-
Notifications
You must be signed in to change notification settings - Fork 1
/
surface_distance.py
244 lines (178 loc) · 7.92 KB
/
surface_distance.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
import numpy as np
from scipy.ndimage import morphology
import time
import copy
import matplotlib.pyplot as plt
import seg_utils
try:
import seg_utils.myutil as futil
except:
import pip
pip.main(['install', '--user', 'seg-utils'])
import seg_utils.myutil as futil
def metrics_per_chn(gdth, pred):
gdth = gdth.astype(int)
pred = pred.astype(int)
fp_array = copy.deepcopy(pred)
fn_array = copy.deepcopy(gdth)
gdth_sum = np.sum(gdth)
pred_sum = np.sum(pred)
intersection = np.bitwise_and(gdth, pred)
union = np.bitwise_or(gdth, pred)
intersection_sum = np.sum(intersection)
union_sum = np.sum(union)
tp_array = intersection
tmp = pred - gdth
fp_array[tmp<1]=0
tmp2 = gdth - pred
fn_array[tmp2<1]=0
tn_array = np.ones(gdth.shape) - union
tp, fp, fn, tn = np.sum(tp_array), np.sum(fp_array), np.sum(fn_array), np.sum(tn_array)
smooth = 0.001
precision_per_chn = tp / (pred_sum + smooth)
recall_per_chn = tp / (gdth_sum + smooth)
false_positive_rate_per_chn = fp / (fp + tn + smooth)
jaccard_per_chn = intersection_sum / (union_sum + smooth)
dice_per_chn = 2 * intersection_sum / (gdth_sum + pred_sum + smooth)
return recall_per_chn, precision_per_chn, false_positive_rate_per_chn, jaccard_per_chn, dice_per_chn
def show_array(img, idx=300):
plt.figure()
plt.imshow(img[idx])
plt.show()
return None
def metrics_all(gdth, pred):
recalls, precisions, false_positive_rates, jaccards, dices = [], [], [], [], []
for gdth_per_chn, pred_per_chn in zip(gdth, pred):
recall_per_chn, precision_per_chn, false_positive_rate_per_chn, jaccard_per_chn, dice_per_chn = metrics_per_chn(gdth_per_chn, pred_per_chn)
recalls.append(recall_per_chn)
precisions.append(precision_per_chn)
false_positive_rates.append(false_positive_rate_per_chn)
jaccards.append(jaccard_per_chn)
dices.append(dice_per_chn)
return recalls, precisions, false_positive_rates, jaccards, dices
def metrics_ave(gdth, pred):
recalls, precisions, false_positive_rates, jaccards, dices = metrics_all(gdth, pred)
recall_ave, precision_ave, false_positive_rate_ave, jaccard_ave, dice_ave = np.average(recalls, precisions, false_positive_rates, jaccards, dices)
return recall_ave, precision_ave, false_positive_rate_ave, jaccard_ave, dice_ave
def surfd(input1, input2, spacing=1, connectivity=1):
input_1 = np.atleast_1d(input1.astype(np.bool))
input_2 = np.atleast_1d(input2.astype(np.bool))
conn = morphology.generate_binary_structure(input_1.ndim, connectivity)
input1_erosion = morphology.binary_erosion(input_1, conn).astype(int)
# show_array(input1_erosion)
input2_erosion = morphology.binary_erosion(input_2, conn).astype(int)
# show_array(input2_erosion)
S = input_1 - input1_erosion
Sprime = input_2 - input2_erosion
S_sum = np.sum(S)
Sprime_sum = np.sum(Sprime)
# show_array(S)
# show_array(Sprime)
S = S.astype(np.bool)
Sprime = Sprime.astype(np.bool)
dta = morphology.distance_transform_edt(~S, spacing)
dtb = morphology.distance_transform_edt(~Sprime, spacing)
# show_array(dta)
# show_array(dtb)
dta_multiply_Sprime = dta[Sprime != 0] # 365703
dtb_multiply_S = dtb[S != 0] # 428467
# show_array(dta_multiply_Sprime)
# show_array(dtb_multiply_S)
count11 = np.count_nonzero(dta_multiply_Sprime)
count22 = np.count_nonzero(dtb_multiply_S)
ds1 = np.ravel(dta[Sprime != 0])
ds2 = np.ravel(dtb[S != 0])
count1 = np.count_nonzero(ds1)
count2 = np.count_nonzero(ds2)
sds = np.concatenate([ds1, ds2])
surface_distance = ds1
msd = surface_distance.mean()
rms = np.sqrt((surface_distance ** 2).mean())
hd = surface_distance.max()
hd95 = np.percentile(surface_distance, 95)
median = np.median(surface_distance)
std = np.std(surface_distance)
surface_distance = ds2
msd = surface_distance.mean()
rms = np.sqrt((surface_distance ** 2).mean())
hd = surface_distance.max()
hd95 = np.percentile(surface_distance, 95)
median = np.median(surface_distance)
std = np.std(surface_distance)
return sds
def one_hot_encode_3D(patch, labels):
# assert len(patch.shape)==5 # (5, 128, 128, 64, 1)
labels = np.array(labels) # i.e. [0,4,5,6,7,8]
N_classes = labels.size # 6, similiar with len(labels)
if len(patch.shape) == 5: # (5, 128, 128, 64, 1)
patch = np.reshape(patch, (patch.shape[0], patch.shape[1], patch.shape[2], patch.shape[3]))
elif len(patch.shape) == 4: # (128, 128, 64, 1)
patch = np.reshape(patch, (patch.shape[0], patch.shape[1], patch.shape[2]))
patches = []
# print('patch.shape', patch.shape)
for i, l in enumerate(labels):
a = np.where(patch != l, 0, 1)
patches.append(a)
patches = np.array(patches)
patches = np.rollaxis(patches, 0, len(patches.shape)) # from [6, 64, 128, 128] to [64, 128, 128, 6]?
# print('patches.shape after on hot encode 3D', patches.shape)
return np.float64(patches)
task='lobe'
model = '1585046531.9177752_0.00011a_o_0.5ds2dr1bn1fs16ptsz144ptzsz64'
for file_name in ['26', '27', '28', '29', '30']:
pred_file_name= '/data/jjia/new/results/lobe/valid/pred/GLUCOLD/' + model + '/GLUCOLD_patients_' + file_name + '.mhd'
gdth_file_name = '/data/jjia/mt/data/lobe/valid/gdth_ct/GLUCOLD/GLUCOLD_patients_' + file_name + '.nrrd'
# gdth_file_name = '/data/jjia/mt/data/vessel/valid/gdth_ct/SSc/SSc_patient_51.mhd'
# pred_file_name = '/data/jjia/practice/SSc_patient_51.mhd'
gdth, _, gdth_spacing = futil.load_itk(gdth_file_name)
pred, _, pred_spacing = futil.load_itk(pred_file_name)
gdth = np.expand_dims(gdth, axis=-1)
pred = np.expand_dims(pred, axis=-1)
if task=='lobe':
labels=[4,5,6,7,8]
else:
labels=[1]
gdth_encode = one_hot_encode_3D(gdth, labels=labels)
pred_encode = one_hot_encode_3D(pred, labels=labels)
msd_list, rms_list, hd_list, hd95_list, median_list, std_list = [], [], [], [], [], []
for i in range(len(labels)):
print('start a loop of ', i)
time1 = time.time()
pred = pred_encode[..., i]
gdth = gdth_encode[..., i]
metrics = metrics_per_chn(gdth, pred)
print('recall_per_chn, precision_per_chn, false_positive_rate_per_chn, jaccard_per_chn, dice_per_chn')
print(metrics)
surface_distance = surfd(pred, gdth, spacing=gdth_spacing)
msd = surface_distance.mean()
rms = np.sqrt((surface_distance ** 2).mean())
hd = surface_distance.max()
hd95 = np.percentile(surface_distance, 95)
median = np.median(surface_distance)
std = np.std(surface_distance)
msd_list.append(msd)
rms_list.append(rms)
hd_list.append(hd)
hd95_list.append(hd95)
median_list.append(median)
std_list.append(std)
print('with spacing: msd:{}, rms:{}, hd:{}, hd95:{}, median:{}, std:{}'.format(msd, rms, hd, hd95, median, std))
time2 = time.time()
print('cost time', time2 - time1)
print('-------')
print(msd_list, rms_list, hd_list, hd95_list, median_list, std_list)
print(np.array(msd_list).mean(), np.array(rms_list).mean(), np.array(hd_list).mean(), np.array(hd95_list).mean(), np.array(median_list).mean(), np.array(std_list).mean())
# surface_distance = surfd(pred, gdth, 1)
#
#
# msd = surface_distance.mean()
# rms = np.sqrt((surface_distance ** 2).mean())
# hd = surface_distance.max()
# hd95 = np.percentile(surface_distance, 95)
# median = np.median(surface_distance)
# std = np.std(surface_distance)
# print('without spacing: msd:{}, rms:{}, hd:{}, hd95:{}, median:{}, std:{}'.format(msd, rms, hd, hd95, median, std))
# time3 = time.time()
# print('cost time', time3 - time2)
# print('-------------------------------------------------------')
print('ok')