def fit_ica(raw,
picks,
reject,
ecg_ch,
eog_hor,
eog_ver,
flow_ecg,
fhigh_ecg,
flow_eog,
fhigh_eog,
ecg_thresh,
eog_thresh,
use_jumeg=True,
random_state=42):
"""
author: C.Kiefer; [email protected]
Fit an ICA object to the raw file. Identify cardiac and ocular components
and mark them for removal.
Parameters:
-----------
inst : instance of Raw, Epochs or Evoked
Raw measurements to be decomposed.
picks : array-like of int
Channels to be included. This selection remains throughout the
initialized ICA solution. If None only good data channels are used.
reject : dict | None
Rejection parameters based on peak-to-peak amplitude.
Valid keys are 'grad', 'mag', 'eeg', 'seeg', 'ecog', 'eog', 'ecg',
'hbo', 'hbr'.
If reject is None then no rejection is done. Example::
reject = dict(grad=4000e-13, # T / m (gradiometers)
mag=4e-12, # T (magnetometers)
eeg=40e-6, # V (EEG channels)
eog=250e-6 # V (EOG channels)
)
It only applies if `inst` is of type Raw.
ecg_ch : array-like | ch_name | None
ECG channel to which the sources shall be compared. It has to be
of the same shape as the sources. If some string is supplied, a
routine will try to find a matching channel. If None, a score
function expecting only one input-array argument must be used,
for instance, scipy.stats.skew (default).
eog_hor : array-like | ch_name | None
Horizontal EOG channel to which the sources shall be compared.
It has to be of the same shape as the sources. If some string
is supplied, a routine will try to find a matching channel. If
None, a score function expecting only one input-array argument
must be used, for instance, scipy.stats.skew (default).
eog_ver : array-like | ch_name | None
Vertical EOG channel to which the sources shall be compared.
It has to be of the same shape as the sources. If some string
is supplied, a routine will try to find a matching channel. If
None, a score function expecting only one input-array argument
must be used, for instance, scipy.stats.skew (default).
flow_ecg : float
Low pass frequency for ECG component identification.
fhigh_ecg : float
High pass frequency for ECG component identification.
flow_eog : float
Low pass frequency for EOG component identification.
fhigh_eog : float
High pass frequency for EOG component identification.
ecg_thresh : float
Threshold for ECG component idenfication.
eog_thresh : float
Threshold for EOG component idenfication.
use_jumeg : bool
Use the JuMEG scoring method for the identification of
artifact components.
random_state : None | int | instance of np.random.RandomState
np.random.RandomState to initialize the FastICA estimation.
As the estimation is non-deterministic it can be useful to
fix the seed to have reproducible results. Defaults to None.
Returns:
--------
ica : mne.preprocessing.ICA
ICA object for raw file with ECG and EOG components marked for removal.
"""
# increased iteration to make it converge
# fix the number of components to 40, depending on your application you
# might want to raise the number
# 'extended-infomax', 'fastica', 'picard'
logger.info('Start ICA FIT: init ICA object')
ica = ICA(method='fastica',
n_components=40,
random_state=random_state,
max_pca_components=None,
max_iter=5000,
verbose=False)
logger.debug('ICA FIT: apply ICA.fit\n reject: {} \n picks: {}'.format(
reject, picks))
ica.fit(raw, picks=picks, decim=None, reject=reject, verbose=True)
logger.info('Done ICA FIT')
#######################################################################
# identify bad components
#######################################################################
if use_jumeg:
logger.info("JuMEG Computing scores and identifying components ...")
#--- get ECG related components using JuMEG
ic_ecg, sc_ecg = get_ics_cardiac(raw,
ica,
flow=flow_ecg,
fhigh=fhigh_ecg,
thresh=ecg_thresh,
tmin=-0.5,
tmax=0.5,
name_ecg=ecg_ch,
use_CTPS=True) #[0]
ic_ecg = list(set(ic_ecg))
ic_ecg.sort()
#--- get EOG related components using JuMEG
ic_eog = get_ics_ocular(raw,
ica,
flow=flow_eog,
fhigh=fhigh_eog,
thresh=eog_thresh,
name_eog_hor=eog_hor,
name_eog_ver=eog_ver,
score_func='pearsonr')
ic_eog = list(set(ic_eog))
ic_eog.sort()
#--- if necessary include components identified by correlation as well
bads_list = []
bads_list.extend(ic_ecg)
bads_list.extend(ic_eog)
bads_list.sort()
ica.exclude = bads_list
msg = [
"JuMEG identified ICA components",
" -> ECG components: {}".format(ic_ecg),
" -> scores: {}".format(sc_ecg[ic_ecg]),
" -> EOG components: {}".format(ic_eog)
]
logger.debug("\n".join(msg))
else:
logger.info("MNE Computing scores and identifying components ...")
ecg_scores = ica.score_sources(raw,
target=ecg_ch,
score_func='pearsonr',
l_freq=flow_ecg,
h_freq=fhigh_ecg,
verbose=False)
# horizontal channel
eog1_scores = ica.score_sources(raw,
target=eog_hor,
score_func='pearsonr',
l_freq=flow_eog,
h_freq=fhigh_eog,
verbose=False)
# vertical channel
eog2_scores = ica.score_sources(raw,
target=eog_ver,
score_func='pearsonr',
l_freq=flow_eog,
h_freq=fhigh_eog,
verbose=False)
# print the top ecg, eog correlation scores
ecg_inds = np.where(np.abs(ecg_scores) > ecg_thresh)[0]
eog1_inds = np.where(np.abs(eog1_scores) > eog_thresh)[0]
eog2_inds = np.where(np.abs(eog2_scores) > eog_thresh)[0]
highly_corr = list(
set(np.concatenate((ecg_inds, eog1_inds, eog2_inds))))
highly_corr.sort()
highly_corr_ecg = list(set(ecg_inds))
highly_corr_eog1 = list(set(eog1_inds))
highly_corr_eog2 = list(set(eog2_inds))
highly_corr_ecg.sort()
highly_corr_eog1.sort()
highly_corr_eog2.sort()
# if necessary include components identified by correlation as well
ica.exclude = highly_corr
msg = [
"MNE Highly correlated artifact components",
" -> ECG : {} ".format(highly_corr_ecg),
" -> EOG 1: {} ".format(highly_corr_eog1),
" -> EOG 2: {} ".format(highly_corr_eog2)
]
logger.debug("\n".join(msg))
logger.info("done ICA FIT\n -> excluded ICs: {}\n".format(ica.exclude))
return ica
if model_scores[idx][0] > MLICA_threshold:
bads_MLICA.append(idx)
# visualisation
# ica.exclude = bads_MLICA
# ica.plot_sources(raw_ds_chop, block=True)
# compare MLICA to results from correlation and ctps analysis
ica.exclude = []
print 'Identifying components..'
# get ECG/EOG related components using JuMEG
ic_ecg = get_ics_cardiac(
raw_filtered_chop,
ica,
flow=flow_ecg,
fhigh=fhigh_ecg,
thresh=ecg_thresh,
tmin=-0.5,
tmax=0.5,
name_ecg=ecg_ch,
use_CTPS=True)[0] # returns both ICs and scores (take only ICs)
ic_eog = get_ics_ocular(raw_filtered_chop,
ica,
flow=flow_eog,
fhigh=fhigh_eog,
thresh=eog_thresh,
name_eog_hor=eog1_ch,
name_eog_ver=eog2_ch,
score_func='pearsonr')
bads_corr_ctps = list(ic_ecg) + list(ic_eog)
bads_corr_ctps = list(set(bads_corr_ctps)) # remove potential duplicates
# -------------------------------------------------------
#ica_method = 'fastica'
ica_method = 'extended-infomax'
decim = 3
ica = ICA(method=ica_method, n_components=0.99)
ica.fit(raw, picks=picks_meg, decim=decim)
# -------------------------------------------------------
# before visual inspection of ICA components
# try the automatic version to see what we get
# -------------------------------------------------------
# search for ECG artifact components (should already be removed by OCARTA)
idx_ecg = get_ics_cardiac(raw,
ica,
flow=10,
fhigh=20,
tmin=-0.3,
tmax=0.3,
name_ecg='ECG 001',
use_CTPS=True,
thresh=0.3)
# search for EOG artifact components
idx_eog1 = get_ics_ocular(raw,
ica,
flow=1,
fhigh=10,
name_eog_hor='EOG 001',
score_func='pearsonr',
thresh=0.25)
idx_eog2 = get_ics_ocular(raw,
ica,
flow=1,
verbose=False)
# print the top ecg, eog correlation scores
ecg_inds = np.where(np.abs(ecg_scores) > ecg_thresh)[0]
eog1_inds = np.where(np.abs(eog1_scores) > eog_thresh)[0]
eog2_inds = np.where(np.abs(eog2_scores) > eog_thresh)[0]
highly_corr = list(
set(np.concatenate((ecg_inds, eog1_inds, eog2_inds))))
print 'Highly correlated artefact components are', highly_corr
# get ECG/EOG related components using JUMEG
ic_ecg = get_ics_cardiac(raw_filt_chop,
ica,
flow=flow_ecg,
fhigh=fhigh_ecg,
thresh=ecg_thresh,
tmin=-0.5,
tmax=0.5,
name_ecg=ecg_ch,
score_func='pearsonr',
use_CTPS=True)
ic_eog = get_ics_ocular(raw_filt_chop,
ica,
flow=flow_eog,
fhigh=fhigh_eog,
thresh=eog_thresh,
name_eog_hor=eog1_ch,
name_eog_ver=eog2_ch,
score_func='pearsonr')
print 'Identified ECG components are: ', ic_ecg
print 'Identified EOG components are: ', ic_eog