def test_bad_by_deviation(raw_tmp):
"""Test detection of channels with relatively high or low amplitudes."""
# Set scaling factors for high and low deviation test channels
low_dev_factor = 0.1
high_dev_factor = 4.0
# Make the signal for a random channel have a very high amplitude
n_chans = raw_tmp._data.shape[0]
high_dev_idx = int(RNG.randint(0, n_chans, 1))
raw_tmp._data[high_dev_idx, :] *= high_dev_factor
# Test detection of abnormally high-amplitude channels
nd = NoisyChannels(raw_tmp, do_detrend=False)
nd.find_bad_by_deviation()
assert nd.bad_by_deviation == [raw_tmp.ch_names[high_dev_idx]]
# Make the signal for a different channel have a very low amplitude
low_dev_idx = (high_dev_idx - 1) if high_dev_idx > 0 else 1
raw_tmp._data[low_dev_idx, :] *= low_dev_factor
# Test detection of abnormally low-amplitude channels
# NOTE: The default z-score threshold (5.0) is too strict to allow detection
# of abnormally low-amplitude channels in some datasets. Using a relaxed Z
# threshold of 3.29 (p < 0.001, two-tailed) until a better solution is found.
nd = NoisyChannels(raw_tmp, do_detrend=False)
nd.find_bad_by_deviation(deviation_threshold=3.29)
bad_by_dev_idx = [low_dev_idx, high_dev_idx]
assert nd.bad_by_deviation == [raw_tmp.ch_names[i] for i in bad_by_dev_idx]
def test_findnoisychannels(raw, montage):
"""Test find noisy channels."""
# Set a random state for the test
rng = np.random.RandomState(30)
raw.set_montage(montage)
nd = NoisyChannels(raw, random_state=rng)
nd.find_all_bads(ransac=True)
bads = nd.get_bads()
iterations = (
10 # remove any noisy channels by interpolating the bads for 10 iterations
)
for iter in range(0, iterations):
if len(bads) == 0:
continue
raw.info["bads"] = bads
raw.interpolate_bads()
nd = NoisyChannels(raw, random_state=rng)
nd.find_all_bads(ransac=True)
bads = nd.get_bads()
# make sure no bad channels exist in the data
raw.drop_channels(ch_names=bads)
# Test for NaN and flat channels
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
# Insert a nan value for a random channel and make another random channel
# completely flat (ones)
idxs = rng.choice(np.arange(m), size=2, replace=False)
rand_chn_idx1 = idxs[0]
rand_chn_idx2 = idxs[1]
rand_chn_lab1 = raw_tmp.ch_names[rand_chn_idx1]
rand_chn_lab2 = raw_tmp.ch_names[rand_chn_idx2]
raw_tmp._data[rand_chn_idx1, n - 1] = np.nan
raw_tmp._data[rand_chn_idx2, :] = np.ones(n)
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_nan_flat()
assert nd.bad_by_nan == [rand_chn_lab1]
assert nd.bad_by_flat == [rand_chn_lab2]
# Test for high and low deviations in EEG data
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
# Now insert one random channel with very low deviations
rand_chn_idx = int(rng.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
raw_tmp._data[rand_chn_idx, :] = raw_tmp._data[rand_chn_idx, :] / 10
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_deviation()
assert rand_chn_lab in nd.bad_by_deviation
# Inserting one random channel with a high deviation
raw_tmp = raw.copy()
rand_chn_idx = int(rng.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
arbitrary_scaling = 5
raw_tmp._data[rand_chn_idx, :] *= arbitrary_scaling
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_deviation()
assert rand_chn_lab in nd.bad_by_deviation
# Test for correlation between EEG channels
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(rng.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# Use cosine instead of sine to create a signal
low = 10
high = 30
n_freq = 5
signal = np.zeros((1, n))
for freq_i in range(n_freq):
freq = rng.randint(low, high, n)
signal[0, :] += np.cos(2 * np.pi * raw.times * freq)
raw_tmp._data[rand_chn_idx, :] = signal * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_correlation()
assert rand_chn_lab in nd.bad_by_correlation
# Test for high freq noise detection
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(rng.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# Use freqs between 90 and 100 Hz to insert hf noise
signal = np.zeros((1, n))
for freq_i in range(n_freq):
freq = rng.randint(90, 100, n)
signal[0, :] += np.sin(2 * np.pi * raw.times * freq)
raw_tmp._data[rand_chn_idx, :] = signal * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_hfnoise()
assert rand_chn_lab in nd.bad_by_hf_noise
# Test for signal to noise ratio in EEG data
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(rng.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# inserting an uncorrelated high frequency (90 Hz) signal in one channel
raw_tmp[rand_chn_idx, :] = np.sin(2 * np.pi * raw.times * 90) * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_SNR()
assert rand_chn_lab in nd.bad_by_SNR
# Test for finding bad channels by RANSAC
raw_tmp = raw.copy()
# Ransac identifies channels that go bad together and are highly correlated.
# Inserting highly correlated signal in channels 0 through 3 at 30 Hz
raw_tmp._data[0:6, :] = np.cos(2 * np.pi * raw.times * 30) * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_ransac()
bads = nd.bad_by_ransac
assert bads == raw_tmp.ch_names[0:6]
# Test for finding bad channels by channel-wise RANSAC
raw_tmp = raw.copy()
# Ransac identifies channels that go bad together and are highly correlated.
# Inserting highly correlated signal in channels 0 through 3 at 30 Hz
raw_tmp._data[0:6, :] = np.cos(2 * np.pi * raw.times * 30) * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_bad_by_ransac(channel_wise=True)
bads = nd.bad_by_ransac
assert bads == raw_tmp.ch_names[0:6]
# Test not-enough-memory and n_samples type exceptions
raw_tmp = raw.copy()
raw_tmp._data[0:6, :] = np.cos(2 * np.pi * raw.times * 30) * 1e-6
nd = NoisyChannels(raw_tmp, random_state=rng)
# Set n_samples very very high to trigger a memory error
n_samples = int(1e100)
with pytest.raises(MemoryError):
nd.find_bad_by_ransac(n_samples=n_samples)
# Set n_samples to a float to trigger a type error
n_samples = 35.5
with pytest.raises(TypeError):
nd.find_bad_by_ransac(n_samples=n_samples)
# Test IOError when not enough channels for ransac predictions
raw_tmp = raw.copy()
# Make flat all channels except 2
num_bad_channels = raw._data.shape[0] - 2
raw_tmp._data[0:num_bad_channels, :] = np.zeros_like(
raw_tmp._data[0:num_bad_channels, :]
)
nd = NoisyChannels(raw_tmp, random_state=rng)
nd.find_all_bads(ransac=False)
with pytest.raises(IOError):
nd.find_bad_by_ransac()
def test_findnoisychannels(raw, montage):
raw.set_montage(montage)
nd = NoisyChannels(raw)
nd.find_all_bads(ransac=True)
bads = nd.get_bads()
iterations = (
10 # remove any noisy channels by interpolating the bads for 10 iterations
)
for iter in range(0, iterations):
raw.info["bads"] = bads
raw.interpolate_bads()
nd = NoisyChannels(raw)
nd.find_all_bads(ransac=True)
bads = nd.get_bads()
# make sure no bad channels exist in the data
raw.drop_channels(ch_names=bads)
# Test for NaN and flat channels
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
# Insert a nan value for a random channel
rand_chn_idx1 = int(np.random.randint(0, m, 1))
rand_chn_idx2 = int(np.random.randint(0, m, 1))
rand_chn_lab1 = raw_tmp.ch_names[rand_chn_idx1]
rand_chn_lab2 = raw_tmp.ch_names[rand_chn_idx2]
raw_tmp._data[rand_chn_idx1, n - 1] = np.nan
raw_tmp._data[rand_chn_idx2, :] = np.ones(n)
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_nan_flat()
assert nd.bad_by_nan == [rand_chn_lab1]
assert nd.bad_by_flat == [rand_chn_lab2]
# Test for high and low deviations in EEG data
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
# Now insert one random channel with very low deviations
rand_chn_idx = int(np.random.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
raw_tmp._data[rand_chn_idx, :] = raw_tmp._data[rand_chn_idx, :] / 10
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_deviation()
assert rand_chn_lab in nd.bad_by_deviation
# Inserting one random channel with a high deviation
raw_tmp = raw.copy()
rand_chn_idx = int(np.random.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
arbitrary_scaling = 5
raw_tmp._data[rand_chn_idx, :] *= arbitrary_scaling
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_deviation()
assert rand_chn_lab in nd.bad_by_deviation
# Test for correlation between EEG channels
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(np.random.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# Use cosine instead of sine to create a signal
low = 10
high = 30
n_freq = 5
signal = np.zeros((1, n))
for freq_i in range(n_freq):
freq = np.random.randint(low, high, n)
signal[0, :] += np.cos(2 * np.pi * raw.times * freq)
raw_tmp._data[rand_chn_idx, :] = signal * 1e-6
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_correlation()
assert rand_chn_lab in nd.bad_by_correlation
# Test for high freq noise detection
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(np.random.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# Use freqs between 90 and 100 Hz to insert hf noise
signal = np.zeros((1, n))
for freq_i in range(n_freq):
freq = np.random.randint(90, 100, n)
signal[0, :] += np.sin(2 * np.pi * raw.times * freq)
raw_tmp._data[rand_chn_idx, :] = signal * 1e-6
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_hfnoise()
assert rand_chn_lab in nd.bad_by_hf_noise
# Test for signal to noise ratio in EEG data
raw_tmp = raw.copy()
m, n = raw_tmp._data.shape
rand_chn_idx = int(np.random.randint(0, m, 1))
rand_chn_lab = raw_tmp.ch_names[rand_chn_idx]
# inserting an uncorrelated high frequency (90 Hz) signal in one channel
raw_tmp[rand_chn_idx, :] = np.sin(2 * np.pi * raw.times * 90) * 1e-6
nd = NoisyChannels(raw_tmp)
nd.find_bad_by_SNR()
assert rand_chn_lab in nd.bad_by_SNR
# Test for finding bad channels by RANSAC
raw_tmp = raw.copy()
# Ransac identifies channels that go bad together and are highly correlated.
# Inserting highly correlated signal in channels 0 through 3 at 30 Hz
raw_tmp._data[0:6, :] = np.cos(2 * np.pi * raw.times * 30) * 1e-6
nd = NoisyChannels(raw_tmp)
np.random.seed(30)
nd.find_bad_by_ransac()
bads = nd.bad_by_ransac
assert bads == raw_tmp.ch_names[0:6]
def preproc(self, event_dict, baseline_start, stim_dur, montage, out_dir='preproc', subjects='all', tasks='all',
hp_cutoff=1, lp_cutoff=50, line_noise=60, seed=42, eog_chan='none'):
'''Preprocesses a single EEG file. Assigns a list of epoched data to Dataset instance,
where each entry in the list is a subject with concatenated task data. Here is the basic
structure of the preprocessing workflow:
- Set the montage
- Band-pass filter (high-pass filter by default)
- Automatically detect bad channels
- Notch filter out line-noise
- Reference data to average of all EEG channels
- Automated removal of eye-related artifacts using ICA
- Spherical interpolation of detected bad channels
- Extract events and epoch the data accordingly
- Align the events based on type (still need to implement this!)
- Create a list of epoched data, with subject as the element concatenated across tasks
Parameters
----------
event_dict: dict
Maps integers to semantic labels for events within the experiment
baseline_start: int or float
Specify start of the baseline period (in seconds)
stim_dur: int or float
Stimulus duration (in seconds)
Note: may need to make more general to allow various durations
montage: mne.channels.montage.DigMontage
Maps sensor locations to coordinates
subjects: list or 'all'
Specify which subjects to iterate through
tasks: list or 'all'
Specify which tasks to iterate through
hp_cutoff: int or float
The low frequency bound for the highpass filter in Hz
line_noise: int or float
The frequency of electrical noise to filter out in Hz
seed: int
Set the seed for replicable results
eog_chan: str
If there are no EOG channels present, select an EEG channel
near the eyes for eye-related artifact detection
'''
missing = [] # initialize missing file list
subj_iter = self.gen_iter(subjects, self.n_subj) # get subject iterator
task_iter = self.gen_iter(tasks, self.n_task) # get task iterator
# Iterate through subjects (initialize subject epoch list)
epochs_subj = []
for subj in subj_iter:
# Iterate through tasks (initialize within-subject task epoch list)
epochs_task = []
for task in task_iter:
# Specify the file format
self.get_file_format(subj, task)
try: # Handles missing files
raw = self.wget_raw_edf() # read
except:
print(f'---\nThis file does not exist: {self.file_path}\n---')
# Need to write the missing file list out
missing.append(self.file_path)
break
# Standardize montage
mne.datasets.eegbci.standardize(raw)
# Set montage and strip channel names of "." characters
raw.set_montage(montage)
raw.rename_channels(lambda x: x.strip('.'))
# Apply high-pass filter
np.random.seed(seed)
raw.filter(l_freq=hp_cutoff, h_freq=lp_cutoff, picks=['eeg', 'eog'])
# Instantiate NoisyChannels object
noise_chans = NoisyChannels(raw, do_detrend=False)
# Detect bad channels through multiple methods
noise_chans.find_bad_by_nan_flat()
noise_chans.find_bad_by_deviation()
noise_chans.find_bad_by_SNR()
# Set the bad channels in the raw object
raw.info['bads'] = noise_chans.get_bads()
print(f'Bad channels detected: {noise_chans.get_bads()}')
# Define the frequencies for the notch filter (60Hz and its harmonics)
#notch_filt = np.arange(line_noise, raw.info['sfreq'] // 2, line_noise)
# Apply notch filter
#print(f'Apply notch filter at {line_noise} Hz and its harmonics')
#raw.notch_filter(notch_filt, picks=['eeg', 'eog'], verbose=False)
# Reference to the average of all the good channels
# Automatically excludes raw.info['bads']
raw.set_eeg_reference(ref_channels='average')
# Instantiate ICA object
ica = mne.preprocessing.ICA(max_iter=1000)
# Run ICA
ica.fit(raw)
# Find which ICs match the EOG pattern
if eog_chan == 'none':
eog_indices, eog_scores = ica.find_bads_eog(raw, verbose=False)
else:
eog_indices, eog_scores = ica.find_bads_eog(raw, eog_chan, verbose=False)
# Apply the IC blink removals (if any)
ica.apply(raw, exclude=eog_indices)
print(f'Removed IC index {eog_indices}')
# Interpolate bad channels
raw.interpolate_bads()
# Specify pre-processing directory
preproc_dir = Path(f'{out_dir}/ {subj}')
# If directory does not exist, one will be created.
if not os.path.isdir(preproc_dir):
os.makedirs(preproc_dir)
# raw.save(Path(preproc_dir, f'subj{subj}_task{task}_raw.fif'),
# overwrite=True)
# Find events
events = mne.events_from_annotations(raw)[0]
# Epoch the data
preproc_epoch = mne.Epochs(raw, events, tmin=baseline_start, tmax=stim_dur,
event_id=event_dict, event_repeated='error',
on_missing='ignore', preload=True)
# Equalize event counts
preproc_epoch.equalize_event_counts(event_dict.keys())
# Rearrange and align the epochs
align = [preproc_epoch[i] for i in event_dict.keys()]
align_epoch = mne.concatenate_epochs(align)
# Add to epoch list
epochs_task.append(align_epoch)
# Assuming some data exists for a subject
# Concatenate epochs within subject
concat_epoch = mne.concatenate_epochs(epochs_task)
epochs_subj.append(concat_epoch)
# Attaches a list with each entry corresponding to epochs for a subject
self.epoch_list = epochs_subj