-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
321 lines (269 loc) · 11.6 KB
/
utils.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
import os
import numpy as np
import pandas as pd
from collections import defaultdict
class Mappings(object):
def __init__(self, num_subs=23):
# fixed mapping between binary coding and physical dims
# it's fixed because DCNN finetuned is unique.
dcnn_mapping = {
0: {'0': 'thin leg', '1': 'thick leg'},
1: {'0': 'thick antenna', '1': 'thin antenna'},
2: {'0': 'pincer mandible', '1': 'shovel mandible'}
}
# '12' and '21' are coding scheme
# 12: is the original coding scheme that is
# the same as the provided stimulus set coding.
# 21: is the flipped coding scheme that is
# the opposite as the provided stimulus set coding.
coding_scheme = {
12: {'0': '0','1': '1'},
21: {'0': '1','1': '0'}
}
behaviour_columns = {
0: 'trial',
1: 'task',
2: 'run',
3: 'dim1',
4: 'dim2',
5: 'dim3',
6: 'answer',
7: 'response',
8: 'RT',
9: 'accuracy'
}
trial_columns = {
0: 'subject',
1: 'trial',
2: 'task',
3: 'run',
4: 'stimulus onset',
5: 'feedback onset'
}
sub2assignment_n_scheme = {
'02': [2,1,3,12,12,12],
'03': [3,1,2,12,12,12],
'04': [1,2,3,21,21,12],
'05': [3,2,1,12,12,21],
'06': [3,1,2,21,12,21],
'07': [3,1,2,12,21,12],
'08': [3,2,1,21,12,12],
'09': [1,2,3,12,21,21],
'10': [2,3,1,12,12,12],
'11': [1,3,2,21,12,21],
'12': [3,2,1,12,12,21],
'13': [1,2,3,21,21,21],
'14': [2,3,1,12,12,21],
'15': [1,2,3,12,12,21],
'16': [2,3,1,12,21,21],
'17': [3,1,2,12,12,21],
'18': [2,1,3,21,21,12],
'19': [2,1,3,21,12,12],
'20': [3,1,2,12,12,12],
'21': [1,3,2,21,21,12],
'22': [1,3,2,21,12,21],
'23': [2,3,1,12,21,12],
'24': [2,1,3,12,21,21],
}
self.dcnn_mapping = dcnn_mapping
self.coding_scheme = coding_scheme
self.behaviour_columns = behaviour_columns
self.trial_columns = trial_columns
self.sub2assignment_n_scheme = sub2assignment_n_scheme
def convert_dcnnCoding_to_subjectCoding(sub):
"""
In order to extract stimulus-specific activations (for RSA later),
we need to first establish the mapping between stimulus coding of
DCNN (which is fixed due to finetuning) and stimulus coding of
subjects (which differs for each subject due to random scheme).
Since the ultimate purpose is to find brain activation of each stimulus
(in terms of DCNN coding), we need to find the coding used by each subject
of every stimulus.
e.g. In terms of DCNN coding, stimulus 000 (thin leg, thick antenna, pincer mandible)
so what is the coding for this subject? i.e. thin leg=? thick antenna=? pincer mandible=?
For different subs, this coding is different, depending on '12' or '21' random scheme:
e.g. if DCNN=101,
with assignment 213 and scheme 12, 12, 12
101 -> 011 -> 011
with assignment 312 and scheme 12, 21, 12
101 -> 110 -> 100
return:
-------
Given all DCNN stimuli, return the conversion ordering for a given subject.
E.g. for subject a, the orders of the stimuli should be [6, 1, 7, 5, 4, 2, 3, 0]
where 6 corresponds to 000 in DCNN coding but 110 in subject coding.
"""
sub2assignment_n_scheme = Mappings().sub2assignment_n_scheme
coding_scheme = Mappings().coding_scheme
conversion_ordering = []
stimulus2order_mapping = {
'000': 0, '001': 1, '010': 2, '011': 3,
'100': 4, '101': 5, '110': 6, '111': 7,
}
for dcnn_stimulus in ['000', '001', '010', '011', '100', '101', '110', '111']:
sub_stimulus = [i for i in dcnn_stimulus]
# print(f'\n\n--------------------------------')
# print(f'[Check] DCNN stimulus {sub_stimulus}')
# assignment (flip three dims)
assignment_n_scheme = sub2assignment_n_scheme[sub]
new_stimulus_0 = sub_stimulus[assignment_n_scheme[0]-1]
new_stimulus_1 = sub_stimulus[assignment_n_scheme[1]-1]
new_stimulus_2 = sub_stimulus[assignment_n_scheme[2]-1]
sub_stimulus[0] = new_stimulus_0
sub_stimulus[1] = new_stimulus_1
sub_stimulus[2] = new_stimulus_2
# print(f'[Check] sub{sub}, assignment stimulus {sub_stimulus}')
# scheme (flip binary codings)
dim1_scheme = assignment_n_scheme[3]
dim2_scheme = assignment_n_scheme[4]
dim3_scheme = assignment_n_scheme[5]
sub_stimulus[0] = coding_scheme[dim1_scheme][sub_stimulus[0]]
sub_stimulus[1] = coding_scheme[dim2_scheme][sub_stimulus[1]]
sub_stimulus[2] = coding_scheme[dim3_scheme][sub_stimulus[2]]
# print(f'[Check] sub{sub}, scheme stimulus {sub_stimulus}')
conversion_ordering.append(
stimulus2order_mapping[
''.join(sub_stimulus)
]
)
return np.array(conversion_ordering)
def reorder_RDM_entries_into_chunks():
"""
For each subject, the groud true label for each stimulus coding is different.
When visualising RDMs of conditions, we want to make sure that
rows and columns are grouped (in chunk) by their labels. This asks
for a mapping from each subject's stimulus coding to their labels in each
task.
Impl:
-----
We create a dictionary like:
mapping = {sub1:
{task1: [orderOf(000), orderOf(001), ...],
task2: [orderOf(000), orderOf(001), ...],
...
Notice, within each task (across runs), the labels are the same so we can
simply use run1.
Notice, there is one extra conversion inside to get the orderOf(..).
That is, after we get a list of labels correspond to 000, 001, ..., 111, which look
like [1, 2, 2, 1, 2, ...], we argsort them to get indices which we will return as
`conversion_ordering` that is going to reorder the RDM entries into desired chunks.
return:
-------
`mapping` explained above
"""
mapping = defaultdict(lambda: defaultdict(list))
num_subs = 23
tasks = [1, 2, 3]
subs = [f'{i:02d}' for i in range(2, num_subs+2)]
stimuli = ['000', '001', '010', '011', '100', '101', '110', '111']
behaviour_path = f'Mack-Data/behaviour'
for sub in subs:
for task in tasks:
behaviour = pd.read_csv(
f'{behaviour_path}/subject_{sub}/{sub}_study{task}_run1.txt',
header=None
).to_numpy()
i = 0
temp_mapping = dict()
# search exactly all stimuli once and stop.
while len(temp_mapping.keys()) != len(stimuli):
behaviour_i = behaviour[i][0].split('\t')
stimulus = ''.join(behaviour_i[3:6])
label = behaviour_i[6] # 6 - ground true answer
print(f'stimulus = {stimulus}, label = {label}')
temp_mapping[stimulus] = int(label)
i += 1
# this is to reorder the stimuli as 000, 001, ..., 111
# so the corresponding list of labels match the order.
labels = []
for stimulus in stimuli:
labels.append(temp_mapping[stimulus])
# sort the labels and get indices in asc order
grouped_labels_indices = np.argsort(labels)
print(f'sub{sub}, task{task}, labels={labels}, order={grouped_labels_indices}')
print('----------------------------------------------------------------------')
mapping[sub][task].extend(grouped_labels_indices)
# mapping[sub][task] = a list of indices that will be used to sort the RDM entries.
return mapping
def prepare_events_table(sub, task, run, save_dir):
"""
Given a subject, task and run, produce a
table which contains:
onset duration weight stimulus (i.e. trial_type)
----- -------- ------ --------
- Both `onset` and `duration` are from 'trialtiming/'
- weight is 1
- trial_type is from 'behaviour/'
Note, each presentation (i.e. repetition) of the stimulus,
is considered a unique condition hence has its own regressor and beta weight.
"""
output_dir = f'{save_dir}/events'
if not os.path.exists(output_dir):
os.makedirs(output_dir)
trialtiming_path = f'Mack-Data/trialtiming/{sub}_study{task}_run{run}.txt'
behaviour_path = f'Mack-Data/behaviour/subject_{sub}/{sub}_study{task}_run{run}.txt'
trialtiming = pd.read_csv(trialtiming_path, header=None).to_numpy()
behaviour = pd.read_csv(behaviour_path, header=None).to_numpy()
onsets = ['onset']
durations = ['duration']
weights = ['weight']
stimuli = ['stimulus']
# all 32 trials of 1 run
for i in range(len(trialtiming)):
# 1, 2, 3, 4
repetition = i // 8 + 1
trialtiming_i = trialtiming[i][0].split('\t')
behaviour_i = behaviour[i][0].split('\t')
# stimulus events
stimulus_onset = int(trialtiming_i[4])
stimulus_duration = 3.5
stimulus = ''.join(behaviour_i[3:6]) # convert ['1', '0', '1'] to '101'
stimulus = f'{stimulus}_rp{repetition}'
onsets.append(stimulus_onset)
durations.append(stimulus_duration)
stimuli.append(stimulus)
# feedback events
feedback_onset = int(trialtiming_i[5])
feedback_duration = 2.0
stimulus_fb = f'{stimulus}_fb'
onsets.append(feedback_onset)
durations.append(feedback_duration)
stimuli.append(stimulus_fb)
onsets = np.array(onsets)
durations = np.array(durations)
stimuli = np.array(stimuli)
df = np.vstack((onsets, durations, stimuli)).T
pd.DataFrame(df).to_csv(
f"{output_dir}/sub-{sub}_task-{task}_run-{run}_events.tsv",
sep='\t', index=False, header=False
)
print(f'[Check] Saved events tsv.')
def prepare_motion_correction_params(sub, task, run, save_dir):
"""
For fitting GLM, we will need to extract
mc params from the preprocessed data.
"""
output_dir = f'{save_dir}/mc_params'
if not os.path.exists(output_dir):
os.makedirs(output_dir)
confounds_path = f'Mack-Data/derivatives/sub-{sub}/func/' \
f'sub-{sub}_task-{task}_run-{run}_desc-confounds_timeseries.tsv'
mc_params = pd.read_table(
confounds_path,
usecols=[
'trans_x', 'trans_y', 'trans_z',
'rot_x', 'rot_y', 'rot_z'
]
)
pd.DataFrame(mc_params).to_csv(
f"{output_dir}/sub-{sub}_task-{task}_run-{run}_mc_params.tsv",
sep='\t', index=False, header=False
)
print(f'[Check] Saved mc_params.')
if __name__ == '__main__':
# convert_dcnnCoding_to_subjectCoding(dcnn_stimulus='101', sub='23')
# mapping = reorder_RDM_entries_into_chunks()
# print(mapping['02'][1])
# print(mapping['03'][2])
base_dir = 'glm_trial-estimate'
prepare_events_table(sub='03', task=2, run=4, save_dir=base_dir)