-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathhelper_datadict.py
876 lines (662 loc) · 49.8 KB
/
helper_datadict.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
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
from ModelArchitecture.Transformations import *
from ImageLoader.ImageLoader3D import ImageLoader3D
from ImageLoader.ImageLoader2D import ImageLoader2D
from ModelArchitecture.UNet import NestedUNet,UNet,HalfUNet,ResUNet,NestedUNetDropout
from ModelArchitecture.UNet2D import NestedUNet2D,UNet2D,HalfUNet2D,ResUNet2D
from ModelArchitecture.UNETR import VITForSegmentation
from ModelArchitecture.UNETR2D import VITForSegmentation2D
from ModelArchitecture.Losses import WBCE_DICELoss,WBCE_FOCALDICELoss,FocalLoss,FOCAL_DICELoss
from ModelArchitecture.Losses import LogCoshDiceLoss,BCE_Loss_Weighted
from ModelArchitecture.Losses_truenet import DiceLoss
from ModelArchitecture.Losses_AdaptiveRegionSpecific import Adaptive_Region_Specific_TverskyLoss
from ModelArchitecture.metrics import Dice_Score
import sys
import os
import nibabel as nib
import torch.nn as nn
import glob
import numpy as np
from tqdm import tqdm
import skimage
from torch.utils.data import DataLoader,ConcatDataset
import json
CURRENT_DIRECTORY = os.getcwd()
PARENT_DIRECTORY = os.path.dirname(CURRENT_DIRECTORY)
sys.path.insert(1, PARENT_DIRECTORY+'/Slim-UNETR')
from src.SlimUNETR.SlimUNETR import SlimUNETR
from src.SlimUNETR2D.SlimUNETR2D import SlimUNETR2D
composed_transform = transforms.Compose([
RandomRotation3D([-10,10],p=0.15),
RandomHorizontalFlip3D(p=0.15),
RandomNoise3D(p=0.15),
#RandomBrightness3D(),
ToTensor3D(True)])
composed_transform_2d = transforms.Compose([
RandomRotation2D([-10,10],p=0.15),
RandomHorizontalFlip2D(p=0.15),
RandomNoise2D(p=0.15),
#RandomBrightness2D(),
ToTensor2D(True)])
def helper_resize(image,output,output_wothresh,label,shape=(),crop_para=[]):
image=image.cpu().numpy().squeeze()
output=output.cpu().numpy().squeeze()
output_wothresh = output_wothresh.cpu().numpy().squeeze()
label=label.cpu().numpy().squeeze()
if(len(label.shape)<=2):
return skimage.transform.resize(image, output_shape = label.shape,order=1,preserve_range=True),skimage.transform.resize(output, output_shape = label.shape,order=0,preserve_range=True),skimage.transform.resize(output_wothresh, output_shape = label.shape,order=1,preserve_range=True),label
shape = (int(shape[0]),int(shape[1]),int(shape[2]))
#print(shape,label.shape)
image = skimage.transform.resize(image, output_shape = shape,order=1,preserve_range=True)
output_wothresh = skimage.transform.resize(output_wothresh, output_shape = shape,order=1,preserve_range=True)
output = skimage.transform.resize(output, output_shape = shape,order=0,preserve_range=True)
actual_image = np.zeros_like(label)
actual_output = np.zeros_like(label)
actual_output_wothresh = np.zeros_like(label)
actual_image[crop_para[0]:crop_para[0] + crop_para[1], crop_para[2]:crop_para[2] + crop_para[3], crop_para[4]:crop_para[4] + crop_para[5]] = image
actual_output[crop_para[0]:crop_para[0] + crop_para[1], crop_para[2]:crop_para[2] + crop_para[3], crop_para[4]:crop_para[4] + crop_para[5]] = output
actual_output_wothresh[crop_para[0]:crop_para[0] + crop_para[1], crop_para[2]:crop_para[2] + crop_para[3], crop_para[4]:crop_para[4] + crop_para[5]] = output_wothresh
return actual_image,actual_output,actual_output_wothresh,label
def helper_transformation(no_aug):
if(no_aug):
global composed_transform
composed_transform = ToTensor3D(True)
global composed_transform_2d
composed_transform_2d = ToTensor2D(True)
def helper_path_configuration(indexes,data_path):
new_header_string = data_path
z = {}
names_list = list(indexes.keys())
for name in names_list:
z[name] = [new_header_string + indexes[name][i][41:] for i in range(len(indexes[name]))]
return z
def helper_supervised(system_data_path,which_data='brats',size=(128,128,128),size2=(128,128,48),no_crop=False,combined_brats_samples=250):
if(which_data=='wmh'):
# White Matter Hyperintensities
wmh_indexes = np.load('./Data_splits/Train_val_test_42_6_x/wmh_indexes.npy', allow_pickle=True).item()
wmh_indexes = helper_path_configuration(wmh_indexes,system_data_path)
datadict_train = ImageLoader3D(wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'],image_size=size,type_of_imgs='nifty',transform = composed_transform,data='wmh',no_crop=no_crop)
datadict_val = ImageLoader3D(wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),data='wmh',no_crop=no_crop)
datadict_test = ImageLoader3D(wmh_indexes['test_names_flair'],wmh_indexes['test_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),data='wmh',no_crop=no_crop,return_size=True,return_orig=True)
elif(which_data=='brats'):
# Brain Tumor Segmentation Challenge
brats_indexes = np.load('./Data_splits/Train_val_test_42_6_x/brats_indexes.npy', allow_pickle=True).item()
brats_indexes = helper_path_configuration(brats_indexes,system_data_path)
datadict_train = ImageLoader3D(brats_indexes['train_names_flair'],brats_indexes['train_names_seg'],image_size=size,type_of_imgs='nifty',transform = composed_transform,data='brats',no_crop=no_crop)
datadict_val = ImageLoader3D(brats_indexes['val_names_flair'],brats_indexes['val_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),data='brats',no_crop=no_crop)
datadict_test = ImageLoader3D(brats_indexes['test_names_flair'],brats_indexes['test_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),data='brats',no_crop=no_crop,return_size=True,return_orig=True)
elif(which_data=='lits'):
# Liver Tumor Segmentation
liver_indexes = np.load('./Data_splits/Train_val_test_42_6_x/liver_indexes.npy', allow_pickle=True).item()
liver_indexes = helper_path_configuration(liver_indexes,system_data_path)
datadict_train = ImageLoader3D(liver_indexes['train_names_flair'],liver_indexes['train_names_seg'],type_of_imgs='nifty',transform = composed_transform,data='liver')
datadict_val = ImageLoader3D(liver_indexes['val_names_flair'],liver_indexes['val_names_seg'],type_of_imgs='nifty', transform = ToTensor3D(True),data='liver')
datadict_test = ImageLoader3D(liver_indexes['test_names_flair'],liver_indexes['test_names_seg'],type_of_imgs='nifty', transform = ToTensor3D(True),data='liver',no_crop=no_crop,return_size=True,return_orig=True)
elif(which_data=='busi'):
# Brain Tumor Segmentation Challenge
busi_indexes = np.load('./Data_splits/Train_val_test_42_6_x/busi_indexes.npy', allow_pickle=True).item()
busi_indexes = helper_path_configuration(busi_indexes,system_data_path)
datadict_train = ImageLoader2D(busi_indexes['train_names_flair'],busi_indexes['train_names_seg'],image_size=size,type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(busi_indexes['val_names_flair'],busi_indexes['val_names_seg'],image_size=size,type_of_imgs='png', transform = ToTensor2D(True),data='busi')
datadict_test = ImageLoader2D(busi_indexes['test_names_flair'],busi_indexes['test_names_seg'],image_size=size,type_of_imgs='png', transform = ToTensor2D(True),data='busi',return_size=True,return_orig=True)
elif(which_data=='idrid'):
# Brain Tumor Segmentation Challenge
idrid_indexes = np.load('./Data_splits/Train_val_test_42_6_x/idrid_indexes.npy', allow_pickle=True).item()
idrid_indexes = helper_path_configuration(idrid_indexes,system_data_path)
datadict_train = ImageLoader2D(idrid_indexes['train_names_flair'],idrid_indexes['train_names_seg'],image_size=512,type_of_imgs='png',transform = composed_transform_2d,data='idrid')
datadict_val = ImageLoader2D(idrid_indexes['val_names_flair'],idrid_indexes['val_names_seg'],image_size=512,type_of_imgs='png', transform = ToTensor2D(True),data='idrid')
datadict_test = ImageLoader2D(idrid_indexes['test_names_flair'],idrid_indexes['test_names_seg'],image_size=512,type_of_imgs='png', transform = ToTensor2D(True),data='idrid',return_size=True,return_orig=True)
print(len(datadict_train),len(datadict_val),len(datadict_test))
return datadict_train,datadict_val,datadict_test
def helper_pre_training(which_data='brats',scale_factor=1.0,sim_path_other=None,size=(128,128,128),no_crop=False):
sim_path = sim_path_other
print('Using simulation path:{} for data {}'.format(sim_path_other,which_data))
if(which_data=='wmh'):
# WMH
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty',image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
print(len(train_names_flair),len(train_names_seg))
elif(which_data=='brats'):
# BraTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty', image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='lits'):
# BraTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty', image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='total3d'):
# BraTS
train_names_flair = []
train_names_seg = []
val_names_flair = []
val_names_seg = []
for which_data in ['bratsonsim','wmhonsim','litsonsim']:
train_names_flair += sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))
train_names_seg += sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))
val_names_flair += sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))
val_names_seg += sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty', image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='busi'):
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*image.png')))
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*mask.png')))
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*image.png')))
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*mask.png')))
datadict_train = ImageLoader2D(train_names_flair,train_names_seg,image_size=(512,512),type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(val_names_flair,val_names_seg,image_size=(512,512),type_of_imgs='png', transform = ToTensor2D(True),data='busi')
return datadict_train,datadict_val
def helper_self_supervised(which_data='brats',scale_factor=1.0,sim_path_other=None,size=(128,128,128),no_crop=False):
sim_path = sim_path_other
print('Using simulation path:{} for data {}'.format(sim_path_other,which_data))
train_size = np.ceil(42*scale_factor).astype(np.int16)
val_size = np.ceil(6*scale_factor).astype(np.int16)
if(which_data=='wmh'):
# WMH
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz'))) [:val_size]
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty',image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
print(len(train_names_flair),len(train_names_seg))
elif(which_data=='brats'):
# BraTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))[:val_size]
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty', image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='lits'):
# BraTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))[:val_size]
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty', image_size=size, transform = composed_transform,no_crop=no_crop)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', image_size=size, transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='busi'):
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*image.png')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*mask.png')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*image.png')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*mask.png')))[:val_size]
datadict_train = ImageLoader2D(train_names_flair,train_names_seg,image_size=(512,512),type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(val_names_flair,val_names_seg,image_size=(512,512),type_of_imgs='png', transform = ToTensor2D(True),data='busi')
elif(which_data=='idrid'):
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*image.png')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*mask.png')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*image.png')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*mask.png')))[:val_size]
datadict_train = ImageLoader2D(train_names_flair,train_names_seg,image_size=(512,512),type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(val_names_flair,val_names_seg,image_size=(512,512),type_of_imgs='png', transform = ToTensor2D(True),data='busi')
return datadict_train,datadict_val
def helper_data_augmentation(system_data_path,which_data='brats',factor=1.0,scale_factor=1.0,sim_path_other=None,size=(128,128,128),no_crop=False):
sim_path = sim_path_other
print('Using simulation path:{} for data {}'.format(sim_path_other,which_data))
augementation_factor = factor
train_size = 42
val_size = 6
if(which_data=='wmh'):
# White Matter Hyperintensities
wmh_indexes = np.load('./Data_splits/Train_val_test_42_6_x/wmh_indexes.npy', allow_pickle=True).item()
wmh_indexes = helper_path_configuration(wmh_indexes,system_data_path)
wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'] = wmh_indexes['train_names_flair'][:np.int16(np.ceil(augementation_factor*train_size))],wmh_indexes['train_names_seg'][:np.int16(np.ceil(augementation_factor*train_size))]
wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'] = wmh_indexes['val_names_flair'][:np.int16(np.ceil(augementation_factor*val_size))],wmh_indexes['val_names_seg'][:np.int16(np.ceil(augementation_factor*val_size))]
real_datadict_train = ImageLoader3D(wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'],image_size=size,type_of_imgs='nifty',transform = composed_transform,no_crop=no_crop)
real_datadict_val = ImageLoader3D(wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='brats'):
# BraTS
brats_indexes = np.load('./Data_splits/Train_val_test_42_6_x/brats_indexes.npy', allow_pickle=True).item()
brats_indexes = helper_path_configuration(brats_indexes,system_data_path)
brats_indexes['train_names_flair'],brats_indexes['train_names_seg'] = brats_indexes['train_names_flair'][:np.int16(np.ceil(augementation_factor*train_size))],brats_indexes['train_names_seg'][:np.int16(np.ceil(augementation_factor*train_size))]
brats_indexes['val_names_flair'],brats_indexes['val_names_seg'] = brats_indexes['val_names_flair'][:np.int16(np.ceil(augementation_factor*val_size))],brats_indexes['val_names_seg'][:np.int16(np.ceil(augementation_factor*val_size))]
real_datadict_train = ImageLoader3D(brats_indexes['train_names_flair'],brats_indexes['train_names_seg'],image_size=size,type_of_imgs='nifty',transform = composed_transform,no_crop=no_crop)
real_datadict_val = ImageLoader3D(brats_indexes['val_names_flair'],brats_indexes['val_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),no_crop=no_crop)
elif(which_data=='lits'):
# LiTS
liver_indexes = np.load('./Data_splits/Train_val_test_42_6_x/liver_indexes.npy', allow_pickle=True).item()
liver_indexes = helper_path_configuration(liver_indexes,system_data_path)
liver_indexes['train_names_flair'],liver_indexes['train_names_seg'] = liver_indexes['train_names_flair'][:np.int16(np.ceil(augementation_factor*train_size))],liver_indexes['train_names_seg'][:np.int16(np.ceil(augementation_factor*train_size))]
liver_indexes['val_names_flair'],liver_indexes['val_names_seg'] = liver_indexes['val_names_flair'][:np.int16(np.ceil(augementation_factor*val_size))],liver_indexes['val_names_seg'][:np.int16(np.ceil(augementation_factor*val_size))]
real_datadict_train = ImageLoader3D(liver_indexes['train_names_flair'],liver_indexes['train_names_seg'],image_size=size,type_of_imgs='nifty',transform = composed_transform,no_crop=no_crop,data='liver')
real_datadict_val = ImageLoader3D(liver_indexes['val_names_flair'],liver_indexes['val_names_seg'],image_size=size,type_of_imgs='nifty', transform = ToTensor3D(True),no_crop=no_crop,data='liver')
elif(which_data=='busi'):
busi_indexes = np.load('./Data_splits/Train_val_test_42_6_x/busi_indexes.npy', allow_pickle=True).item()
busi_indexes = helper_path_configuration(busi_indexes,system_data_path)
real_datadict_train = ImageLoader2D(busi_indexes['train_names_flair'],busi_indexes['train_names_seg'],image_size=(512,512),type_of_imgs='nifty',transform = composed_transform_2d,data='busi')
real_datadict_val = ImageLoader2D(busi_indexes['val_names_flair'],busi_indexes['val_names_seg'],image_size=(512,512),type_of_imgs='nifty', transform = ToTensor2D(True),data='busi')
elif(which_data=='idrid'):
idrid_indexes = np.load('./Data_splits/Train_val_test_42_6_x/idrid_indexes.npy', allow_pickle=True).item()
idrid_indexes = helper_path_configuration(idrid_indexes,system_data_path)
real_datadict_train = ImageLoader2D(idrid_indexes['train_names_flair'],idrid_indexes['train_names_seg'],image_size=(512,512),type_of_imgs='nifty',transform = composed_transform_2d,data='busi')
real_datadict_val = ImageLoader2D(idrid_indexes['val_names_flair'],idrid_indexes['val_names_seg'],image_size=(512,512),type_of_imgs='nifty', transform = ToTensor2D(True),data='busi')
train_size = np.ceil(42*scale_factor).astype(np.int16)
val_size = np.ceil(6*scale_factor).astype(np.int16)
if(which_data=='wmh'):
# WMH
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
train_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*.json')))[:train_size]
if(train_names_json!=[]):
new_train_names_flair = []
new_train_names_seg = []
for i in range(len(train_names_json)):
with open(train_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if system_data_path + clean_path[41:] not in wmh_indexes['train_names_flair']:
pass
else:
new_train_names_flair.append(train_names_flair[i])
new_train_names_seg.append(train_names_seg[i])
train_names_flair = new_train_names_flair
train_names_seg = new_train_names_seg
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))[:val_size]
val_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*.json')))[:val_size]
if(val_names_json!=[]):
new_val_names_flair = []
new_val_names_seg = []
for i in range(len(val_names_json)):
with open(val_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if (system_data_path + clean_path[41:] not in wmh_indexes['val_names_flair']) and (system_data_path + clean_path[41:] not in wmh_indexes['train_names_flair']):
pass
else:
new_val_names_flair.append(val_names_flair[i])
new_val_names_seg.append(val_names_seg[i])
val_names_flair = new_val_names_flair
val_names_seg = new_val_names_seg
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = composed_transform)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = ToTensor3D(True))
print(len(datadict_train),len(datadict_val))
elif(which_data=='brats'):
# BraTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
train_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*.json')))[:train_size]
if(train_names_json!=[]):
new_train_names_flair = []
new_train_names_seg = []
for i in range(len(train_names_json)):
with open(train_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if system_data_path + clean_path[41:] not in brats_indexes['train_names_flair']:
pass
else:
new_train_names_flair.append(train_names_flair[i])
new_train_names_seg.append(train_names_seg[i])
train_names_flair = new_train_names_flair
train_names_seg = new_train_names_seg
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))[:val_size]
val_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*.json')))[:val_size]
if(val_names_json!=[]):
new_val_names_flair = []
new_val_names_seg = []
for i in range(len(val_names_json)):
with open(val_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if (system_data_path + clean_path[41:] not in brats_indexes['val_names_flair']) and (system_data_path + clean_path[41:] not in brats_indexes['train_names_flair']):
pass
else:
new_val_names_flair.append(val_names_flair[i])
new_val_names_seg.append(val_names_seg[i])
val_names_flair = new_val_names_flair
val_names_seg = new_val_names_seg
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = composed_transform)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = ToTensor3D(True))
print(len(datadict_train),len(datadict_val))
elif(which_data=='lits'):
# LiTS
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*FLAIR.nii.gz')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*manualmask.nii.gz')))[:train_size]
train_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*.json')))[:train_size]
if(train_names_json!=[]):
new_train_names_flair = []
new_train_names_seg = []
for i in range(len(train_names_json)):
with open(train_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if system_data_path + clean_path[41:] not in liver_indexes['train_names_flair']:
pass
else:
new_train_names_flair.append(train_names_flair[i])
new_train_names_seg.append(train_names_seg[i])
train_names_flair = new_train_names_flair
train_names_seg = new_train_names_seg
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*FLAIR.nii.gz')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*manualmask.nii.gz')))[:val_size]
val_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*.json')))[:val_size]
if(val_names_json!=[]):
new_val_names_flair = []
new_val_names_seg = []
for i in range(len(val_names_json)):
with open(val_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if (system_data_path + clean_path[41:] not in liver_indexes['val_names_flair']) and (system_data_path + clean_path[41:] not in liver_indexes['train_names_flair']):
pass
else:
new_val_names_flair.append(val_names_flair[i])
new_val_names_seg.append(val_names_seg[i])
val_names_flair = new_val_names_flair
val_names_seg = new_val_names_seg
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = composed_transform)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty',image_size=size,no_crop=no_crop, transform = ToTensor3D(True))
print(len(datadict_train),len(datadict_val))
elif(which_data=='busi'):
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*image.png')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*mask.png')))[:train_size]
train_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*.json')))[:train_size]
new_train_names_flair = []
new_train_names_seg = []
for i in range(len(train_names_json)):
with open(train_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if system_data_path + clean_path[41:] not in busi_indexes['train_names_flair']:
pass
else:
new_train_names_flair.append(train_names_flair[i])
new_train_names_seg.append(train_names_seg[i])
if(len(train_names_json)):
train_names_flair = new_train_names_flair
train_names_seg = new_train_names_seg
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*image.png')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*mask.png')))[:val_size]
val_names_json = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*.json')))[:val_size]
new_val_names_flair = []
new_val_names_seg = []
for i in range(len(val_names_json)):
with open(val_names_json[i]) as json_file:
data = json.load(json_file)
clean_path = data['clean_path']
if (system_data_path + clean_path[41:] not in busi_indexes['val_names_flair']) and (system_data_path + clean_path[41:] not in busi_indexes['train_names_flair']):
pass
else:
new_val_names_flair.append(val_names_flair[i])
new_val_names_seg.append(val_names_seg[i])
if(len(val_names_json)):
val_names_flair = new_val_names_flair
val_names_seg = new_val_names_seg
datadict_train = ImageLoader2D(train_names_flair,train_names_seg,image_size=(512,512),type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(val_names_flair,val_names_seg,image_size=(512,512),type_of_imgs='png', transform = ToTensor2D(True),data='busi')
print(len(datadict_train),len(datadict_val))
elif(which_data=='idrid'):
train_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*image.png')))[:train_size]
train_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/TrainSet/*mask.png')))[:train_size]
val_names_flair = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*image.png')))[:val_size]
val_names_seg = sorted(glob.glob((sim_path+'/'+which_data+'/ValSet/*mask.png')))[:val_size]
datadict_train = ImageLoader2D(train_names_flair,train_names_seg,image_size=(512,512),type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(val_names_flair,val_names_seg,image_size=(512,512),type_of_imgs='png', transform = ToTensor2D(True),data='busi')
datadict_train = torch.utils.data.ConcatDataset([real_datadict_train,datadict_train])
datadict_val = torch.utils.data.ConcatDataset([real_datadict_val,datadict_val])
return datadict_train,datadict_val
def helper_fine_tuning(system_data_path,which_data='brats',factor=1.0,scale_factor=1.0,size=(128,128,128),no_crop=False):
fine_tuning_factor = factor
train_size = 42
val_size = 6
if(which_data=='wmh'):
# White Matter Hyperintensities
wmh_indexes = np.load('./Data_splits/Train_val_test_42_6_x/wmh_indexes.npy', allow_pickle=True).item()
wmh_indexes = helper_path_configuration(wmh_indexes,system_data_path)
wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'] = wmh_indexes['train_names_flair'][:np.int16(np.ceil(fine_tuning_factor*train_size))],wmh_indexes['train_names_seg'][:np.int16(np.ceil(fine_tuning_factor*train_size))]
wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'] = wmh_indexes['val_names_flair'][:np.int16(np.ceil(fine_tuning_factor*val_size))],wmh_indexes['val_names_seg'][:np.int16(np.ceil(fine_tuning_factor*val_size))]
datadict_train = ImageLoader3D(wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty',transform = composed_transform)
datadict_val = ImageLoader3D(wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty', transform = ToTensor3D(True))
elif(which_data=='brats'):
# Brain Tumor Segmentation Challenge
brats_indexes = np.load('./Data_splits/Train_val_test_42_6_x/brats_indexes.npy', allow_pickle=True).item()
brats_indexes = helper_path_configuration(brats_indexes,system_data_path)
brats_indexes['train_names_flair'],brats_indexes['train_names_seg'] = brats_indexes['train_names_flair'][:np.int16(np.ceil(fine_tuning_factor*train_size))],brats_indexes['train_names_seg'][:np.int16(np.ceil(fine_tuning_factor*train_size))]
brats_indexes['val_names_flair'],brats_indexes['val_names_seg'] = brats_indexes['val_names_flair'][:np.int16(np.ceil(fine_tuning_factor*val_size))],brats_indexes['val_names_seg'][:np.int16(np.ceil(fine_tuning_factor*val_size))]
datadict_train = ImageLoader3D(brats_indexes['train_names_flair'],brats_indexes['train_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty',transform = composed_transform)
datadict_val = ImageLoader3D(brats_indexes['val_names_flair'],brats_indexes['val_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty', transform = ToTensor3D(True))
elif(which_data=='lits'):
# Brain Tumor Segmentation Challenge
liver_indexes = np.load('./Data_splits/Train_val_test_42_6_x/liver_indexes.npy', allow_pickle=True).item()
liver_indexes = helper_path_configuration(liver_indexes,system_data_path)
liver_indexes['train_names_flair'],liver_indexes['train_names_seg'] = liver_indexes['train_names_flair'][:np.int16(np.ceil(fine_tuning_factor*train_size))],liver_indexes['train_names_seg'][:np.int16(np.ceil(fine_tuning_factor*train_size))]
liver_indexes['val_names_flair'],liver_indexes['val_names_seg'] = liver_indexes['val_names_flair'][:np.int16(np.ceil(fine_tuning_factor*val_size))],liver_indexes['val_names_seg'][:np.int16(np.ceil(fine_tuning_factor*val_size))]
datadict_train = ImageLoader3D(liver_indexes['train_names_flair'],liver_indexes['train_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty',transform = composed_transform,data='liver')
datadict_val = ImageLoader3D(liver_indexes['val_names_flair'],liver_indexes['val_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty', transform = ToTensor3D(True),data='liver')
elif(which_data=='busi'):
busi_indexes = np.load('./Data_splits/Train_val_test_42_6_x/busi_indexes.npy', allow_pickle=True).item()
busi_indexes = helper_path_configuration(busi_indexes,system_data_path)
datadict_train = ImageLoader2D(busi_indexes['train_names_flair'],busi_indexes['train_names_seg'],image_size=512,type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(busi_indexes['val_names_flair'],busi_indexes['val_names_seg'],image_size=512,type_of_imgs='png', transform = ToTensor2D(True),data='busi')
elif(which_data=='idrid'):
idrid_indexes = np.load('./Data_splits/Train_val_test_42_6_x/idrid_indexes.npy', allow_pickle=True).item()
idrid_indexes = helper_path_configuration(idrid_indexes,system_data_path)
datadict_train = ImageLoader2D(idrid_indexes['train_names_flair'],idrid_indexes['train_names_seg'],image_size=512,type_of_imgs='png',transform = composed_transform_2d,data='busi')
datadict_val = ImageLoader2D(idrid_indexes['val_names_flair'],idrid_indexes['val_names_seg'],image_size=512,type_of_imgs='png', transform = ToTensor2D(True),data='busi')
return datadict_train,datadict_val
def helper_ss_data_adaptation(which_data='brats',factor=1.0,adapt_path=None,adapt_save_path=None,model=None,hyper_parameters=None,device=0,size=(128,128,128),no_crop=False):
sim_path = './simulation_data/Full_sim_22_11_23/'
fine_tuning_factor = factor
if(which_data=='wmh'):
# White Matter Hyperintensities
train_size = 42
val_size = 6
test_size = 12
wmh_indexes = np.load('./Data_splits/Train_val_test_42_6_x/wmh_indexes.npy', allow_pickle=True).item()
wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'] = wmh_indexes['train_names_flair'][:np.int16(np.ceil(fine_tuning_factor*train_size))],wmh_indexes['train_names_seg'][:np.int16(np.ceil(fine_tuning_factor*train_size))]
wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'] = wmh_indexes['val_names_flair'][:np.int16(np.ceil(fine_tuning_factor*val_size))],wmh_indexes['val_names_seg'][:np.int16(np.ceil(fine_tuning_factor*val_size))]
datadict_train = ImageLoader3D(wmh_indexes['train_names_flair'],wmh_indexes['train_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty',transform = composed_transform,return_orig=True)
datadict_val = ImageLoader3D(wmh_indexes['val_names_flair'],wmh_indexes['val_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty', transform = ToTensor3D(True),return_orig=True)
elif(which_data=='brats'):
# Brain Tumor Segmentation Challenge
train_size = 876
val_size = 125
test_size = 250
brats_indexes = np.load('./Data_splits/Train_val_test_42_6_x/brats_indexes.npy', allow_pickle=True).item()
brats_indexes['train_names_flair'],brats_indexes['train_names_seg'] = brats_indexes['train_names_flair'][:np.int16(np.ceil(fine_tuning_factor*train_size))],brats_indexes['train_names_seg'][:np.int16(np.ceil(fine_tuning_factor*train_size))]
brats_indexes['val_names_flair'],brats_indexes['val_names_seg'] = brats_indexes['val_names_flair'][:np.int16(np.ceil(fine_tuning_factor*val_size))],brats_indexes['val_names_seg'][:np.int16(np.ceil(fine_tuning_factor*val_size))]
datadict_train = ImageLoader3D(brats_indexes['train_names_flair'],brats_indexes['train_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty',transform = composed_transform,return_orig=True)
datadict_val = ImageLoader3D(brats_indexes['val_names_flair'],brats_indexes['val_names_seg'],image_size=size,no_crop=no_crop,type_of_imgs='nifty', transform = ToTensor3D(True),return_orig=True)
adapt_model = helper_model(model,which_data=which_data,hyper_parameters=hyper_parameters,device=device)
adapt_model.load_state_dict(torch.load(adapt_path)['model_state_dict'])
adapt_model.eval()
trainloader = DataLoader(datadict_train, batch_size=1, shuffle=False,num_workers=1)
valloader = DataLoader(datadict_val, batch_size=1, shuffle=False,num_workers=1)
os.makedirs(adapt_save_path+'TrainSet/',exist_ok=True)
os.makedirs(adapt_save_path+'ValSet/',exist_ok=True)
print('--------------------------------------Adapting Train----------------------------------------')
with tqdm(range(len(trainloader))) as pbar:
for i, data in zip(pbar, trainloader):
with torch.no_grad():
torch.cuda.empty_cache()
image = data['input'].to(device)
output = adapt_model.forward(image) > 0.5
image1 = nib.Nifti1Image(image[0,0].cpu().numpy(), affine=data['affine'][0])
nib.save(image1, adapt_save_path+'TrainSet/Training'+str(i)+'_FLAIR.nii.gz')
output1 = nib.Nifti1Image(output[0,0].cpu().numpy().astype(np.single), affine=data['affine'][0])
nib.save(output1, adapt_save_path+'TrainSet/Training'+str(i)+'_manualmask.nii.gz')
print('--------------------------------------Adapting Val----------------------------------------')
with tqdm(range(len(valloader))) as pbar:
for i, data in zip(pbar, valloader):
with torch.no_grad():
torch.cuda.empty_cache()
image = data['input'].to(device)
output = adapt_model.forward(image) > 0.5
image1 = nib.Nifti1Image(image[0,0].cpu().numpy(), affine=data['affine'][0])
nib.save(image1, adapt_save_path+'ValSet/Validation'+str(i)+'_FLAIR.nii.gz')
output1 = nib.Nifti1Image(output[0,0].cpu().numpy().astype(np.single), affine=data['affine'][0])
nib.save(output1, adapt_save_path+'ValSet/Valdiation'+str(i)+'_manualmask.nii.gz')
train_names_flair = sorted(glob.glob((adapt_save_path+'TrainSet/*FLAIR.nii.gz')))
train_names_seg = sorted(glob.glob((adapt_save_path+'TrainSet/*manualmask.nii.gz')))
val_names_flair = sorted(glob.glob((adapt_save_path+'ValSet/*FLAIR.nii.gz')))
val_names_seg = sorted(glob.glob((adapt_save_path+'ValSet/*manualmask.nii.gz')))
datadict_train = ImageLoader3D(train_names_flair,train_names_seg,type_of_imgs='nifty',transform = composed_transform,data=which_data)
datadict_val = ImageLoader3D(val_names_flair,val_names_seg,type_of_imgs='nifty', transform = ToTensor3D(True),data=which_data)
return datadict_train,datadict_val
def helper_model(model_type,which_data,hyper_parameters,device=0,size=(128,128,128)):
if(model_type == 'nestedunet'):
model = NestedUNet(out_channels=1,**hyper_parameters).to(device)
if(which_data=='busi' or which_data=='idrid'):
model = NestedUNet2D(out_channels=1,**hyper_parameters).to(device)
elif(model_type == 'nestedunet_dropout'):
model = NestedUNetDropout(out_channels=1,**hyper_parameters).to(device)
if(which_data=='busi' or which_data=='idrid'):
model = NestedUNet2D(out_channels=1,**hyper_parameters).to(device)
elif(model_type == 'halfunet'):
model = HalfUNet(out_channels=1).to(device)
if(which_data=='busi' or which_data=='idrid'):
model = HalfUNet2D(out_channels=1,**hyper_parameters).to(device)
elif(model_type == 'resunet'):
model = ResUNet(out_channels=1).to(device)
if(which_data=='busi' or which_data=='idrid'):
model = ResUNet2D(out_channels=1,**hyper_parameters).to(device)
elif(model_type == 'unet'):
model = UNet(out_channels=1,**hyper_parameters).to(device)
if(which_data=='busi' or which_data=='idrid'):
model = UNet2D(out_channels=1,**hyper_parameters).to(device)
elif(model_type == 'unetr'):
config = {
"patch_size": 16, # Input image size: 128x128 -> 8x8 patches
"hidden_size": 768,
"num_hidden_layers": 12,
"num_attention_heads": 12,
"intermediate_size": 4 * 768, # 4 * hidden_size
"hidden_dropout_prob": 0.0,
"attention_probs_dropout_prob": 0.0,
"initializer_range": 0.02,
"image_size": (128,128,128),
"num_classes": 1, # num_classes of binary
"num_channels": 1,
"qkv_bias": True,
"use_faster_attention": True,
}
config['image_size'] = size
if(which_data=='busi' or which_data=='idrid'):
model = VITForSegmentation2D(config=config).to(device)
else:
model = VITForSegmentation(config=config).to(device)
elif(model_type == 'slimunetr'):
if(which_data=='busi' or which_data=='idrid'):
model = SlimUNETR2D(in_channels=1, out_channels=1,**hyper_parameters).to(device)
else:
model = SlimUNETR(in_channels=1, out_channels=1,**hyper_parameters).to(device)
return model
def helper_criterion(criterion_type='dice',device=0):
if(criterion_type == 'wbce'):
criterion= BCE_Loss_Weighted(weight=5).to(device)
elif(criterion_type == 'bce'):
criterion = nn.BCELoss().to(device)
elif(criterion_type == 'dice'):
criterion = DiceLoss().to(device)
elif(criterion_type == 'logcoshdice'):
criterion = LogCoshDiceLoss().to(device)
elif(criterion_type == 'adaptive'):
criterion = Adaptive_Region_Specific_TverskyLoss().to(device)
elif(criterion_type == 'wbce + dice'):
criterion = WBCE_DICELoss().to(device)
elif(criterion_type == 'wbce + focaldice'):
criterion = WBCE_FOCALDICELoss().to(device)
elif(criterion_type == 'focal'):
criterion = FocalLoss().to(device)
elif(criterion_type == 'focal + dice'):
criterion = FOCAL_DICELoss().to(device)
elif(criterion_type == 'wbce + focal'):
criterion = FOCAL_DICELoss().to(device)
return criterion
def izipmerge(a, b):
for i, j in zip(a,b):
yield i
yield j
def helper_train(epoch,model,criterion,optimizer,scheduler,train_losses,train_dices,val_losses,val_dices,trainloader,valloader,adapt_model=None,device=0):
epoch_loss = 0
epoch_dice = 0
model.train()
train_size = len(trainloader)
val_size = len(valloader)
if(type(trainloader) == tuple):
train_size = len(trainloader[0])*2
val_size = len(valloader[0])*2
trainloader = izipmerge(*trainloader)
valloader = izipmerge(*valloader)
with tqdm(range(train_size)) as pbar:
for i, data in zip(pbar, trainloader):
torch.cuda.empty_cache()
err = 0
image = data['input'].to(device)
output = model.forward(image)
label = data['gt'].to(device)
if(adapt_model=='unet_synth_tum'):
# print(output[:,2:,:,:,:].shape)
err = criterion(output[:,2,:,:,:],label[:,0,:,:,:])
dice = Dice_Score(torch.sigmoid(output[:,2,:,:,:]).cpu().detach().numpy(),label[:,0,:,:,:].cpu().detach().numpy())
else:
err = criterion(output,label,wt = label)
dice = Dice_Score(output.cpu().detach().numpy(),label.cpu().detach().numpy())
model.zero_grad()
err.backward()
optimizer.step()
pbar.set_postfix(Train_Loss = np.round(err.cpu().detach().numpy().item(), 5),Train_Dice = np.round(dice, 5))
pbar.update(0)
epoch_loss += err.item()
epoch_dice += dice
del label
del err
train_losses.append([epoch_loss/train_size])
train_dices.append([epoch_dice/train_size])
print('Training Loss and Dice at epoch {} is : Total {} and {}'.format(epoch,*train_losses[-1],*train_dices[-1]))
epoch_loss = 0
epoch_dice = 0
model.eval()
with tqdm(range(val_size)) as pbar:
for i, data in zip(pbar, valloader):
torch.cuda.empty_cache()
err = 0
with torch.no_grad():
image = data['input'].to(device)
output = model.forward(image)
label = data['gt'].to(device)
if(adapt_model=='unet_synth_tum'):
# print(output[:,2:,:,:,:].shape)
err = criterion(output[:,2,:,:,:],label[:,0,:,:,:])
dice = Dice_Score(torch.sigmoid(output[:,2,:,:,:]).cpu().detach().numpy(),label[:,0,:,:,:].cpu().detach().numpy())
else:
err = criterion(output,label,wt = label)
dice = Dice_Score(output.cpu().detach().numpy(),label.cpu().detach().numpy())
del image
del label
pbar.set_postfix(Val_Loss = np.round(err.cpu().detach().numpy().item(), 5),Val_Dice=np.round(dice, 5))
pbar.update(0)
epoch_loss += err.item()
epoch_dice += dice
del err
val_losses.append([epoch_loss/val_size])
val_dices.append([epoch_dice/val_size])
print('Validation Loss and Dice at epoch {} is : Total {} and {}'.format(epoch,*val_losses[-1],*val_dices[-1]))
scheduler.step(*val_dices[-1])
return epoch_loss,epoch_dice
def helper_save_model(epoch,model,optimizer,epoch_loss,epoch_dice,scheduler,model_type,best_dice=False):
if(not best_dice):
loss_type = '_state_dict'
else:
loss_type = '_state_dict_best_dice'
torch.save({
'epoch': epoch,
'model_state_dict': model.state_dict(),
'optimizer_state_dict': optimizer.state_dict(),
'loss': epoch_loss,
'dice':epoch_dice,
'lr_scheduler_state_dict':scheduler.state_dict(),
}, model_type+loss_type+str(epoch)+'.pth')