1
|
|
|
import abc |
2
|
|
|
from pathlib import Path |
3
|
|
|
from typing import Optional |
4
|
|
|
|
5
|
|
|
from ..typing import TypePath |
6
|
|
|
from ..transforms import Transform |
7
|
|
|
from ..download import download_and_extract_archive |
8
|
|
|
from .. import SubjectsDataset, Subject, ScalarImage, LabelMap |
9
|
|
|
|
10
|
|
|
|
11
|
|
|
class BITE(SubjectsDataset, abc.ABC): |
12
|
|
|
base_url = 'http://www.bic.mni.mcgill.ca/uploads/Services/' |
13
|
|
|
|
14
|
|
|
def __init__( |
15
|
|
|
self, |
16
|
|
|
root: TypePath, |
17
|
|
|
transform: Optional[Transform] = None, |
18
|
|
|
download: bool = False, |
19
|
|
|
**kwargs, |
20
|
|
|
): |
21
|
|
|
root = Path(root).expanduser().absolute() |
22
|
|
|
if download: |
23
|
|
|
self._download(root) |
24
|
|
|
subjects_list = self._get_subjects_list(root) |
25
|
|
|
self.kwargs = kwargs |
26
|
|
|
super().__init__(subjects_list, transform=transform, **kwargs) |
27
|
|
|
|
28
|
|
|
def _download(self, root: Path): |
29
|
|
|
raise NotImplementedError |
30
|
|
|
|
31
|
|
|
def _get_subjects_list(self, root: Path): |
32
|
|
|
raise NotImplementedError |
33
|
|
|
|
34
|
|
|
|
35
|
|
|
class BITE3(BITE): |
36
|
|
|
dirname = 'group3' |
37
|
|
|
"""Pre- and post-resection MR images in BITE. |
38
|
|
|
|
39
|
|
|
*The goal of BITE is to share in vivo medical images of patients wtith |
40
|
|
|
brain tumors to facilitate the development and validation of new image |
41
|
|
|
processing algorithms.* |
42
|
|
|
|
43
|
|
|
Please check the `BITE website`_ for more information and |
44
|
|
|
acknowledgments instructions. |
45
|
|
|
|
46
|
|
|
.. _BITE website: http://nist.mni.mcgill.ca/?page_id=672 |
47
|
|
|
|
48
|
|
|
Args: |
49
|
|
|
root: Root directory to which the dataset will be downloaded. |
50
|
|
|
transform: An instance of |
51
|
|
|
:class:`~torchio.transforms.transform.Transform`. |
52
|
|
|
download: If set to ``True``, will download the data into :attr:`root`. |
53
|
|
|
""" |
54
|
|
|
def _download(self, root: Path): |
55
|
|
|
if (root / self.dirname).is_dir(): |
56
|
|
|
return |
57
|
|
|
root.mkdir(exist_ok=True, parents=True) |
58
|
|
|
filename = f'{self.dirname}.tar.gz' |
59
|
|
|
url = self.base_url + filename |
60
|
|
|
download_and_extract_archive( |
61
|
|
|
url, |
62
|
|
|
download_root=root, |
63
|
|
|
md5='e415b63887c40b727c45552614b44634', |
64
|
|
|
) |
65
|
|
|
(root / filename).unlink() # cleanup |
66
|
|
|
|
67
|
|
|
def _get_subjects_list(self, root: Path): |
68
|
|
|
subjects_dir = root / self.dirname |
69
|
|
|
subjects = [] |
70
|
|
|
for i in range(1, 15): |
71
|
|
|
if i == 13: |
72
|
|
|
continue # no MRI for this subject |
73
|
|
|
subject_id = f'{i:02d}' |
74
|
|
|
subject_dir = subjects_dir / subject_id |
75
|
|
|
preop_path = subject_dir / f'{subject_id}_preop_mri.mnc' |
76
|
|
|
postop_path = subject_dir / f'{subject_id}_postop_mri.mnc' |
77
|
|
|
images_dict = {} |
78
|
|
|
images_dict['preop'] = ScalarImage(preop_path) |
79
|
|
|
images_dict['postop'] = ScalarImage(postop_path) |
80
|
|
|
for fp in subject_dir.glob('*tumor*'): |
81
|
|
|
images_dict[fp.stem[3:]] = LabelMap(fp) |
82
|
|
|
subject = Subject(images_dict) |
83
|
|
|
subjects.append(subject) |
84
|
|
|
return subjects |
85
|
|
|
|