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
264
265
266
267
268
269
270
271
272
273
274
275
276
277 | #coding: utf-8
import pandas as pd
import tqdm
import SimpleITK as sitk
import numpy as np
import cv2
from scipy.misc import toimage, imsave
from glob import glob
import os
import json
from data.dm_preprocess import DMImagePreprocessor as imprep
import sys
json_dir = '/home/HDD1/mammogram/DDSM/json_v3/'
root_dir = '/home/HDD1/mammogram/DDSM/'
out_dir = '/home/HDD1/mammogram/DDSM/patch/all_318'
calc_train_dir = 'Calc-Test_Full_Mammogram_Images/'
calc_train_roi_dir = 'Calc-Test_ROI_and_Cropped_Images/'
bkg_dir=os.path.join(out_dir, 'background')
calc_pos_dir=os.path.join(out_dir, 'calc_mal')
calc_neg_dir=os.path.join(out_dir, 'calc_ben')
mass_pos_dir=os.path.join(out_dir, 'mass_mal')
mass_neg_dir=os.path.join(out_dir, 'mass_ben')
target_height = 1344
patch_size = 318 #224
nb_abn = 10
nb_bkg = 10
nb_hns = 5
pos_cutoff = 0.9
neg_cutoff = 0.2
verbose = False
rng = np.random.RandomState(12345)
problem_case = []
def read_img(fname):
image_array = []
if fname.split('.')[-1] == 'dcm':
image = sitk.ReadImage(fname)
image_array = sitk.GetArrayFromImage(image)
image_array = np.squeeze(image_array)
return image_array
def crop_img(img, bbox):
x_min,y_min,x_max,y_max = bbox
return img[x_min:x_max, y_min:y_max]
def add_img_margins(img, margin_size):
'''Add all zero margins to an image
'''
enlarged_img = np.zeros((img.shape[0]+margin_size*2,
img.shape[1]+margin_size*2))
enlarged_img[margin_size:margin_size+img.shape[0],
margin_size:margin_size+img.shape[1]] = img
return enlarged_img
def crop_val(v, minv, maxv):
v = v if v >= minv else minv
v = v if v <= maxv else maxv
return v
def sample_blob_negatives(img, roi_mask, basename, blob_detector, nb_bkg=10, start_sample_nb=0):
key_pts = blob_detector.detect((img/img.max()*255).astype('uint8'))
key_pts = rng.permutation(key_pts)
sampled_bkg = 0
for kp in key_pts:
if sampled_bkg >= nb_bkg:
break
x, y = int(kp.pt[0]), int(kp.pt[1])
if not overlap_patch_roi((x, y), patch_size, roi_mask, cutoff=neg_cutoff):
patch = img[y-patch_size/2:y + patch_size/2,
x-patch_size/2:x + patch_size/2]
# 全黑去掉
if patch.mean() / patch.max() > 10.0/255.0:
patch = patch.astype('int32')
patch_img = toimage(patch, high=patch.max(), low=patch.min(), mode='I')
filename = basename + "_%02d" % (start_sample_nb + sampled_bkg) + ".png"
fullname = os.path.join(bkg_dir, filename)
imsave(fullname, patch_img)
if verbose:
print "sampled a blob patch at (x,y) center=", (x,y)
sampled_bkg += 1
return sampled_bkg
def overlap_patch_roi(patch_center, patch_size, roi_mask,
add_val=1000, cutoff=.5):
x1,y1 = (patch_center[0] - patch_size/2,
patch_center[1] - patch_size/2)
x2,y2 = (patch_center[0] + patch_size/2,
patch_center[1] + patch_size/2)
x1 = crop_val(x1, 0, roi_mask.shape[1])
y1 = crop_val(y1, 0, roi_mask.shape[0])
x2 = crop_val(x2, 0, roi_mask.shape[1])
y2 = crop_val(y2, 0, roi_mask.shape[0])
roi_area = (roi_mask > 0).sum()
roi_patch_added = roi_mask.copy()
roi_patch_added[y1:y2, x1:x2] += add_val
patch_area = (roi_patch_added>=add_val).sum()
inter_area = (roi_patch_added>add_val).sum().astype('float32')
return (inter_area/roi_area > cutoff or inter_area/patch_area > cutoff)
def create_blob_detector(roi_size=(128,128), blob_min_area=3,
blob_min_int=.5, blob_max_int=.95, blob_th_step=10):
params = cv2.SimpleBlobDetector_Params()
params.filterByArea = True
params.minArea = blob_min_area
params.maxArea = roi_size[0]*roi_size[1]
params.filterByCircularity = False
params.filterByColor = False
params.filterByConvexity = False
params.filterByInertia = False
# blob detection only works with "uint8" images.
params.minThreshold = int(blob_min_int*255)
params.maxThreshold = int(blob_max_int*255)
params.thresholdStep = blob_th_step
ver = (cv2.__version__).split('.')
if int(ver[0]) < 3:
return cv2.SimpleBlobDetector(params)
else:
return cv2.SimpleBlobDetector_create(params)
def sample_patchs(img, roi_mask, basename, pos,
nb_bkg=10,
start_sample_nb=0, itype='mass', pos_cut=.75, neg_cut=.35):
print(basename)
if pos == 1:
if itype == 'mass':
roi_out = mass_pos_dir
else:
roi_out = calc_pos_dir
else:
if itype == 'mass':
roi_out = mass_neg_dir
else:
roi_out = calc_neg_dir
# 根据图像的矩计算重心
roi_mask_8u = roi_mask.astype('uint8')
ver = (cv2.__version__).split('.')
if int(ver[0]) < 3:
contours, _ = cv2.findContours(roi_mask_8u.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
else:
_, contours, _ = cv2.findContours(roi_mask_8u.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cont_areas = [cv2.contourArea(cont) for cont in contours]
if len(cont_areas) < 1:
problem_case.append(basename)
return False
idx = np.argmax(cont_areas) # find the largest contour.
rx,ry,rw,rh = cv2.boundingRect(contours[idx])
if verbose:
M = cv2.moments(contours[idx])
try:
cx = int(M['m10'] / M['m00'])
cy = int(M['m01'] / M['m00'])
print "ROI centroid=", (cx, cy); sys.stdout.flush()
except ZeroDivisionError:
cx = rx + int(rw / 2)
cy = ry + int(rh / 2)
print "ROI centroid=Unknown, use b-box center=", (cx, cy)
sampled_abn = 0
nb_try = 0
while sampled_abn < nb_abn:
if nb_abn > 1:
x = rng.randint(rx, rx + rw)
y = rng.randint(ry, ry + rh)
nb_try += 1
if nb_try >= 1000:
print "Nb of trials reached maximum, decrease overlap cutoff by 0.05"
pos_cut -= .05
nb_try = 0
print pos_cut
if pos_cut <= .0:
raise Exception("overlap cutoff becomes non-positive, "
"check roi mask input.")
else:
x = cx
y = cy
if nb_abn == 1 or overlap_patch_roi((x, y), patch_size, roi_mask,
cutoff=pos_cut):
patch = img[y - patch_size / 2:y + patch_size / 2,
x - patch_size / 2:x + patch_size / 2]
patch = patch.astype('int32')
patch_img = toimage(patch, high=patch.max(), low=patch.min(),
mode='I')
filename = basename + "_%02d" % (sampled_abn) + ".png"
fullname = os.path.join(roi_out, filename)
imsave(fullname, patch_img)
sampled_abn += 1
nb_try = 0
# Sample background
sampled_bkg = start_sample_nb
while sampled_bkg < start_sample_nb + nb_bkg:
x = rng.randint(patch_size / 2, img.shape[1] - patch_size / 2)
y = rng.randint(patch_size / 2, img.shape[0] - patch_size / 2)
if not overlap_patch_roi((x, y), patch_size, roi_mask, cutoff=neg_cut):
patch = img[y - patch_size / 2:y + patch_size / 2,
x - patch_size / 2:x + patch_size / 2]
# 全黑去掉
if patch.mean() / patch.max() > 10.0/255.0:
patch = patch.astype('int32')
patch_img = toimage(patch, high=patch.max(), low=patch.min(),
mode='I')
filename = basename + "_%02d" % (sampled_bkg) + ".png"
fullname = os.path.join(bkg_dir, filename)
imsave(fullname, patch_img)
sampled_bkg += 1
if verbose:
print "Sampled a bkg patch at (x,y) center=", (x,y)
sys.stdout.flush()
return True
def sampling(json_file):
blob_detector = create_blob_detector(roi_size=(patch_size, patch_size))
img_id = json_file.split('/')[-1].split('.')[0]
f = open(json_file, 'r')
data = json.load(f)
f.close()
bbox = np.array(data['other_info']['bbox'])
image_path = str(data['other_info']['dataPath'])
bkg_sampled = False
for nodule in data['nodes']:
pos = nodule['attr']['malignant']
points = np.array(nodule['rois'][0]['edge'])
abn = nodule['name']
itype = abn.split('_')[0]
basename = '_'.join([img_id, abn])
image = read_img(image_path)
roi_mask = np.zeros(image.shape, np.uint8)
cv2.fillPoly(roi_mask, (points,), (255, 255, 255))
roi_mask = crop_img(roi_mask, bbox)
image = crop_img(image, bbox)
# Resize到固定高度
if target_height is not None:
target_width = int(float(target_height) / image.shape[0] * image.shape[1])
roi_mask = cv2.resize(roi_mask, dsize=(target_width, target_height),
interpolation=cv2.INTER_CUBIC)
image = cv2.resize(image, dsize=(target_width, target_height),
interpolation=cv2.INTER_CUBIC)
roi_mask = add_img_margins(roi_mask, patch_size / 2)
image = add_img_margins(image, patch_size / 2)
nb_hns_ = nb_hns if not bkg_sampled else 0
if nb_hns_ > 0:
hns_sampled = sample_blob_negatives(image, roi_mask, basename, blob_detector, nb_hns_, 0)
else:
hns_sampled = 0
nb_bkg_ = nb_bkg - hns_sampled if not bkg_sampled else 0
sample_patchs(image, roi_mask, basename, pos, nb_bkg_, hns_sampled, itype, pos_cutoff, neg_cutoff)
bkg_sampled = True
def main():
json_files = glob(json_dir + '*.txt')
i = 0
for json_file in json_files:
sampling(json_file)
print('index: %d' % i)
i += 1
print(problem_case)
if __name__=='__main__':
main()
# json_file = glob(json_dir + '00047_LEFT_MLO.txt')[0]
# sampling(json_file)
|