Chris Oswald
commited on
Commit
·
81b3fb4
1
Parent(s):
76cb606
cleaned up comments
Browse files
SPIDER.py
CHANGED
@@ -11,20 +11,17 @@
|
|
11 |
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
# See the License for the specific language governing permissions and
|
13 |
# limitations under the License.
|
14 |
-
|
15 |
-
"""TODO: Add a description here."""
|
16 |
|
17 |
# Import packages
|
18 |
import csv
|
19 |
-
import json
|
20 |
import os
|
21 |
-
from typing import Dict, List, Mapping, Optional,
|
22 |
|
23 |
import numpy as np
|
24 |
import pandas as pd
|
25 |
|
26 |
import datasets
|
27 |
-
import PIL
|
28 |
import skimage
|
29 |
import SimpleITK as sitk
|
30 |
|
@@ -59,14 +56,17 @@ MAX_IVD = 9
|
|
59 |
DEFAULT_SCAN_TYPES = ['t1', 't2', 't2_SPACE']
|
60 |
DEFAULT_RESIZE = (512, 512, 30)
|
61 |
|
62 |
-
# TODO: Add BibTeX citation
|
63 |
-
# Find for instance the citation on arxiv or on the dataset repo/website
|
64 |
_CITATION = """\
|
65 |
-
@
|
66 |
-
title
|
67 |
-
author={
|
68 |
-
|
69 |
-
|
|
|
|
|
|
|
|
|
|
|
70 |
}
|
71 |
"""
|
72 |
|
@@ -81,9 +81,6 @@ _HOMEPAGE = "https://zenodo.org/records/10159290"
|
|
81 |
_LICENSE = """Creative Commons Attribution 4.0 International License \
|
82 |
(https://creativecommons.org/licenses/by/4.0/legalcode)"""
|
83 |
|
84 |
-
# TODO: Add link to the official dataset URLs here
|
85 |
-
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
86 |
-
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
87 |
_URLS = {
|
88 |
"images":"https://zenodo.org/records/10159290/files/images.zip",
|
89 |
"masks":"https://zenodo.org/records/10159290/files/masks.zip",
|
@@ -111,45 +108,11 @@ class CustomBuilderConfig(datasets.BuilderConfig):
|
|
111 |
class SPIDER(datasets.GeneratorBasedBuilder):
|
112 |
"""TODO: Short description of my dataset."""
|
113 |
|
|
|
114 |
DEFAULT_WRITER_BATCH_SIZE = 16 # PyArrow default is too large for image data
|
115 |
-
|
116 |
VERSION = datasets.Version("1.1.0")
|
117 |
-
|
118 |
BUILDER_CONFIG_CLASS = CustomBuilderConfig
|
119 |
|
120 |
-
# BUILDER_CONFIGS = [
|
121 |
-
# CustomBuilderConfig(
|
122 |
-
# name="all_scan_types",
|
123 |
-
# version=VERSION,
|
124 |
-
# description="Use images of all scan types (t1, t2, t2 SPACE)",
|
125 |
-
# scan_types=['t1', 't2', 't2_SPACE'],
|
126 |
-
# resize_shape=DEFAULT_RESIZE,
|
127 |
-
# ),
|
128 |
-
# CustomBuilderConfig(
|
129 |
-
# name="t1_scan_types",
|
130 |
-
# version=VERSION,
|
131 |
-
# description="Use images of t1 scan types only",
|
132 |
-
# scan_types=['t1'],
|
133 |
-
# resize_shape=DEFAULT_RESIZE,
|
134 |
-
# ),
|
135 |
-
# CustomBuilderConfig(
|
136 |
-
# name="t2_scan_types",
|
137 |
-
# version=VERSION,
|
138 |
-
# description="Use images of t2 scan types only",
|
139 |
-
# scan_types=['t2'],
|
140 |
-
# resize_shape=DEFAULT_RESIZE,
|
141 |
-
# ),
|
142 |
-
# CustomBuilderConfig(
|
143 |
-
# name="t2_SPACE_scan_types",
|
144 |
-
# version=VERSION,
|
145 |
-
# description="Use images of t2 SPACE scan types only",
|
146 |
-
# scan_types=['t2_SPACE'],
|
147 |
-
# resize_shape=DEFAULT_RESIZE,
|
148 |
-
# ),
|
149 |
-
# ]
|
150 |
-
|
151 |
-
# DEFAULT_CONFIG_NAME = "all_scan_types"
|
152 |
-
|
153 |
def __init__(
|
154 |
self,
|
155 |
*args,
|
@@ -162,10 +125,9 @@ class SPIDER(datasets.GeneratorBasedBuilder):
|
|
162 |
self.resize_shape = resize_shape
|
163 |
|
164 |
def _info(self):
|
165 |
-
"""
|
166 |
-
|
167 |
-
|
168 |
-
"""
|
169 |
image_size = self.config.resize_shape
|
170 |
features = datasets.Features({
|
171 |
"patient_id": datasets.Value("string"),
|
@@ -227,32 +189,16 @@ class SPIDER(datasets.GeneratorBasedBuilder):
|
|
227 |
})
|
228 |
|
229 |
return datasets.DatasetInfo(
|
230 |
-
# This is the description that will appear on the datasets page.
|
231 |
description=_DESCRIPTION,
|
232 |
-
|
233 |
-
features=features, # Here we define them above because they are different between the two configurations
|
234 |
-
# If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
|
235 |
-
# specify them. They'll be used if as_supervised=True in builder.as_dataset.
|
236 |
-
# supervised_keys=("sentence", "label"),
|
237 |
-
# Homepage of the dataset for documentation
|
238 |
homepage=_HOMEPAGE,
|
239 |
-
# License for the dataset if available
|
240 |
license=_LICENSE,
|
241 |
-
# Citation for the dataset
|
242 |
citation=_CITATION,
|
243 |
)
|
244 |
|
245 |
def _split_generators(self, dl_manager):
|
246 |
-
"""
|
247 |
-
This method is tasked with downloading/extracting the data
|
248 |
-
and defining the splits depending on the configuration
|
249 |
-
If several configurations are possible (listed in BUILDER_CONFIGS),
|
250 |
-
the configuration selected by the user is in self.config.name
|
251 |
-
"""
|
252 |
|
253 |
-
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
|
254 |
-
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
255 |
-
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
256 |
paths_dict = dl_manager.download_and_extract(_URLS)
|
257 |
return [
|
258 |
datasets.SplitGenerator(
|
@@ -373,13 +319,6 @@ class SPIDER(datasets.GeneratorBasedBuilder):
|
|
373 |
overview_dict[key] = {
|
374 |
k:v for k,v in item.items() if k not in exclude_vars
|
375 |
}
|
376 |
-
|
377 |
-
# # Determine maximum number of radiological gradings per patient
|
378 |
-
# max_ivd = 0
|
379 |
-
# for temp_dict_1 in grades_dict.values():
|
380 |
-
# for temp_dict_2 in temp_dict_1:
|
381 |
-
# if int(temp_dict_2['IVD label']) > max_ivd:
|
382 |
-
# max_ivd = int(temp_dict_2['IVD label'])
|
383 |
|
384 |
# Merge patient records for radiological gradings data
|
385 |
grades_dict = {}
|
@@ -478,10 +417,7 @@ class SPIDER(datasets.GeneratorBasedBuilder):
|
|
478 |
# Load .mha image file
|
479 |
image_path = os.path.join(paths_dict['images'], 'images', example)
|
480 |
image = sitk.ReadImage(image_path)
|
481 |
-
|
482 |
-
# # Rescale image intensities to [0, 255] and cast as UInt8 type
|
483 |
-
# image = sitk.Cast(sitk.RescaleIntensity(image), sitk.sitkUInt8)
|
484 |
-
|
485 |
# Convert .mha image to original size numeric array
|
486 |
image_array_original = sitk.GetArrayFromImage(image)
|
487 |
|
@@ -491,14 +427,9 @@ class SPIDER(datasets.GeneratorBasedBuilder):
|
|
491 |
resize_shape,
|
492 |
)
|
493 |
|
494 |
-
# NOTE: since the original array shape is not standardized, cannot return in dataset
|
495 |
-
|
496 |
# Load .mha mask file
|
497 |
mask_path = os.path.join(paths_dict['masks'], 'masks', example)
|
498 |
mask = sitk.ReadImage(mask_path)
|
499 |
-
|
500 |
-
# # Rescale mask intensities to [0, 255] and cast as UInt8 type
|
501 |
-
# mask = sitk.Cast(sitk.RescaleIntensity(mask), sitk.sitkUInt8)
|
502 |
|
503 |
# Convert .mha mask to original size numeric array
|
504 |
mask_array_original = sitk.GetArrayFromImage(mask)
|
|
|
11 |
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
# See the License for the specific language governing permissions and
|
13 |
# limitations under the License.
|
14 |
+
"""TODO: Add a description here.""" #TODO
|
|
|
15 |
|
16 |
# Import packages
|
17 |
import csv
|
|
|
18 |
import os
|
19 |
+
from typing import Dict, List, Mapping, Optional, Sequence, Tuple, Union
|
20 |
|
21 |
import numpy as np
|
22 |
import pandas as pd
|
23 |
|
24 |
import datasets
|
|
|
25 |
import skimage
|
26 |
import SimpleITK as sitk
|
27 |
|
|
|
56 |
DEFAULT_SCAN_TYPES = ['t1', 't2', 't2_SPACE']
|
57 |
DEFAULT_RESIZE = (512, 512, 30)
|
58 |
|
|
|
|
|
59 |
_CITATION = """\
|
60 |
+
@misc{vandergraaf2023lumbar,
|
61 |
+
title={Lumbar spine segmentation in MR images: a dataset and a public benchmark},
|
62 |
+
author={Jasper W. van der Graaf and Miranda L. van Hooff and \
|
63 |
+
Constantinus F. M. Buckens and Matthieu Rutten and \
|
64 |
+
Job L. C. van Susante and Robert Jan Kroeze and \
|
65 |
+
Marinus de Kleuver and Bram van Ginneken and Nikolas Lessmann},
|
66 |
+
year={2023},
|
67 |
+
eprint={2306.12217},
|
68 |
+
archivePrefix={arXiv},
|
69 |
+
primaryClass={eess.IV}
|
70 |
}
|
71 |
"""
|
72 |
|
|
|
81 |
_LICENSE = """Creative Commons Attribution 4.0 International License \
|
82 |
(https://creativecommons.org/licenses/by/4.0/legalcode)"""
|
83 |
|
|
|
|
|
|
|
84 |
_URLS = {
|
85 |
"images":"https://zenodo.org/records/10159290/files/images.zip",
|
86 |
"masks":"https://zenodo.org/records/10159290/files/masks.zip",
|
|
|
108 |
class SPIDER(datasets.GeneratorBasedBuilder):
|
109 |
"""TODO: Short description of my dataset."""
|
110 |
|
111 |
+
# Class attributes
|
112 |
DEFAULT_WRITER_BATCH_SIZE = 16 # PyArrow default is too large for image data
|
|
|
113 |
VERSION = datasets.Version("1.1.0")
|
|
|
114 |
BUILDER_CONFIG_CLASS = CustomBuilderConfig
|
115 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
116 |
def __init__(
|
117 |
self,
|
118 |
*args,
|
|
|
125 |
self.resize_shape = resize_shape
|
126 |
|
127 |
def _info(self):
|
128 |
+
"""Specify datasets.DatasetInfo object containing information and typing
|
129 |
+
for the dataset."""
|
130 |
+
|
|
|
131 |
image_size = self.config.resize_shape
|
132 |
features = datasets.Features({
|
133 |
"patient_id": datasets.Value("string"),
|
|
|
189 |
})
|
190 |
|
191 |
return datasets.DatasetInfo(
|
|
|
192 |
description=_DESCRIPTION,
|
193 |
+
features=features,
|
|
|
|
|
|
|
|
|
|
|
194 |
homepage=_HOMEPAGE,
|
|
|
195 |
license=_LICENSE,
|
|
|
196 |
citation=_CITATION,
|
197 |
)
|
198 |
|
199 |
def _split_generators(self, dl_manager):
|
200 |
+
"""Download and extract data and define splits based on configuration."""
|
|
|
|
|
|
|
|
|
|
|
201 |
|
|
|
|
|
|
|
202 |
paths_dict = dl_manager.download_and_extract(_URLS)
|
203 |
return [
|
204 |
datasets.SplitGenerator(
|
|
|
319 |
overview_dict[key] = {
|
320 |
k:v for k,v in item.items() if k not in exclude_vars
|
321 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
|
323 |
# Merge patient records for radiological gradings data
|
324 |
grades_dict = {}
|
|
|
417 |
# Load .mha image file
|
418 |
image_path = os.path.join(paths_dict['images'], 'images', example)
|
419 |
image = sitk.ReadImage(image_path)
|
420 |
+
|
|
|
|
|
|
|
421 |
# Convert .mha image to original size numeric array
|
422 |
image_array_original = sitk.GetArrayFromImage(image)
|
423 |
|
|
|
427 |
resize_shape,
|
428 |
)
|
429 |
|
|
|
|
|
430 |
# Load .mha mask file
|
431 |
mask_path = os.path.join(paths_dict['masks'], 'masks', example)
|
432 |
mask = sitk.ReadImage(mask_path)
|
|
|
|
|
|
|
433 |
|
434 |
# Convert .mha mask to original size numeric array
|
435 |
mask_array_original = sitk.GetArrayFromImage(mask)
|