|
1
|
|
|
#!/usr/bin/env python |
|
2
|
|
|
|
|
3
|
|
|
"""Tests for `utils` package.""" |
|
4
|
|
|
|
|
5
|
|
|
import unittest |
|
6
|
|
|
import torch |
|
7
|
|
|
import numpy as np |
|
8
|
|
|
import SimpleITK as sitk |
|
9
|
|
|
from torchio import LABEL, INTENSITY, RandomFlip |
|
10
|
|
|
from torchio.utils import ( |
|
11
|
|
|
to_tuple, |
|
12
|
|
|
get_stem, |
|
13
|
|
|
guess_type, |
|
14
|
|
|
nib_to_sitk, |
|
15
|
|
|
sitk_to_nib, |
|
16
|
|
|
apply_transform_to_file, |
|
17
|
|
|
) |
|
18
|
|
|
from .utils import TorchioTestCase |
|
19
|
|
|
|
|
20
|
|
|
|
|
21
|
|
|
class TestUtils(TorchioTestCase): |
|
22
|
|
|
"""Tests for `utils` module.""" |
|
23
|
|
|
|
|
24
|
|
|
def test_to_tuple(self): |
|
25
|
|
|
assert to_tuple(1) == (1,) |
|
26
|
|
|
assert to_tuple((1,)) == (1,) |
|
27
|
|
|
assert to_tuple(1, length=3) == (1, 1, 1) |
|
28
|
|
|
assert to_tuple((1, 2)) == (1, 2) |
|
29
|
|
|
assert to_tuple((1, 2), length=3) == (1, 2) |
|
30
|
|
|
assert to_tuple([1, 2], length=3) == (1, 2) |
|
31
|
|
|
|
|
32
|
|
|
def test_get_stem(self): |
|
33
|
|
|
assert get_stem('/home/image.nii.gz') == 'image' |
|
34
|
|
|
assert get_stem('/home/image.nii') == 'image' |
|
35
|
|
|
assert get_stem('/home/image.nrrd') == 'image' |
|
36
|
|
|
|
|
37
|
|
|
def test_guess_type(self): |
|
38
|
|
|
assert guess_type('None') is None |
|
39
|
|
|
assert isinstance(guess_type('1'), int) |
|
40
|
|
|
assert isinstance(guess_type('1.5'), float) |
|
41
|
|
|
assert isinstance(guess_type('(1, 3, 5)'), tuple) |
|
42
|
|
|
assert isinstance(guess_type('(1,3,5)'), tuple) |
|
43
|
|
|
assert isinstance(guess_type('[1,3,5]'), list) |
|
44
|
|
|
assert isinstance(guess_type('test'), str) |
|
45
|
|
|
|
|
46
|
|
|
def test_check_consistent_shape(self): |
|
47
|
|
|
good_sample = self.sample |
|
48
|
|
|
bad_sample = self.get_inconsistent_sample() |
|
49
|
|
|
good_sample.check_consistent_shape() |
|
50
|
|
|
with self.assertRaises(ValueError): |
|
51
|
|
|
bad_sample.check_consistent_shape() |
|
52
|
|
|
|
|
53
|
|
|
def test_apply_transform_to_file(self): |
|
54
|
|
|
transform = RandomFlip() |
|
55
|
|
|
apply_transform_to_file( |
|
56
|
|
|
self.get_image_path('input'), |
|
57
|
|
|
transform, |
|
58
|
|
|
self.get_image_path('output'), |
|
59
|
|
|
verbose=True, |
|
60
|
|
|
) |
|
61
|
|
|
|
|
62
|
|
|
def test_sitk_to_nib(self): |
|
63
|
|
|
data = np.random.rand(10, 10) |
|
64
|
|
|
image = sitk.GetImageFromArray(data) |
|
65
|
|
|
tensor, affine = sitk_to_nib(image) |
|
66
|
|
|
self.assertAlmostEqual(data.sum(), tensor.sum()) |
|
67
|
|
|
|
|
68
|
|
|
|
|
69
|
|
|
class TestNibabelToSimpleITK(TorchioTestCase): |
|
70
|
|
|
def setUp(self): |
|
71
|
|
|
super().setUp() |
|
72
|
|
|
self.affine = np.eye(4) |
|
73
|
|
|
|
|
74
|
|
|
def test_wrong_dims(self): |
|
75
|
|
|
with self.assertRaises(ValueError): |
|
76
|
|
|
nib_to_sitk(np.random.rand(10, 10), self.affine) |
|
77
|
|
|
|
|
78
|
|
|
def test_2d_single(self): |
|
79
|
|
|
data = np.random.rand(1, 1, 10, 12) |
|
80
|
|
|
image = nib_to_sitk(data, self.affine) |
|
81
|
|
|
assert image.GetDimension() == 2 |
|
82
|
|
|
assert image.GetSize() == (10, 12) |
|
83
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 1 |
|
84
|
|
|
|
|
85
|
|
|
def test_2d_multi(self): |
|
86
|
|
|
data = np.random.rand(5, 1, 10, 12) |
|
87
|
|
|
image = nib_to_sitk(data, self.affine) |
|
88
|
|
|
assert image.GetDimension() == 2 |
|
89
|
|
|
assert image.GetSize() == (10, 12) |
|
90
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 5 |
|
91
|
|
|
|
|
92
|
|
|
def test_2d_3d_single(self): |
|
93
|
|
|
data = np.random.rand(1, 1, 10, 12) |
|
94
|
|
|
image = nib_to_sitk(data, self.affine, force_3d=True) |
|
95
|
|
|
assert image.GetDimension() == 3 |
|
96
|
|
|
assert image.GetSize() == (1, 10, 12) |
|
97
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 1 |
|
98
|
|
|
|
|
99
|
|
|
def test_2d_3d_multi(self): |
|
100
|
|
|
data = np.random.rand(5, 1, 10, 12) |
|
101
|
|
|
image = nib_to_sitk(data, self.affine, force_3d=True) |
|
102
|
|
|
assert image.GetDimension() == 3 |
|
103
|
|
|
assert image.GetSize() == (1, 10, 12) |
|
104
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 5 |
|
105
|
|
|
|
|
106
|
|
|
def test_3d_single(self): |
|
107
|
|
|
data = np.random.rand(1, 8, 10, 12) |
|
108
|
|
|
image = nib_to_sitk(data, self.affine) |
|
109
|
|
|
assert image.GetDimension() == 3 |
|
110
|
|
|
assert image.GetSize() == (8, 10, 12) |
|
111
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 1 |
|
112
|
|
|
|
|
113
|
|
|
def test_3d_multi(self): |
|
114
|
|
|
data = np.random.rand(5, 8, 10, 12) |
|
115
|
|
|
image = nib_to_sitk(data, self.affine) |
|
116
|
|
|
assert image.GetDimension() == 3 |
|
117
|
|
|
assert image.GetSize() == (8, 10, 12) |
|
118
|
|
|
assert image.GetNumberOfComponentsPerPixel() == 5 |
|
119
|
|
|
|
fepegar /
torchio
