-
Notifications
You must be signed in to change notification settings - Fork 0
/
FPVS_Compute_ICA_sweep.py
executable file
·516 lines (348 loc) · 16.8 KB
/
FPVS_Compute_ICA_sweep.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
#!/imaging/local/software/miniconda/envs/mne0.20/bin/python
"""
Compute ICA for FPVS Frequency Sweep.
Based on Fiff_Compute_ICA.py.
==========================================
OH July 2019
"""
# TO DO: change to compute ICA across concatenated raw files/epochs
# often only ~6 blinks per run
# Subject 18 has not detectable eye-blinks in at least one run, lfhf1
import sys
from os import path as op
import numpy as np
# import matplotlib
# matplotlib.use('Agg') # for running graphics on cluster ### EDIT
from matplotlib import pyplot as plt
from importlib import reload
import mne
from mne.preprocessing import ICA, create_eog_epochs, create_ecg_epochs
from mne.report import Report
import config_sweep as config
reload(config)
print('MNE Version: %s\n\n' % mne.__version__) # just in case
print(mne)
# whether to show figures on screen or just write to file
show = False
# conditions
conds = config.do_conds
# "emulate" the args from ArgParser in Fiff_Compute_ICA.py
# filenames depend on subject, the rest are variables
class create_args:
def __init__(self, FileRaw, FileICA, FileHTML):
self.FileRaw = FileRaw
self.FileICA = FileICA
self.FileHTML = FileHTML
self.EOG = ['EOG062']
self.ECG = ['ECG063'] # test 'synth' option if no ECG channel present
self.maxEOG = 2
self.maxECG = 2
self.ECGmeth = 'ctps'
self.EOGthresh = 2.5
self.ECGthresh = 0.05
self.ChanTypes = ['eeg', 'meg']
self.RejEEG = config.reject['eeg']
self.RejGrad = config.reject['grad']
self.RejMag = config.reject['mag']
self.n_pca_comps = '0.99' # string required
self.method = 'infomax'
def run_Compute_ICA(sbj_id):
"""Compute ICA for one subject."""
# path to subject's data
sbj_path = op.join(config.data_path, config.map_subjects[sbj_id][0])
# raw-filename mappings for this subject
tmp_fnames = config.sss_map_fnames[sbj_id][1]
# only use files for correct conditions
sss_map_fnames = []
for cond in conds:
for [fi, ff] in enumerate(tmp_fnames):
if cond in ff:
sss_map_fnames.append(ff)
print(sss_map_fnames)
# Concatenate raws. These raw files are not huge.
raw_all = [] # will contain list of all raw files
for raw_stem_in in sss_map_fnames:
FileRaw = op.join(sbj_path, raw_stem_in[:-7] + 'sss_f_raw')
# Read raw data info
raw1 = mne.io.read_raw_fif(FileRaw + '.fif', preload=True)
raw_all.append(raw1)
# concatenate raws
print('Concatenating %d raw files.' % len(raw_all))
raw = mne.concatenate_raws(raw_all)
# -ica.fif will be appended
FileICA = op.join(sbj_path, config.map_subjects[sbj_id][0] + '_sss_f_raw')
# -ica.html will be appended
FileHTML = op.join(sbj_path, config.map_subjects[sbj_id][0] + '_sss_f_raw')
# define variables for the following ICA pipeline
# this would be from command line arguments of Fiff_Compute_ICA.py
args = create_args(FileRaw, FileICA, FileHTML)
# If a channel for ECG detection explicity specified, use it
if config.ECG_channels[sbj_id] != '':
args.ECG = [config.ECG_channels[sbj_id]]
# otherwise use ECG from data, but if not present, dont' do ICA for ECG
elif not raw.__contains__('ecg'):
args.ECG = []
print('###\nNo ECG found in raw data, so I am not doing it.\n###')
# Now turn the "fake command line parameters" into variables for the
# analysis pipeline
# if float, select n_components by explained variance of PCA
if '.' in args.n_pca_comps:
n_components = float(args.n_pca_comps)
print('Number of PCA components by fraction of variance (%f)' %
n_components)
else:
n_components = int(args.n_pca_comps)
print('Number of PCA components: %d.' % n_components)
method = args.method # for comparison with EEGLAB "extended-infomax"
print('\nUsing ICA method %s.' % method)
decim = 3 # downsample data to save time
# same random state for each ICA (not sure if beneficial?)
random_state = 23
# raw data input filename, not needed here for concatenated raws
raw_fname_in = args.FileRaw + '.fif'
# filename for ICA output
if args.FileICA == '':
ica_fname_out = args.FileRaw + '-ica.fif'
else:
ica_fname_out = args.FileICA + '-ica.fif'
# filename for ICA output
if args.FileHTML == '':
fname_html = args.FileRaw + '-ica.html'
else:
fname_html = args.FileHTML + '-ica.html'
###
# START ICA
###
report = Report(subject=config.map_subjects[sbj_id][0], title='ICA:')
# print('###\nReading raw file %s.' % raw_fname_in)
# # Read raw data
# raw = mne.io.read_raw_fif(raw_fname_in, preload=True)
# check if EEG in raw data
if not raw.__contains__('eeg'):
args.ChanTypes = ['meg']
print('###\nNo EEG found in raw data, continuing with MEG only.\n###')
# They say high-pass filtering helps
print('Band-pass filtering raw data between 1 and 40 Hz.')
raw.filter(1., 40., fir_design='firwin')
# which channel types to use
to_pick = {'meg': False, 'eeg': False, 'eog': False, 'stim': False,
'exclude': 'bads'}
# pick channel types as specified
print('Using channel types: ')
for chtype in args.ChanTypes:
print(chtype + ' ')
to_pick[chtype.lower()] = True
picks_meg_eeg_eog = mne.pick_types(raw.info, meg=to_pick['meg'],
eeg=to_pick['eeg'],
eog=True, ecg=True,
stim=to_pick['stim'],
exclude=to_pick['exclude'])
# to remove non-physiological artefacts (parameters based on MNE example)
reject = {}
if to_pick['meg'] == True:
reject['mag'] = args.RejMag
reject['grad'] = args.RejGrad
print('Thresholds for MEG: Grad %.1e, Mag %.1e.' % (reject['grad'], reject['mag']))
if to_pick['eeg'] == True:
reject['eeg'] = args.RejEEG
print('Threshold for EEG: %.1e.' % reject['eeg'])
picks_meg = mne.pick_types(raw.info, meg=to_pick['meg'],
eeg=to_pick['eeg'],
eog=to_pick['eog'], stim=to_pick['stim'],
exclude=to_pick['exclude'])
# Compute ICA model ###################################################
print('###\nDefine the ICA object instance using %s. \
Number of PCA components based on: %s.' %
(method, str(n_components)))
ica = ICA(n_components=n_components, method=method,
random_state=random_state)
print('Fitting ICA.')
ica.fit(raw, picks=picks_meg, decim=decim, reject=reject)
print(ica)
print('Plotting ICA components.')
# plot for specified channel types
for ch_type in reject.keys():
fig_ic = ica.plot_components(ch_type=ch_type, show=show)
captions = [ch_type.upper() + ' Components' for i in fig_ic]
report.add_figs_to_section(fig_ic, captions=captions,
section='ICA Components', scale=1)
# indices of ICA components to be removed across EOG and ECG
ica_inds = []
###
# EOG COMPONENTS
###
# for all specified EOG channels
eog_inds = [] # ICA components found to be bad for EOG
eog_scores = [] # corresponding ICA scores
for eog_ch in args.EOG:
print('\n###\nFinding components for EOG channel %s.\n' % eog_ch)
# get single EOG trials
eog_epochs = create_eog_epochs(raw, ch_name=eog_ch, reject=reject)
eog_average = eog_epochs.average() # average EOG epochs
# find via correlation
inds, scores = ica.find_bads_eog(eog_epochs, ch_name=eog_ch,
threshold=args.EOGthresh)
if inds != []: # if some bad components found
print('###\nEOG components and scores for channel %s:\n' % eog_ch)
for [ee, ss] in zip(inds, scores):
print('%d: %.2f\n' % (ee, ss))
# look at r scores of components
fig_sc = ica.plot_scores(scores, exclude=inds, show=show)
report.add_figs_to_section(fig_sc, captions='%s Scores' %
eog_ch, section='%s ICA component \
scores' % eog_ch, scale=1)
print('Plotting raw ICA sources.')
fig_rc = ica.plot_sources(raw, show=show)
report.add_figs_to_section(fig_rc, captions='%s Sources' %
eog_ch, section='%s raw ICA sources'
% eog_ch, scale=1)
print('Plotting EOG average sources.')
# look at source time course
fig_so = ica.plot_sources(eog_average, show=show)
report.add_figs_to_section(fig_so, captions='%s Sources' %
eog_ch, section='%s ICA Sources' %
eog_ch, scale=1)
print('Plotting EOG epochs properties.')
fig_pr = ica.plot_properties(eog_epochs, picks=inds,
psd_args={'fmax': 35.},
image_args={'sigma': 1.},
show=show)
txt_str = '%s Properties' % eog_ch
captions = [txt_str for i in fig_pr]
report.add_figs_to_section(fig_pr, captions=captions,
section='%s ICA Properties' %
eog_ch, scale=1)
print(ica.labels_)
# Remove ICA components #######################################
fig_ov = ica.plot_overlay(eog_average, exclude=inds, show=show)
# red -> before, black -> after.
report.add_figs_to_section(fig_ov, captions='%s Overlay' %
eog_ch, section='%s ICA Overlay' %
eog_ch, scale=1)
plt.close('all')
eog_inds += inds # keep bad ICA components
# keep scores for bad ICA components
eog_scores += list(scores[inds])
else:
print('\n!!!Nothing bad found for %s!!!\n' % eog_ch)
if eog_inds != []: # if there are bad ECG components
# deal with case where there are more bad ICA components than
# specified
n_comps = np.min([args.maxEOG, len(eog_inds)])
print('\n###\nUsing %d out of %d detected ICA components for EOG.'
% (n_comps, len(eog_inds)))
for [c, s] in zip(eog_inds, eog_scores):
print('Component %d with score %f.' % (c, s))
# sort to find ICA components with highest scores
idx_sort = np.argsort(np.abs(eog_scores))
# only keep desired number of bad ICA components with high scores
ica_inds += [eog_inds[idx] for idx in idx_sort[-n_comps:]]
#
# ECG COMPONENTS
#
# for all specified EOG channels
ecg_inds = [] # ICA components found to be bad for ECG
ecg_scores = [] # corresponding ICA scores
for ecg_ch in args.ECG:
if ecg_ch == 'synth':
print('Creating synthetic ECG channel.')
# check which channel, if any, is ECG
ecg_idx = np.where(['ECG' in ch for ch in
raw.info['ch_names']])[0]
# if there is an ECG channel, change it
if not ecg_idx.shape[0] == 0:
ecg_name = raw.info['ch_names'][ecg_idx[0]]
raw.set_channel_types({ecg_name: 'misc'})
# create synthetic ECG channel across MEG channels
ecg_ch_name = None # for create_ecg_epochs
ecg_find_name = 'ECG-SYN'
keep_ecg = True
else:
ecg_ch_name = ecg_ch
ecg_find_name = ecg_ch
keep_ecg = False
print('\n###\nFinding components for ECG channel %s.\n' % ecg_ch)
# get single ECG trials
ecg_epochs = create_ecg_epochs(raw, ch_name=ecg_ch_name,
keep_ecg=keep_ecg, reject=reject)
ecg_average = ecg_epochs.average() # average ECG epochs
# find via cross-trial phase statistics
inds, scores = ica.find_bads_ecg(ecg_epochs, ch_name=ecg_find_name,
method=args.ECGmeth,
threshold=args.ECGthresh)
if inds != []: # if some bad components found
print('ECG components and scores:\n')
for [ee, ss] in zip(inds, scores):
print('%d: %.2f\n' % (ee, ss))
# look at r scores of components
fig_sc = ica.plot_scores(scores, show=show)
report.add_figs_to_section(fig_sc, captions='%s Scores' %
ecg_ch, section='%s component \
scores' % ecg_ch, scale=1)
print('Plotting raw ICA sources.')
fig_rc = ica.plot_sources(raw, show=show)
report.add_figs_to_section(fig_rc, captions='%s Sources' %
ecg_ch, section='%s raw sources'
% ecg_ch, scale=1)
print('Plotting ECG average sources.')
# look at source time course
fig_so = ica.plot_sources(ecg_average, show=show)
report.add_figs_to_section(fig_so, captions='%s Sources' %
ecg_ch, section='%s ICA Sources' %
ecg_ch, scale=1)
print('Plotting ECG epochs properties.')
fig_pr = ica.plot_properties(ecg_epochs, picks=inds,
psd_args={'fmax': 35.},
image_args={'sigma': 1.},
show=show)
txt_str = '%s Properties' % ecg_ch
captions = [txt_str for i in fig_pr]
report.add_figs_to_section(fig_pr, captions=captions,
section='%s ICA Properties' %
ecg_ch, scale=1)
print(ica.labels_)
# Remove ICA components #######################################
fig_ov = ica.plot_overlay(ecg_average, exclude=inds, show=show)
# red -> before, black -> after. Yes! We remove quite a lot!
report.add_figs_to_section(fig_ov, captions='%s Overlay' %
ecg_ch, section='%s ICA Overlay' %
ecg_ch, scale=1)
plt.close('all')
ecg_inds += inds # keep bad ICA components
ecg_scores += list(scores[inds]) # keep bad ICA components
else:
print('\n!!!Nothing bad found for %s!!!\n' % ecg_ch)
if ecg_inds != []: # if there are bad ECG components
# deal with case where there are more bad ICA components than
# specified
n_comps = np.min([args.maxECG, len(ecg_inds)])
print('\n###\nUsing %d out of %d detected ICA components for ECG.'
% (n_comps, len(ecg_inds)))
for [c, s] in zip(ecg_inds, ecg_scores):
print('Component %d with score %f.' % (c, s))
# sort to find ICA components with highest scores
idx_sort = np.argsort(np.abs(ecg_scores))
# only keep desired number of bad ICA components with high scores
ica_inds += [ecg_inds[idx] for idx in idx_sort[-n_comps:]]
if ica_inds == []:
print('###\nNo bad components found anywhere.\n###')
# specify components to be removed
ica.exclude = ica_inds
###
# SAVE ICA
###
# from now on the ICA will reject this component even if no exclude
# parameter is passed, and this information will be stored to disk
# on saving
print('\nSaving ICA to %s' % (ica_fname_out))
ica.save(ica_fname_out)
print('Saving HTML report to {0}'.format(fname_html))
report.save(fname_html, overwrite=True, open_browser=True)
# get all input arguments except first
if len(sys.argv) == 1:
sbj_ids = np.arange(0,len(config.map_subjects)) + 1
else:
# get list of subjects IDs to process
sbj_ids = [int(aa) for aa in sys.argv[1:]]
for ss in sbj_ids:
run_Compute_ICA(ss)