-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
316 lines (259 loc) · 9.8 KB
/
utils.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
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
# -*- coding: utf-8 -*-
"""
@author: serdarhelli
"""
import numpy as np
import math
import cv2
import pydicom
from pydicom.pixel_data_handlers.util import apply_voi_lut
def find_center(img):
thresh=(img)*(255/np.max(img))
thresh = thresh.astype(np.uint8)
kernel =( np.ones((5,5), dtype=np.float32))
ret,thresh = cv2.threshold(thresh, 0, 255, cv2.THRESH_BINARY)
thresh=cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel,iterations=1 )
thresh=cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel,iterations=1 )
thresh=cv2.erode(thresh,kernel,iterations =1)
contours, hierarchy = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
if len(contours)!=0:
c_area=np.zeros([len(contours)])
for i in range(len(contours)):
c_area[i]= cv2.contourArea(contours[i])
c_1=contours[np.argmax(c_area)]
M = cv2.moments(c_1)
cX = int(M["m10"] / M["m00"])
cY = int(M["m01"] / M["m00"])
return cX,cY
else:
return 0,0
def points_center_mass(predict):
points=np.zeros([6,2])
for i in range(6):
points[i,:]=find_center(predict[0,:,:,i])
return np.int32(points)
def points_max_value(predict):
points=np.zeros([6,2])
for i in range(6):
pre=predict[0,:,:,i]
points[i,:]=np.where(pre == pre.max())
return np.fliplr(np.int32(points))
def read_dicom(path, voi_lut = True, fix_monochrome = True):
dicom = pydicom.read_file(path)
# VOI LUT (if available by DICOM device) is used to transform raw DICOM data to "human-friendly" view
if voi_lut:
data = apply_voi_lut(dicom.pixel_array, dicom)
else:
data = dicom.pixel_array
# depending on this value, X-ray may look inverted - fix that:
if fix_monochrome and dicom.PhotometricInterpretation == "MONOCHROME1":
data = np.amax(data) - data
# data=data*255
# data = np.uint8(data)
try:
PatientName=str(dicom.PatientName.components[0])
except:
PatientName="Empty"
pass
try:
PatientID=str(dicom.PatientID)
except:
PatientID="Empty"
pass
try:
SOPInstanceUID=str(dicom.SOPInstanceUID.name)
except:
SOPInstanceUID="Empty"
pass
try:
StudyDate=str(dicom.StudyDate)
except:
StudyDate="Empty"
pass
try:
InstitutionAddress=str(dicom.InstitutionName)
except:
InstitutionAddress="Empty"
pass
try:
PatientAge=str(dicom.PatientAge)
except:
PatientAge="Empty"
pass
try:
PatientSex=str(dicom.PatientSex)
except:
PatientSex="Empty"
pass
#data -> np.uint16
return data,PatientName,PatientID,SOPInstanceUID,StudyDate,InstitutionAddress,PatientAge,PatientSex
def modification_cropping(roi):
if roi.shape[0]!=roi.shape[1]:
if roi.shape[0]>roi.shape[1]:
img2=np.zeros([roi.shape[0],roi.shape[0]])
add=(roi.shape[0]-roi.shape[1])
a1=add//2
a2=add-a1
img2[:,a1:(roi.shape[0]-a2)]=roi
if roi.shape[1]>roi.shape[0]:
img2=np.zeros([roi.shape[1],roi.shape[1]])
add=(roi.shape[1]-roi.shape[0])
a1=add//2
a2=add-a1
img2[a1:(roi.shape[1]-a2),:]=roi
else:
img2=roi
return img2
def croping(img,x, y, w, h):
if y<0:
y=0
if abs(w)<abs(h):
z=np.abs(h-w)
if img.shape[1]<x+w+(z//2):
if x-(z//2)>0:
img2=img[y:y+h, x-(z//2):img.shape[1]].copy()
else:
img2=img[y:y+h, 0:img.shape[1]].copy()
else:
if x-(z//2)>0:
img2=img[y:y+h, x-(z//2):x+w+(z//2)].copy()
else:
img2=img[y:y+h, 0:x+w+(z//2)].copy()
if abs(h)<abs(w):
z=np.abs(h-w)
if img.shape[0]<y+h+(z//2):
if y-(z//2)>0:
img2=img[y-(z//2):img.shape[0], x:x+w].copy()
else:
img2=img[0:img.shape[0], x:x+w].copy()
else:
if y-(z//2)>0:
img2=img[y-(z//2):y+h+(z//2), x:x+w].copy()
else:
img2=img[0:y+h+(z//2), x:x+w].copy()
if abs(h)==abs(w):
img2=img[y:y + h, x:x + w].copy()
return img2
def crop_resize(path):
try:
data,PatientName,PatientID,SOPInstanceUID,StudyDate,InstitutionAddress,PatientAge,PatientSex=read_dicom(path,False,True)
except:
data,PatientName,PatientID,SOPInstanceUID,StudyDate,InstitutionAddress,PatientAge,PatientSex=read_dicom(path,True,True)
pass
img = np.copy(data)
#Denoise Image
kernel =( np.ones((5,5), dtype=np.float32))
img2=cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel,iterations=2 )
img2=cv2.erode(img2,kernel,iterations =2)
if len(img2.shape)==3:
img2=img2[:,:,0]
#Threshhold 100- 4096
ret,thresh = cv2.threshold(img2,100, 4096, cv2.THRESH_BINARY)
#To Thresh uint8 becasue "findContours" doesnt accept uint16
thresh =((thresh/np.max(thresh))*255).astype('uint8')
a1,b1=thresh.shape
#Find Countours
contours, hierarchy = cv2.findContours(thresh.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
#If There is no countour
if len(contours)==0:
return thresh,PatientName,PatientID,SOPInstanceUID,StudyDate,InstitutionAddress,PatientAge,PatientSex
#Get Areas
c_area=np.zeros([len(contours)])
for i in range(len(contours)):
c_area[i]= cv2.contourArea(contours[i])
#Find Max Countour
cnts=contours[np.argmax(c_area)]
x, y, w, h = cv2.boundingRect(cnts)
#Posibble Square
roi = croping(data, x, y, w, h)
# Absolute Square
roi=modification_cropping(roi)
# Resize to 256x256 with Inter_Nearest
roi=cv2.resize(roi,(256,256),interpolation=cv2.INTER_NEAREST)
return roi,PatientName,PatientID,SOPInstanceUID,StudyDate,InstitutionAddress,PatientAge,PatientSex
def put_text_point(original_img,heatpoint):
original_img =((original_img/np.max(original_img))*255).astype('uint8')
color = (0, 51, 204)
img = cv2.cvtColor(original_img, cv2.COLOR_BGR2RGB)
for i in range(6):
if heatpoint[i,0]<=0 and heatpoint[i,1]<=0:
print("L"+str(i)+" There is no Point")
else :
if i>2:
coordx=0
coordy=-(i*3)
else:
coordx=-(i*3)
coordy=+(i*3)+10
img=cv2.putText(img, "L"+str(i),(heatpoint[i,0]+coordx,heatpoint[i,1]+coordy), cv2.FONT_HERSHEY_SIMPLEX,0.35, color, 1)
img = cv2.circle(img, (heatpoint[i,0],heatpoint[i,1]), radius=2, color=color, thickness=-1)
return img
def get_vector(pt1,pt2):
vec=np.zeros([2])
vec[1]=(pt2[1]-pt1[1])
vec[0]=(pt2[0]-pt1[0])
return vec
def dotproduct(v1, v2):
return sum((a*b) for a, b in zip(v1, v2))
def length(v):
return math.sqrt(dotproduct(v, v))
def getAngle(v1, v2):
if length(v1)==0 or length(v2)==0:
return "Failed"
return math.degrees(math.acos(dotproduct(v1, v2) / (length(v1) * length(v2))))
def bisector_vector(v1,v2):
if length(v1)==0 or length(v2) ==0:
return [0,0]
v1=v1/(length(v1))
v2=v2/(length(v2))
v3=(v1+v2)
return v3
#magnitude 50 length to l1 to l3
def angle_patellercongruence(heatpoint,magnitude=50):
v1=get_vector(heatpoint[1,:],heatpoint[2,:])
v2=get_vector(heatpoint[1,:],heatpoint[0,:])
v3=get_vector(heatpoint[1,:],heatpoint[3,:])
v4=bisector_vector(v1,v2)
v=np.int32(v4*magnitude)
coord=v+heatpoint[1,:]
if length(v3)==0:
return "Failed",[0,0]
angle_patellercongruence=getAngle(v3/(length(v3)),v4)
return angle_patellercongruence,coord
def angle_paraleltilt_displacement(heatpoint):
v1=get_vector(heatpoint[4,:],heatpoint[5,:])
v2=get_vector(heatpoint[0,:],heatpoint[2,:])
angle_paraleltilt=getAngle(v1,v2)
return angle_paraleltilt
def draw_angle(img,heatpoint):
color = (255, 26, 26)
color2=(255, 255, 0)
color3=(51, 255, 51)
if np.min(heatpoint[0:3,:])<=0:
patellercongruence,angle_paraleltilt="Failed"
return img
if np.min(heatpoint[3:,:])<=0:
angle_paraleltilt="Failed"
v1=get_vector(heatpoint[1,:],heatpoint[2,:])
v2=get_vector(heatpoint[1,:],heatpoint[0,:])
angle=getAngle(v1,v2)
patellercongruence,coord=angle_patellercongruence(heatpoint)
angle_paraleltilt=angle_paraleltilt_displacement(heatpoint)
img=cv2.line(img,tuple( (heatpoint[1,:])), tuple((heatpoint[2,:])), color, thickness=1, lineType=8)
img=cv2.line(img, tuple((heatpoint[1,:])), tuple((heatpoint[0,:])), color, thickness=1, lineType=8)
img=cv2.line(img, tuple((heatpoint[1,:])), tuple((heatpoint[3,:])), color2, thickness=1, lineType=8)
img=cv2.line(img, tuple((heatpoint[4,:])), tuple((heatpoint[5,:])), color3, thickness=1, lineType=8)
img=cv2.line(img, tuple((heatpoint[0,:])), tuple((heatpoint[2,:])), color3, thickness=1, lineType=8)
img=cv2.line(img,tuple( (heatpoint[1,:])), tuple(coord), color2, thickness=1, lineType=8)
img=cv2.putText(img,"Pateller Congruence Angle :"+str(round(patellercongruence,2)),(25,25), cv2.FONT_HERSHEY_SIMPLEX,0.35, color2, 1)
img=cv2.putText(img,"Paralel Tilt Angle :"+str(round(angle_paraleltilt,2)),(50,50), cv2.FONT_HERSHEY_SIMPLEX,0.35, color3, 1)
img=cv2.putText(img, "Angle :"+str(round(angle,2)),(heatpoint[1,0]+10,heatpoint[1,1]+15), cv2.FONT_HERSHEY_SIMPLEX,0.35, color,1)
return img,patellercongruence,angle_paraleltilt
def predict(img,model):
#Normalization
img=np.float32(img/(np.max(img)))
img=np.reshape(img,(1,256,256,1))
predictions=model.predict(img)
#Get Final Prediction
pre=predictions[-1]
return pre