-
Notifications
You must be signed in to change notification settings - Fork 5
/
computs.py
763 lines (647 loc) · 36.9 KB
/
computs.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
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
#!/usr/bin/env python3
from nibabel.processing import resample_from_to
import scipy
import scipy.spatial
import nibabel as nib
import numpy as np
import matplotlib as mpl
import matplotlib.pyplot as plt
import pickle as pkl
import inspect
from nilearn.image import math_img
from statistics import mean
import scipy.stats as stat
from scipy.signal import resample
import os
import pandas as pd
import itertools
############# COMPUTE DIFF.
def compute_rel(f1, file_name=None, f2=None):
# Compute relative diff in BT
if f2 is not None:
f1 = nib.load(f1)
f1_data = f1.get_fdata()
if type(f2) == str : f2 = nib.load(f2)
f2_data = f2.get_fdata()
img_rel = (f1_data - f2_data)
# activated regions have nonzero values
img_rel = np.where((f1_data == 0) & (f2_data == 0), np.nan, img_rel)
nft_img = nib.Nifti1Image(img_rel, f1.affine, header=f1.header)
nib.save(nft_img, os.path.join('data/diff/', '{}-rel.nii.gz'.format(file_name)))
# Compute relative diff in WT
else:
file_ = f1.replace('MCA', '1')
file_ = nib.load(file_)
f1_data = file_.get_fdata()
max_ = np.zeros(f1_data.shape)
for pair_ in list(itertools.combinations(range(1, 11), 2)):
file_n1 = f1.replace('MCA', str(pair_[0]))
file_ = nib.load(file_n1)
f1_data = file_.get_fdata()
file_n2 = f1.replace('MCA', str(pair_[1]))
file_ = nib.load(file_n2)
f2_data = file_.get_fdata()
img_rel = (f1_data - f2_data)
img_rel = np.where((f1_data == 0) & (f2_data == 0), np.nan, img_rel)
if np.nanstd(img_rel) > np.nanstd(max_):
max_ = img_rel
f1_name = file_n1
f2_name = file_n2
print('{} and {} at {}'.format(f1_name, f2_name, file_name))
nft_img_rel = nib.Nifti1Image(max_, file_.affine, header=file_.header)
nib.save(nft_img_rel, os.path.join('data/diff/', '{}-rel.nii.gz'.format(file_name)))
return nft_img_rel
def combine_diffs(f1, f2, meta_, file_name):
var_f2_res = resample_from_to(f2, f1, order=0)
# to combine two image variances, we use: var(x+y) = var(x) + var(y) + 2*cov(x,y)
# and since the correlation between two arrays are so weak, we droped `2*cov(x,y)` from the formula
f1t = f1.get_fdata()
f2t = var_f2_res.get_fdata()
wt_diff = (np.nan_to_num(f1.get_fdata()) + np.nan_to_num(var_f2_res.get_fdata()))
# nan bg images
wt_diff = np.where((np.isnan(f1t)) | (np.isnan(f2t)), np.nan, wt_diff)
nft_img = nib.Nifti1Image(wt_diff, meta_.affine, header=meta_.header)
nib.save(nft_img, os.path.join('data/diff/', '{}-rel.nii.gz'.format(file_name)))
def var_between_tool(tool_results):
for type_ in ['unthresh', 'thresh']:
f = 'stat_file'
if type_ == 'thresh': f = 'act_deact'
fsl_ = tool_results['fsl'][f]
afni_ = tool_results['afni'][f]
spm_ = tool_results['spm'][f]
# resampling first image on second image
spm_res = resample_from_to(nib.load(spm_), nib.load(fsl_), order=0)
afni_res = resample_from_to(nib.load(afni_), nib.load(fsl_), order=0)
# compute abs diff
compute_rel(fsl_, 'fsl-afni-{}'.format(type_), f2=afni_res)
compute_rel(fsl_, 'fsl-spm-{}'.format(type_), f2=spm_res)
spm_res = resample_from_to(nib.load(spm_), nib.load(afni_), order=0)
compute_rel(afni_, 'afni-spm-{}'.format(type_), f2=spm_res)
## unthresholded subject-level
for i in range(1, 17):
fsl_ = tool_results['fsl']['SBJ'].replace('NUM', '%.2d' % i )
afni_ = tool_results['afni']['SBJ'].replace('NUM', '%.2d' % i )
spm_ = tool_results['spm']['SBJ'].replace('NUM', '%.2d' % i )
# resampling first image on second image
spm_res = resample_from_to(nib.load(spm_), nib.load(fsl_), order=0)
afni_res = resample_from_to(nib.load(afni_), nib.load(fsl_), order=0)
# compute abd diff
compute_rel(fsl_, 'sbj{}-fsl-afni-unthresh'.format('%.2d' % i), f2=afni_res)
compute_rel(fsl_, 'sbj{}-fsl-spm-unthresh'.format('%.2d' % i), f2=spm_res)
spm_res = resample_from_to(nib.load(spm_), nib.load(afni_), order=0)
compute_rel(afni_, 'sbj{}-afni-spm-unthresh'.format('%.2d' % i), f2=spm_res)
def var_between_fuzzy(mca_results):
for type_ in ['thresh', 'unthresh']:
f = 'stat_file'
if type_ == 'thresh': f = 'act_deact'
# compute abs diff
fsl_abs = compute_rel(mca_results['fsl'][f], 'fuzzy-fsl-{}'.format(type_), f2=None)
afni_abs = compute_rel(mca_results['afni'][f], 'fuzzy-afni-{}'.format(type_), f2=None)
spm_abs = compute_rel(mca_results['spm'][f], 'fuzzy-spm-{}'.format(type_), f2=None)
combine_diffs(fsl_abs, afni_abs, nib.load(mca_results['fsl'][f].replace('MCA', '1')),
'fuzzy-fsl-afni-{}'.format(type_))
combine_diffs(fsl_abs, spm_abs, nib.load(mca_results['fsl'][f].replace('MCA', '1')),
'fuzzy-fsl-spm-{}'.format(type_))
combine_diffs(afni_abs, spm_abs, nib.load(mca_results['afni'][f].replace('MCA', '1')),
'fuzzy-afni-spm-{}'.format(type_))
# unthresholded subject-level
for i in range(1, 17):
fsl_ = mca_results['fsl']['SBJ'].replace('NUM', '%.2d' % i )
afni_ = mca_results['afni']['SBJ'].replace('NUM', '%.2d' % i )
spm_ = mca_results['spm']['SBJ'].replace('NUM', '%.2d' % i )
# compute abs diff
fsl_abs = compute_rel(fsl_, 'sbj{}-fuzzy-fsl-unthresh'.format('%.2d' % i), f2=None)
afni_abs = compute_rel(afni_, 'sbj{}-fuzzy-afni-unthresh'.format('%.2d' % i), f2=None)
spm_abs = compute_rel(spm_, 'sbj{}-fuzzy-spm-unthresh'.format('%.2d' % i), f2=None)
combine_diffs(fsl_abs, afni_abs, nib.load(fsl_.replace("MCA", '1')),
'sbj{}-fuzzy-fsl-afni-unthresh'.format('%.2d' % i))
combine_diffs(fsl_abs, spm_abs, nib.load(fsl_.replace("MCA", '1')),
'sbj{}-fuzzy-fsl-spm-unthresh'.format('%.2d' % i))
combine_diffs(afni_abs, spm_abs, nib.load(afni_.replace("MCA", '1')),
'sbj{}-fuzzy-afni-spm-unthresh'.format('%.2d' % i))
############# PLOT CORR. DIFF.
def plot_diff_corr_group(sbj=False):
### Plot correlation of differences between BT and WT
num=2
if sbj == True: num=17
for ind1, type_ in enumerate(['unthresh']):
for i in range(1, num):
fig, ax = plt.subplots(nrows=2,ncols=3,figsize=(24, 12))
for ind_, pair_ in enumerate(['fsl-spm', 'fsl-afni', 'afni-spm']):
# read BT
if sbj == True: btool = nib.load('./data/diff/sbj{}-{}-{}-rel.nii.gz'.format('%.2d' % i, pair_, type_))
else: btool = nib.load('./data/diff/{}-{}-rel.nii.gz'.format(pair_, type_))
bt_ = btool.get_fdata()
data2 = np.reshape(bt_, -1)
data2 = np.nan_to_num(data2)
# read WT
for ind_2, tool_ in enumerate(pair_.split('-')):
if sbj == True: bfuzzy = nib.load('./data/diff/sbj{}-fuzzy-{}-{}-rel.nii.gz'.format('%.2d' % i, pair_, type_))
else: bfuzzy = nib.load('./data/diff/fuzzy-{}-{}-rel.nii.gz'.format(tool_, type_))
wt_ = bfuzzy.get_fdata()
if btool.shape != bfuzzy.shape:
res_wt = resample_from_to(bfuzzy, btool, order=0)
wt_ = res_wt.get_fdata()
data1 = np.reshape(wt_, -1)
data1 = np.nan_to_num(data1)
# r, p = scipy.stats.pearsonr(data1, data2)
# print("P-value: {} and R: {}".format(p, r))
slope, intercept, r2, p2, stderr = scipy.stats.linregress(data1, data2)
line = 'Regression line'
y = intercept + slope * data1
label_r = "Regression line\np={:.3f}, r={:.3f}".format(p2, r2)
hist = ax[ind_2][ind_].hist2d(data1, data2, bins=20, norm=mpl.colors.LogNorm(), cmap=plt.cm.Reds)
ax[ind_2][ind_].plot(data1, y, color='black', label=label_r)
ax[ind_2][ind_].set_title('')
ax[ind_2][ind_].set_xlabel('')
# if ind_ == 0:
ax[ind_2][ind_].set_ylabel(f'BT({pair_.upper()})', fontsize=14)
ax[ind_2][ind_].set_xlabel(f'WT({tool_.upper()})', fontsize=14)
# ax[ind_2][ind_].set_ylim([-7.5, 7.5])
# ax[ind_2][ind_].set_xlim([-3.5, 3.5])
# #ax[ind_].set_xticklabels(fontsize=14)
# #ax[ind_].set_yticklabels(fontsize=14)
ax[ind_2][ind_].legend(facecolor='white', loc='upper right', fontsize=12)
ax[ind_2][ind_].tick_params(axis='both', labelsize=12)
#ax[ind_].set_xscale('log')
plt.subplots_adjust(wspace=0.3)
if sbj == True: fig.savefig('./paper/figures/abs/corr/sbj{}-rel-corr-{}-plot.png'.format('%.2d' % i, type_), bbox_inches='tight')
else: fig.savefig('./paper/figures/abs/corr/rel-corr-{}-plot.png'.format(type_), bbox_inches='tight')
############# COMPUTE and PLOT DICES
def dump_variable(var_, name_):
with open('./data/{}.pkl'.format(str(name_)),'wb') as f:
pkl.dump(var_, f)
def load_variable(name_):
with open('./data/{}.pkl'.format(name_),'rb') as fr:
return pkl.load(fr)
def retrieve_name(var):
callers_local_vars = inspect.currentframe().f_back.f_locals.items()
return [var_name for var_name, var_val in callers_local_vars if var_val is var]
def compute_dice(data1_file, data2_file):
# Load nifti images
data1_img = nib.load(data1_file)
data2_img = nib.load(data2_file)
# Load data from images
data2 = data2_img.get_data()
data1 = data1_img.get_data()
# Get asbolute values (positive and negative blobs are of interest)
data2 = np.absolute(data2)
data1 = np.absolute(data1)
# Resample data2 on data1 using nearest neighbours
data2_resl_img = resample_from_to(data2_img, data1_img, order=0)
data2_res = data2_resl_img.get_data()
data2_res = np.absolute(data2_res)
# Masking (compute Dice using intersection of both masks)
# bg_ = np.logical_or(np.isnan(data1), np.isnan(data2_res))
# Binarizing data
data1 = np.nan_to_num(data1)
data1[data1>0] = 1
data2_res = np.nan_to_num(data2_res)
data2_res[data2_res>0] = 1
num_activated_1 = np.sum(data1 > 0)
num_activated_res_2 = np.sum(data2_res>0)
# Masking intersections
# data1[bg_] = 0
# data2_res[bg_] = 0
# Vectorize
data1 = np.reshape(data1, -1)
data2_res = np.reshape(data2_res, -1)
# similarity = 1.0 - dissimilarity
dice_score = 1-scipy.spatial.distance.dice(data1>0, data2_res>0)
return (dice_score, num_activated_1, num_activated_res_2)
def keep_roi(stat_img, reg_, image_parc, filename=None):
parc_img = nib.load(image_parc)
data_img = nib.load(stat_img)
# Resample parcellation on data1 using nearest neighbours
parc_img_res = resample_from_to(parc_img, data_img, order=0)
parc_data_res = parc_img_res.get_fdata(dtype=np.float32)
colls = np.where(parc_data_res != reg_)
parc_data_res[colls] = np.nan
# data_img_nan = nib.Nifti1Image(parc_data_res, img_.affine, header=img_.header)
# nib.save(data_img_nan, 'parce_region{}.nii.gz'.format(reg_))
data_orig = data_img.get_data()
# If there are NaNs in data_file remove them (to mask using parcelled data only)
data_orig = np.nan_to_num(data_orig)
# Replace background by NaNs
data_nan = data_orig.astype(float)
data_nan[np.isnan(parc_data_res)] = np.nan
data_orig[np.isnan(parc_data_res)] = 0.0
if np.all(data_orig == 0.0):
print("There is not activation in R")
s = 0
else:
print("There is activated voxel in R")
s = 1
# Save as image
data_img_nan = nib.Nifti1Image(data_nan, data_img.get_affine())
if filename is None:
filename = stat_img.replace('.nii', '_nan.nii')
nib.save(data_img_nan, filename)
return(filename, s)
def get_dice_values(regions_txt, image_parc, tool_results, mca_results):
# read parcellation data
parc_img = nib.load(image_parc)
parc_data = parc_img.get_fdata(dtype=np.float32)
regions = {}
with open(regions_txt) as f:
for line in f:
(key, val) = line.split()
regions[int(key)+180] = val
regions[int(key)] = 'R_' + '_'.join(val.split('_')[1:])
dices_ = {}
file_cols = ["Region", "FSL IEEE", "SPM IEEE", "AFNI IEEE",
"FSL MCA1", "FSL MCA2", "FSL MCA3", "FSL MCA4", "FSL MCA5", "FSL MCA6",
"FSL MCA7","FSL MCA8", "FSL MCA9", "FSL MCA10",
"SPM MCA1", "SPM MCA2", "SPM MCA3", "SPM MCA4", "SPM MCA5", "SPM MCA6",
"SPM MCA7", "SPM MCA8", "SPM MCA9", "SPM MCA10",
"AFNI MCA1", "AFNI MCA2", "AFNI MCA3", "AFNI MCA4", "AFNI MCA5", "AFNI MCA6",
"AFNI MCA7", "AFNI MCA8", "AFNI MCA9", "AFNI MCA10"]
dframe = pd.DataFrame(np.array([['R', 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]]), columns=file_cols)
#for act_ in ['exc_set_file', 'exc_set_file_neg', 'act_deact', 'stat_file']:
for act_ in ['act_deact']:
### Print global Dice values
bt1 = compute_dice(tool_results['fsl'][act_], tool_results['afni'][act_])[0]
bt2 = compute_dice(tool_results['fsl'][act_], tool_results['spm'][act_])[0]
bt3 = compute_dice(tool_results['afni'][act_], tool_results['spm'][act_])[0]
print("Global Dice in BT for FSL-AFNI {}, FSL-SPM {}, and AFNI-SPM {}".format(bt1, bt2, bt3))
for tool_ in ['fsl', 'spm', 'afni']:
dice_wt_list = []
for pair_ in list(itertools.combinations(range(1, 11), 2)):
f1_data = mca_results[tool_][act_].replace('MCA', str(pair_[0]))
f2_data = mca_results[tool_][act_].replace('MCA', str(pair_[1]))
dice_wt_list.append(compute_dice(f1_data, f2_data)[0])
# wt2 = compute_dice(mca_results[tool_][1][act_], mca_results[tool_][3][act_])[0]
# wt3 = compute_dice(mca_results[tool_][2][act_], mca_results[tool_][3][act_])[0]
print("Global Dice in WT for {} {}".format(tool_, np.mean(dice_wt_list)))
masked_regions = {}
for r in regions.keys():
act_bin = []
act_bin.append(regions[r])
colls = np.where(parc_data == r)
if regions[r] not in dices_.keys():
dices_[regions[r]] = {}
dices_[regions[r]]['size'] = len(colls[0])
dices_[regions[r]][act_] = {}
dices_[regions[r]][act_]['tool'] = {}
dices_[regions[r]][act_]['mca'] = {}
else:
dices_[regions[r]][act_] = {}
dices_[regions[r]][act_]['tool'] = {}
dices_[regions[r]][act_]['mca'] = {}
masked_regions['fsl'], s = keep_roi(tool_results['fsl'][act_], r, image_parc)
act_bin.append(s)
masked_regions['spm'], s = keep_roi(tool_results['spm'][act_], r, image_parc)
act_bin.append(s)
masked_regions['afni'], s = keep_roi(tool_results['afni'][act_], r, image_parc)
act_bin.append(s)
dices_[regions[r]][act_]['tool']['fsl-afni'] = compute_dice(masked_regions['fsl'], masked_regions['afni'])[0]
dices_[regions[r]][act_]['tool']['fsl-spm'] = compute_dice(masked_regions['fsl'], masked_regions['spm'])[0]
dices_[regions[r]][act_]['tool']['afni-spm'] = compute_dice(masked_regions['afni'], masked_regions['spm'])[0]
# dices = (dice_res_1, dark_dice_1[1], dark_dice_2[1], num_activated_1, num_activated_2)
for tool_ in mca_results.keys():
for i in range(1, 11):
masked_regions['{}{}'.format(tool_, i)], s = keep_roi(mca_results[tool_][act_].replace('MCA', str(i)), r, image_parc)#, '{}1'.format(tool_))
act_bin.append(s)
# masked_regions['{}2'.format(tool_)], s = keep_roi(mca_results[tool_][2][act_], r, image_parc)#, '{}2'.format(tool_))
# act_bin.append(s)
# masked_regions['{}3'.format(tool_)], s = keep_roi(mca_results[tool_][3][act_], r, image_parc)#, '{}3'.format(tool_))
# act_bin.append(s)
# dices_[regions[r]][act_]['mca']['{}1'.format(tool_)] = compute_dice(masked_regions['{}1'.format(tool_)], masked_regions['{}2'.format(tool_)])[0]
# dices_[regions[r]][act_]['mca']['{}2'.format(tool_)] = compute_dice(masked_regions['{}1'.format(tool_)], masked_regions['{}3'.format(tool_)])[0]
# dices_[regions[r]][act_]['mca']['{}3'.format(tool_)] = compute_dice(masked_regions['{}2'.format(tool_)], masked_regions['{}3'.format(tool_)])[0]
dframe2 = pd.DataFrame(np.array([act_bin]),
columns=file_cols)
dframe = dframe.append(dframe2, ignore_index=True)
dframe.to_csv('./data/active_regions.csv')
dump_variable(dices_, retrieve_name(dices_)[0])
return dices_
def plot_dices(dices_):
fig, ax = plt.subplots(nrows=1,ncols=1,figsize=(16, 12))
marker_ = ['o', '*', 'v', '>', '1', '2', '3', '4', 'P']
colors = ['red', 'blue', 'green']
for ind1_, act_ in enumerate(['act_deact']):
for ind_, tool_ in enumerate(['fsl-afni', 'fsl-spm', 'afni-spm']):
tool_list = []
mca_list = []
for reg_ in dices_.keys():
tool_list.append(dices_[reg_][act_]['tool'][tool_])
f1, f2 = tool_.split('-')
mca_mean = []
for i in [1, 2, 3]:
mca_mean.append(dices_[reg_][act_]['mca']['{}{}'.format(f1, i)])
mca_mean.append(dices_[reg_][act_]['mca']['{}{}'.format(f2, i)])
mca_list.append(mean(mca_mean))
# Get region sizes
# tool_list = np.nan_to_num(tool_list)
r_sizes = []
r_name = []
for i in range(len(tool_list)):
l_ = list(dices_.keys())
s = dices_[l_[i]]['size']
r_sizes.append(s)
r_name.append(l_[i])
# Plot Dice values by region size
tool_dice = np.array(tool_list)
mca_dice = np.array(mca_list)
tool_dice_nonz = [ tool_dice[i] for i in range(len(tool_dice)) if not (np.isnan(tool_dice[i]) or np.isnan(mca_dice[i]))]
mca_dice_nonz = [ mca_dice[i] for i in range(len(tool_dice)) if not (np.isnan(tool_dice[i]) or np.isnan(mca_dice[i]))]
sizes_nonz = [ r_sizes[i] for i in range(len(tool_dice)) if not (np.isnan(tool_dice[i]) or np.isnan(mca_dice[i]))]
tools_nonz = [ tool_list[i] for i in range(len(tool_dice)) if tool_dice[i] != 0 and mca_dice[i] != 0]
tool_dice = tool_dice_nonz
mca_dice = mca_dice_nonz
# sizes_nonz = r_sizes
print(len(tool_dice), len(mca_dice))
# Is there a correlation between tool_dice and region size?
_, _, r, p, _ = scipy.stats.linregress(tool_dice, sizes_nonz)
print(f'BT Dice and region size are correlated (p={p})')
# Is there a correlation between mca_dice and region size?
_, _, r, p, _ = scipy.stats.linregress(mca_dice, sizes_nonz)
print(f'WT Dice and region size are correlated (p={p})')
slope, intercept, r, p, stderr = scipy.stats.linregress(mca_dice, tool_dice)
line = f'Regression line: y={intercept:.2f}+{slope:.2f}x (r={r:.2f}, p={p})'
y = intercept + slope * np.array(mca_dice)
print(sorted(mca_dice))
ax.plot(mca_dice, tool_dice, linewidth=0, alpha=.5, color=colors[ind_], marker='o', label='{}'.format(tool_.upper()))
ax.plot(mca_dice, y, color=colors[ind_], alpha=.7, label=line)
ax.set_xlabel('WT Dice', fontsize=22)
ax.set_ylabel('BT Dice', fontsize=22)
ax.legend(fontsize=16)
ax.tick_params(axis='both', labelsize=18)
#plt.show()
name = './paper/figures/dices_corr.png'
print(f'Dice plot saved in {name}')
plt.savefig(name, bbox_inches='tight')
############# PRINT STATS
def print_gl_stats(path_):
# Compute stats of variabilities (ignore NaNs as bg)
for type_ in ['unthresh', 'thresh']:
for pair_ in ['fsl-spm', 'fsl-afni', 'afni-spm']:
bt_ = nib.load('{}{}-{}-rel.nii.gz'.format(path_, pair_, type_))
bt_abs_data = bt_.get_fdata()
bt_abs_mean = np.nanmean(bt_abs_data)
bt_abs_std = np.nanstd(bt_abs_data)
print('BT variability of tstats in {} {}:\nMean {}\nStd. {}'.format(pair_, type_, bt_abs_mean, bt_abs_std))
# one-sample t-test for zero centering
bg_ = np.where((np.isnan(bt_abs_data)) , False, True)
bt_img = np.nan_to_num(bt_abs_data)[bg_]
zt_stat1, zp_val1 = stat.ttest_1samp(resample(bt_img, int(1e4)), 0.01)
print("Zero centering one-sample t-test in BT({}) is {} and p-value {}".format(pair_, zt_stat1, zp_val1))
for tool_ in ['fsl', 'spm', 'afni']:
wt2_ = nib.load('{}fuzzy-{}-{}-rel.nii.gz'.format(path_, tool_, type_))
wt2_abs_data = wt2_.get_fdata()
wt2_abs_mean = np.nanmean(wt2_abs_data)
wt2_abs_std = np.nanstd(wt2_abs_data)
print('WT variability of tstats in {} {}:\nMean {}\nStd. {}'.format(tool_, type_, wt2_abs_mean, wt2_abs_std))
# one-sample t-test for zero centering
bg_ = np.where((np.isnan(wt2_abs_data)) , False, True)
wt_img = np.nan_to_num(wt2_abs_data)[bg_]
zt_stat1, zp_val1 = stat.ttest_1samp(resample(wt_img, int(1e4)), 0.01)
print("Zero centering one-sample t-test in WT({}) is {} and p-value {}".format(tool_, zt_stat1, zp_val1))
# FSL stats at global virtual precision t=17
# if type_ == 'unthresh':
# img_ = './data/abs/FL-FSL/p17_fsl_unthresh_abs.nii.gz'
# wt2_ = nib.load(img_)
# wt2_abs_data = wt2_.get_fdata()
# wt2_abs_mean = np.nanmean(wt2_abs_data)
# wt2_abs_std = np.nanstd(wt2_abs_data)
# print('WT variability of tstats in FSL {} at precision t=17 bits:\nMean {}\nStd. {}'.format(type_, wt2_abs_mean, wt2_abs_std))
print("stop")
def print_sl_stats(path_):
# Compute Mean of std over subjects (ignore NaNs as bg)
min_sbj = 1000
max_sbj = 0
all_wt_list = {}
all_bt_list = {}
for i in range(1, 17):
wt_list = []
bt_list = []
for pair_ in ['fsl-spm', 'fsl-afni', 'afni-spm']:
bt_ = nib.load(os.path.join(path_, 'sbj{}-{}-unthresh-rel.nii.gz'.format('%.2d' % i, pair_)))
bt_abs_data = bt_.get_fdata()
# one-sample t-test for zero centering
bg_ = np.where((np.isnan(bt_abs_data)) , False, True)
bt_img = np.nan_to_num(bt_abs_data)[bg_]
zt_stat1, zp_val1 = stat.ttest_1samp(resample(bt_img, int(1e4)), 0.01)
print("Sbj {}, Zero centering one-sample t-test in BT({}) is {} and p-value {}".format(i, pair_, zt_stat1, zp_val1))
bt_abs_mean = (np.nanmean(bt_abs_data))
# print("Sbj {}, Mean of diff {} is {}".format(i, pair_, bt_abs_mean))
bt_abs_std = np.nanstd(bt_abs_data)
bt_list.append(bt_abs_std)
# print("Sbj {}, STD of diff {} is {}".format(i, pair_, bt_abs_std))
if pair_ in all_bt_list.keys():
tmp = all_bt_list[pair_]['mean']
all_bt_list[pair_]['mean'] = bt_abs_mean + tmp
tmp = all_bt_list[pair_]['std']
all_bt_list[pair_]['std'] = bt_abs_std + tmp
else:
all_bt_list[pair_] = {}
all_bt_list[pair_]['mean'] = bt_abs_mean
all_bt_list[pair_]['std'] = bt_abs_std
for tool_ in ['fsl', 'spm', 'afni']:
wt_ = nib.load(os.path.join(path_, 'sbj{}-fuzzy-{}-unthresh-rel.nii.gz'.format('%.2d' % i, tool_)))
wt_abs_data = wt_.get_fdata()
# one-sample t-test for zero centering
bg_ = np.where((np.isnan(wt_abs_data)) , False, True)
wt_img = np.nan_to_num(wt_abs_data)[bg_]
zt_stat1, zp_val1 = stat.ttest_1samp(resample(wt_img, int(1e4)), 0.01)
print("Sbj {}, Zero centering one-sample t-test in WT({}) is {} and p-value {}".format(i, tool_, zt_stat1, zp_val1))
wt_abs_mean = (np.nanmean(wt_abs_data))
# print("Sbj {}, Mean of diff {} is {}".format(i, tool_, wt_abs_mean))
wt_abs_std = np.nanstd(wt_abs_data)
wt_list.append(wt_abs_std)
# print("Sbj {}, STD of diff {} is {}".format(i, tool_, wt_abs_std))
if tool_ in all_wt_list.keys():
tmp = all_wt_list[tool_]['mean']
all_wt_list[tool_]['mean'] = wt_abs_mean + tmp
tmp = all_wt_list[tool_]['std']
all_wt_list[tool_]['std'] = wt_abs_std + tmp
else:
all_wt_list[tool_] = {}
all_wt_list[tool_]['mean'] = wt_abs_mean
all_wt_list[tool_]['std'] = wt_abs_std
if mean(wt_list) > max_sbj:
wt_list_fff = wt_list
max_sbj = mean(wt_list)
i_max = i
max_sbj_bt = mean(bt_list)
print('Subject {} has the highest WT variability with average differences of {}\n BT avg diff is {}'.format(i_max, max_sbj, max_sbj_bt))
for pair_ in ['fsl-spm', 'fsl-afni', 'afni-spm']:
print('BT variability of tstats in {} unthresholded:\nMean {}\nStd. {}'.format(pair_, all_bt_list[pair_]['mean']/16, all_bt_list[pair_]['std']/16))
for tool_ in ['fsl', 'spm', 'afni']:
print('WT variability of tstats in {} unthresholded:\nMean {}\nStd. {}'.format(tool_, all_wt_list[tool_]['mean']/16, all_wt_list[tool_]['std']/16))
print('stop')
def compute_stat_test(path_):
for type_ in ['unthresh', 'thresh']:
for pair_ in ['fsl-spm', 'fsl-afni', 'afni-spm']:
gvp = False
num_sample = int(1e4)
bt_ = nib.load('{}{}-{}-rel.nii.gz'.format(path_, pair_, type_))
bt_abs_data = bt_.get_fdata()
bg_ = np.where((np.isnan(bt_.get_fdata())) , False, True)
bt_img = np.nan_to_num(bt_abs_data)[bg_]
t1, t2 = pair_.split('-')
wt1_ = nib.load('{}fuzzy-{}-{}-rel.nii.gz'.format(path_, t1, type_))
wt1_abs_data = wt1_.get_fdata()
bg_ = np.where((np.isnan(wt1_.get_fdata())) , False, True)
img1 = np.nan_to_num(wt1_abs_data)[bg_]
wt2_ = nib.load('{}fuzzy-{}-{}-rel.nii.gz'.format(path_, t2, type_))
wt2_abs_data = wt2_.get_fdata()
bg_ = np.where((np.isnan(wt2_.get_fdata())) , False, True)
img2 = np.nan_to_num(wt2_abs_data)[bg_]
res_bt = resample(bt_img, num_sample)
res1_ = resample(img1, num_sample)
res2_ = resample(img2, num_sample)
t_stat1, p_val1 = stat.wilcoxon(res_bt, res1_)
t_stat2, p_val2 = stat.wilcoxon(res_bt, res2_)
print("wilcoxon-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t1, t_stat1, p_val1))
print("wilcoxon-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t2, t_stat2, p_val2))
t_stat1, p_val1 = stat.ttest_ind(res_bt, res1_)
t_stat2, p_val2 = stat.ttest_ind(res_bt, res2_)
print("t-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t1, t_stat1, p_val1))
print("t-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t2, t_stat2, p_val2))
print('stop')
def compute_sbj_stat_test(path_):
for i in range(1, 17):
print(f"Subject{'%.2d' % i}:")
for pair_ in ['fsl-spm', 'fsl-afni', 'afni-spm']:
num_sample = int(1e4)
bt_ = nib.load('{}sbj{}-{}-unthresh-rel.nii.gz'.format(path_, '%.2d' % i, pair_))
bt_abs_data = bt_.get_fdata()
bg_ = np.where((np.isnan(bt_.get_fdata())) , False, True)
bt_img = np.nan_to_num(bt_abs_data)[bg_]
t1, t2 = pair_.split('-')
wt1_ = nib.load('{}sbj{}-fuzzy-{}-unthresh-rel.nii.gz'.format(path_, '%.2d' % i, t1))
wt1_abs_data = wt1_.get_fdata()
bg_ = np.where((np.isnan(wt1_.get_fdata())) , False, True)
img1 = np.nan_to_num(wt1_abs_data)[bg_]
wt2_ = nib.load('{}sbj{}-fuzzy-{}-unthresh-rel.nii.gz'.format(path_, '%.2d' % i, t2))
wt2_abs_data = wt2_.get_fdata()
bg_ = np.where((np.isnan(wt2_.get_fdata())) , False, True)
img2 = np.nan_to_num(wt2_abs_data)[bg_]
# print('WT img1 mean {} and img2 mean {} and bt_img mean {} '.format(mean(img1), mean(img2), mean(bt_img)))
res_bt = resample(bt_img, num_sample)
res1_ = resample(img1, num_sample)
res2_ = resample(img2, num_sample)
t_stat1, p_val1 = stat.wilcoxon(res_bt, res1_)
t_stat2, p_val2 = stat.wilcoxon(res_bt, res2_)
print("wilcoxon-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t1, t_stat1, p_val1))
print("wilcoxon-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t2, t_stat2, p_val2))
t_stat1, p_val1 = stat.ttest_ind(res_bt, res1_)
t_stat2, p_val2 = stat.ttest_ind(res_bt, res2_)
print("t-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t1, t_stat1, p_val1))
print("t-test between BT({}) and WT({}) is {} and p-value {}".format(pair_, t2, t_stat2, p_val2))
print('stop')
def combine_thresh(tool_results, mca_results):
## Combine activation and deactivation of thresholded maps ###
# between tool
for i in tool_results.keys():
path_ = os.path.dirname(tool_results[i]['exc_set_file'])
n = nib.load(tool_results[i]['exc_set_file'])
d = n.get_data()
exc_set_nonan = nib.Nifti1Image(np.nan_to_num(d), n.affine, header=n.header)
n = nib.load(tool_results[i]['exc_set_file_neg'])
d = n.get_data()
exc_set_neg_nonan = nib.Nifti1Image(np.nan_to_num(d), n.affine, header=n.header)
to_display = math_img("img1-img2", img1=exc_set_nonan, img2=exc_set_neg_nonan)
nib.save(to_display, os.path.join(path_, '{}.nii.gz'.format(i)))
# within tool
# fsl_ = tool_results['fsl']['SBJ'].replace('NUM', '%.2d' % i )
for tool_ in mca_results.keys():
dic_ = mca_results[tool_]
for i in range(1, 11):
stat_file = dic_['exc_set_file'].replace('MCA', str(i))
path_ = os.path.dirname(stat_file)
n = nib.load(stat_file)
d = n.get_data()
exc_set_nonan = nib.Nifti1Image(np.nan_to_num(d), n.affine, header=n.header)
n = nib.load(dic_['exc_set_file_neg'].replace('MCA', str(i)))
d = n.get_data()
exc_set_neg_nonan = nib.Nifti1Image(np.nan_to_num(d), n.affine, header=n.header)
to_display = math_img("img1-img2", img1=exc_set_nonan, img2=exc_set_neg_nonan)
nib.save(to_display, os.path.join(path_, '{}_s{}.nii.gz'.format(tool_, i)))
def main(args=None):
tool_results = {}
tool_results['fsl'] = {}
tool_results['fsl']['exc_set_file'] = './results/ds000001/tools/FSL/thresh_zstat1.nii.gz'
tool_results['fsl']['exc_set_file_neg'] = './results/ds000001/tools/FSL/thresh_zstat2.nii.gz'
tool_results['fsl']['stat_file'] = './results/ds000001/tools/FSL/tstat1.nii.gz'
tool_results['fsl']['act_deact'] = './results/ds000001/tools/FSL/fsl.nii.gz'
tool_results['fsl']['SBJ'] = './results/ds000001/tools/FSL/subject_level/sbjNUM_tstat1.nii.gz' # NUM will replace with the sbj number
tool_results['fsl']['SBJ-AR'] = 'results/ds000001/tools/FSL/subject_level/LEVEL1/sub-NUM/run-01.feat/reg/highres2standard.nii.gz' # anatomical registration image
tool_results['fsl']['SBJ-FR'] = 'results/ds000001/tools/FSL/subject_level/LEVEL1/sub-NUM/run-01.feat/reg/example_func2standard.nii.gz' # functional registration image
tool_results['spm'] = {}
tool_results['spm']['exc_set_file'] = './results/ds000001/tools/SPM/Octave/spm_exc_set.nii.gz'
tool_results['spm']['exc_set_file_neg'] = './results/ds000001/tools/SPM/Octave/spm_exc_set_neg.nii.gz'
tool_results['spm']['stat_file'] = './results/ds000001/tools/SPM/Octave/spm_stat.nii.gz'
# tool_results['spm']['exc_set_file'] = './results/ds000001/tools/SPM/Matlab2016b/spm_exc_set.nii.gz'
# tool_results['spm']['exc_set_file_neg'] = './results/ds000001/tools/SPM/Matlab2016b/spm_exc_set_neg.nii.gz'
# tool_results['spm']['stat_file'] = './results/ds000001/tools/SPM/Matlab2016b/spm_stat.nii.gz'
tool_results['spm']['act_deact'] = './results/ds000001/tools/SPM/Octave/spm.nii.gz'
tool_results['spm']['SBJ'] = './results/ds000001/tools/SPM/Octave/subject_level/sub-NUM/spm_stat.nii.gz' # NUM will replace with the sbj number
tool_results['afni'] = {}
tool_results['afni']['exc_set_file'] = './results/ds000001/tools/AFNI/Positive_clustered_t_stat.nii.gz'
tool_results['afni']['exc_set_file_neg'] = './results/ds000001/tools/AFNI/Negative_clustered_t_stat.nii.gz'
tool_results['afni']['stat_file'] = './results/ds000001/tools/AFNI/3dMEMA_result_t_stat_masked.nii.gz'
tool_results['afni']['act_deact'] = './results/ds000001/tools/AFNI/afni.nii.gz'
tool_results['afni']['SBJ'] = './results/ds000001/tools/AFNI/subject_level/tstats/sbjNUM_result_t_stat_masked.nii.gz' # NUM will replace with the sbj number
tool_results['afni']['SBJ-AR'] = './results/original/AFNI/subject_level/registrations/anatQQ_images/anatQQ.sub-NUM.nii'
fsl_mca = {}
fsl_mca['exc_set_file'] = "./results/ds000001/fuzzy/p53/FSL/runMCA/thresh_zstat1_53_runMCA.nii.gz"
fsl_mca['exc_set_file_neg'] = "./results/ds000001/fuzzy/p53/FSL/runMCA/thresh_zstat2_53_runMCA.nii.gz"
fsl_mca['stat_file'] = "./results/ds000001/fuzzy/p53/FSL/runMCA/tstat1_53_runMCA.nii.gz"
fsl_mca['act_deact'] = "./results/ds000001/fuzzy/p53/FSL/runMCA/fsl_sMCA.nii.gz"
fsl_mca['SBJ'] = './results/ds000001/fuzzy/p53/FSL/subject_level/runMCA/sbjNUM_tstat1.nii.gz'
spm_mca = {}
spm_mca['exc_set_file'] = "./results/ds000001/fuzzy/p53/SPM-Octace/runMCA/spm_exc_set.nii.gz"
spm_mca['exc_set_file_neg'] = "./results/ds000001/fuzzy/p53/SPM-Octace/runMCA/spm_exc_set_neg.nii.gz"
spm_mca['stat_file'] = "./results/ds000001/fuzzy/p53/SPM-Octace/runMCA/spm_stat.nii.gz"
spm_mca['act_deact'] = "./results/ds000001/fuzzy/p53/SPM-Octace/runMCA/spm_sMCA.nii.gz"
spm_mca['SBJ'] = './results/ds000001/fuzzy/p53/SPM-Octace/subject_level/runMCA/sub-NUM/spm_stat.nii.gz'
afni_mca = {}
afni_mca['exc_set_file'] = "./results/ds000001/fuzzy/p53/AFNI/runMCA/Positive_clustered_t_stat.nii.gz"
afni_mca['exc_set_file_neg'] = "./results/ds000001/fuzzy/p53/AFNI/runMCA/Negative_clustered_t_stat.nii.gz"
afni_mca['stat_file'] = "./results/ds000001/fuzzy/p53/AFNI/runMCA/3dMEMA_result_t_stat_masked.nii.gz"
afni_mca['act_deact'] = "./results/ds000001/fuzzy/p53/AFNI/runMCA/afni_sMCA.nii.gz"
afni_mca['SBJ'] = './results/ds000001/fuzzy/p53/AFNI/subject_level/tstats-runMCA/sbjNUM_result_t_stat_masked.nii.gz'
mca_results = {}
mca_results['fsl'] = {}
mca_results['fsl'] = fsl_mca
mca_results['spm'] = {}
mca_results['spm'] = spm_mca
mca_results['afni'] = {}
mca_results['afni'] = afni_mca
rel_path = 'data/diff/'
#joint histogram
### Combine activation and deactivation maps
# combine_thresh(tool_results, mca_results)
### Create abs diff images
# var_between_tool(tool_results) #BT
# var_between_fuzzy(mca_results) #WT
### Print stats
# print_gl_stats(rel_path)
# print_sl_stats(rel_path)
# compute_stat_test(rel_path)
# compute_sbj_stat_test(rel_path)
### Plot correlation of variances between BT and FL (Fig 4)
# plot_diff_corr_group(sbj=False)
# plot_diff_corr_group(sbj=True)
### Compute Dice scores and then plot
# image_parc = './data/MNI-parcellation/HCPMMP1_on_MNI152_ICBM2009a_nlin_resampled.splitLR.nii.gz'
# regions_txt = './data/MNI-parcellation/HCP-MMP1_on_MNI152_ICBM2009a_nlin.txt'
# if os.path.exists('./data/dices2_.pkl'):
# dices_ = load_variable('dices_')
# # plot_dices(dices_)
# else:
# dices_ = get_dice_values(regions_txt, image_parc, tool_results, mca_results)
# # plot_dices(dices_)
### abs diff in different precisions in WT
# path_ = './results/ds000001/fuzzy/'
# compute_abs_WT(path_)
### Global nearest precision
# p_nearest, all_rmse = global_nearest_precision()
# print(p_nearest)
# plot_rmse_nearest(all_rmse)
if __name__ == '__main__':
main()