-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_on_epoch_end.py
177 lines (133 loc) · 8.5 KB
/
test_on_epoch_end.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
from __future__ import print_function
import os
import shutil
import numpy as np
import SimpleITK as sitk
from cropping import corrected_crop
def test(model, img_path, label_path, subject_list, patch_size, batch_size=1, numClasses=2, epoch=0, current_best=0.):
# the following parameters need to be assigned values before training
patch_size = np.array(patch_size) # very important
num_of_downpooling = 4 # very important
patch_stride_regulator = np.array([2, 2, 2]) # this value determines the stride in each dimension when getting the patch; if value = 2, the stride in that dimension is half the value of patch_size
assert np.all(np.mod(patch_size, 2**num_of_downpooling)) == 0
stride = patch_size/patch_stride_regulator
tmp_path = os.path.join(os.getcwd(), 'tmp')
best_path = os.path.join(os.getcwd(), 'best_results')
dice_coef_list = []
for j in range(len(subject_list)):
subject_index = subject_list[j]
### compute basic parameter for usage later
image = sitk.ReadImage(img_path + 'image{:04d}.nii.gz'.format(subject_index))
image = sitk.GetArrayFromImage(image)
image = ( image - np.min(image) ) / ( np.max(image) - np.min(image) )
image_size = np.array(np.shape(image))
expanded_image_size = (np.ceil(image_size/(1.0*stride))*stride).astype(np.int32)
expanded_image = np.zeros(expanded_image_size, dtype=np.float32)
expanded_image[0:image_size[0], 0:image_size[1], 0:image_size[2]] = image
predicted_seg = np.zeros([numClasses, expanded_image_size[0], expanded_image_size[1], expanded_image_size[2]], dtype=np.float32)
count_matrix_seg = np.zeros([numClasses, expanded_image_size[0], expanded_image_size[1], expanded_image_size[2]], dtype=np.float32)
num_of_patch_with_overlapping = expanded_image_size/stride - patch_stride_regulator + 1
total_num_of_patches = np.prod(num_of_patch_with_overlapping)
num_patch_z = int(num_of_patch_with_overlapping[0]) # used for get patches
num_patch_y = int(num_of_patch_with_overlapping[1]) # used for get patches
num_patch_x = int(num_of_patch_with_overlapping[2]) # used for get patches
center = np.zeros([total_num_of_patches.astype(np.int32), 3])
patch_index = 0
for ii in range(0, num_patch_z):
for jj in range(0, num_patch_y):
for kk in range(0, num_patch_x):
center[patch_index] = np.array([int(ii*stride[0] + patch_size[0]//2),
int(jj*stride[1] + patch_size[1]//2),
int(kk*stride[2] + patch_size[2]//2)])
patch_index += 1
modulo=np.mod(total_num_of_patches.astype(np.int32), batch_size)
if modulo!=0:
num_to_add=batch_size-modulo
inds_to_add=np.random.randint(0, int(total_num_of_patches.astype(np.int32)), int(num_to_add)) ## the return value is a ndarray
to_add = center[inds_to_add]
new_center = np.vstack((center, to_add))
else:
new_center = center
np.random.shuffle(new_center)
for i_batch in range(int(new_center.shape[0]/batch_size)):
label = 0
subvertex = new_center[i_batch*batch_size:(i_batch+1)*batch_size]
for count in range(batch_size):
## 96*96*96 ##
image_one = np.zeros([int(patch_size[0]), int(patch_size[1]), int(patch_size[2])], dtype=np.float32)
z_lower_bound = int(subvertex[count][0] - patch_size[0]//2)
z_higher_bound = int(subvertex[count][0] + patch_size[0]//2)
y_lower_bound = int(subvertex[count][1] - patch_size[1]//2)
y_higher_bound = int(subvertex[count][1] + patch_size[1]//2)
x_lower_bound = int(subvertex[count][2] - patch_size[2]//2)
x_higher_bound = int(subvertex[count][2] + patch_size[2]//2)
virgin_range = np.array([z_lower_bound, z_higher_bound, y_lower_bound, y_higher_bound, x_lower_bound, x_higher_bound])
copy_from, copy_to = corrected_crop(virgin_range, expanded_image_size)
cf_z_lower_bound = int(copy_from[0])
if copy_from[1] is not None:
cf_z_higher_bound = int(copy_from[1])
else:
cf_z_higher_bound = None
cf_y_lower_bound = int(copy_from[2])
if copy_from[3] is not None:
cf_y_higher_bound = int(copy_from[3])
else:
cf_y_higher_bound = None
cf_x_lower_bound = int(copy_from[4])
if copy_from[5] is not None:
cf_x_higher_bound = int(copy_from[5])
else:
cf_x_higher_bound = None
image_one[int(copy_to[0]):copy_to[1],
int(copy_to[2]):copy_to[3],
int(copy_to[4]):copy_to[5]] = \
expanded_image[cf_z_lower_bound:cf_z_higher_bound,
cf_y_lower_bound:cf_y_higher_bound,
cf_x_lower_bound:cf_x_higher_bound]
image_one = np.expand_dims(image_one, axis=0)
image_1 = np.expand_dims(image_one, axis=0)
if label == 0:
Img_1 = image_1
label += 1
else:
Img_1 = np.vstack((Img_1, image_1))
label += 1
predicted_one, _ = model(Img_1)
for idx in range(batch_size):
predicted_seg[:,
int(subvertex[idx][0] - patch_size[0]//2):int(subvertex[idx][0] + patch_size[0]//2),
int(subvertex[idx][1] - patch_size[1]//2):int(subvertex[idx][1] + patch_size[1]//2),
int(subvertex[idx][2] - patch_size[2]//2):int(subvertex[idx][2] + patch_size[2]//2)] += predicted_one[idx]
count_matrix_seg[:,
int(subvertex[idx][0] - patch_size[0]//2):int(subvertex[idx][0] + patch_size[0]//2),
int(subvertex[idx][1] - patch_size[1]//2):int(subvertex[idx][1] + patch_size[1]//2),
int(subvertex[idx][2] - patch_size[2]//2):int(subvertex[idx][2] + patch_size[2]//2)] += 1.0
predicted_seg_ = predicted_seg/(1.0*count_matrix_seg)
output_seg = predicted_seg_[:, 0:image_size[0], 0:image_size[1], 0:image_size[2]]
output_label = np.argmax(output_seg, axis=0)
label1 = np.zeros_like(image, dtype=np.float32)
label1[np.where(output_label == 1)] = 1.
groundtruth = sitk.ReadImage(label_path + 'label{:04d}.nii.gz'.format(subject_index))
groundtruth = sitk.GetArrayFromImage(groundtruth)
groundtruth1 = np.zeros_like(groundtruth, dtype=np.float32)
groundtruth1[np.where(groundtruth == 1)] = 1.
intersection = np.logical_and(label1, groundtruth1)
im_sum = label1.sum() + groundtruth1.sum()
dice_coef_list.append(2. * intersection.sum() / im_sum)
output_image_to_save = sitk.GetImageFromArray(output_label.astype(np.float32))
if not os.path.exists(os.path.join(tmp_path, 'subject{0}'.format(subject_index))):
os.makedirs(os.path.join(tmp_path, 'subject{0}'.format(subject_index)))
sitk.WriteImage(output_image_to_save, os.path.join(tmp_path, 'subject{0}/pred{0:04d}.nii.gz'.format(subject_index)))
print('dice_coef_list: ', dice_coef_list, '\nmean: ', np.mean(dice_coef_list), ' std: ', np.std(dice_coef_list))
current_mean = np.mean(dice_coef_list)
if current_mean > current_best:
if os.path.exists(best_path):
shutil.rmtree(best_path)
shutil.copytree(tmp_path, best_path)
model.save_weights('./checkpoints/supervised_model_best.ckpt')
model.save_weights('./checkpoints/supervised_model_epoch{0}_avg{1:5f}.ckpt'.format(epoch+1, current_mean))
return current_mean
else:
return current_best