|
ds = 'N170'
|
|
for subj in ['014']:
|
|
data = load_unprocessed_subject(subj, ds)
|
|
folder = "Dataset\\" + ds + "\\sub-" + subj + "\\ses-" + ds + "\\eeg\\"
|
|
filepath = folder + "sub-" + subj + "_task-" + ds
|
|
ann = mne.read_annotations(filepath + "_" + "badannotations.csv")
|
|
data.annotations.append(ann.onset, ann.duration, ann.description)
|
|
data.load_data()
|
|
data.set_channel_types({'HEOG_left': 'eog', 'HEOG_right': 'eog', 'VEOG_lower': 'eog'})
|
|
data.set_montage('standard_1020', match_case=False)
|
|
ica_raw = data.copy()
|
|
ica_raw.filter(l_freq=1, h_freq=None)
|
|
|
|
# Then run ICA
|
|
ica = mne.preprocessing.ICA(method="fastica",
|
|
random_state=123) # Use a random state for reproducable results #TODO Old Random state 123 or new one?
|
|
ica.fit(ica_raw, verbose=True)
|
|
ica_raw.load_data()
|
|
# ica.plot_components(inst=ica_raw, ch_type='eeg', contours=0, topomap_args={'extrapolate': 'head'},
|
|
# psd_args={'fmin': 0, 'fmax': 80})
|
|
ica.plot_sources(ica_raw)
|
|
ica.plot_properties(inst=ica_raw, dB=False, topomap_args={'extrapolate': 'head', 'contours': 0},
|
|
psd_args={'fmin': 3, 'fmax': 50}, picks=['eeg'])
|
|
|
|
|
|
def plot_joint_eog_plots():
|
|
ds = 'N170'
|
|
for subj in ['014']:
|
|
data = load_unprocessed_subject(subj, ds)
|
|
data.load_data()
|
|
data.set_channel_types({'HEOG_left': 'eog', 'HEOG_right': 'eog', 'VEOG_lower': 'eog'})
|
|
data.set_montage('standard_1020', match_case=False)
|
|
eog_evoked = create_eog_epochs(data).average()
|
|
eog_evoked.apply_baseline(baseline=(None, -0.2))
|
|
eog_evoked.plot_joint()
|
|
|
|
|
|
# plot_ica()
|
|
# plot_joint_eog_plots()
|
|
plot_filter_data_epoched('003')
|