-
Notifications
You must be signed in to change notification settings - Fork 0
/
biopore_detect.py
313 lines (244 loc) · 12.7 KB
/
biopore_detect.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
# coding: utf-8
import numpy as np
import matplotlib.pyplot as plt
import scipy.ndimage as snd
import seaborn as sns
from skimage import img_as_float, morphology, measure
from skimage.color import rgb2hsv
from skimage.morphology import reconstruction
from skimage.exposure import rescale_intensity
from skimage.measure import label
from astropy.table import Table
from scipy import spatial
from skimage.filters import sobel
from skimage.feature import peak_local_max
def biop_det(fi, mp_threshold, patch_threshold, perc,px, plot=True, morph=False,testing=True):
"""
Function for detecting biopores to analyse their spatial arrangement & matrix interaction
Line 105:134 are adapted from the preprocessor of the echoRD-model by C. Jackisch.
For further informations see: https://github.com/cojacoo/echoRD_model/tree/master/echoRD
file "macropore_ini.py"
Parameters
----------
fi : input image ('.png'-format, either as rgb or rgba image)
mp_threshold : lower limit for removing small macropores
patch_threshold : min [0] and max [1] of the desired patch
size limits (usually min=100,max=10000)
perc : value up to which percentile gray values among to biopores
(0.125 shows good results for brighter soil matrix)
px : actual length of one pixel in input image [mm]
plot : True/False: whether results should be plotted (default:True)
morph : if True the morphology of detected biopores will be plotted, otherwise pores are displayed as
scatterplot and distinguished whether stained or not (default)
testing : if True no distances are calculated and only the detected macropores are
plotted to reduce computing time during threshold adjustment (default),
otherwise all distances are computed
Output
------
Dictionary with following keys:
'biopores' : labeled biopores
'biopores_centroidxy' : x/y-coordinates of detected biopores
'biopores_stained_centroidxy' : x/y-coordinates of detected stained biopores
'biopores_area' : area of detected biopores (number of pixels)
'biopores_diameter' : diameter of detected biopores (diameter of circle with same area [mm])
'distance_matrix_biopore' : distance of each image pixel to nearest biopore [mm]
'distance_matrix_stained_biopore' : distance of each image to nearest stained biopore [mm]
'biopore_matrix_interaction' : distance of pixels from stained patches including at least one
biopore to nearest stained biopore [mm] (estimation of biopore-matrix interaction)
'stained_patches' : labeled blue-stained patches
'patches_with_biopores' : detected blue-stained patches including at least one biopore
'table' : summary table with number and main propertiesd of detected biopores
'stained_index' : index of stained biopores
'unstained_index' : index of unstained biopores
"""
im_raw = snd.imread(fi) # load image
sim = np.shape(im_raw)
if sim[2]==4:
imrgb=im_raw[:,:,:3]
else:
imrgb=im_raw
imhsv = rgb2hsv(imrgb) # convert RGB image to HSV color-space
img = imhsv[:,:,2] # extract value channel
im = img_as_float(imrgb) # load image as float for detection of stained patches
sim = np.shape(im) # extract dimensions of input image
# morphological reconstruction for detecting holes inside the picture (according to general example
# "filling holes and detecting peaks" from scikit-image http://scikit-image.org/docs/dev/auto_examples/features_detection/plot_holes_and_peaks.html#sphx-glr-auto-examples-features-detection-plot-holes-and-peaks-py)
seed = np.copy(img)
seed[1:-1, 1:-1] = img.max()
mask = img
filled = reconstruction(seed, mask, method='erosion')
holes=img-filled
# rescale and extract macropores
holes_resc=rescale_intensity(holes,out_range=(0.0,1))
thresh=np.percentile(holes_resc,perc)
holes_resc[holes_resc>thresh]=1
holes_resc[holes_resc<thresh]=0
bp_label=label(holes_resc,neighbors=8, background=1)
bp_label[bp_label==-1]=0
# remove objects smaller than threshold
bp_label_clean = morphology.remove_small_objects(bp_label, min_size=mp_threshold)
# detect and label blue stained patches
# calculate difference of channels to extract blue stained patches
dim=abs(im[:,:,1]-im[:,:,0])
# discard low contrasts
dim[dim<0.2]=0.0
# filter to local maxima for further segmentation
# process segmentation according to sobel function of skimage
image_max = snd.maximum_filter(dim, size=5, mode='constant')
elevation_map = sobel(dim)
markers = np.zeros_like(dim)
markers[image_max < 0.1] = 2
markers[image_max > 0.2] = 1
segmentation = morphology.watershed(elevation_map, markers)
segmentation = snd.binary_fill_holes(1-(segmentation-1))
# clean patches below theshold
patches_cleaned = morphology.remove_small_objects(segmentation, patch_threshold[0])
labeled_patches = label(patches_cleaned)
sizes = np.bincount(labeled_patches.ravel())[1:] #first entry (background) discarded
# reanalyse for large patches and break them by means of watershed segmentation
idx=np.where(sizes>patch_threshold[1])[0]+1
labeled_patches_large=labeled_patches*0
idy=np.in1d(labeled_patches,idx).reshape(np.shape(labeled_patches))
labeled_patches_large[idy]=labeled_patches[idy]
distance = snd.distance_transform_edt(labeled_patches_large)
footp=int(np.round(np.sqrt(patch_threshold[1])/100)*100)
local_maxi = peak_local_max(distance, indices=False, footprint=np.ones((footp, footp)),labels=labeled_patches_large)
markers = snd.label(local_maxi)[0]
labels_broken_large = morphology.watershed(-distance, markers, mask=labeled_patches_large)
labeled_patches[idy]=labels_broken_large[idy]+np.max(labeled_patches)
# measure regionproperties of biopores
meas_bp=measure.regionprops(bp_label_clean, intensity_image=None)
bp_labels = np.unique(bp_label_clean)[1:]
bp_centroidx = bp_labels.astype(np.float64)
bp_centroidy = bp_labels.astype(np.float64)
bp_area = bp_labels.astype(np.float64)
bp_diameter = bp_labels.astype(np.float64)
# extract regionprops for each labeled biopore
for i in np.arange(len(bp_labels)):
bp_centroidx[i], bp_centroidy[i]=meas_bp[i]['centroid']
bp_area[i]=(meas_bp[i]['area'])
bp_diameter[i]=(meas_bp[i]['equivalent_diameter'])*px
bp_centroidxy = np.stack((bp_centroidx,bp_centroidy), axis=-1)
# extract biopores inside stained areas = "stained biopores"
stain_info=np.zeros(len(bp_centroidxy))
rbp_centroidxy=np.around(bp_centroidxy).astype(int)
for i in np.arange(len(bp_centroidxy)):
if labeled_patches[rbp_centroidxy[i,0],rbp_centroidxy[i,1]]>0:
stain_info[i]=1
else:
stain_info[i]=2
stained=np.where(stain_info==1)
unstained=np.where(stain_info==2)
# select value of stained patches including an biopore
bp_stained=np.around(bp_centroidxy[stained]).astype(int)
label_value=np.zeros(len(bp_stained)).astype(int)
for i in np.arange(len(bp_stained)):
label_value[i]=labeled_patches[bp_stained[i,0], bp_stained[i,1]]
# remove labeled patches without any biopore
label_withbp=np.copy(labeled_patches)
for i in np.arange(len(label_value)):
label_withbp[label_withbp==label_value[i]]=-1
label_withbp[label_withbp!=-1]=0
label_withbp[label_withbp==-1]=1
# distance calculations
if testing==False:
# Compute Euclidian distance for each pixel to nearest biopore
m_bp_dist = np.zeros((sim[0],sim[1]))
for i in np.arange(sim[0]):
for j in np.arange(sim[1]):
matrixp1=[i,j]
m_bp_dist[i,j]=spatial.KDTree(bp_centroidxy).query(matrixp1,p=2)[0]
# compute Euclidian distance for each pixel to nearest stained biopore
m_stbp_dist=np.zeros((sim[0],sim[1]))
for i in np.arange(sim[0]):
for j in np.arange(sim[1]):
matrixp1=[i,j]
m_stbp_dist[i,j]=spatial.KDTree(bp_centroidxy[stained]).query(matrixp1,p=2)[0]
# compute Euclidian distance to nearest stained biopore for each pixel of stained areas including a biopore ~ biopore-matrix interaction
stp_stbp_dist = np.zeros((sim[0],sim[1]))
for i in np.arange(sim[0]):
for j in np.arange(sim[1]):
if label_withbp[i,j]!=0:
matrixp3=[i,j]
stp_stbp_dist[i,j,]=spatial.KDTree(bp_centroidxy[stained]).query(matrixp3,p=2)[0]
else:
stp_stbp_dist[i,j]=np.nan
# table for comparison
sbp_diameter=bp_diameter[stained]
t1='All','Stained'
t2=len(bp_diameter),len(sbp_diameter)
t3 = len(bp_diameter[bp_diameter<2]),len(sbp_diameter[sbp_diameter<2])
t4 = len(bp_diameter[bp_diameter>=6]),len(sbp_diameter[sbp_diameter>=6])
t5 =len(bp_diameter[bp_diameter>=2]),len(sbp_diameter[sbp_diameter>=2])
attr=[t1,t2,t3,np.subtract(t5,t4),t4]
bp_t=Table(attr,names=('Properties','Sum','<2mm','2-6mm','>6mm'),meta=None)
# plot results
if plot==True:
#colors for plot
from matplotlib.colors import ListedColormap
ghostwhite=(248/255,248/255,255/255)
blue=(31/255,119/255,180/255)
cmap=ListedColormap([ghostwhite, blue])
if testing==False:
# flatten arrays for kernel density estimate plot
m_bp_distall=np.ravel(m_bp_dist*px)
m_stbp_distall=np.ravel(m_stbp_dist*px)
stp_stbp_distall=np.ravel(stp_stbp_dist*px)
#plot
sns.set_style("white")
plt.figure(figsize=(15,4))
ax1=plt.subplot(131)
plt.imshow(imrgb)
plt.axis('off')
plt.title('Input image')
plt.subplot(132,sharex=ax1, sharey=ax1)
plt.imshow(labeled_patches, vmin=0, vmax=1, cmap=cmap)
plt.imshow(imrgb, alpha=0.5)
if morph==True:
plt.imshow(bp_label_clean, vmin=0, vmax=1,cmap='binary', alpha=0.5)
else:
plt.scatter(bp_centroidxy[unstained][:,1],bp_centroidxy[unstained][:,0] ,color='black', s=10,label='unstained')
plt.scatter(bp_centroidxy[stained][:,1], bp_centroidxy[stained][:,0] ,color='red', s=15, label='stained')
plt.legend(bbox_to_anchor=[0.8,0], ncol=2)
plt.axis('off')
plt.title('Labeled patches & Biopores')
plt.subplot(133)
sns.kdeplot(m_bp_distall, cut=0, label='All pores')
if len(stained[0])>0:
sns.kdeplot(m_stbp_distall, cut=0, label='Stained pores' ,alpha=0.5)
sns.kdeplot(stp_stbp_distall[~np.isnan(stp_stbp_distall)], cut=0, label='Biopore-matrix interaction' ,alpha=0.5)
plt.title('Frequency distribution of calculated distances')
plt.show()
print(bp_t)
else:
#plot
sns.set_style("white")
plt.figure(figsize=(12,5))
ax1=plt.subplot(121)
plt.imshow(imrgb)
plt.axis('off')
plt.title('Input image')
plt.subplot(122, sharex=ax1, sharey=ax1)
plt.imshow(labeled_patches, vmin=0, vmax=1, cmap=cmap)
plt.imshow(imrgb, alpha=0.5)
if morph==True:
plt.imshow(bp_label_clean, vmin=0, vmax=1,cmap='binary', alpha=0.5)
else:
plt.scatter(bp_centroidxy[unstained][:,1],bp_centroidxy[unstained][:,0] ,color='black', s=10,label='unstained')
plt.scatter(bp_centroidxy[stained][:,1], bp_centroidxy[stained][:,0] ,color='red', s=15, label='stained')
plt.legend(bbox_to_anchor=[0.8,0], ncol=2)
plt.axis('off')
plt.title('Labeled patches & Biopores')
plt.show()
print(bp_t)
# results for output
bp_res={}
bp_res['biopores'], bp_res['stained_patches'],bp_res['patches_with_biopores']=bp_label_clean, labeled_patches, label_withbp
bp_res['biopores_centroidxy']=bp_centroidxy
bp_res['biopores_stained_centroidxy']=bp_centroidxy[stained]
bp_res['biopores_area'], bp_res['biopores_diameter']=bp_area, bp_diameter
if testing==False:
bp_res['distance_matrix_biopore'], bp_res['distance_matrix_stained_biopore'], bp_res['biopore_matrix_interaction']=m_bp_dist*px, m_stbp_dist*px, stp_stbp_dist*px
bp_res['table']=bp_t
bp_res['stained_index'], bp_res['unstained_index']=stained, unstained
return bp_res