-
Notifications
You must be signed in to change notification settings - Fork 0
/
readSubjectsfif.py
336 lines (291 loc) · 12.9 KB
/
readSubjectsfif.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
import mne
import numpy as np
import os
import matplotlib.pyplot as plt
from scipy import io
from scipy.signal import hilbert, chirp
import time
import warnings
import pandas as pd
from importlib import reload
from libs import utils,preprocessing
from collections import OrderedDict
import multiprocessing as mp
import re
import json
from libs import utils
from tqdm import tqdm
from importlib import reload # Not needed in Python 2
#train_path_meta = r'C:\Users\MarcWin\Desktop\googleDriveSnycRoot\biomag21'
def get_condition(subject: str) -> str:
"""
returns:
-----
condition: one of 'mci','dementia','control'
"""
return re.split(r'(\D+)', subject)[1]
def get_bad_times(condition,subject,signal_length=76800,frame_length=256,bad_samples_path=''):
from skimage import transform
bads = utils.load_bad_samples(condition,subject,path=bad_samples_path)
out = transform.resize(bads,[160,signal_length],anti_aliasing=False,preserve_range=True)
bads_indices = np.where(np.sum(out,axis=0)>30)[0]
earlier = np.clip(bads_indices-frame_length,0,signal_length-frame_length)
merged_bads = np.unique(earlier)
return merged_bads
def get_bad_times_one_hot(condition,subject,signal_length=76800,bad_samples_path="."):
merged_bads = get_bad_times(condition,subject,signal_length,bad_samples_path)
really_bad = np.zeros(signal_length)
really_bad[merged_bads]=1
return really_bad
def resample_goods(merged_bads,signal_length=76800,frame_length=256):
"""
We want equal length arrays with good samples
"""
if len(merged_bads)==0:
return np.arange(signal_length-frame_length)
spacing = (signal_length-frame_length)//len(merged_bads)
goods = np.concatenate([np.setdiff1d(np.arange(signal_length-frame_length),merged_bads),
(np.arange(signal_length-frame_length)[::spacing])[:len(merged_bads)]])
return goods
def detect_wss_sections(x,length_wss_section,wss_th):
"""
x np.ndarray num_Channels, num_samples
Many connectivity measures rely on Wide-Sense-Stationarity (wss).
The given MEG data was measured in a Resting State setting with eyes closed, so
we suppose constant statistical moments. Deviations are results of artificats.
-----
return is_wss, array 0 or 1, shape (num_time_steps/length_wss_section)
"""
assert (x.shape[1]%length_wss_section)==0
return np.ones(int(x.shape[1]/length_wss_section))
def subjectsId_from_subjects(subjects):
return [len(subjects['mci']),len(subjects['dementia']),len(subjects['control'])]
def make_empty_X(num_subjects,fs,num_bands=None):
#print(num_subjects,fs)
if num_bands is not None:
return np.empty((num_subjects,160,num_bands,5*60*fs))#default dtype is np.float64
else:
return np.empty((num_subjects,160,5*60*fs))#default dtype is np.float64
#
class DataLoader():
"""
todo: rename to DataLoaderFIF and build baseclase DataLoader that also has the Numpy loader
DataLoader was never used.
Compared to v2, this loader is more flexibel in order to try different preprocessing.
"""
def __init__(self,
train_path_meta:str,
data_dir:str,
ch_matching=None,
site_as_label=False,**kwargs):
"""
train_path_meta:
data_dir:
ch_matching is ignored
"""
self.train_path_meta = train_path_meta
self.controlGroup_meta = os.path.join(self.train_path_meta,'control')
self.dementiaGroup_meta = os.path.join(self.train_path_meta,'dementia')
self.mciGroup_meta = os.path.join(self.train_path_meta,'mci')
self.data_dir = data_dir
self.site_as_label =site_as_label
if site_as_label:
warnings.warn('site as label')
def readData(self,file,**kwargs):
"""
params:
condition: mci,dementia or control
subject_id: nbr of subject
-------
returns:
--------
subjectData, Dictionary
keys: {'fidpts','data','wss_section','condition','site'}
"""
filename = os.path.join(self.data_dir,file)
condition,subject_id = utils.parse_filename(filename)
# split channels
use_filter = False
if 'l_freq' in kwargs.keys():
l_freq = kwargs['l_freq']
use_filter = True
else:
l_freq = 0
if 'h_freq' in kwargs.keys():
h_freq = kwargs['h_freq']
use_filter = True
else:
h_freq = 100
if 'frame_length' in kwargs.keys():
frame_length = kwargs['frame_length']
else:
frame_length = 256
if 'num_channels' in kwargs.keys():
num_channels = kwargs['num_channels']
else:
# for KIT/Yokogawa
num_channels = 160
#if 'num_samples' in kwargs.keys():
# num_samples = kwargs['num_samples']
#else:
# # for KIT/Yokogawa at 256Hz with 5 minute recording
# num_samples = 76800
if 'fs' in kwargs.keys():
fs = kwargs['fs']
if 'utility_data' in kwargs.keys():
utility_data_path = kwargs['utility_data']
else:
utility_data_path = "."
if 'verbose' in kwargs.keys():
verbose = kwargs['verbose']
else:
verbose = 0
X = mne.io.read_raw_fif(filename,verbose=verbose)
if fs is not None and X.info['sfreq']!=fs:
X.load_data()
X.resample(fs)
# https://mne.tools/stable/auto_examples/time_frequency/plot_time_frequency_global_field_power.html?highlight=bands
if hasattr(l_freq, '__iter__'): #implies use_filter
assert len(l_freq)==len(h_freq)
#logging.info('Extracting multiple frequency bands. This can take a long time.')
bands = np.zeros((num_channels,len(l_freq),X.n_times))
for i,(lf,hf) in enumerate(zip(l_freq,h_freq)):
Xcopy = X.copy()
Xcopy.load_data()
#logging.info('Frequency Band Extraction uses a small transition bandwidth of .5 Hz.')
#logging.info('Using non causal filter. This is unwanted for evoked response detection.')
Xcopy.filter(l_freq=lf,h_freq=hf,verbose=verbose,l_trans_bandwidth=.5,h_trans_bandwidth=.5)
bands[:,i] = Xcopy.get_data()
num_samples = X.n_times
X = mne.time_frequency.AverageTFR(Xcopy.info,bands,
times=np.arange(0,num_samples)/Xcopy.info['sfreq'],
freqs=np.stack([l_freq,h_freq],axis=1),
nave=1)
elif use_filter:
#print('Load and Filter: ',condition, ' ', subject_id)
X.load_data()
X.filter(l_freq=l_freq,h_freq=h_freq,verbose=verbose,l_trans_bandwidth=.5,h_trans_bandwidth=.5)
if hasattr(X,'n_times'):
signal_length = X.n_times
else:
signal_length = X.times.shape
if 'bad_samples_path' in kwargs.keys():
bad_samples_path = kwargs['bad_samples_path']
merged_bads = get_bad_times(condition,subject_id,signal_length,bad_samples_path=bad_samples_path,
frame_length=frame_length)
goods = resample_goods(merged_bads,signal_length=signal_length,frame_length=frame_length)
else:
goods = None#==np.arange(len(signal_length))
site = utils.get_site_from_condition_number(condition,
subject_id,
direc= utility_data_path)
subjectData = {}
subjectData['data'] = X
subjectData['site'] = site
#subjectData['artifacts'] = is_wss
subjectData['condition'] = condition
subjectData['filename'] = file
subjectData['id'] = subject_id
subjectData['good_samples'] = goods
return subjectData
def make_Keras_data(self,subjects: dict,
fs:int,
use_multiprocessing = False,
preload=True,
**readDatakwargs):
"""
Creates data arrays which can be primarily used with a Keras Model.
params:
---------
subjects: dict
keys: 'dementia','mci','control'
values: list of int representing subject number
train_path: str
ch_matching: str
path to json file that matches channels in site A and site B
Raises:
------
Returns:
--------
X: (MEGArray | list of mne.RawArray)
if preload=True MEGArray (WIP might change to plain numpy array)
contains also the label e.g.: 'mci5', shape: (subjects,channels,time_steps),
if preload=False: returns list of ,mne.RawArray
y: numpy array, contains the label [0,1,2]
0 means control
1 means mci
2 means dementia
meta: dictionary
keys: siteAmeta, siteBmeta, mci{i}, control{j}, dementia{k}
"""
# todo: If subjects length(keys)<3, and keys in right group.
# then add keys with empty list as value
readDatakwargs = {**readDatakwargs,**{'fs':fs}}
conditions = ['mci','dementia','control']
print(dict(zip(conditions,subjectsId_from_subjects(subjects))))
siteAmeta = None#io.loadmat(os.path.join(self.dementiaGroup_meta,'hokuto_dementia{}.mat'.format(1)))# subject from site A
siteBmeta = None#io.loadmat(os.path.join(self.dementiaGroup_meta,'hokuto_dementia{}.mat'.format(2)))# subject from site B
num_subjects = len(subjects['mci'])+len(subjects['dementia'])+len(subjects['control'])
# compare input subjects with files in data_dir
allfiles = [f for f in os.listdir(self.data_dir) if f.endswith('.fif')]
files = []
for f in allfiles:
condition,subid = utils.parse_filename(f)
if subid in subjects[condition]:
files+=[f]
if len(files)==0:
raise ValueError('Cant find data files')
# merges constant readDataknwargs and variable subject ids.
readDatadicts = []
for f in files:
readDatadicts += [{**readDatakwargs,'file':f}]
if preload:
if 'l_freq' in readDatakwargs.keys() and hasattr(readDatakwargs['l_freq'],'__iter__'):
l_freq = readDatakwargs['l_freq']
X = make_empty_X(num_subjects,fs,num_bands=len(l_freq)).astype(np.float32)
else:
X = make_empty_X(num_subjects,fs).astype(np.float32)
else:
X = []
print('')
if use_multiprocessing:
warnings.warn('Multiprocessing has not been tested.')
with mp.Pool(processes=2) as pool:
results = pool.starmap(self.readData,readDatadicts)
else:
results = [self.readData(**subjectkwargs) for subjectkwargs in tqdm(readDatadicts,position=0,leave=True)]
meta = {'siteAmeta':siteAmeta,'siteBmeta':siteBmeta}
meta['subjects']=[]
idxes = []
y = np.empty(len(results)).astype(np.int8)
for idx,r in enumerate(results):
# todo: add check: r['data'] is mne.Raw instance
# also meta must be filled
if preload:
if hasattr(r['data'],'get_data'):
X[idx]=r['data'].get_data()*1e15
else:
X[idx]=r['data'].data*1e15
else:
X+=[r['data']]
if self.site_as_label:
y[idx] = r['site']=='B'
else:
y[idx]=utils.condition_to_digit(r['condition'])
r.pop('data')
meta['subjects']+=[r]
# check subjects delivered:
assert len(meta['subjects'])==num_subjects,('Is: {} should: {}'.format(len(meta['subjects']),num_subjects))
return X,y,meta
def load_info(self,condition,subject_id):
if condition=='dementia':
info = io.loadmat(os.path.join(self.dementiaGroup_meta,'hokuto_dementia{}.mat'.format(subject_id)))
elif condition=='mci':
info = io.loadmat(os.path.join(self.mciGroup_meta,'hokuto_mci{}.mat'.format(subject_id)))
elif condition=='control':
info = io.loadmat(os.path.join(self.controlGroup_meta,'hokuto_control{}.mat'.format(subject_id)))
else:
raise ValueError("Wrong condition provided")
return info
#loader = readSubjectsv2.DataLoaderNumpy(readSubjectsv2.train_path_meta,r'E:\2021_Biomag_Dementia_NUMPY\float32\rawfloat32','A_B_graph.json')
#out = loader.make_Keras_data({'dementia':[1],'mci':[1],'control':[1,23]},fs=128)