def btnVesselSegmentation(self):
"""
Function calls segmentation.vesselSegmentation function.
"""
outputSegmentation = segmentation.vesselSegmentation(
self.oseg.data3d,
self.oseg.segmentation,
threshold=-1,
inputSigma=0.15,
dilationIterations=2,
nObj=1,
biggestObjects=False,
interactivity=True,
binaryClosingIterations=2,
binaryOpeningIterations=0)
# rint outputSegmentation
# rint np.unique(outputSegmentation)
# rint self.oseg.slab
slab = {'porta': 2}
slab.update(self.oseg.slab)
# rom PyQt4.QtCore import pyqtRemoveInputHook
# yqtRemoveInputHook()
# mport ipdb; ipdb.set_trace() # BREAKPOINT
self.oseg.slab = slab
self.oseg.segmentation[outputSegmentation == 1] = slab['porta']
def data_to_binar(self):
# ## Median filter
filteredData = scipy.ndimage.filters.median_filter(self.data3d, size=2)
# ## Segmentation
data3d_thr = segmentation.vesselSegmentation(
filteredData, # self.data3d,
segmentation=np.ones(filteredData.shape, dtype='int8'),
threshold=self.threshold,
inputSigma=0, # 0.15,
dilationIterations=2,
nObj=1,
biggestObjects=False,
useSeedsOfCompactObjects=False,
interactivity=not self.nogui,
binaryClosingIterations=self.binaryClosing, # noqa 5,
binaryOpeningIterations=self.binaryOpening, # 1 # noqa
qapp=self.qapp
)
del(filteredData)
# ## Zalepeni der
scipy.ndimage.morphology.binary_fill_holes(data3d_thr,
output=data3d_thr)
self.data3d_thr = data3d_thr
def hepaticVeinsSegmentation(self):
import segmentation
outputSegmentation = segmentation.vesselSegmentation(
self.data3d,
self.segmentation,
threshold=-1,
inputSigma=0.15,
dilationIterations=10,
nObj=1,
biggestObjects=False,
useSeedsOfCompactObjects=True,
interactivity=True,
binaryClosingIterations=2,
binaryOpeningIterations=0)
slab = {'hepatic_veins': 3}
slab.update(self.slab)
# rom PyQt4.QtCore import pyqtRemoveInputHook
# yqtRemoveInputHook()
# mport ipdb; ipdb.set_trace() # BREAKPOINT
self.slab = slab
self.segmentation[outputSegmentation == 1] = slab['hepatic_veins']
# skeletonizace
self.__vesselTree(outputSegmentation, 'hepatic_veins')
def main4():
data = misc.obj_from_file('c:\_bp_data\d5\org-liver-orig004.mhd-3mm_alpha45.pklz', filetype = 'pickle')
outputTmp = sg.vesselSegmentation(
data['data3d'],
segmentation = data['segmentation'],
threshold = -1,
inputSigma = 0.15,
dilationIterations = 2,
interactivity = True,
biggestObjects = False,
binaryClosingIterations = 2,
binaryOpeningIterations = 0)
def main4():
data = misc.obj_from_file('c:\_bp_data\d5\org-liver-orig004.mhd-3mm_alpha45.pklz', filetype = 'pickle')
outputTmp = sg.vesselSegmentation(
data['data3d'],
segmentation = data['segmentation'],
threshold = -1,
inputSigma = 0.15,
dilationIterations = 2,
interactivity = True,
biggestObjects = False,
binaryClosingIterations = 2,
binaryOpeningIterations = 0)
def main2():
slab = {'none':0, 'liver':1, 'porta':2}
voxelsize_mm = np.array([1.0, 1.0, 1.2])
segm = np.zeros([256, 256, 80], dtype = np.uint8)
# liver
segm[70:180, 40:190, 30:60] = slab['liver']
# port
segm[120:130, 70:190, 40:45] = slab['porta']
segm[80:130, 100:110, 40:45] = slab['porta']
segm[120:170, 130:135, 40:44] = slab['porta']
data3d = np.zeros(segm.shape)
data3d[segm == slab['liver']] = 156
data3d[segm == slab['porta']] = 206
#noise = (np.random.rand(segm.shape[0], segm.shape[1], segm.shape[2])*30).astype(np.int16)
noise = (np.random.normal(0, 30, segm.shape))#.astype(np.int16)
data3d = (data3d + noise).astype(np.uint8)
# @TODO je tam bug, prohl??e? neum? korektn? pracovat s doubly
# app = QApplication(sys.argv)
# #pyed = QTSeedEditor(noise )
# pyed = QTSeedEditor(data3d)
# pyed.exec_()
# #img3d = np.zeros([256,256,80], dtype=np.int16)
# pyed = sed3.sed3(data3d)
# pyed.show()
outputTmp = sg.vesselSegmentation(
data3d,
segmentation = segm == slab['liver'],
voxelsize_mm = voxelsize_mm,
threshold = 180,
inputSigma = 0.15,
dilationIterations = 2,
nObj = 1,
interactivity = True,
biggestObjects = False,
binaryClosingIterations = 5,
binaryOpeningIterations = 1)
if outputTmp == None:
print('Final output is None')
def main2():
slab = {'none':0, 'liver':1, 'porta':2}
voxelsize_mm = np.array([1.0, 1.0, 1.2])
segm = np.zeros([256, 256, 80], dtype = np.uint8)
# liver
segm[70:180, 40:190, 30:60] = slab['liver']
# port
segm[120:130, 70:190, 40:45] = slab['porta']
segm[80:130, 100:110, 40:45] = slab['porta']
segm[120:170, 130:135, 40:44] = slab['porta']
data3d = np.zeros(segm.shape)
data3d[segm == slab['liver']] = 156
data3d[segm == slab['porta']] = 206
#noise = (np.random.rand(segm.shape[0], segm.shape[1], segm.shape[2])*30).astype(np.int16)
noise = (np.random.normal(0, 30, segm.shape))#.astype(np.int16)
data3d = (data3d + noise).astype(np.uint8)
# @TODO je tam bug, prohl??e? neum? korektn? pracovat s doubly
# app = QApplication(sys.argv)
# #pyed = QTSeedEditor(noise )
# pyed = QTSeedEditor(data3d)
# pyed.exec_()
# #img3d = np.zeros([256,256,80], dtype=np.int16)
# pyed = sed3.sed3(data3d)
# pyed.show()
outputTmp = sg.vesselSegmentation(
data3d,
segmentation = segm == slab['liver'],
voxelsize_mm = voxelsize_mm,
threshold = 180,
inputSigma = 0.15,
dilationIterations = 2,
nObj = 1,
interactivity = True,
biggestObjects = False,
binaryClosingIterations = 5,
binaryOpeningIterations = 1)
if outputTmp == None:
print 'Final output is None'
def data_to_binar(self):
### Median filter
filteredData = scipy.ndimage.filters.median_filter(self.data3d, size=2)
### Segmentation
data3d_thr = segmentation.vesselSegmentation(
filteredData,
segmentation=np.ones(self.data3d.shape, dtype='int8'),
threshold=self.threshold, #-1,
inputSigma=0, #0.15,
dilationIterations=2,
nObj=1,
biggestObjects= False,
interactivity= not self.nogui,
binaryClosingIterations=2, #5, # TODO !!! - vytvari na stranach oblasti ktere se pak nenasegmentuji
binaryOpeningIterations=0 #1
)
### NoGUI segmentation fix
# Pri pouziti nogui modu segmentation.py nepouzije funkci getPriorityObjects jelikoz nema seedy => odlisne vysledky.
# Proto musim rucne zjistit vhodny seed a zavolat funkci az ted.
if self.nogui:
# Get seed that is inside of big object
seed_data = scipy.ndimage.morphology.binary_erosion(data3d_thr, iterations=2) # hledam pouze velke cevy
seed = None
for i in xrange(seed_data.shape[0]/3,seed_data.shape[0]-1): # shape/3 aby se zaclo hledat kolem prostredku
for n in xrange(seed_data.shape[1]/3,seed_data.shape[1]-1):
for j in xrange(seed_data.shape[2]/3,seed_data.shape[2]-1):
if seed_data[i][n][j]==1:
seed = (np.array([i]),np.array([n]),np.array([j]))
logger.debug('automatic seed -> '+str(seed)+' value -> '+str(seed_data[i][n][j]))
break
if seed is not None:
break
if seed is not None:
break
# Zaruci ze ve vystupu nebudou cevy ktere vedou od nikud nikam
data3d_thr = thresholding_functions.getPriorityObjects(data3d_thr, nObj=1, seeds=seed)
### Zalepeni der
data3d_thr = scipy.ndimage.morphology.binary_fill_holes(data3d_thr)
return data3d_thr
#oseg.data3d = oseg.data3d[cri[0][0]:cri[0][1],cri[1][0]:cri[1][1],cri[2][0]:cri[2][1]]
#pyed = sed3.sed3(oseg.data3d, contour = oseg.orig_scale_segmentation)
print('slab', data['slab'])
#import ipdb; ipdb.set_trace() # BREAKPOINT
#pyed = sed3.sed3(data['data3d'], contour = data['segmentation'])
#pyed.show()
#import pdb; pdb.set_trace()
outputTmp = segmentation.vesselSegmentation(
data['data3d'],
segmentation=data['segmentation'],
#segmentation = oseg.orig_scale_segmentation,
threshold=-1,
inputSigma=0.15,
dilationIterations=2,
nObj=1,
biggestObjects=args.biggest,
# dataFiltering = True,
interactivity=True,
binaryClosingIterations=2,
binaryOpeningIterations=0)
# Uvolneni pameti
garbage.collect()
import inspector
inspect = inspector.inspector(outputTmp)
output = inspect.run()
# Uvolneni pameti
def test_synthetic_data_segmentation(self):
"""
Function uses organ_segmentation for synthetic box object
segmentation.
"""
#dcmdir = os.path.join(path_to_script,'./../sample_data/matlab/examples/sample_data/DICOM/digest_article/')
# data
slab = {'none':0, 'liver':1, 'porta':2}
voxelsize_mm = np.array([1.0,1.0,1.2])
segm = np.zeros([256,256,80], dtype=np.int16)
# liver
segm[70:180,40:190,30:60] = slab['liver']
# port
segm[120:130,70:190,40:45] = slab['porta']
segm[80:130,100:110,40:45] = slab['porta']
segm[120:170,130:135,40:44] = slab['porta']
data3d = np.zeros(segm.shape)
data3d[segm== slab['liver']] = 156
data3d[segm== slab['porta']] = 206
#noise = (np.random.rand(segm.shape[0], segm.shape[1], segm.shape[2])*30).astype(np.int16)
noise = (np.random.normal(0,30,segm.shape))#.astype(np.int16)
data3d = (data3d + noise ).astype(np.int16)
# @TODO je tam bug, prohlížeč neumí korektně pracovat s doubly
# app = QApplication(sys.argv)
# #pyed = QTSeedEditor(noise )
# pyed = QTSeedEditor(data3d)
# pyed.exec_()
# #img3d = np.zeros([256,256,80], dtype=np.int16)
# pyed = py3DSeedEditor.py3DSeedEditor(data3d)
# pyed.show()
outputTmp = segmentation.vesselSegmentation(
data3d, #.astype(np.uint8),
segmentation = (segm==slab['liver']), #.astype(np.uint8),
#segmentation = oseg.orig_scale_segmentation,
voxelsize_mm = voxelsize_mm,
threshold = 180,
inputSigma = 0.15,
dilationIterations = 2,
nObj = 1,
interactivity = False,
biggestObjects=True,
binaryClosingIterations = 5,
binaryOpeningIterations = 1)
# ověření výsledku
#pyed = py3DSeedEditor.py3DSeedEditor(outputTmp, contour=segm==slab['porta'])
#pyed.show()
# @TODO opravit chybu v vesselSegmentation
outputTmp = (outputTmp == 2)
errim = np.abs(outputTmp.astype(np.int) - (segm == slab['porta']).astype(np.int))
# ověření výsledku
#pyed = py3DSeedEditor.py3DSeedEditor(errim, contour=segm==slab['porta'])
#pyed.show()
#evaluation
sum_of_wrong_voxels = np.sum(errim)
sum_of_voxels = np.prod(segm.shape)
#print "wrong ", sum_of_wrong_voxels
#print "voxels", sum_of_voxels
errorrate = sum_of_wrong_voxels/sum_of_voxels
#import pdb; pdb.set_trace()
self.assertLess(errorrate,0.1)
def portalVeinSegmentation(self, **inparams):
"""
Segmentation of vein in specified volume. It is given by label "liver".
Usualy it is number 1. If there is no specified volume all image is
used.
If this function is used repeatedly (if there is some segmentation in
this image) all segmentation labeled as 'porta' is removed and setted
to 'liver' before processing.
You can use additional parameters from vesselSegmentation()
For example interactivity=False, biggestObjects=True, ...
"""
import segmentation
logger.info('segmentation max label ' + str(np.max(self.segmentation)))
# if there is no organ segmentation, use all image
slab = {'porta': 2}
slab.update(self.slab)
logger.debug('self.slab')
logger.debug(str(self.slab))
self.slab = slab
# if there is no liver segmentation, use whole image
if np.max(self.segmentation) == 0:
self.segmentation = self.segmentation + 1
# remove prev segmentation
self.segmentation[self.segmentation == slab['porta']] = slab['liver']
params = {
'threshold': -1,
'inputSigma': 0.15,
'dilationIterations': 10,
'nObj': 1,
'biggestObjects': False,
'useSeedsOfCompactObjects': True,
'interactivity': True,
'binaryClosingIterations': 2,
'binaryOpeningIterations': 0
}
params.update(inparams)
outputSegmentation = segmentation.vesselSegmentation(
self.data3d,
self.segmentation,
**params
# threshold=-1,
# inputSigma=0.15,
# dilationIterations=10,
# nObj=1,
# biggestObjects=False,
# useSeedsOfCompactObjects=True,
# # useSeedsOfCompactObjects=False,
# interactivity=True,
# binaryClosingIterations=2,
# binaryOpeningIterations=0
)
# rom PyQt4.QtCore import pyqtRemoveInputHook
# yqtRemoveInputHook()
# mport ipdb; ipdb.set_trace() # BREAKPOINT
self.segmentation[outputSegmentation == 1] = slab['porta']
def _interactivity_end(self, igc):
"""
This is called after processing step. All data are rescaled to original
resolution.
"""
logger.debug('_interactivity_end()')
self.__resize_to_orig(igc.seeds)
self.organ_interactivity_counter = igc.interactivity_counter
logger.debug("org inter counter " +
str(self.organ_interactivity_counter))
logger.debug('nonzero segm ' + str(np.nonzero(self.segmentation)))
# if False:
if False:
# TODO dodělat postprocessing PV
import segmentation
outputSegmentation = segmentation.vesselSegmentation( # noqa
self.data3d,
self.segmentation,
threshold=-1,
inputSigma=0.15,
dilationIterations=10,
nObj=1,
biggestObjects=False,
seeds=(self.segmentation > 0).astype(np.int8),
useSeedsOfCompactObjects=True,
interactivity=True,
binaryClosingIterations=2,
binaryOpeningIterations=0)
self._segmentation_postprocessing()
# rint 'autocrop', self.autocrop
if self.autocrop is True:
# rint
# mport pdb; pdb.set_trace()
tmpcrinfo = qmisc.crinfo_from_specific_data(
self.segmentation,
self.autocrop_margin)
self.crop(tmpcrinfo)
if self.texture_analysis not in (None, False):
import texture_analysis
# doplnit nějaký kód, parametry atd
# elf.orig_scale_segmentation =
# texture_analysis.segmentation(self.data3d,
# self.orig_scale_segmentation, params = self.texture_analysis)
self.segmentation = texture_analysis.segmentation(
self.data3d,
self.segmentation,
self.voxelsize_mm
)
# set label number
# !! pomaly!!!
# @TODO make faster
self.segmentation[self.segmentation == 1] = self.output_label
#
logger.debug('self.slab')
logger.debug(str(self.slab))
self.processing_time = (
datetime.datetime.now() - self.time_start).total_seconds()
logger.debug('processing_time = ' + str(self.processing_time))
#oseg.data3d = oseg.data3d[cri[0][0]:cri[0][1],cri[1][0]:cri[1][1],cri[2][0]:cri[2][1]]
#pyed = sed3.sed3(oseg.data3d, contour = oseg.orig_scale_segmentation)
print 'slab', data['slab']
#import ipdb; ipdb.set_trace() # BREAKPOINT
#pyed = sed3.sed3(data['data3d'], contour = data['segmentation'])
#pyed.show()
#import pdb; pdb.set_trace()
outputTmp = segmentation.vesselSegmentation(
data['data3d'],
segmentation = data['segmentation'],
#segmentation = oseg.orig_scale_segmentation,
threshold = -1,
inputSigma = 0.15,
dilationIterations = 2,
nObj = 1,
biggestObjects = args.biggest,
# dataFiltering = True,
interactivity = True,
binaryClosingIterations = 2,
binaryOpeningIterations = 0)
# Uvolneni pameti
garbage.collect()
import inspector
inspect = inspector.inspector(outputTmp)
output = inspect.run()
# Uvolneni pameti
def test_synthetic_data_vessel_tree_evaluation(self):
"""
Generovani umeleho stromu do 3D a jeho evaluace.
V testu dochazi ke kontrole predpokladaneho objemu a deky cev
"""
from gen_volume_tree import TreeVolumeGenerator
from histology_analyser import HistologyAnalyser
from histology_report import HistologyReport
import segmentation
import misc
# generate 3d data from yaml for testing
tvg = TreeVolumeGenerator()
yaml_path = os.path.join(path_to_script, "./hist_stats_test.yaml")
tvg.importFromYaml(yaml_path)
tvg.voxelsize_mm = [1,1,1]
tvg.shape = [100,100,100]
tvg.generateTree()
# init histology Analyser
metadata = {'voxelsize_mm': tvg.voxelsize_mm}
data3d = tvg.data3d*10
threshold = 2.5
ha = HistologyAnalyser(data3d, metadata, threshold)
# modifited ha.data_to_skeleton() function
data3d_thr = segmentation.vesselSegmentation(
data3d,
segmentation=np.ones(tvg.data3d.shape, dtype='int8'),
threshold=threshold,
inputSigma=0.15,
dilationIterations=2,
nObj=1,
biggestObjects=False,
interactivity=False,
binaryClosingIterations=5,
binaryOpeningIterations=1)
data3d_skel = ha.binar_to_skeleton(data3d_thr)
# get statistics
ha.skeleton_to_statistics(data3d_thr, data3d_skel)
yaml_new = "hist_stats_new.yaml"
ha.writeStatsToYAML(filename=yaml_new)
# get histology reports
hr = HistologyReport()
hr.importFromYaml(yaml_path)
hr.generateStats()
stats_orig = hr.stats
hr = HistologyReport()
hr.importFromYaml(os.path.join(path_to_script, yaml_new))
hr.generateStats()
stats_new = hr.stats
# compare
self.assertGreater(stats_orig['Total length mm'],stats_new['Total length mm']*0.9)
self.assertLess(stats_orig['Total length mm'],stats_new['Total length mm']*1.1)
self.assertGreater(stats_orig['Avg length mm'],stats_new['Avg length mm']*0.9)
self.assertLess(stats_orig['Avg length mm'],stats_new['Avg length mm']*1.1)
self.assertGreater(stats_orig['Avg radius mm'],stats_new['Avg radius mm']*0.9)
self.assertLess(stats_orig['Avg radius mm'],stats_new['Avg radius mm']*1.1)
