forked from jackyko1991/vnet-tensorflow
-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
1108 lines (925 loc) · 45.3 KB
/
model.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
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf
import NiftiDataset3D
import NiftiDataset2D
import sys
import datetime
import numpy as np
import networks
import math
import SimpleITK as sitk
import multiprocessing
from tqdm import tqdm
def grayscale_to_rainbow(image):
# grayscale to rainbow colormap, convert to HSV (H = reversed grayscale from 0:2/3, S and V are all 1)
# then convert to RGB
H = tf.squeeze((1. - image)*2./3., axis=-1)
SV = tf.ones(H.get_shape())
HSV = tf.stack([H,SV,SV], axis=3)
RGB = tf.image.hsv_to_rgb(HSV)
return RGB
def dice_coe(output, target, loss_type='jaccard', axis=(1, 2, 3), weighted=False, smooth=1e-5):
"""Soft dice (Sørensen or Jaccard) coefficient for comparing the similarity
of two batch of data, usually be used for binary image segmentation
i.e. labels are binary. The coefficient between 0 to 1, 1 means totally match.
Parameters
-----------
output : Tensor
A distribution with shape: [batch_size, ....], (any dimensions).
target : Tensor
The target distribution, format the same with `output`.
loss_type : str
``jaccard`` or ``sorensen``, default is ``jaccard``.
axis : tuple of int
All dimensions are reduced, default ``[1,2,3]``.
weight : bool
Boolean option for generalized dice loss
smooth : float
This small value will be added to the numerator and denominator.
- If both output and target are empty, it makes sure dice is 1.
- If either output or target are empty (all pixels are background), dice = ```smooth/(small_value + smooth)``, then if smooth is very small, dice close to 0 (even the image values lower than the threshold), so in this case, higher smooth can have a higher dice.
Examples
---------
>>> import tensorlayer as tl
>>> outputs = tl.act.pixel_wise_softmax(outputs)
>>> dice_loss = 1 - tl.cost.dice_coe(outputs, y_)
References
-----------
- `Wiki-Dice <https://en.wikipedia.org/wiki/Sørensen–Dice_coefficient>`__
"""
inse = tf.reduce_sum(output * target, axis=axis)
if loss_type == 'jaccard':
l = tf.reduce_sum(output * output, axis=axis)
r = tf.reduce_sum(target * target, axis=axis)
elif loss_type == 'sorensen':
l = tf.reduce_sum(output, axis=axis)
r = tf.reduce_sum(target, axis=axis)
else:
raise Exception("Unknown loss_type")
if weighted:
w = 1/(tf.reduce_sum(target*target, axis=axis) + smooth)
dice = tf.reduce_sum(2.* w * inse +smooth, axis=-1)/tf.reduce_sum(w*(l + r + smooth),axis=-1)
dice = tf.reduce_mean(dice, name='dice_coe')
else:
# old axis=[0,1,2,3]
# dice = 2 * (inse) / (l + r)
# epsilon = 1e-5
# dice = tf.clip_by_value(dice, 0, 1.0-epsilon) # if all empty, dice = 1
# new haodong
dice = (2. * inse + smooth) / (l + r + smooth)
dice = tf.reduce_mean(dice, name='dice_coe')
return dice
def prepare_batch(image_ijk_patch_indices_dict):
# image_batches = []
# for batch in ijk_patch_indices:
# image_batch = []
# for patch in batch:
# image_patch = images[patch[0]:patch[1],patch[2]:patch[3],patch[4]:patch[5],:]
# image_batch.append(image_patch)
# image_batch = np.asarray(image_batch)
# image_batches.append(image_batch)
images, ijk_patch_indices = image_ijk_patch_indices_dict['images'], image_ijk_patch_indices_dict['indexes']
# return image_batches
image_batch = []
for patch in ijk_patch_indices:
image_patch = images[patch[0]:patch[1],patch[2]:patch[3],patch[4]:patch[5],:]
image_batch.append(image_patch)
image_batch = np.asarray(image_batch)
return image_batch
class image2label(object):
def __init__(self,sess,config):
"""
Args:
sess: Tensorflow session
config: Model configuration
"""
self.sess = sess
self.config = config
self.model = None
self.graph = tf.Graph()
self.graph.as_default()
self.epoches = 999999999999999999
def read_config(self):
print("{}: Reading configuration file...".format(datetime.datetime.now()))
# training config
self.input_channel_num = len(self.config['TrainingSetting']['Data']['ImageFilenames'])
self.output_channel_num = len(self.config['TrainingSetting']['SegmentationClasses'])
self.label_classes = self.config['TrainingSetting']['SegmentationClasses']
self.train_data_dir = self.config['TrainingSetting']['Data']['TrainingDataDirectory']
self.test_data_dir = self.config['TrainingSetting']['Data']['TestingDataDirectory']
self.image_filenames = self.config['TrainingSetting']['Data']['ImageFilenames']
self.label_filename = self.config['TrainingSetting']['Data']['LabelFilename']
self.batch_size = self.config['TrainingSetting']['BatchSize']
self.patch_shape = self.config['TrainingSetting']['PatchShape']
self.dimension = len(self.config['TrainingSetting']['PatchShape'])
self.image_log = self.config['TrainingSetting']['ImageLog']
self.testing = self.config['TrainingSetting']['Testing']
self.test_step = self.config['TrainingSetting']['TestStep']
self.restore_training = self.config['TrainingSetting']['Restore']
self.log_dir = self.config['TrainingSetting']['LogDir']
self.ckpt_dir = self.config['TrainingSetting']['CheckpointDir']
self.epoches = self.config['TrainingSetting']['Epoches']
self.max_itr = self.config['TrainingSetting']['MaxIterations']
self.log_interval = self.config['TrainingSetting']['LogInterval']
self.network_name = self.config['TrainingSetting']['Networks']['Name']
self.dropout_rate = self.config['TrainingSetting']['Networks']['Dropout']
self.optimizer_name = self.config['TrainingSetting']['Optimizer']['Name']
self.initial_learning_rate = self.config['TrainingSetting']['Optimizer']['InitialLearningRate']
self.decay_factor = self.config['TrainingSetting']['Optimizer']['Decay']['Factor']
self.decay_steps = self.config['TrainingSetting']['Optimizer']['Decay']['Steps']
self.spacing = self.config['TrainingSetting']['Spacing']
self.drop_ratio = self.config['TrainingSetting']['DropRatio']
self.min_pixel = self.config['TrainingSetting']['MinPixel']
self.loss_name = self.config['TrainingSetting']['Loss']
# evaluation config
self.checkpoint_path = self.config['EvaluationSetting']['CheckpointPath']
self.evaluate_data_dir = self.config['EvaluationSetting']['Data']['EvaluateDataDirectory']
self.evaluate_image_filenames = self.config['EvaluationSetting']['Data']['ImageFilenames']
self.evaluate_label_filename = self.config['EvaluationSetting']['Data']['LabelFilename']
self.evaluate_probability_filename = self.config['EvaluationSetting']['Data']['ProbabilityFilename']
self.evaluate_stride = self.config['EvaluationSetting']['Stride']
self.evaluate_batch = self.config['EvaluationSetting']['BatchSize']
self.evaluate_probability_output = self.config['EvaluationSetting']['ProbabilityOutput']
print("{}: Reading configuration file complete".format(datetime.datetime.now()))
def placeholder_inputs(self, input_batch_shape, output_batch_shape):
"""Generate placeholder variables to represent the the input tensors.
These placeholders are used as inputs by the rest of the model building
code and will be fed from the downloaded ckpt in the .run() loop, below.
Args:
patch_shape: The patch_shape will be baked into both placeholders.
Returns:
images_placeholder: Images placeholder.
labels_placeholder: Labels placeholder.
"""
# Note that the shapes of the placeholders match the shapes of the full
# image and label tensors, except the first dimension is now batch_size
# rather than the full size of the train or test ckpt sets.
# batch_size = -1
images_placeholder = tf.placeholder(tf.float32, shape=input_batch_shape, name="images_placeholder")
labels_placeholder = tf.placeholder(tf.int32, shape=output_batch_shape, name="labels_placeholder")
return images_placeholder, labels_placeholder
def dataset_iterator(self, data_dir, transforms, train=True):
if self.dimension==2:
Dataset = NiftiDataset2D.NiftiDataset(
data_dir=data_dir,
image_filenames=self.image_filenames,
label_filename=self.label_filename,
transforms3D=transforms['3D'],
transforms2D=transforms['2D'],
train=train,
labels=self.label_classes
)
else:
Dataset = NiftiDataset3D.NiftiDataset(
data_dir=data_dir,
image_filenames=self.image_filenames,
label_filename=self.label_filename,
transforms=transforms,
train=train,
labels=self.label_classes
)
dataset = Dataset.get_dataset()
if self.dimension == 2:
dataset = dataset.shuffle(buffer_size=5)
else:
dataset = dataset.shuffle(buffer_size=3)
dataset = dataset.batch(self.batch_size,drop_remainder=True)
return dataset.make_initializable_iterator()
def build_model_graph(self):
print("{}: Start to build model graph...".format(datetime.datetime.now()))
self.global_step_op = tf.train.get_or_create_global_step()
if self.dimension == 2:
input_batch_shape = (None, self.patch_shape[0], self.patch_shape[1], self.input_channel_num)
output_batch_shape = (None, self.patch_shape[0], self.patch_shape[1], 1)
elif self.dimension == 3:
input_batch_shape = (None, self.patch_shape[0], self.patch_shape[1], self.patch_shape[2], self.input_channel_num)
output_batch_shape = (None, self.patch_shape[0], self.patch_shape[1], self.patch_shape[2], 1)
else:
sys.exit('Invalid Patch Shape (length should be 2 or 3)')
self.images_placeholder, self.labels_placeholder = self.placeholder_inputs(input_batch_shape,output_batch_shape)
# plot input and output images to tensorboard
if self.image_log:
if self.dimension == 2:
for image_channel in range(self.input_channel_num):
image_log = tf.cast(self.images_placeholder[:,:,:,image_channel:image_channel+1], dtype=tf.uint8)
tf.summary.image(self.image_filenames[image_channel], image_log, max_outputs=self.batch_size)
if 0 in self.label_classes:
labels_log = tf.cast(self.labels_placeholder*math.floor(255/(self.output_channel_num-1)), dtype=tf.uint8)
else:
labels_log = tf.cast(self.labels_placeholder*math.floor(255/self.output_channel_num), dtype=tf.uint8)
tf.summary.image("label",labels_log, max_outputs=self.batch_size)
else:
for batch in range(self.batch_size):
for image_channel in range(self.input_channel_num):
image_log = tf.cast(self.images_placeholder[batch:batch+1,:,:,:,image_channel], dtype=tf.uint8)
tf.summary.image(self.image_filenames[image_channel]+"_batch"+str(batch), tf.transpose(image_log,[3,1,2,0]),max_outputs=self.patch_shape[-1])
if 0 in self.label_classes:
labels_log = tf.cast(self.labels_placeholder[batch:batch+1,:,:,:,0]*math.floor(255/(self.output_channel_num-1)),dtype=tf.uint8)
else:
labels_log = tf.cast(self.labels_placeholder[batch:batch+1,:,:,:,0]*math.floor(255/self.output_channel_num), dtype=tf.uint8)
tf.summary.image("label"+"_batch"+str(batch), tf.transpose(labels_log,[3,1,2,0]),max_outputs=self.patch_shape[-1])
# Get images and labels
# create transformations to image and labels
# Force input pipepline to CPU:0 to avoid operations sometimes ended up at GPU and resulting a slow down
with tf.device('/cpu:0'):
if self.dimension == 2:
train_transforms_3d = []
train_transforms_2d = [
NiftiDataset2D.ManualNormalization(0,300),
NiftiDataset2D.Resample(self.spacing),
NiftiDataset2D.Padding(self.patch_shape),
NiftiDataset2D.RandomCrop(self.patch_shape)
]
test_transforms_3d = []
test_transforms_2d = [
NiftiDataset2D.ManualNormalization(0,300),
NiftiDataset2D.Resample(self.spacing),
NiftiDataset2D.Padding(self.patch_shape),
NiftiDataset2D.RandomCrop(self.patch_shape)
]
trainTransforms = {"3D": train_transforms_3d, "2D": train_transforms_2d}
testTransforms = {"3D": test_transforms_3d, "2D": test_transforms_2d}
else:
trainTransforms = [
# NiftiDataset.Normalization(),
# NiftiDataset3D.ExtremumNormalization(0.1),
NiftiDataset3D.ManualNormalization(0,300),
# NiftiDataset3D.StatisticalNormalization(2.5),
NiftiDataset3D.Resample((self.spacing[0],self.spacing[1],self.spacing[2])),
NiftiDataset3D.Padding((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2])),
NiftiDataset3D.RandomCrop((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2]),self.drop_ratio, self.min_pixel),
# NiftiDataset.ConfidenceCrop((FLAGS.patch_size*3, FLAGS.patch_size*3, FLAGS.patch_layer*3),(0.0001,0.0001,0.0001)),
# NiftiDataset.BSplineDeformation(randomness=2),
# NiftiDataset.ConfidenceCrop((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2]),(0.5,0.5,0.5)),
# NiftiDataset3D.ConfidenceCrop2((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2]),rand_range=32,probability=0.8),
# NiftiDataset3D.RandomFlip([True, False, False]),
NiftiDataset3D.RandomNoise()
]
# use random crop for testing
testTransforms = [
# NiftiDataset.Normalization(),
# NiftiDataset3D.ExtremumNormalization(0.1),
NiftiDataset3D.ManualNormalization(0,300),
# NiftiDataset3D.StatisticalNormalization(2.5),
NiftiDataset3D.Resample((self.spacing[0],self.spacing[1],self.spacing[2])),
NiftiDataset3D.Padding((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2])),
NiftiDataset3D.RandomCrop((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2]),self.drop_ratio, self.min_pixel)
# NiftiDataset.ConfidenceCrop((FLAGS.patch_size*2, FLAGS.patch_size*2, FLAGS.patch_layer*2),(0.0001,0.0001,0.0001)),
# NiftiDataset.BSplineDeformation(),
# NiftiDataset.ConfidenceCrop((self.patch_shape[0], self.patch_shape[1], self.patch_shape[2]),(0.75,0.75,0.75)),
# NiftiDataset.ConfidenceCrop2((FLAGS.patch_size, FLAGS.patch_size, FLAGS.patch_layer),rand_range=32,probability=0.8),
# NiftiDataset.RandomFlip([True, False, False]),
]
# get input and output datasets
self.train_iterator = self.dataset_iterator(self.train_data_dir, trainTransforms)
self.next_element_train = self.train_iterator.get_next()
if self.testing:
self.test_iterator = self.dataset_iterator(self.test_data_dir, testTransforms)
self.next_element_test = self.test_iterator.get_next()
print("{}: Dataset pipeline complete".format(datetime.datetime.now()))
# network models:
if self.network_name == "FCN":
sys.exit("Network to be developed")
elif self.network_name == "UNet":
self.network = networks.UNet(
num_output_channels=self.output_channel_num,
dropout_rate=0.01,
num_channels=4,
num_levels=4,
num_convolutions=2,
bottom_convolutions=2,
is_training=True,
activation_fn="relu"
)
elif self.network_name =="VNet":
self.network = networks.VNet(
num_classes=self.output_channel_num,
dropout_rate=self.dropout_rate,
num_channels=16,
num_levels=4,
num_convolutions=(1, 2, 3, 3),
bottom_convolutions=3,
is_training = True,
activation_fn="prelu"
)
else:
sys.exit("Invalid Network")
print("{}: Core network complete".format(datetime.datetime.now()))
self.logits = self.network.GetNetwork(self.images_placeholder)
# softmax op
self.softmax_op = tf.nn.softmax(self.logits,name="softmax")
if self.image_log:
if self.dimension == 2:
for output_channel in range(self.output_channel_num):
# softmax_log = grayscale_to_rainbow(self.softmax_op[:,:,:,output_channel:output_channel+1])
softmax_log = self.softmax_op[:,:,:,output_channel:output_channel+1]
softmax_log = tf.cast(softmax_log*255, dtype = tf.uint8)
tf.summary.image("softmax_" + str(self.label_classes[output_channel]),softmax_log,max_outputs=self.batch_size)
else:
for batch in range(self.batch_size):
for output_channel in range(self.output_channel_num):
softmax_log = grayscale_to_rainbow(tf.transpose(self.softmax_op[batch:batch+1,:,:,:,output_channel],[3,1,2,0]))
softmax_log = tf.cast(softmax_log*255,dtype=tf.uint8)
tf.summary.image("softmax_" + str(self.label_classes[output_channel])+"_batch"+str(batch),softmax_log,max_outputs=self.patch_shape[-1])
print("{}: Output layers complete".format(datetime.datetime.now()))
# loss function
with tf.name_scope("loss"):
# """
# Tricks for faster converge: Here we provide two calculation methods, first one will ignore to classical dice formula
# method 1: exclude the 0-th label in dice calculation. to use this method properly, you must set 0 as the first value in SegmentationClasses in config.json
# method 2: dice will be average on all classes
# """
if self.dimension == 2:
labels = tf.one_hot(self.labels_placeholder[:,:,:,0], depth=self.output_channel_num)
else:
labels = tf.one_hot(self.labels_placeholder[:,:,:,:,0], depth=self.output_channel_num)
# if 0 in self.label_classes:
# ################### method 1 ###################
# if self.dimension ==2:
# labels = labels[:,:,:,1:]
# softmax = self.softmax_op[:,:,:,1:]
# else:
# labels = labels[:,:,:,:,1:]
# softmax = self.softmax_op[:,:,:,:,1:]
# else:
# ################### method 2 ###################
# labels = labels
# softmax = self.softmax_op
labels = labels
softmax = self.softmax_op
if (self.loss_name == "xent"):
self.loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels,logits=self.logits))
elif (self.loss_name == "sorensen"):
if self.dimension == 2:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen',axis=(1,2))
else:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen')
self.loss_op = 1. - sorensen
elif (self.loss_name == "weighted_sorensen"):
if self.dimension == 2:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen', axis=(1,2), weighted=True)
else:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen', weighted=True)
self.loss_op = 1. - sorensen
elif (self.loss_name == "jaccard"):
if self.dimension == 2:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard',axis=(1,2))
else:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard')
self.loss_op = 1. - jaccard
elif (self.loss_name == "weighted_jaccard"):
if self.dimension == 2:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard',axis=(1,2), weighted=True)
else:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard', weighted=True)
self.loss_op = 1. - jaccard
elif (self.loss_name == "mixed_sorensen"):
xent = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels,logits=self.logits))
if self.dimension == 2:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen',axis=(1,2))
else:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen')
self.loss_op = (1. - sorensen) + xent
elif (self.loss_name == "mixed_weighted_sorensen"):
xent = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels,logits=self.logits))
if self.dimension == 2:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen', axis=(1,2), weighted=True)
else:
sorensen = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='sorensen', weighted=True)
self.loss_op = (1. - sorensen) + xent
elif (self.loss_name == "mixed_jaccard"):
xent = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels,logits=self.logits))
if self.dimension == 2:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard',axis=(1,2))
else:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard')
self.loss_op = (1. - jaccard) + xent
elif (self.loss_name == "mixed_weighted_jaccard"):
xent = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels,logits=self.logits))
if self.dimension == 2:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard',axis=(1,2), weighted=True)
else:
jaccard = dice_coe(softmax,tf.cast(labels,dtype=tf.float32), loss_type='jaccard', weighted=True)
self.loss_op = (1. - jaccard) + xent
else:
sys.exit("Invalid loss function")
tf.summary.scalar('loss', self.loss_op)
print("{}: Loss function complete".format(datetime.datetime.now()))
# argmax op
with tf.name_scope("predicted_label"):
self.pred_op = tf.argmax(self.logits, axis=-1 , name="prediction")
if self.image_log:
if self.dimension == 2:
if 0 in self.label_classes:
pred_log = tf.cast(self.pred_op*math.floor(255/(self.output_channel_num-1)),dtype=tf.uint8)
else:
pred_log = tf.cast(self.pred_op*math.floor(255/self.output_channel_num),dtype=tf.uint8)
pred_log = tf.expand_dims(pred_log,axis=-1)
tf.summary.image("pred", pred_log, max_outputs=self.batch_size)
else:
for batch in range(self.batch_size):
if 0 in self.label_classes:
pred_log = tf.cast(self.pred_op[batch:batch+1,:,:,:]*math.floor(255/(self.output_channel_num-1)), dtype=tf.uint8)
else:
pred_log = tf.cast(self.pred_op[batch:batch+1,:,:,:]*math.floor(255/(self.output_channel_num)), dtype=tf.uint8)
tf.summary.image("pred"+"_batch"+str(batch), tf.transpose(pred_log,[3,1,2,0]),max_outputs=self.patch_shape[-1])
# accuracy of the model
with tf.name_scope("metrics"):
correct_pred = tf.equal(tf.expand_dims(self.pred_op,-1), tf.cast(self.labels_placeholder,dtype=tf.int64))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
tf.summary.scalar('accuracy', accuracy)
# confusion matrix
if self.dimension == 2:
label_one_hot = tf.one_hot(self.labels_placeholder[:,:,:,0], depth=self.output_channel_num)
pred_one_hot = tf.one_hot(self.pred_op, depth=self.output_channel_num)
else:
label_one_hot = tf.one_hot(self.labels_placeholder[:,:,:,:,0],depth=self.output_channel_num)
pred_one_hot = tf.one_hot(self.pred_op[:,:,:,:], depth=self.output_channel_num)
for i in range(self.output_channel_num):
if i == 0:
continue
else:
if self.dimension == 2:
tp, tp_op = tf.metrics.true_positives(label_one_hot[:,:,:,i], pred_one_hot[:,:,:,i], name="true_positives_"+str(self.label_classes[i]))
tn, tn_op = tf.metrics.true_negatives(label_one_hot[:,:,:,i], pred_one_hot[:,:,:,i], name="true_negatives_"+str(self.label_classes[i]))
fp, fp_op = tf.metrics.false_positives(label_one_hot[:,:,:,i], pred_one_hot[:,:,:,i], name="false_positives_"+str(self.label_classes[i]))
fn, fn_op = tf.metrics.false_negatives(label_one_hot[:,:,:,i], pred_one_hot[:,:,:,i], name="false_negatives_"+str(self.label_classes[i]))
else:
tp, tp_op = tf.metrics.true_positives(label_one_hot[:,:,:,:,i], pred_one_hot[:,:,:,:,i], name="true_positives_"+str(self.label_classes[i]))
tn, tn_op = tf.metrics.true_negatives(label_one_hot[:,:,:,:,i], pred_one_hot[:,:,:,:,i], name="true_negatives_"+str(self.label_classes[i]))
fp, fp_op = tf.metrics.false_positives(label_one_hot[:,:,:,:,i], pred_one_hot[:,:,:,:,i], name="false_positives_"+str(self.label_classes[i]))
fn, fn_op = tf.metrics.false_negatives(label_one_hot[:,:,:,:,i], pred_one_hot[:,:,:,:,i], name="false_negatives_"+str(self.label_classes[i]))
sensitivity_op = tf.divide(tf.cast(tp_op,tf.float32),tf.cast(tf.add(tp_op,fn_op),tf.float32))
specificity_op = tf.divide(tf.cast(tn_op,tf.float32),tf.cast(tf.add(tn_op,fp_op),tf.float32))
dice_op = 2.*tp_op/(2.*tp_op+fp_op+fn_op)
tf.summary.scalar('sensitivity_'+str(self.label_classes[i]), sensitivity_op)
tf.summary.scalar('specificity_'+str(self.label_classes[i]), specificity_op)
tf.summary.scalar('dice_'+str(self.label_classes[i]), dice_op)
print("{}: Metrics complete".format(datetime.datetime.now()))
print("{}: Build graph complete".format(datetime.datetime.now()))
def train(self):
# read config to class variables
self.read_config()
"""Train image2label model"""
self.build_model_graph()
# learning rate
with tf.name_scope("learning_rate"):
learning_rate = tf.train.exponential_decay(self.initial_learning_rate, self.global_step_op,
self.decay_steps,self.decay_factor,staircase=False)
tf.summary.scalar('learning_rate', learning_rate)
# optimizer
with tf.name_scope("training"):
# optimizer
if self.optimizer_name == "SGD":
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
elif self.optimizer_name == "Adam":
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
elif self.optimizer_name == "Momentum":
optimizer = tf.train.MomentumOptimizer(learning_rate=learning_rate, momentum=self.momentum)
elif self.optimizer_name == "NesterovMomentum":
optimizer = tf.train.MomentumOptimizer(learning_rate=learning_rate, momentum=self.momentum, use_nesterov=True)
else:
sys.exit("Invalid optimizer");
train_op = optimizer.minimize(
loss = self.loss_op,
global_step=self.global_step_op)
# the update op is required by batch norm layer: https://www.tensorflow.org/api_docs/python/tf/layers/batch_normalization
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
train_op = tf.group([train_op, update_ops])
start_epoch = tf.get_variable("start_epoch", shape=[1], initializer=tf.zeros_initializer, dtype=tf.int32)
start_epoch_inc = start_epoch.assign(start_epoch+1)
# actual training cycle
# initialize all variables
self.sess.run(tf.initializers.global_variables())
print("{}: Start training...".format(datetime.datetime.now()))
# saver
print("{}: Setting up Saver...".format(datetime.datetime.now()))
if not self.restore_training:
# clear log directory
if os.path.exists(self.log_dir):
shutil.rmtree(self.log_dir)
os.makedirs(self.log_dir)
# clear checkpoint directory
if os.path.exists(self.ckpt_dir):
shutil.rmtree(self.ckpt_dir)
os.makedirs(self.ckpt_dir)
saver = tf.train.Saver(keep_checkpoint_every_n_hours=5)
checkpoint_prefix = os.path.join(self.ckpt_dir,"checkpoint")
else:
saver = tf.train.Saver(keep_checkpoint_every_n_hours=5)
checkpoint_prefix = os.path.join(self.ckpt_dir,"checkpoint")
# check if checkpoint exists
if os.path.exists(checkpoint_prefix+"-latest"):
print("{}: Last checkpoint found at {}, loading...".format(datetime.datetime.now(),self.ckpt_dir))
latest_checkpoint_path = tf.train.latest_checkpoint(self.ckpt_dir,latest_filename="checkpoint-latest")
saver.restore(self.sess, latest_checkpoint_path)
print("{}: Last checkpoint epoch: {}".format(datetime.datetime.now(),start_epoch.eval(session=self.sess)[0]))
print("{}: Last checkpoint global step: {}".format(datetime.datetime.now(),tf.train.global_step(self.sess, self.global_step_op)))
summary_op = tf.summary.merge_all()
# summary writer for tensorboard
train_summary_writer = tf.summary.FileWriter(self.log_dir + '/train', self.sess.graph)
if self.testing:
test_summary_writer = tf.summary.FileWriter(self.log_dir + '/test', self.sess.graph)
# testing initializer need to execute outside training loop
if self.testing:
self.sess.run(self.test_iterator.initializer)
# loop over epochs
for epoch in np.arange(start_epoch.eval(session=self.sess), self.epoches):
print("{}: Epoch {} starts...".format(datetime.datetime.now(),epoch+1))
# initialize iterator in each new epoch
self.sess.run(self.train_iterator.initializer)
# print("{}: Dataset iterator initialize ok".format(datetime.datetime.now()))
# training phase
loss_sum = 0
count = 0
while True:
if self.global_step_op.eval() > self.max_itr:
sys.exit("{}: Reach maximum iteration steps, training abort.".format(datetime.datetime.now()))
try:
self.sess.run(tf.local_variables_initializer())
# print("{}: Local variable initialize ok".format(datetime.datetime.now()))
self.network.is_training = True
print("{}: Set network to training ok".format(datetime.datetime.now()))
image, label = self.sess.run(self.next_element_train)
print("{}: Get next element train ok".format(datetime.datetime.now()))
if self.dimension == 2:
label = label[:,:,:,np.newaxis]
else:
label = label[:,:,:,:,np.newaxis]
train, summary, loss = self.sess.run([train_op,summary_op,self.loss_op], feed_dict={
self.images_placeholder: image,
self.labels_placeholder: label,
self.network.train_phase: True
})
print('{}: Segmentation training loss: {}'.format(datetime.datetime.now(), str(loss)))
loss_sum += loss
count += 1
train_summary_writer.add_summary(summary,global_step=tf.train.global_step(self.sess,self.global_step_op))
train_summary_writer.flush()
# save checkpoint
if self.global_step_op.eval()%self.log_interval == 0:
print("{}: Saving checkpoint of step {} at {}...".format(datetime.datetime.now(),self.global_step_op.eval(),self.ckpt_dir))
if not (os.path.exists(self.ckpt_dir)):
os.makedirs(self.ckpt_dir,exist_ok=True)
saver.save(self.sess, checkpoint_prefix,
global_step=tf.train.global_step(self.sess, self.global_step_op),
latest_filename="checkpoint-latest")
# testing phase
if self.testing and (self.global_step_op.eval()%self.test_step == 0):
self.sess.run(tf.local_variables_initializer())
print("{}: Set network to training ok".format(datetime.datetime.now()))
self.network.is_training = False
try:
image, label = self.sess.run(self.next_element_test)
except tf.errors.OutOfRangeError:
self.sess.run(self.test_iterator.initializer)
image, label = self.sess.run(self.next_element_test)
print("{}: Get next element test ok".format(datetime.datetime.now()))
if self.dimension == 2:
label = label[:,:,:,np.newaxis]
else:
label = label[:,:,:,:,np.newaxis]
summary, loss = self.sess.run([summary_op, self.loss_op],feed_dict={
self.images_placeholder: image,
self.labels_placeholder: label,
self.network.train_phase: False
})
print('{}: Segmentation testing loss: {}'.format(datetime.datetime.now(), str(loss)))
test_summary_writer.add_summary(summary, global_step=tf.train.global_step(self.sess, self.global_step_op))
test_summary_writer.flush()
except tf.errors.OutOfRangeError:
print("{}: Training of epoch {} complete, epoch loss: {}".format(datetime.datetime.now(),epoch+1,loss_sum/count))
start_epoch_inc.op.run()
self.network.is_training = False;
# print(start_epoch.eval())
# save the model at end of each epoch training
print("{}: Saving checkpoint of epoch {} at {}...".format(datetime.datetime.now(),epoch+1,self.ckpt_dir))
if not (os.path.exists(self.ckpt_dir)):
os.makedirs(self.ckpt_dir,exist_ok=True)
saver.save(self.sess, checkpoint_prefix,
global_step=tf.train.global_step(self.sess, self.global_step_op),
latest_filename="checkpoint-latest")
print("{}: Saving checkpoint succeed".format(datetime.datetime.now()))
break
# close tensorboard summary writer
train_summary_writer.close()
if self.testing:
test_summary_writer.close()
def evaluate_single_3D(self, sample, transforms):
input_origin = sample['image'][0].GetOrigin()
input_direction = sample['image'][0].GetDirection()
input_spacing = sample['image'][0].GetSpacing()
input_size = sample['image'][0].GetSize()
for transform in transforms:
sample = transform(sample)
images = sample['image']
label = sample['label']
if self.evaluate_probability_output:
# create empty softmax image in pair with transformed image
softmax_tfm = sitk.Image(images[0].GetSize(),sitk.sitkFloat32)
softmax_tfm.SetOrigin(images[0].GetOrigin())
softmax_tfm.SetDirection(images[0].GetDirection())
softmax_tfm.SetSpacing(images[0].GetSpacing())
# convert image to numpy array
for channel in range(self.input_channel_num):
image_ = sitk.GetArrayFromImage(images[channel])
if channel == 0:
images_np = image_[:,:,:,np.newaxis]
else:
images_np = np.append(images_np, image_[:,:,:,np.newaxis], axis=-1)
images_np = np.asarray(images_np,np.float32)
label_np = sitk.GetArrayFromImage(label)
label_np = np.asarray(label_np,np.int32)
if self.evaluate_probability_output:
softmax_np = sitk.GetArrayFromImage(softmax_tfm)
softmax_np = np.asarray(softmax_np,np.float32)
# unify numpy and sitk orientation
images_np = np.transpose(images_np,(2,1,0,3))
label_np = np.transpose(label_np,(2,1,0))
if self.evaluate_probability_output:
softmax_np = np.transpose(softmax_np,(2,1,0))
# a weighting matrix will be used for averaging the overlapped region
weight_np = np.zeros(label_np.shape)
# prepare image batch indices
inum = int(math.ceil((images_np.shape[0]-self.patch_shape[0])/float(self.evaluate_stride[0]))) + 1
jnum = int(math.ceil((images_np.shape[1]-self.patch_shape[1])/float(self.evaluate_stride[1]))) + 1
knum = int(math.ceil((images_np.shape[2]-self.patch_shape[2])/float(self.evaluate_stride[2]))) + 1
patch_total = 0
image_ijk_patch_indices_dicts = []
ijk_patch_indicies_tmp = []
for i in range(inum):
for j in range(jnum):
for k in range (knum):
if patch_total % self.evaluate_batch == 0:
ijk_patch_indicies_tmp = []
istart = i * self.evaluate_stride[0]
if istart + self.patch_shape[0] > images_np.shape[0]: #for last patch
istart = images_np.shape[0] - self.patch_shape[0]
iend = istart + self.patch_shape[0]
jstart = j * self.evaluate_stride[1]
if jstart + self.patch_shape[1] > images_np.shape[1]: #for last patch
jstart = images_np.shape[1] - self.patch_shape[1]
jend = jstart + self.patch_shape[1]
kstart = k * self.evaluate_stride[2]
if kstart + self.patch_shape[2] > images_np.shape[2]: #for last patch
kstart = images_np.shape[2] - self.patch_shape[2]
kend = kstart + self.patch_shape[2]
ijk_patch_indicies_tmp.append([istart, iend, jstart, jend, kstart, kend])
if patch_total % self.evaluate_batch == 0:
image_ijk_patch_indices_dicts.append({'images': images_np, 'indexes':ijk_patch_indicies_tmp})
patch_total += 1
# for last batch
image_ijk_patch_indices_dicts.append({'images': images_np, 'indexes':ijk_patch_indicies_tmp})
p = multiprocessing.Pool(multiprocessing.cpu_count())
batches = p.map(prepare_batch,image_ijk_patch_indices_dicts)
p.close()
p.join()
# acutal segmentation
for i in tqdm(range(len(batches))):
batch = batches[i]
if self.evaluate_probability_output:
[pred, softmax] = self.sess.run(['predicted_label/prediction:0','softmax:0'], feed_dict={
'images_placeholder:0': batch,
'train_phase_placeholder:0': False})
else:
pred = self.sess.run('predicted_label/prediction:0', feed_dict={
'images_placeholder:0': batch,
'train_phase_placeholder:0': False})
for j in range(pred.shape[0]):
istart = image_ijk_patch_indices_dicts[i]['indexes'][j][0]
iend = image_ijk_patch_indices_dicts[i]['indexes'][j][1]
jstart = image_ijk_patch_indices_dicts[i]['indexes'][j][2]
jend = image_ijk_patch_indices_dicts[i]['indexes'][j][3]
kstart = image_ijk_patch_indices_dicts[i]['indexes'][j][4]
kend = image_ijk_patch_indices_dicts[i]['indexes'][j][5]
label_np[istart:iend,jstart:jend,kstart:kend] += pred[j,:,:,:]
if self.evaluate_probability_output:
softmax_np[istart:iend,jstart:jend,kstart:kend] += softmax[j,:,:,:,1]
weight_np[istart:iend,jstart:jend,kstart:kend] += 1.0
print("{}: Evaluation complete".format(datetime.datetime.now()))
# eliminate overlapping region using the weighted value
label_np = np.rint(np.float32(label_np)/np.float32(weight_np) + 0.01)
if self.evaluate_probability_output:
softmax_np = softmax_np/np.float32(weight_np)
# convert back to sitk space
label_np = np.transpose(label_np,(2,1,0))
if self.evaluate_probability_output:
softmax_np = np.transpose(softmax_np,(2,1,0))
# convert label numpy back to sitk image
label_tfm = sitk.GetImageFromArray(label_np)
label_tfm.SetOrigin(images[0].GetOrigin())
label_tfm.SetDirection(images[0].GetDirection())
label_tfm.SetSpacing(images[0].GetSpacing())
if self.evaluate_probability_output:
softmax_tfm = sitk.GetImageFromArray(softmax_np)
softmax_tfm.SetOrigin(images[0].GetOrigin())
softmax_tfm.SetDirection(images[0].GetDirection())
softmax_tfm.SetSpacing(images[0].GetSpacing())
# resample the label back to original space
resampler = sitk.ResampleImageFilter()
resampler.SetInterpolator(sitk.sitkNearestNeighbor)
resampler.SetOutputSpacing(input_spacing)
resampler.SetSize(input_size)
resampler.SetOutputOrigin(input_origin)
resampler.SetOutputDirection(input_direction)
print("{}: Resampling label back to original image space...".format(datetime.datetime.now()))
label = resampler.Execute(label_tfm)
if not self.evaluate_probability_output:
return label
if self.evaluate_probability_output:
resampler.SetInterpolator(sitk.sitkLinear)
print("{}: Resampling probability map back to original image space...".format(datetime.datetime.now()))
prob = resampler.Execute(softmax_tfm)
return label, prob
def evaluate_single_2D(self,sample, transforms):
input_origin = sample['image'][0].GetOrigin()
input_direction = sample['image'][0].GetDirection()
input_spacing = sample['image'][0].GetSpacing()
input_size = sample['image'][0].GetSize()
for transform in transforms['3D']:
sample = transform(sample)
images = sample['image']
label = sample['label']
if self.evaluate_probability_output:
# create empty softmax image in pair with transformed image
prob = sitk.Image(images[0].GetSize(),sitk.sitkFloat32)
prob.SetOrigin(images[0].GetOrigin())
prob.SetDirection(images[0].GetDirection())
prob.SetSpacing(images[0].GetSpacing())
# loop over slices
for layer in tqdm(range(images[0].GetSize()[2])):
# print(str(layer),"/",str(images[0].GetSize()[2]))
extractor = sitk.ExtractImageFilter()
size = [images[0].GetSize()[0],images[0].GetSize()[1],0]
index = [0,0,layer]
extractor.SetSize(size)
extractor.SetIndex(index)
# extract a slice and convert to numpy array
images_slice = []
for channel in range(len(images)):
images_slice.append(extractor.Execute(images[channel]))
label_slice = extractor.Execute(label)
sample = {'image':images_slice, 'label':label_slice}
input_slice_spacing = label_slice.GetSpacing()
input_slice_direction = label_slice.GetDirection()
input_slice_size = label_slice.GetSize()
input_slice_origin = label_slice.GetOrigin()
for transform in transforms['2D']:
sample = transform(sample)
images_slice = sample['image']
label_slice = sample['label']
for channel in range(len(images_slice)):
image_ = sitk.GetArrayFromImage(images_slice[channel])
if channel == 0:
images_np = image_[:,:,np.newaxis]
else:
images_np = np.append(images_np, image_[:,:,np.newaxis],axis=-1)
images_np = np.asarray(images_np, np.float32)
label_np = sitk.GetArrayFromImage(label_slice)
label_np = np.asarray(label_np, np.int32)
if self.evaluate_probability_output:
softmax_np = np.zeros(label_np.shape)
softmax_np = np.asarray(softmax_np,np.float32)
# a weighting matrix will be used for averaging the overlapped region
weight_np = np.zeros(label_np.shape)
# prepare image batch indices
inum = int(math.ceil((images_np.shape[0]-self.patch_shape[0])/float(self.evaluate_stride[0]))) + 1
jnum = int(math.ceil((images_np.shape[1]-self.patch_shape[1])/float(self.evaluate_stride[1]))) + 1
patch_total = 0
image_ij_patch_indices_dicts = []
ij_patch_indicies_tmp = []
for i in range(inum):
for j in range(jnum):
if patch_total % self.evaluate_batch == 0:
ij_patch_indicies_tmp = []
istart = i * self.evaluate_stride[0]
if istart + self.patch_shape[0] > images_np.shape[0]: #for last patch
istart = images_np.shape[0] - self.patch_shape[0]
iend = istart + self.patch_shape[0]
jstart = j * self.evaluate_stride[1]
if jstart + self.patch_shape[1] > images_np.shape[1]: #for last patch
jstart = images_np.shape[1] - self.patch_shape[1]
jend = jstart + self.patch_shape[1]
image_patch = images_np[istart:iend,jstart:jend,:]
image_batch = image_patch[np.newaxis,:,:,:]
[pred, softmax] = self.sess.run(['predicted_label/prediction:0','softmax:0'], feed_dict={
'images_placeholder:0': image_batch,
'train_phase_placeholder:0': False})
label_np[istart:iend, jstart:jend] += pred[0,:,:]
if self.evaluate_probability_output:
softmax_np[istart:iend,jstart:jend] += softmax[0,:,:,1]
weight_np[istart:iend,jstart:jend] += 1.0
# eliminate overlapping region using the weighted value
label_np = np.rint(np.float32(label_np)/np.float32(weight_np) + 0.01)
if self.evaluate_probability_output:
softmax_np = softmax_np/np.float32(weight_np)
# convert label numpy back to sitk image
label_slice = sitk.GetImageFromArray(label_np)
label_slice.SetOrigin(images_slice[0].GetOrigin())
label_slice.SetDirection(images_slice[0].GetDirection())
label_slice.SetSpacing(images_slice[0].GetSpacing())
# resample the label back to original space
resampler = sitk.ResampleImageFilter()
resampler.SetInterpolator(sitk.sitkNearestNeighbor)
resampler.SetOutputSpacing(input_slice_spacing)
resampler.SetSize(input_slice_size)
resampler.SetOutputOrigin(input_slice_origin)
resampler.SetOutputDirection(input_slice_direction)
label_slice = resampler.Execute(label_slice)
label_slice = sitk.JoinSeries(label_slice)
castFilter = sitk.CastImageFilter()
castFilter.SetOutputPixelType(sitk.sitkUInt8)
label_slice = castFilter.Execute(label_slice)
label = sitk.Paste(label,label_slice, label_slice.GetSize(), destinationIndex=[0,0,layer])
if self.evaluate_probability_output:
prob_slice = sitk.GetImageFromArray(softmax_np)
prob_slice.SetOrigin(images_slice[0].GetOrigin())
prob_slice.SetDirection(images_slice[0].GetDirection())