-
Notifications
You must be signed in to change notification settings - Fork 2
/
dataset.py
232 lines (172 loc) · 7.5 KB
/
dataset.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
import pickle
import torch.utils.data as data
from os import listdir
from os.path import join
import numpy as np
import nibabel as nib
import glob
import cv2
from scipy import ndimage
from helper import *
import os
import matplotlib.pyplot as plt
class TrainDatasetACDC(data.Dataset):
def __init__(self, data_path):
super(TrainDatasetACDC, self).__init__()
self.data_path = data_path
self.filename = [f for f in sorted(listdir(self.data_path))]
def __getitem__(self, index):
input, fix_seg, fix_seg_myomask, mov_seg = load_data_ACDC(self.data_path, self.filename[index], size=96, rand_frame=None, load_seg_mask=True)
mov = input[:1]
fix = input[1:]
# print(image.shape)
# print(image_pred.shape)
return mov, fix, mov_seg, fix_seg, fix_seg_myomask
def __len__(self):
return len(self.filename)
class TestDatasetACDC(data.Dataset):
def __init__(self, data_path):
super(TestDatasetACDC, self).__init__()
self.data_path = data_path
self.filename = [f for f in sorted(listdir(self.data_path))]
def __getitem__(self, index):
input, fix_seg, fix_seg_myomask, mov_seg = load_data_ACDC(self.data_path, self.filename[index], size=96, rand_frame=0, load_seg_mask=True)
mov = input[:1]
fix = input[1:]
file_name = self.filename[index]
return mov, fix, mov_seg, fix_seg, fix_seg_myomask, file_name
def __len__(self):
return len(self.filename)
def load_data_ACDC(data_path, filename, size, rand_frame=None, load_seg_mask=False):
# Load configuration file to get ED and ES frame number
info_file_path = os.path.join(data_path, filename, 'Info.cfg')
with open(info_file_path) as f:
line1 = f.readline()
line2 = f.readline()
ED_idx = '{:02d}'.format(int(line1.split(':')[1]))
ES_idx = '{:02d}'.format(int(line2.split(':')[1])) # type str
# Load images and labels
nim = nib.load(join(data_path, filename, filename+'_frame'+ED_idx+'.nii.gz')) # load ED image (source)
image = nim.get_data()[:, :, :] # h x w x z
image = np.array(image, dtype='float32')
# generate random index for t and z dimension
if rand_frame is not None:
# rand_t = rand_frame
# rand_t = 1
rand_z = rand_frame
# print('this is rand_t', rand_t)
# print('this is rand_z', rand_z)
else:
# rand_t = np.random.randint(0, image.shape[3])
rand_z = np.random.randint(1, image.shape[2]-1)
# preprocessing
image_max = np.max(np.abs(image))
image /= image_max
image_ED = image[..., rand_z]
image_ED = image_ED[np.newaxis]
nim = nib.load(join(data_path, filename, filename+'_frame'+ES_idx+'.nii.gz')) # load ES image (target)
image = nim.get_data()[:, :, :]
image = np.array(image, dtype='float32')
nim_seg = nib.load(join(data_path, filename, filename+'_frame'+ES_idx+'_gt.nii.gz')) # load ES mask (target)
seg = nim_seg.get_data()[:, :, :]
image_ES = image[..., rand_z]
image_ES /= image_max
seg_ES = seg[..., rand_z]
slice = (seg[..., image.shape[2]//2] == 2).astype(np.uint8) # get the middle slice in the z stack
centre = ndimage.measurements.center_of_mass(slice)
centre = np.round(centre).astype(np.uint8)
# print(centre)
image_ES = image_ES[np.newaxis]
seg_ES = seg_ES[np.newaxis]
image_bank = np.concatenate((image_ED, image_ES), axis=0)
image_bank = centre_crop(image_bank, size, centre)
seg_ES = centre_crop(seg_ES, size, centre)
image_bank = np.transpose(image_bank, (0, 2, 1))
seg_ES = np.transpose(seg_ES, (0, 2, 1))
image_bank = np.array(image_bank, dtype='float32')
seg_ES = np.array(seg_ES, dtype='int16')
mask = (seg_ES == 2).astype(np.uint8)
# mask_new = mask
# print(mask.shape)
# mask = centre_crop(mask, size, centre)
# print(np.max(mask_new - mask))
kernel = np.ones((3, 3), np.uint8)
mask = cv2.dilate(mask[0], kernel, iterations=3)
mask = np.array(mask[np.newaxis], dtype='int16')
if load_seg_mask:
nim_seg_ED = nib.load(join(data_path, filename, filename+'_frame'+ED_idx+'_gt.nii.gz')) # load ED mask (source)
seg_ED = nim_seg_ED.get_data()[:, :, :]
seg_ED = seg_ED.copy()[..., rand_z]
seg_ED = seg_ED[np.newaxis]
seg_ED = centre_crop(seg_ED, size, centre)
seg_ED = np.transpose(seg_ED, (0, 2, 1))
seg_ED = np.array(seg_ED, dtype='int16')
if not load_seg_mask:
return image_bank, seg_ES, mask # (mov, fix), fix_seg, fix_seg_myo
if load_seg_mask:
return image_bank, seg_ES, mask, seg_ED # (mov, fix), fix_seg, fix_seg_myo, mov_seg
class TrainDatasetLVQuant(data.Dataset):
def __init__(self, data_path):
super(TrainDatasetLVQuant, self).__init__()
self.data_path = data_path
self.filename = [f for f in sorted(listdir(self.data_path))]
def __getitem__(self, index):
input, fix_seg, fix_seg_myomask, mov_seg = load_data_LVQuant(self.data_path, self.filename[index], size=96, load_seg_mask=True)
mov = input[:1]
fix = input[1:]
# print(image.shape)
# print(image_pred.shape)
return mov, fix, mov_seg, fix_seg, fix_seg_myomask
def __len__(self):
return len(self.filename)
class TestDatasetLVQuant(data.Dataset):
def __init__(self, data_path):
super(TestDatasetLVQuant, self).__init__()
self.data_path = data_path
self.filename = [f for f in sorted(listdir(self.data_path))]
def __getitem__(self, index):
input, fix_seg, fix_seg_myomask, mov_seg = load_data_LVQuant(self.data_path, self.filename[index], size=96, load_seg_mask=True)
mov = input[:1]
fix = input[1:]
file_name = self.filename[index]
return mov, fix, mov_seg, fix_seg, fix_seg_myomask, file_name
def __len__(self):
return len(self.filename)
def load_data_LVQuant(data_path, filename, size, load_seg_mask=False):
# load file
with open(os.path.join(data_path, filename), 'rb') as f:
file = pickle.load(f)
mov_image = np.array(file['mov'], dtype='float32')
fix_image = np.array(file['fix'], dtype='float32')
mov_seg = file['mov_seg']
fix_seg = file['fix_seg']
# preprocessing
mov_image /= np.max(np.abs(mov_image))
mov_image = mov_image[np.newaxis]
mov_seg = mov_seg[np.newaxis]
fix_image /= np.max(np.abs(fix_image))
fix_image = fix_image[np.newaxis]
fix_seg = fix_seg[np.newaxis]
myo_slice = (fix_seg[0] == 2).astype(np.uint8)
# plt.imshow(myo_slice, cmap='gray')
# plt.show()
centre = ndimage.measurements.center_of_mass(myo_slice) # myo_slice has to be two-dimensional
centre = np.round(centre).astype(np.uint8)
# print(centre)
image_bank = np.concatenate((mov_image, fix_image), axis=0)
fix_seg = centre_crop(fix_seg, size, centre)
image_bank = centre_crop(image_bank, size, centre)
fix_seg = np.transpose(fix_seg, (0,2,1))
image_bank = np.transpose(image_bank, (0,2,1))
mask = (fix_seg == 2).astype(np.uint8)
kernel = np.ones((3, 3), np.uint8)
mask = cv2.dilate(mask[0], kernel, iterations=3)
mask = np.array(mask[np.newaxis], dtype='int16')
if load_seg_mask:
mov_seg = centre_crop(mov_seg, size, centre)
mov_seg = np.transpose(mov_seg, (0, 2, 1))
mov_seg = np.array(mov_seg, dtype='int16')
if not load_seg_mask:
return image_bank, fix_seg, mask
if load_seg_mask:
return image_bank, fix_seg, mask, mov_seg