-
Notifications
You must be signed in to change notification settings - Fork 1
/
runmainWithIsolatedMask.py
341 lines (272 loc) · 9.93 KB
/
runmainWithIsolatedMask.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
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
# -*- coding: utf-8 -*-
"""RunMain.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1kCQxSaqlZvUz19ZfPYekVR6uZCFHbtF6
"""
from google.colab import drive
drive.mount('/content/gdrive')
import numpy as np
import cv2
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import scipy
import os
os.chdir('/content/gdrive/My Drive/MP2/')
def cvt2LAB(img, show):
lab= cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
plt.imshow(lab)
if(show):
plt.show()
l, a, b = cv2.split(lab)
plt.imshow(l, cmap ='gray')
if(show):
plt.show()
plt.imshow(a, cmap= 'gray')
if(show):
plt.show()
plt.imshow(b, cmap = 'gray')
if(show):
plt.show()
labm = cv2.medianBlur(lab, 9)
plt.imshow(labm)
if(show):
plt.show()
return lab, labm, l, a, b
def extractGabor(img, ksizeRange, sigmaRange, thetaRange, gammaRange, lamdaRange, show):
features = []
fmasks = []
dim = len(img.shape)
for ksize in np.arange(ksizeRange[0], ksizeRange[1], ksizeRange[2]):
for sigma in np.arange(sigmaRange[0], sigmaRange[1], sigmaRange[2]):
for lamda in np.arange(lamdaRange[0], lamdaRange[1], lamdaRange[2]):
for gamma in np.arange(gammaRange[0], gammaRange[1], gammaRange[2]):
for theta in np.arange(thetaRange[0], thetaRange[1], thetaRange[2]):
k = cv2.getGaborKernel((ksize, ksize), sigma, theta, lamda, gamma, 0, ktype=cv2.CV_32F)
fimg = cv2.filter2D(img, cv2.CV_8UC1, k)
fimg = cv2.medianBlur(fimg, 5)
#fimg = cv2.GaussianBlur(fimg, (51,51), 5, 5)
if(show):
plt.figure(figsize = (8,8))
plt.imshow(fimg, cmap = 'gray')
plt.show()
print(ksize, sigma, gamma, lamda, theta)
if(dim == 2):
#ret,fmask = cv2.threshold(fimg,254,1,cv2.THRESH_BINARY_INV)
#fmask = fmask.reshape((fimg.shape[0]*fimg.shape[1],))
fimg = fimg.reshape((fimg.shape[0]*fimg.shape[1],))
features.append(fimg)
#features.append(fmask)
elif(dim == 3):
fimg = fimg.reshape((fimg.shape[0]*fimg.shape[1],3))
features.append(fimg[:,0])
features.append(fimg[:,1])
features.append(fimg[:,2])
else:
print("Channels error")
return 0
features = np.array(features)
features = features.T
return features
def addLABfeatures(features, labm):
features = np.hstack((features,labm[:,:,0].reshape((labm.shape[0]*labm.shape[1]),1)))
features = np.hstack((features,labm[:,:,1].reshape((labm.shape[0]*labm.shape[1]),1)))
features = np.hstack((features,labm[:,:,2].reshape((labm.shape[0]*labm.shape[1]),1)))
return features
def kmeansClustering(features, n_clusters):
kmeans = KMeans(n_clusters=n_clusters, init = 'k-means++')
kmeans.fit(features)
y = kmeans.predict(features)
return kmeans, y
#Find Background Class as most populated class and setting it to zero
def bgToZero(img_seg):
counts = np.bincount(img_seg.flatten())
background = np.argmax(counts)
if(background):
print("Changing BG")
img_seg[img_seg == background] = 255
img_seg[img_seg == 0] = background
img_seg[img_seg == 255] = 0
return img_seg
#Finding Cell and Cytoplasm Clusters
def findCorrectLabel(img_seg, lab):
clusters = np.unique(img_seg)
clusters = clusters[1:] #As label zero is background
#Masks for each label
mask1 = np.zeros(img_seg.shape)
mask2 = np.zeros(img_seg.shape)
mask1[img_seg == clusters[0]] = 255
mask2[img_seg == clusters[1]] = 255
#Finding histogram for a space
hist1 = cv2.calcHist([lab],[1],np.uint8(mask1),[255],[0,256])
peak1 = np.argsort(-hist1.flatten())[0]
hist2 = cv2.calcHist([lab],[1],np.uint8(mask2),[255],[0,256])
peak2 = np.argsort(-hist2.flatten())[0]
if(peak1 > peak2):
cell = clusters[0]
cyto = clusters[1]
else:
cell = clusters[1]
cyto = clusters[0]
return cell, cyto
def segmentLabel(img, img_seg, cell, cyto, background): #Re -labels the image to group tumors and isolated cells separately
#Input no. of malignant and benign
print("Malignant:")
mal = input()
print("Benign: ")
ben = input()
img_seg_labelled = findTumors(img_seg, cell, cyto, background, int(mal), int(ben))
#print(img_seg_labelled.dtype)
#plt.imshow(img_seg_labelled, cmap = 'gray')
#plt.show()
return img_seg_labelled, mal, ben
def findTumors(img_seg, cell, cyto, background, malignant, benign):
#Finding contours and respective areas
contours, heirarchy = cv2.findContours(np.uint8(img_seg), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
n = len(contours)
areas = []
for cnt in contours:
area = cv2.contourArea(cnt)
areas.append(area)
#Sorting areas in descending order
areas = np.array(areas)
ind = np.argsort(-1*areas)
#Getting indices of tumours
numTumors = malignant + benign
tumorInd = ind[0:numTumors]
z = np.zeros(img_seg.shape)
#Creating a mask with tumors of largest area
for i in tumorInd:
cv2.drawContours( z, contours[i], -1, (255,255,255), 3)
masked_image = scipy.ndimage.morphology.binary_fill_holes(z)
z[masked_image] = img_seg[masked_image]
z = np.int64(z)
counts = np.bincount(z.flatten())
if(len(counts)-1 < cell or len(counts)-1 <cyto):
counts = np.append(counts, [0])
if(counts[cell] <= counts[cyto]):
img_seg[z == cyto] = background
img_seg = findTumors(img_seg, cell, cyto, background, malignant, benign)
else:
img_seg[img_seg == cyto] = background
#if enhancing
img_seg[img_seg == cell] = background
img_seg[masked_image] = cyto
return img_seg
def enhance(lab):
l,a,b = cv2.split(lab)
clahe = cv2.createCLAHE(clipLimit=7.0, tileGridSize=(8,8))
cla = clahe.apply(a)
#plt.imshow(cla)
#plt.show()
#-----Merge the CLAHE enhanced L-channel with the a and b channel-----------
limg = cv2.merge((l,cla,b))
#plt.imshow(limg)
#plt.show()
return cv2.cvtColor(limg, cv2.COLOR_LAB2BGR)
def getIsolatedCells(imgEnh, *args):
if(len(args)):
thresh = arg[0]
else:
thresh = 50 #After Histogram Analysis
isoMask = imgEnh[:,:,1] < 50
plt.imshow(isoMask)
plt.show()
return isoMask
def diceScore(num_clusters, img_seg, img_gt):
num_clusters = 4
gray_values=[]
img_shape = img_gt_gray.shape
img_gt_gray.reshape(img_shape[0]*img_shape[1])
isolatedMask.reshape(img_shape[0]*img_shape[1])
for i in range(num_clusters-1):
gray_values.append(int(np.mean(img_gt_gray[(img_gt_gray>=255*i/4) & (img_gt_gray<255*(i+1)/4)])))
gray_values.append(255)
print(gray_values)
for i in range(num_clusters):
img_gt_gray[img_gt_gray==gray_values[i]]=i
gt_pixel_ratios=[]
seg_pixel_ratios=[]
for i in range(num_clusters):
seg_pixel_ratios.append(np.sum(isolatedMask[isolatedMask==i]==i))
gt_pixel_ratios.append(np.sum(img_gt_gray[img_gt_gray==i]==i))
print(seg_pixel_ratios)
print(gt_pixel_ratios)
seg_order = np.argsort(seg_pixel_ratios)
print(seg_order)
gt_order = np.argsort(gt_pixel_ratios)
print(gt_order)
for i in range(num_clusters):
isolatedMask[isolatedMask==seg_order[i]]= -gt_order[i]
for i in range(num_clusters):
isolatedMask[isolatedMask==(-i)]= i
isolatedMask.reshape((img_shape[0],img_shape[1]))
plt.imshow(isolatedMask,cmap='gray')
img_gt_gray.reshape((img_shape[0],img_shape[1]))
dice = []
for k in range(num_clusters):
dice.append(np.sum(isolatedMask[img_gt_gray==k]==k)*2.0 / (np.sum(isolatedMask[isolatedMask==k]==k) + np.sum(img_gt_gray[img_gt_gray==k]==k)))
print(dice)
def main(new, img_fname, *args):
#Reading images
img = cv2.imread(img_fname)
if(not new): #If not a new image
img_gt_fname = args[0]
img_gt = cv2.imread(img_gt_fname)
img_gt = np.array(img_gt)
#Convert to Lab Space
# Outputs : lab space, median filtered lab images, l, a, b channels separately
# Inputs : rgb image, show = 1, to show the lab images. Show = 0 to not display
lab,labm, l, a, b = cvt2LAB(img,0)
#Extract Gabor Features
# FN: extractGabor(img, ksizeRange, sigmaRange, thetaRange, gammaRange, lamdaRange, show)
features = extractGabor(lab, [9,10,2], [1,2,1], [0,1,1], [0.5,1.25,0.25], [3.25,4.1,0.25], 0)
#Feature Selection
features = np.hstack([features[:,0:12], features[:,18:36]])
#Adding LAB Colour channels to feature space
#features = addLABfeatures(features, labm)
#Kmeans - inputs : feature vectors, no. of clusters
#Kmeans - outputs : kmeans object, label vector
[kmeans, y] = kmeansClustering(features, 3)
img_seg = y.reshape((img.shape[0], img.shape[1]))
img_seg = cv2.medianBlur(np.uint8(img_seg), 9)
#Display Image
plt.figure()
plt.imshow(img)
plt.title('Original image', fontsize = 20)
plt.show()
plt.figure()
plt.imshow(img_seg , cmap = 'gray')
plt.title('Segmented image after kmeans', fontsize = 20)
plt.show()
#Re-labelling the image
img_seg = bgToZero(img_seg)
cell, cyto = findCorrectLabel(img_seg, lab) #Finds the correct labels for cell and cytoplasm clusters
#Finding the tumors
img_seg_copy = np.copy(img_seg)
img_seg_withTumor, no_malig, no_benign = segmentLabel(img, img_seg_copy, cell, cyto, 0)
#Find Isolated Cells
imgEnh = enhance(lab)
isoMask = getIsolatedCells(imgEnh)
tumorMask = img_seg_withTumor == cyto
img_seg_withTumor[isoMask] = cell
img_seg_withTumor[tumorMask] = cyto
#Dice Score
if(not new):
fig2, (a3,a4) = plt.subplots(1,2, figsize=(15,15))
a3.imshow(img_seg_withTumor, cmap = 'gray')
a3.set_title('Final Segmented Image Output', fontsize = 20)
a4.imshow(img_gt)
a4.set_title('Ground Truth', fontsize = 20)
fig2.show()
#if(no_malig):
#dice = diceScore(4, img_seg_withTumor, img_gt)
#else:
#dice = diceScore(3, img_seg_withTumor, img_gt)
else:
plt.figure(figsize = (10,10))
plt.imshow(img_seg_withTumor, cmap = 'gray')
plt.title('Final Segmented Image Output', fontsize = 15)
plt.show()
return 0
main(0, 'benign_4x/_11064.tif', 'benign_4x/_11064_gt.png')