/
dr.py
275 lines (266 loc) · 14.7 KB
/
dr.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
# -*- coding: utf-8 -*-
"""
Created on Sat Jun 13 15:59:29 2015
@author: Saiprasad
"""
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 04 11:53:38 2015
@author: SKotur
"""
from __future__ import division
import cv2
import numpy as np
import copy
import math
import os
import csv
'''
This function gets the radius of the eye in image
'''
def GetRadius(image):
retina = copy.copy(image);
retina_blue , retina_green , retina_red = cv2.split(retina)
# scaling the histogram linearly to normalize all the images
retina_green = cv2.bitwise_not(retina_green);
ret, retina_green = cv2.threshold(retina_green,retina_green.max()-1,255,cv2.THRESH_BINARY);
retina_green = cv2.bitwise_not(retina_green)
retina_green = cv2.medianBlur(retina_green,55)
img,contours,hierarchy = cv2.findContours(retina_green, 1, 2)
area = 0;
for item in contours:
if cv2.contourArea(item) > area :
area = cv2.contourArea(item)
cnt = item
leftmost = tuple(cnt[cnt[:,:,0].argmin()][0])
rightmost = tuple(cnt[cnt[:,:,0].argmax()][0])
topmost = tuple(cnt[cnt[:,:,1].argmin()][0])
bottommost = tuple(cnt[cnt[:,:,1].argmax()][0])
topmost = list(topmost);
bottommost = list(bottommost);
leftmost = list(leftmost);
rightmost = list(rightmost);
centre = [0,0];
centre[1] = (rightmost[0] + leftmost[0] + topmost[0] + bottommost[0])/4 ;
centre[0] = (rightmost[1] + leftmost[1] + topmost[1] + bottommost[1])/4 ;
radius_x = math.pow(rightmost[0]-leftmost[0],2) + math.pow(rightmost[1]-leftmost[1],2);
radius_x = math.sqrt(radius_x)/2;
radius_y = math.pow(topmost[0]-bottommost[0],2) + math.pow(topmost[1]-bottommost[1],2);
radius_y = math.sqrt(radius_y)/2;
radius =( radius_x + radius_y )/2;
return(radius);
'''
this function normalizes all the images to (3000,3000)
and a linear intensity scaling is applied over the images in order to normalize
'''
def Normalize(image,parameter):
retina = copy.copy(image);
retina_blue , retina_green , retina_red = cv2.split(retina)
# scaling the histogram linearly to normalize all the images
if (parameter == 0):
cv2.normalize(retina,retina,0,255,cv2.NORM_MINMAX)
elif(parameter == 1):
retina = cv2.cvtColor(retina,cv2.COLOR_BGR2YCR_CB)
channels = cv2.split(retina);
cv2.equalizeHist(channels[0],channels[0]);
cv2.merge(channels,retina);
retina = cv2.cvtColor(retina,cv2.COLOR_YCR_CB2BGR);
retina_green = cv2.bitwise_not(retina_green);
ret, retina_green = cv2.threshold(retina_green,retina_green.max()-1,255,cv2.THRESH_BINARY);
retina_green = cv2.bitwise_not(retina_green)
retina_green = cv2.medianBlur(retina_green,55)
img,contours,hierarchy = cv2.findContours(retina_green, 1, 2)
area = 0;
for item in contours:
if cv2.contourArea(item) > area :
area = cv2.contourArea(item)
cnt = item
leftmost = tuple(cnt[cnt[:,:,0].argmin()][0])
rightmost = tuple(cnt[cnt[:,:,0].argmax()][0])
topmost = tuple(cnt[cnt[:,:,1].argmin()][0])
bottommost = tuple(cnt[cnt[:,:,1].argmax()][0])
top = list(topmost);
bottom = list(bottommost);
leftmost = list(leftmost);
rightmost = list(rightmost);
leftmost[0] = leftmost[0];
rightmost[0] = rightmost[0];
if topmost[1] > 20 :
top[1] = top[1] - 20;
if bottommost[1] > retina_blue.shape[0] :
bottom[1] = bottom[1] + 20;
image = retina[top[1] : bottom[1],leftmost[0] : rightmost[0]];
resized_image = cv2.resize(image, (3000, 3000));
return (resized_image);
'''
Exudate detection using binary thresholding
this function takes color retinal image as input and then outputs the exudate features
'''
def Exudates(image):
retina = copy.copy(image);
retina_blue , retina_green , retina_red = cv2.split(retina);
#preprocessing of the images
#applying adaptive histogram equalization
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
#preprocessing of the images
retina_green = clahe.apply(retina_green);
test_green = copy.copy(retina_green)
a = retina_green.max();
test_green = copy.copy(retina_green)
retina_g = cv2.equalizeHist(retina_green)
retina_r = cv2.equalizeHist(retina_red)
#detection of optical disc
#finding the brightest spots in red and green channels
ret, green = cv2.threshold(retina_g,retina_g.max()-5,255,cv2.THRESH_BINARY)
ret, red = cv2.threshold(retina_r,retina_r.max()-5,255,cv2.THRESH_BINARY)
# median filtering to remove the unwanted bright spots
red = cv2.medianBlur(red,5)
green = cv2.medianBlur(green,5)
# usually the bright spot is yellow coloured so it should be bright in red and green channels
# so taking bitwise and to remove the unwanted spots from both the channels
output = cv2.bitwise_and(red,green)
# median filtering to remove the residual (unwanted) bright spots;
output = cv2.medianBlur(output,7)
kernel = np.ones((5,5),np.uint8)
output = cv2.morphologyEx(output, cv2.MORPH_CLOSE, kernel)
#copying image before applying findcontour as findcontour modifies the image
img,contours,hierarchy = cv2.findContours(output, 1, 2)
leftmost = retina_blue.shape;
rightmost = [0,0];
topmost = retina_blue.shape;
bottommost = [0,0];
for item in contours:
area = cv2.contourArea(item)
perimeter = cv2.arcLength(item,True);
if(perimeter !=0):
R= 4*np.pi*area/math.pow(perimeter,2);
if (R>0.3):
new_left = tuple(item[item[:,:,0].argmin()][0])
new_right = tuple(item[item[:,:,0].argmax()][0])
new_top = tuple(item[item[:,:,1].argmin()][0])
new_bottom = tuple(item[item[:,:,1].argmax()][0])
if new_left[0] <= leftmost[0] :
leftmost = new_left
if new_right[0] >= rightmost[0] :
rightmost = new_right
if new_top[1] <= topmost[1] :
topmost = new_top
if new_bottom[1] >= bottommost[1] :
bottommost = new_bottom
centre = [0,0];
centre[1] = int((topmost[1] + bottommost[1])/2)
centre[0] = int((leftmost[0] + rightmost[0])/2)
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(75,75))
radius = math.sqrt(math.pow(topmost[1]-2-centre[1],2) + math.pow(leftmost[0]-2-centre[0],2));
radius = int(radius);
if (radius<25):
radius = 25;
elif (radius>40):
radius = 40;
centre = tuple(centre);
cv2.circle(retina_green,centre, int(radius)+5, (0,0,0), -1);
kernel = np.ones((10,10),np.uint8)
tophat = cv2.morphologyEx(retina_green, cv2.MORPH_TOPHAT, kernel)
tophat = cv2.equalizeHist(tophat)
ret, tophat = cv2.threshold(tophat,tophat.max()-10,255,cv2.THRESH_BINARY)
tophat = cv2.morphologyEx(tophat, cv2.MORPH_CLOSE, kernel)
exudate = copy.copy(tophat);
img,contours,hierarchy = cv2.findContours(tophat, 1, 2)
mask = np.ones(retina_green.shape, dtype="uint8") * 255
# removing false exudates
area = [];
distance = 0;
number = 0;
total_area = 0;
for items in contours:
temp_area = cv2.contourArea(items)
perimeter = cv2.arcLength(items,True)
#print temp_area
if perimeter != 0:
R= 4*np.pi*temp_area/math.pow(perimeter,2);
else :
R=0;
if (R<0.3) or (temp_area>3000) :
cv2.drawContours(mask, [items], -1, 0, -1)
else:
number = number + 1;
M = cv2.moments(items);
if(M['m00'] != 0):
cx = int(M['m10']/M['m00'])
cy = int(M['m01']/M['m00'])
distance = distance + math.sqrt(math.pow(centre[1]-cy,2) + math.pow(centre[0]-cx,2));
area.append(temp_area);
total_area = total_area + temp_area;
exudate = cv2.bitwise_and(exudate,exudate, mask=mask);
exudate = cv2.medianBlur(exudate,5);
test = cv2.bitwise_and(test_green,exudate)
b1 = test.max();
test[(test>(b1-10))] = 0;
b2 = test.max();
test[(test>(b2-10))] = 0;
b3 = test.max();
#tophat = cv2.adaptiveThreshold(tophat,255,cv2.ADAPTIVE_THRESH_MEAN_C,cv2.THRESH_BINARY,301,0)
if(number!=0):
distance = distance/number;
output = [total_area,np.asarray(area).max(),distance,number,b1/a,b2/a,b3/a];
else :
output = [total_area,0,0,number,b1/a,b2/a,b3/a];
return(output);
'''
blood_vessels detection using blackhat operation and then binary thresholding
this function takes coloured retinal image and output the blood vessel features
'''
def blood_vessels(image):
retina = copy.copy(image);
retina_blue , retina_green , retina_red = cv2.split(retina);
clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
retina_green = clahe.apply(retina_green)
kernel = np.ones((10,10),np.uint8)
blackhat = cv2.morphologyEx(retina_green, cv2.MORPH_BLACKHAT, kernel)
#blackhat = cv2.GaussianBlur(blackhat,(15,15),0)
blackhat = cv2.medianBlur(blackhat,5)
blackhat = cv2.equalizeHist(blackhat)
ret, output_green = cv2.threshold(blackhat,blackhat.max()-10,255,cv2.THRESH_BINARY);
blood_vessel = cv2.medianBlur(output_green,5)
vessels = copy.copy(blood_vessel)
mask1 = np.ones(retina_green.shape, dtype="uint8") * 255
img,contours,hierarchy = cv2.findContours(blood_vessel, 1, 2)
for items in contours:
area = cv2.contourArea(items)
if (area < 10):
cv2.drawContours(mask1, [items], -1, 0, -1)
vessel = cv2.bitwise_and(vessels,vessels, mask=mask1);
blood_clot = copy.copy(vessel)
mask2 = np.ones(retina_green.shape, dtype="uint8") * 255
img,contours,hierarchy = cv2.findContours(vessel, 1, 2)
for items in contours:
area = cv2.contourArea(items)
perimeter = cv2.arcLength(items,True)
if perimeter !=0:
R= 4*np.pi*area/math.pow(perimeter,2);
if R>0.5:
cv2.drawContours(mask2, [items], -1, 0, -1)
blood_vessel = cv2.bitwise_and(blood_clot,blood_clot, mask=mask2);
kernel = np.ones((5,5),np.uint8)
blood_vessel = cv2.morphologyEx(blood_vessel, cv2.MORPH_CLOSE, kernel)
hemo = cv2.bitwise_not(mask2);
img,contours,hierarchy = cv2.findContours(blood_vessel, 1, 2)
thickness = [];
number_vessel = len(contours);
total_area = 0;
for items in contours:
area = cv2.contourArea(items)
total_area = total_area + area;
perimeter = cv2.arcLength(items,True)
if(perimeter !=0) :
thickness.append(area/perimeter);
if(len(thickness)!=0):
max_vessel = np.asarray(thickness).max();
else:
max_vessel = 0;
mean = np.average(np.asarray(thickness));
return (mean,max_vessel,number_vessel,total_area);
def Hemorrhages(image):
retina_green = copy.copy(image);
retina_green[(retina_green<20)] = 255;
retina_green = cv2.bitwise_not(retina_green);