-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathdata_loader.py
263 lines (245 loc) · 10.7 KB
/
data_loader.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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
# -*- coding: utf-8 -*-
import SimpleITK as sitk
from torch.utils.data import Dataset
import os
import numpy as np
import random
import cv2
import torch,pydicom
"""
mask[256，256]
Liver: 63 (55<<<70)
"""
class ChaosDataset_Syn_new(Dataset):
def __init__(self, path="../datasets/chaos2019", split='train', modals=('t1','t2','ct'),transforms=None):
super(ChaosDataset_Syn_new, self).__init__()
for modal in modals:
assert modal in {'t1','t2','ct'}
fold = split + "/"
path1 = os.path.join(path, fold+modals[0])
path2 = os.path.join(path, fold + modals[1])
path3 = os.path.join(path, fold + modals[2])
list_path = sorted([os.path.join(path1, x) for x in os.listdir(path1)]) + sorted([os.path.join(path2, x) for x in os.listdir(path2)])
raw_path = []
label_path=[]
for x in list_path:
if "t1" in x:
x += "/T1DUAL"
c = np.array(0)
elif "t2" in x:
x += "/T2SPIR"
c = np.array(1)
for y in os.listdir(x):
if "Ground" in y:
tmp = os.path.join(x, y)
raw_path.append([tmp.replace("Ground", "DICOM_anon"),c])
break
#########
self.raw_dataset = []
self.label_dataset = []
#######
self.transfroms = transforms
for i,c in raw_path:
if c == 0:
i += "/InPhase"
for y in os.listdir(i):
tmp = os.path.join(i, y)
img = sitk.ReadImage(tmp)
img = sitk.GetArrayFromImage(img)[0]
self.raw_dataset.append([raw_preprocess(img, True),c])
a = tmp.replace("DICOM_anon/InPhase", "Ground")
img = sitk.ReadImage(a.replace(".dcm", ".png"))
img = sitk.GetArrayFromImage(img)
self.label_dataset.append(label_preprocess(img))
elif c==1:
for y in os.listdir(i):
tmp = os.path.join(i, y)
img = sitk.ReadImage(tmp)
img = sitk.GetArrayFromImage(img)[0]
self.raw_dataset.append([raw_preprocess(img, True),c])
a = tmp.replace("DICOM_anon", "Ground")
img = sitk.ReadImage(a.replace(".dcm", ".png"))
img = sitk.GetArrayFromImage(img)
self.label_dataset.append(label_preprocess(img))
list_path = sorted([os.path.join(path3, x) for x in os.listdir(path3)])
raw_path =[]
assert len(raw_path)==0
for x in list_path:
c = np.array(2)
for y in os.listdir(x):
if "Ground" in y:
tmp = os.path.join(x, y)
label_path.append(tmp)
raw_path.append([tmp.replace("Ground", "DICOM_anon"), c])
break
for i,c in raw_path:
for y in sorted(os.listdir(i)):
tmp = os.path.join(i, y)
dcm = pydicom.dcmread(tmp)
wc = dcm.WindowCenter[0]
ww = dcm.WindowWidth[0]
slope = dcm.RescaleSlope
intersept = dcm.RescaleIntercept
low = wc - ww // 2
high = wc + ww // 2
img = dcm.pixel_array * slope + intersept
img[img < low] = low
img[img > high] = high
img = (img - low) / (high - low)
shape= img.copy()
shape[shape!=0]=1
self.raw_dataset.append([[img, shape], c])
for i in label_path:
for y in sorted(os.listdir(i)):
img = sitk.ReadImage(os.path.join(i, y))
img = sitk.GetArrayFromImage(img)
data = img.astype(dtype=int)
new_seg = np.zeros(data.shape, data.dtype)
new_seg[data != 0] = 1
self.label_dataset.append(new_seg)
self.split = split
assert len(self.raw_dataset) == len(self.label_dataset)
print("chaos train data load success!")
print("modal:{},fold:{}, total size:{}".format(modals,fold,len(self.raw_dataset)))
def __getitem__(self, item):
img, shape_mask, class_label, seg_mask = self.raw_dataset[item][0][0], self.raw_dataset[item][0][1], \
self.raw_dataset[item][
1], self.label_dataset[item]
if img.shape[0]!=256:
img = cv2.resize(img, (256, 256), interpolation=cv2.INTER_LINEAR)
seg_mask = cv2.resize(seg_mask, (256, 256), interpolation=cv2.INTER_NEAREST)
shape_mask = cv2.resize(shape_mask, (256, 256), interpolation=cv2.INTER_NEAREST)
t_img = img * seg_mask
if self.split == 'train':
if random.random() > 0.5:
img = cv2.flip(img, 1)
seg_mask = cv2.flip(seg_mask, 1)
shape_mask = cv2.flip(shape_mask, 1)
t_img = cv2.flip(t_img, 1)
# scale to [-1,1]
img = (img - 0.5) / 0.5
t_img = (t_img - 0.5) / 0.5
return torch.from_numpy(img).type(torch.FloatTensor).unsqueeze(dim=0), torch.from_numpy(t_img).type(
torch.FloatTensor).unsqueeze(dim=0), \
torch.from_numpy(shape_mask).type(
torch.LongTensor).unsqueeze(dim=0), torch.from_numpy(seg_mask).type(
torch.LongTensor).unsqueeze(dim=0), torch.from_numpy(class_label).type(torch.FloatTensor)
def __len__(self):
return len(self.raw_dataset)
class ChaosDataset_Syn_Test(Dataset):
def __init__(self, path="../datasets/chaos2019", split='test', modal='t1',gan=False, transforms=None):
super(ChaosDataset_Syn_Test, self).__init__()
assert modal in {'t1', 't2','ct'}
fold = split + "/" + modal
path = os.path.join(path, fold)
list_path = sorted([os.path.join(path, x) for x in os.listdir(path)])
raw_path = []
label_path = []
if gan is True:
list_path = list_path[0:1]
for x in list_path:
if modal == "t1":
x += "/T1DUAL"
elif modal == "t2":
x += "/T2SPIR"
for y in os.listdir(x):
if "Ground" in y:
tmp = os.path.join(x, y)
if "ct" in x:
label_path.append(tmp)
raw_path.append(tmp.replace("Ground", "DICOM_anon"))
break
self.transfroms = transforms
self.raw_dataset = []
self.label_dataset = []
self.index = []
if modal == "t1":
for i in raw_path:
i += "/InPhase"
n = 0
for y in os.listdir(i):
tmp = os.path.join(i, y)
img = sitk.ReadImage(tmp)
img = sitk.GetArrayFromImage(img)[0]
self.raw_dataset.append(raw_preprocess(img))
a = tmp.replace("DICOM_anon/InPhase", "Ground")
img = sitk.ReadImage(a.replace(".dcm", ".png"))
img = sitk.GetArrayFromImage(img)
self.label_dataset.append(label_preprocess(img))
n += 1
self.index.append(n)
elif modal=='t2':
for i in raw_path:
n = 0
for y in os.listdir(i):
tmp = os.path.join(i, y)
img = sitk.ReadImage(tmp)
img = sitk.GetArrayFromImage(img)[0]
self.raw_dataset.append(raw_preprocess(img))
a = tmp.replace("DICOM_anon", "Ground")
img = sitk.ReadImage(a.replace(".dcm", ".png"))
img = sitk.GetArrayFromImage(img)
self.label_dataset.append(label_preprocess(img))
n += 1
self.index.append(n)
else:
for i in raw_path:
n = 0
for y in sorted(os.listdir(i)):
tmp = os.path.join(i, y)
dcm = pydicom.dcmread(tmp)
wc = dcm.WindowCenter[0]
ww = dcm.WindowWidth[0]
slope = dcm.RescaleSlope
intersept = dcm.RescaleIntercept
low = wc - ww // 2
high = wc + ww // 2
img = dcm.pixel_array * slope + intersept
img[img < low] = low
img[img > high] = high
img = (img - low) / (high - low)
self.raw_dataset.append(img)
n += 1
self.index.append(n)
for i in label_path:
for y in sorted(os.listdir(i)):
img = sitk.ReadImage(os.path.join(i, y))
img = sitk.GetArrayFromImage(img)
data = img.astype(dtype=int)
new_seg = np.zeros(data.shape, data.dtype)
new_seg[data != 0] = 1
self.label_dataset.append(new_seg)
self.split = split
assert len(self.raw_dataset) == len(self.label_dataset)
print("chaos test data load success!")
print("modal:{},fold:{}, total size:{}".format(modal, fold, len(self.raw_dataset)))
def __getitem__(self, item):
img, mask = self.raw_dataset[item], self.label_dataset[item]
if img.shape[0] != 256:
img = cv2.resize(img, (256, 256), interpolation=cv2.INTER_LINEAR)
mask = cv2.resize(mask, (256, 256), interpolation=cv2.INTER_NEAREST)
img = (img - 0.5) / 0.5
return torch.from_numpy(img).type(torch.FloatTensor).unsqueeze(dim=0), torch.from_numpy(mask).type(torch.LongTensor)
def __len__(self):
return len(self.raw_dataset)
def _getIndex(self):
return self.index
def label_preprocess(data):
data = data.astype(dtype=int)
new_seg = np.zeros(data.shape, data.dtype)
new_seg[(data > 55) & (data <= 70)] = 1
return new_seg
def raw_preprocess(data, get_s=False):
"""
:param data: [155,224,224]
:return:
"""
data = data.astype(dtype=float)
data[data<50] = 0
out = data.copy()
out = (out - out.min()) / (out.max() - out.min())
if get_s:
share_mask = out.copy()
share_mask[share_mask != 0] = 1
return out, share_mask
return out