-
Notifications
You must be signed in to change notification settings - Fork 0
/
roi_rsa.py
539 lines (455 loc) · 18.8 KB
/
roi_rsa.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
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
import os
import scipy
import time
import numpy as np
import multiprocessing
import scipy.stats as stats
from subprocess import call
import matplotlib.pyplot as plt
from matplotlib import colorbar, colors
from matplotlib.colors import ListedColormap
from sklearn.metrics import pairwise_distances
import nibabel as nb
import nilearn as nl
import nipype.interfaces.ants as ants
from nipype.interfaces.fsl import MultiImageMaths
from nilearn.masking import apply_mask
from nilearn import plotting, image
from utils import convert_dcnnCoding_to_subjectCoding, reorder_RDM_entries_into_chunks
"""
1. Extract ROI-level beta weights and compile into RDMs.
2. Perform RSA
"""
def run_ants_command(roi, roi_path, roi_nums, smooth_mask=False):
"""
Helper function that automatically grabs files to prepare
for the final ants command.
"""
maths = MultiImageMaths()
# V1,2,3,1-3,4,LOC (left+right)
if roi_nums is not None:
all_files = []
for roi_num in roi_nums:
all_files.append(f'{roi_path}/perc_VTPM_vol_roi{roi_num}_lh.nii.gz')
all_files.append(f'{roi_path}/perc_VTPM_vol_roi{roi_num}_rh.nii.gz')
maths.inputs.op_string = ''
for _ in range(len(roi_nums) * 2 - 1):
maths.inputs.op_string += '-add %s '
if smooth_mask:
maths.inputs.op_string += f'-s {smooth_mask}'
maths.inputs.op_string += ' -bin '
maths.inputs.in_file = all_files[0]
maths.inputs.operand_files = all_files[1:]
# HPC
elif 'HHPC' in roi:
if 'LH' in roi:
h = 'l'
else:
h = 'r'
maths.inputs.in_file = f'{roi_path}/HIPP_BODY_{h}h.nii.gz'
maths.inputs.op_string = '-add %s -add %s '
if smooth_mask:
maths.inputs.op_string += f'-s {smooth_mask}'
maths.inputs.op_string += ' -bin '
maths.inputs.operand_files = [
f'{roi_path}/HIPP_HEAD_{h}h.nii.gz',
f'{roi_path}/HIPP_TAIL_{h}h.nii.gz'
]
# LHLOC, RHLOC
elif 'HLOC' in roi:
if 'LH' in roi:
h = 'l'
else:
h = 'r'
# WARNING: hardcoded roi_num
maths.inputs.in_file = f'{roi_path}/perc_VTPM_vol_roi14_{h}h.nii.gz'
maths.inputs.op_string = '-add %s '
if smooth_mask:
maths.inputs.op_string += f'-s {smooth_mask}'
maths.inputs.op_string += ' -bin '
maths.inputs.operand_files = [
f'{roi_path}/perc_VTPM_vol_roi15_{h}h.nii.gz'
]
maths.inputs.out_file = f'{roi_path}/mask-{roi}.nii.gz'
runCmd = '/usr/bin/fsl5.0-' + maths.cmdline
print(f'runCmd = {runCmd}')
call(runCmd, shell=True)
def merge_n_smooth_mask(roi, roi_path, smooth_mask):
"""
First processing of the standard ROI masks is to
merge some left&right masks and smooth them.
"""
print(f'[Check] running `merge_n_smooth_mask`')
if not os.path.exists(f'{roi_path}/mask-{roi}.nii.gz'):
if roi in ['LHHPC', 'RHHPC', 'LHLOC', 'RHLOC']:
# If the above ROI, there is no left+right merge
roi_nums = None
else:
# Will need to merge left+right
roi_number_mapping = {
'V1': [1, 2],
'V2': [3, 4],
'V3': [5, 6],
'V4': [7],
'V1-3': [1, 2, 3, 4, 5, 6],
'LOC': [14, 15]
}
roi_nums = roi_number_mapping[roi]
run_ants_command(
roi=roi,
roi_path=roi_path,
roi_nums=roi_nums,
smooth_mask=smooth_mask
)
else:
print(f'[Check] mask-{roi} already done, skip')
def transform_mask_MNI_to_T1(sub, roi, roi_path, root_path):
"""
Given a subject and a ROI,
transform ROI mask from MNI space to subject's T1 space.
This is based on the fact that the ROI masks provided are already
in standard MNI space.
"""
print(f'[Check] running `transform_mask_MNI_to_T1`')
if not os.path.exists(f'{roi_path}/mask-{roi}_T1_sub-{sub}.nii.gz'):
print(f'[Check] transform roi mask to subject{sub} T1 space')
at = ants.ApplyTransforms()
reference_image_path = f'{root_path}/Mack-Data/dropbox/sub-{sub}/anat'
transform_path = f'{root_path}/Mack-Data/derivatives/sub-{sub}/anat'
assert os.path.exists(reference_image_path)
assert os.path.exists(transform_path)
at.inputs.dimension = 3
at.inputs.input_image = f'{roi_path}/mask-{roi}.nii.gz'
at.inputs.reference_image = f'{reference_image_path}/sub-{sub}_T1w.nii.gz'
at.inputs.output_image = f'{roi_path}/mask-{roi}_T1_sub-{sub}.nii.gz'
at.inputs.interpolation = 'NearestNeighbor'
at.inputs.default_value = 0
at.inputs.transforms = [f'{transform_path}/sub-{sub}_from-MNI152NLin2009cAsym_to-T1w_mode-image_xfm.h5']
at.inputs.invert_transform_flags = [False]
runCmd = at.cmdline
call(runCmd, shell=True)
else:
print(f'[Check] mask-{roi}_T1_sub-{sub} Already done, skip')
def applyMask(roi, root_path, glm_path, roi_path, sub, task, run, dataType, condition, smooth_beta):
"""
Apply ROI mask (T1 space) to subject's whole brain beta weights.
return:
-------
per (ROI, subject, task, run, condition) beta weights
"""
# print(f'[Check] apply mask..')
output_path = f'output_run_{run}_sub_{sub}_task_{task}'
if dataType == 'beta':
data_path = f'{root_path}/{glm_path}/work_1st/{output_path}/datasink/' \
f'{glm_path}/datasink/model/{output_path}/{dataType}_{condition}.nii'
elif dataType == 'spmT':
data_path = f'{root_path}/{glm_path}/work_1st/{output_path}/datasink/' \
f'{glm_path}/datasink/1stLevel/{output_path}/{dataType}_{condition}.nii'
imgs = nb.load(data_path)
maskROI = nb.load(f'{roi_path}/mask-{roi}_T1_sub-{sub}.nii.gz')
maskROI = nl.image.resample_img(
maskROI,
target_affine=imgs.affine,
target_shape=imgs.shape[:3],
interpolation='nearest'
)
# print('maskROI.shape = ', maskROI.shape)
fmri_masked = apply_mask(
imgs, maskROI, smoothing_fwhm=smooth_beta
)
# print('fmri_masked.shape, ', fmri_masked.shape) # per ROI & subject & condition beta weights
return roi, maskROI, fmri_masked
def return_RDM(embedding_mtx, sub, task, run, roi, distance, RDM_fpath):
"""
Compute and save RDM or just load of given beta weights and sort
the conditions based on specified ordering.
"""
if not os.path.exists(rdm_path):
os.mkdir(rdm_path)
if len(embedding_mtx) != 0:
print(f'[Check] Computing RDM..')
if distance == 'euclidean':
RDM = pairwise_distances(embedding_mtx, metric='euclidean')
elif distance == 'pearson':
RDM = pairwise_distances(embedding_mtx, metric='correlation')
# rearrange so entries are grouped by two categories.
conversion_ordering = reorder_mapper[sub][task]
print(f'[Check] sub{sub}, task{task}, conversion_ordering={conversion_ordering}')
# reorder both cols and rows based on ordering.
RDM = RDM[conversion_ordering, :][:, conversion_ordering]
np.save(RDM_fpath, RDM)
print(f'[Check] Saved: {RDM_fpath}')
assert RDM.shape == (embedding_mtx.shape[0], embedding_mtx.shape[0])
else:
print(f'[Check] Already exists, {RDM_fpath}')
def applyMask_returnRDM(roi, root_path, glm_path, roi_path, sub, task, run, dataType, conditions, smooth_beta, distance):
"""
Combines `applyMask` and `returnRDM` in one function,
this is done so to enable multiprocessing.
NOTE:
Under `trial-level-glm`, the RDM needs to be saved at
the repetition level. That is, each run, there are 4 RDMs.
"""
for rp in range(1, num_repetitions_per_run+1):
# Check if the RDM of a run's last repetition is non-existent, then
# consider this is a new run entirely. Otherwise just load the saved RDM
RDM_fpath = f'{rdm_path}/sub-{sub}_task-{task}_run-{run}' \
f'_rp-{rp}_roi-{roi}_{distance}_{dataType}.npy'
if not os.path.exists(RDM_fpath):
conditions_of_the_same_rp = conditions[rp-1::num_repetitions_per_run]
assert len(conditions_of_the_same_rp) == 8
# embedding of the same rp
beta_weights_masked = []
for condition in conditions_of_the_same_rp:
# given beta weights from a task & run & condition
# apply the transformed mask
roi, maskROI, fmri_masked = applyMask(
roi=roi,
root_path=root_path,
glm_path=glm_path,
roi_path=roi_path,
sub=sub,
task=task,
run=run,
dataType=dataType,
condition=condition,
smooth_beta=smooth_beta
)
beta_weights_masked.append(fmri_masked)
beta_weights_masked = np.array(beta_weights_masked)
print('beta_weights_masked.shape', beta_weights_masked.shape)
# Either way, return RDM
return_RDM(
embedding_mtx=beta_weights_masked,
sub=sub,
task=task,
run=run,
roi=roi,
distance=distance,
RDM_fpath=RDM_fpath
)
def kendall_a(a, b):
"""Kendalls tau-a
Arguments:
a {array} -- [description]
b {[array]} -- [description]
Returns:
tau -- Kendalls tau-a
E.g.
x1 = np.random.random(10000)
x2 = np.random.random(10000)
print(kendall_a(x1,x2))
"""
a, b = np.array(a), np.array(b)
assert a.size == b.size, 'Both arrays need to be the same size'
K = 0
n = a.size
for k in range(n):
pairRelations_a = np.sign(a[k]-a[k+1:])
pairRelations_b = np.sign(b[k]-b[k+1:])
K = K + np.sum(pairRelations_a * pairRelations_b)
return K/(n*(n-1)/2)
def compute_RSA(RDM_1, RDM_2, method):
"""
Compute spearman correlation between
two RDMs' upper trigular entries
"""
RDM_1_triu = RDM_1[np.triu_indices(RDM_1.shape[0])]
RDM_2_triu = RDM_2[np.triu_indices(RDM_2.shape[0])]
if method == 'spearman':
rho, _ = stats.spearmanr(RDM_1_triu, RDM_2_triu)
elif method == 'kendall_a':
rho = kendall_a(RDM_1_triu, RDM_2_triu)
return rho
def roi_execute(
rois, subs, tasks, runs,
dataType, conditions, distances,
smooth_mask, smooth_beta,
num_processes
):
"""
This is a top-level execution routine that does the following in order:
1. `merge_n_smooth_mask`:
- merge given ROI masks by left+right hemisphere and apply smoothing.
- this step is subject general as it's in the MNI space.
- the merged and smoothed masks are saved in `ROIs/`
2. `transform_mask_MNI_to_T1`
- transform the saved masks from MNI space to subject's T1 space.
- this step is subject specific.
- the transformed masks are saved in `ROIs/*_T1.nii.gz`
3. `applyMask`
- extract beta weights based on ROI masks.
- this step is done for tasks, runs and conditions, one at a time.
4. `return_RDM`
- convert all stimuli beta weights (i.e. embedding matrix) into
RDM based on provided distance metric.
- notice RDM entries need somehow reordered.
"""
with multiprocessing.Pool(num_processes) as pool:
for roi in rois:
if 'HPC' not in roi:
roi_path = 'ROIs/ProbAtlas_v4/subj_vol_all'
else:
roi_path = 'ROIs/HPC'
merge_n_smooth_mask(roi=roi, roi_path=roi_path, smooth_mask=smooth_mask)
for sub in subs:
transform_mask_MNI_to_T1(sub=sub, roi=roi, roi_path=roi_path, root_path=root_path)
for task in tasks:
for run in runs:
for distance in distances:
# Create a single process to produce
# 1 RDM.
results = pool.apply_async(
applyMask_returnRDM,
args=[
roi, root_path, glm_path, roi_path, sub, task, run, dataType,
conditions, smooth_beta, distance
]
)
pool.close()
pool.join()
def visualize_RDM(sub, problem_type, distance, run, repetition, roi):
"""
Visualize subject's RDM given a problem_type
"""
if int(sub) % 2 == 0:
if problem_type == 1:
task = 2
elif problem_type == 2:
task = 3
else:
task = 1
# odd sub: Type1 is task3, Type2 is task2
else:
if problem_type == 1:
task = 3
elif problem_type == 2:
task = 2
else:
task = 1
RDM = np.load(
f'subject_RDMs/sub-{sub}_task-{task}_run-{run}_rp-{repetition}_roi-{roi}_{distance}_beta.npy'
)
fig, ax = plt.subplots()
for i in range(RDM.shape[0]):
for j in range(RDM.shape[0]):
text = ax.text(
j, i, np.round(RDM[i, j], 1),
ha="center", va="center", color="w"
)
ax.set_title(f'sub: {sub}, distance: {distance}, Type {problem_type}, roi: {roi}')
plt.imshow(RDM)
plt.savefig(f'subject_RDMs/sub-{sub}_task-{task}_run-{run}_rp-{repetition}_roi-{roi}_{distance}_beta.pdf')
plt.close()
print(f'[Check] plotted.')
def correlate_against_ideal_RDM(rois, distance, problem_type, num_shuffles, method, dataType, conditions, seed=999):
"""
Correlate each subject's RDM to the ideal RDM
of a given type. To better examine the significance of
the correlations, we build in a shuffle mechanism that
randomly permutes the entries of the RDM to determine if
the correlation we get during un-shuffled is real.
NOTE:
Under `trial-level-glm`, each RSA comparison is done at
repetition level instead of run level.
"""
ideal_RDM = np.ones((8, 8))
ideal_RDM[:4, :4] = 0
ideal_RDM[4:, 4:] = 0
runs = [1, 2, 3, 4]
for roi in rois:
for run in runs:
for rp in range(1, num_repetitions_per_run+1):
np.random.seed(seed)
all_rho = [] # one per subject-run-repetition of a task
for shuffle in range(num_shuffles):
for sub in subs:
# even sub: Type1 is task2, Type2 is task3
if int(sub) % 2 == 0:
if problem_type == 1:
task = 2
elif problem_type == 2:
task = 3
else:
task = 1
# odd sub: Type1 is task3, Type2 is task2
else:
if problem_type == 1:
task = 3
elif problem_type == 2:
task = 2
else:
task = 1
# get one repetition's RDM
sub_RDM = np.load(
f'{rdm_path}/sub-{sub}_task-{task}_run-{run}_rp-{rp}_roi-{roi}_{distance}_{dataType}.npy'
)
if num_shuffles > 1:
shuffle_indices = np.random.choice(
range(sub_RDM.shape[0]),
size=sub_RDM.shape[0],
replace=False
)
sub_RDM = sub_RDM[shuffle_indices, :]
# compute one repetition's correlation to the ideal RDM
rho = compute_RSA(sub_RDM, ideal_RDM, method=method)
# collects all repetitions of a run and of all subjects
all_rho.append(rho)
print(
f'Dist=[{distance}], Type=[{problem_type}], roi=[{roi}], run=[{run}], rp=[{rp}], ' \
f'avg_rho=[{np.mean(all_rho):.2f}], ' \
f'std=[{np.std(all_rho):.2f}], ' \
f't-stats=[{stats.ttest_1samp(a=all_rho, popmean=0)[0]:.2f}], ' \
f'pvalue=[{stats.ttest_1samp(a=all_rho, popmean=0)[1]:.2f}]' \
)
print('------------------------------------------------------------------------')
if __name__ == '__main__':
root_path = '/home/ken/projects/brain_data'
glm_path = 'glm_trial-estimate'
rdm_path = 'subject_RDMs'
rois = ['V1', 'V2', 'V3', 'V1-3', 'V4', 'LOC', 'RHHPC', 'LHHPC']
num_subs = 23
dataType = 'beta'
num_conditions = 64
subs = [f'{i:02d}' for i in range(2, num_subs+2) if i!=9]
num_subs = len(subs)
conditions = [f'{i:04d}' for i in range(1, num_conditions+1)]
tasks = [1, 2, 3]
runs = [1, 2, 3, 4]
distances = ['euclidean', 'pearson']
num_repetitions_per_run = 4
if dataType == 'beta':
# ignore `_rp*_fb` conditions, the remaining are `_rp*` conditions.
conditions = [f'{i:04d}' for i in range(1, num_conditions, 2)]
num_conditions = len(conditions)
# reorder_mapper = reorder_RDM_entries_into_chunks()
# roi_execute(
# rois=rois,
# subs=subs,
# tasks=tasks,
# runs=runs,
# dataType=dataType,
# conditions=conditions,
# distances=distances,
# smooth_mask=0.2,
# smooth_beta=2,
# num_processes=70
# )
# correlate_against_ideal_RDM(
# rois=rois,
# distance='pearson',
# problem_type=1,
# seed=999,
# num_shuffles=1,
# method='spearman',
# dataType='beta',
# conditions=conditions
# )
visualize_RDM(
sub='08',
problem_type=1,
distance='pearson',
run=4,
repetition=4,
roi='LHHPC'
)