|
1
|
|
|
import copy |
|
2
|
|
|
import shutil |
|
3
|
|
|
import random |
|
4
|
|
|
import tempfile |
|
5
|
|
|
import unittest |
|
6
|
|
|
from pathlib import Path |
|
7
|
|
|
import torch |
|
8
|
|
|
import numpy as np |
|
9
|
|
|
import nibabel as nib |
|
10
|
|
|
from torchio.datasets import IXITiny |
|
11
|
|
|
from torchio import INTENSITY, LABEL, DATA, Image, ImagesDataset, Subject |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
class TorchioTestCase(unittest.TestCase): |
|
15
|
|
|
|
|
16
|
|
|
def setUp(self): |
|
17
|
|
|
"""Set up test fixtures, if any.""" |
|
18
|
|
|
self.dir = Path(tempfile.gettempdir()) / '.torchio_tests' |
|
19
|
|
|
self.dir.mkdir(exist_ok=True) |
|
20
|
|
|
random.seed(42) |
|
21
|
|
|
np.random.seed(42) |
|
22
|
|
|
|
|
23
|
|
|
registration_matrix = np.array([ |
|
24
|
|
|
[1, 0, 0, 10], |
|
25
|
|
|
[0, 1, 0, 0], |
|
26
|
|
|
[0, 0, 1.2, 0], |
|
27
|
|
|
[0, 0, 0, 1] |
|
28
|
|
|
]) |
|
29
|
|
|
|
|
30
|
|
|
subject_a = Subject( |
|
31
|
|
|
t1=Image(self.get_image_path('t1_a'), INTENSITY), |
|
32
|
|
|
) |
|
33
|
|
|
subject_b = Subject( |
|
34
|
|
|
t1=Image(self.get_image_path('t1_b'), INTENSITY), |
|
35
|
|
|
label=Image(self.get_image_path('label_b', binary=True), LABEL), |
|
36
|
|
|
) |
|
37
|
|
|
subject_c = Subject( |
|
38
|
|
|
label=Image(self.get_image_path('label_c', binary=True), LABEL), |
|
39
|
|
|
) |
|
40
|
|
|
subject_d = Subject( |
|
41
|
|
|
t1=Image( |
|
42
|
|
|
self.get_image_path('t1_d'), |
|
43
|
|
|
INTENSITY, |
|
44
|
|
|
pre_affine=registration_matrix, |
|
45
|
|
|
), |
|
46
|
|
|
t2=Image(self.get_image_path('t2_d'), INTENSITY), |
|
47
|
|
|
label=Image(self.get_image_path('label_d', binary=True), LABEL), |
|
48
|
|
|
) |
|
49
|
|
|
self.subjects_list = [ |
|
50
|
|
|
subject_a, |
|
51
|
|
|
subject_b, |
|
52
|
|
|
subject_c, |
|
53
|
|
|
subject_d, |
|
54
|
|
|
] |
|
55
|
|
|
self.dataset = ImagesDataset(self.subjects_list) |
|
56
|
|
|
self.sample = self.dataset[-1] |
|
57
|
|
|
|
|
58
|
|
|
def tearDown(self): |
|
59
|
|
|
"""Tear down test fixtures, if any.""" |
|
60
|
|
|
print('Deleting', self.dir) |
|
61
|
|
|
shutil.rmtree(self.dir) |
|
62
|
|
|
|
|
63
|
|
|
def assertTensorEqual(self, a, b): |
|
64
|
|
|
assert torch.all(torch.eq(a, b)) |
|
65
|
|
|
|
|
66
|
|
|
def make_2d(self, sample): |
|
67
|
|
|
sample = copy.deepcopy(sample) |
|
68
|
|
|
for image in sample.get_images(intensity_only=False): |
|
69
|
|
|
image[DATA] = image[DATA][:, 0:1, ...] |
|
|
|
|
|
|
70
|
|
|
return sample |
|
71
|
|
|
|
|
72
|
|
|
def get_inconsistent_sample(self): |
|
73
|
|
|
"""Return a sample containing images of different shape.""" |
|
74
|
|
|
subject = Subject( |
|
75
|
|
|
t1=Image(self.get_image_path('t1_inc'), INTENSITY), |
|
76
|
|
|
t2=Image( |
|
77
|
|
|
self.get_image_path('t2_inc', shape=(10, 20, 31)), INTENSITY), |
|
78
|
|
|
label=Image( |
|
79
|
|
|
self.get_image_path( |
|
80
|
|
|
'label_inc', |
|
81
|
|
|
shape=(8, 17, 25), |
|
82
|
|
|
binary=True, |
|
83
|
|
|
), |
|
84
|
|
|
LABEL, |
|
85
|
|
|
), |
|
86
|
|
|
) |
|
87
|
|
|
subjects_list = [subject] |
|
88
|
|
|
dataset = ImagesDataset(subjects_list) |
|
89
|
|
|
return dataset[0] |
|
90
|
|
|
|
|
91
|
|
|
def get_reference_image_and_path(self): |
|
92
|
|
|
"""Return a reference image and its path""" |
|
93
|
|
|
path = self.get_image_path('ref', shape=(10, 20, 31), spacing=(1, 1, 2)) |
|
94
|
|
|
image = Image(path, INTENSITY) |
|
95
|
|
|
return image, path |
|
96
|
|
|
|
|
97
|
|
|
def get_ixi_tiny(self): |
|
98
|
|
|
root_dir = Path(tempfile.gettempdir()) / 'torchio' / 'ixi_tiny' |
|
99
|
|
|
return IXITiny(root_dir, download=True) |
|
100
|
|
|
|
|
101
|
|
|
def get_image_path( |
|
102
|
|
|
self, |
|
103
|
|
|
stem, |
|
104
|
|
|
binary=False, |
|
105
|
|
|
shape=(10, 20, 30), |
|
106
|
|
|
spacing=(1, 1, 1), |
|
107
|
|
|
): |
|
108
|
|
|
data = np.random.rand(*shape) |
|
109
|
|
|
if binary: |
|
110
|
|
|
data = (data > 0.5).astype(np.uint8) |
|
111
|
|
|
affine = np.diag((*spacing, 1)) |
|
112
|
|
|
suffix = random.choice(('.nii.gz', '.nii', '.nrrd', '.minc', '.img')) |
|
113
|
|
|
path = self.dir / f'{stem}{suffix}' |
|
114
|
|
|
if np.random.rand() > 0.5: |
|
115
|
|
|
path = str(path) |
|
116
|
|
|
image = Image(tensor=data, affine=affine) |
|
117
|
|
|
image.save(path) |
|
118
|
|
|
return path |
|
119
|
|
|
|
fepegar /
torchio
Loading history...