def _automatic_localization(self):
"""
Automatic localization made by Tomas Ryba.
"""
# seeds = self.get_seeds_using_class_1(class1)
liver = self.data3d * (self.segmentation != 0)
print 'analyzing histogram...'
class1 = tools.analyse_histogram(self.data3d,
roi=self.segmentation != 0)
# py3DSeedEditor.py3DSeedEditor(self.data3d, seeds=class1).show()
print 'getting seeds...'
seeds = self.get_seeds_using_prob_class1(
liver,
class1,
thresholdType='percOfMaxDist',
percT=0.3)
# py3DSeedEditor.py3DSeedEditor(self.data3d, seeds=seeds).show()
print('Starting random walker...')
rw = random_walker(liver, seeds, mode='cg_mg')
print('...finished.')
label_l = self.data['slab']['lesions']
lessions = rw == 2
py3DSeedEditor.py3DSeedEditor(self.data3d, contour=lessions).show()
lessions = self.filter_objects(lessions)
self.segmentation = np.where(lessions, label_l, self.segmentation)
def main():
logger = logging.getLogger()
logger.setLevel(logging.WARNING)
ch = logging.StreamHandler()
logger.addHandler(ch)
#logger.debug('input params')
data3d_a_path = None # os.path.join(path_to_script, '../../../data/medical/data_orig/sliver07/training-part1/liver-orig001.mhd')
data3d_b_path = None #os.path.join(path_to_script, '../../../data/medical/data_orig/sliver07/training-part1/liver-seg001.mhd')
parser = argparse.ArgumentParser(
description='Information about pkl/pklz file')
parser.add_argument('pklzFile',
help='path to data' )
args = parser.parse_args()
data = misc.obj_from_file(args.pklzFile, 'pickle')
print data
try:
pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'], contour=data['segmentation'])
pyed.show()
except:
try:
pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'])
pyed.show()
except:
print "Problem with visualization"
def main():
logger = logging.getLogger()
logger.setLevel(logging.WARNING)
ch = logging.StreamHandler()
logger.addHandler(ch)
# logger.debug('input params')
# input parser
parser = argparse.ArgumentParser(
description='Module for segmentation of simple anatomical structures')
parser.add_argument('-i', '--inputfile', default='vessels.pkl',
help='path to data dir')
# args = parser.parse_args()
# horsi kvalita segmentace
# datapath = os.path.join(path_to_script, "../vessels1.pkl")
# hypodenzni meta
# datapath = os.path.join(path_to_script, args.inputfile)
# horsi kvalita segmentace
# datapath = os.path.join(path_to_script, "../organ.pkl")
# data = misc.obj_from_file(args.inputfile, filetype = 'pickle')
dcmdir = '/home/tomas/Dropbox/Work/Data/medical/org-38289898-export1.pklz'
data = misc.obj_from_file(dcmdir, filetype='pickle')
# windowing
data['data3d'] = tools.windowing(data['data3d'], level=50, width=350)
# data['data3d'] = smoothing(data['data3d'], sliceId=0)
# smoothing ----------------
# bilateral
# data['data3d'] = tools.smoothing_bilateral(data['data3d'],
# sigma_space=15, sigma_color=0.05, sliceId=0)
# more bilateral
# data['data3d'] = tools.smoothing_bilateral(data['data3d'],
# sigma_space=15, sigma_color=0.1, sliceId=0)
# total-variation
data['data3d'] = tools.smoothing_tv(data['data3d'], weight=0.05, sliceId=0)
# more total-variation
# data['data3d'] = tools.smoothing_tv(data['data3d'], weight=0.2,
# multichannel=False, sliceId=0)
# py3DSeedEditor.py3DSeedEditor(data['data3d']).show()
# ---------------------------gym
tumory = Lesions()
# tumory.overlay_test()
tumory.import_data(data)
tumory.automatic_localization()
tumors = tumory.segmentation == tumory.data['slab']['lesions']
py3DSeedEditor.py3DSeedEditor(tumory.data3d, contour=tumors).show()
def one_experiment_setting_for_whole_dataset(inputdata, tile_shape,
feature_fcn, classif_fcn, train,
visualization=False):
fvall = []
fv_tiles = []
indata_len = len(inputdata['data'])
#indata_len = 3
for i in range(0, indata_len):
data3d_orig, data3d_seg, voxelsize_mm = read_data_orig_and_seg(
inputdata, i)
if visualization:
pyed = py3DSeedEditor.py3DSeedEditor(data3d_orig,
contour=data3d_seg)
pyed.show()
#import pdb; pdb.set_trace()
#fvall.insert(i, get_features(
# data3d_orig,
# data3d_seg,
# visualization=args.visualization
# ))
#feature_fcn = feat_hist
fv_t = get_features_in_tiles(data3d_orig, data3d_seg, tile_shape,
feature_fcn)
cidxs, features_t, seg_cover_t = fv_t
if train is True:
labels_train_lin_float = np.array(seg_cover_t)
labels_train_lin = labels_train_lin_float > 0.5
#from sklearn import svm
#clf = svm.SVC()
clf = classif_fcn()
clf.fit(features_t, labels_train_lin)
labels_lin = clf.predict(features_t)
d_shp = data3d_orig.shape
labels = arrange_to_tiled_data(cidxs, tile_shape, d_shp,
labels_lin)
#ltl = (labels_train_lin_float * 10).astype(np.int8)
#labels_train = arrange_to_tiled_data(cidxs, tile_shape,
# d_shp, ltl)
#pyed = py3DSeedEditor.py3DSeedEditor(labels_train, contour=labels)
if visualization:
pyed = py3DSeedEditor.py3DSeedEditor(data3d_seg, contour=labels)
pyed.show()
fv_tiles.insert(i, fv_t)
# @TODO vracet něco inteligentního, fvall je prázdný
return fvall
def show(data3d_a_path, sliver_seg_path, ourSegmentation):
reader = datareader.DataReader()
#data3d_a_path = os.path.join(path_to_script, data3d_a_path)
datap_a = reader.Get3DData(data3d_a_path,
dataplus_format=True)
if 'orig_shape' in datap_a.keys():
# pklz
data3d_a = qmisc.uncrop(datap_a['data3d'], datap_a['crinfo'],
datap_a['orig_shape'])
else:
#dicom
data3d_a = datap_a['data3d']
if sliver_seg_path is not None:
sliver_seg_path = os.path.join(path_to_script, sliver_seg_path)
sliver_datap = reader.Get3DData(sliver_seg_path,
dataplus_format=True)
if 'segmentation' in sliver_datap.keys():
sliver_seg = sliver_datap['segmentation']
sliver_seg = qmisc.uncrop(sliver_datap['segmentation'],
sliver_datap['crinfo'],
data3d_a.shape)
else:
sliver_seg = sliver_datap['data3d']
pyed = py3DSeedEditor.py3DSeedEditor(data3d_a, contour=sliver_seg)
print "Sliver07 segmentation"
pyed.show()
if ourSegmentation != None:
ourSegmentation = os.path.join(path_to_script, ourSegmentation)
datap_our = reader.Get3DData(ourSegmentation, dataplus_format=True)
#data_our = misc.obj_from_file(ourSegmentation, 'pickle')
#data3d_our = data_our['segmentation']
our_seg = qmisc.uncrop(datap_our['segmentation'], datap_our['crinfo'],
data3d_a.shape)
if ourSegmentation != None:
pyed = py3DSeedEditor.py3DSeedEditor(data3d_a, contour=our_seg)
print "Our segmentation"
pyed.show()
if (ourSegmentation is not None) and (sliver_seg_path is not None):
diff = (our_seg.astype(np.int8) - sliver_seg)
diff[diff==-1] = 2
#import ipdb; ipdb.set_trace() # BREAKPOINT
pyed = py3DSeedEditor.py3DSeedEditor(data3d_a, contour=our_seg,
seeds=diff)
print "Sliver07 and our segmentation differences"
pyed.show()
def show_segmentation(self, img = None):
if img == None:
img = self.segmentation
pyed = py3DSeedEditor.py3DSeedEditor(img)
pyed.show()
def one_experiment_setting_for_whole_dataset(inputdata, tile_shape,
feature_fcn, classif_fcn, train,
visualization=False):
fvall = []
fv_tiles = []
indata_len = len(inputdata['data'])
indata_len = 3
for i in range(0, indata_len):
data3d_orig, data3d_seg = read_data_orig_and_seg(inputdata, i)
feat_hist_by_segmentation(data3d_orig, data3d_seg, visualization)
if visualization:
pyed = py3DSeedEditor.py3DSeedEditor(data3d_orig,
contour=data3d_seg)
pyed.show()
#import pdb; pdb.set_trace()
#fvall.insert(i, get_features(
# data3d_orig,
#ltl = (labels_train_lin_float * 10).astype(np.int8)
#labels_train = arrange_to_tiled_data(cidxs, tile_shape,
# d_shp, ltl)
#pyed = py3DSeedEditor.py3DSeedEditor(labels_train, contour=labels)
# @TODO vracet něco inteligentního, fvall je prázdný
return fvall
def main():
logger = logging.getLogger()
logger.setLevel(logging.WARNING)
ch = logging.StreamHandler()
logger.addHandler(ch)
#logger.debug('input params')
# input parser
parser = argparse.ArgumentParser(
description='Module for segmentation of simple anatomical structures')
parser.add_argument('-i', '--inputfile',
default='organ.pkl',
help='path to data dir')
args = parser.parse_args()
data = misc.obj_from_file(args.inputfile, filetype = 'pickle')
data3d = data['data3d']
voxelsize_mm = data['voxelsize_mm']
ss = SimpleSegmentation()
simple_seg = ss.simple_segmentation(data3d, voxelsize_mm)
#visualization
pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'], seeds=simple_seg)
pyed.show()
# save
savestring = raw_input('Save output data? (y/n): ')
if savestring in ['Y', 'y']:
misc.obj_to_file(data, "resection.pkl", filetype='pickle')
def resection(data):
#pyed = py3DSeedEditor.py3DSeedEditor(data['segmentation'])
#pyed.show()
# vessels = get_biggest_object(data['segmentation'] == data['slab']['porta'])
vessels = data['segmentation'] == data['slab']['porta']
# ostranění porty z více kusů, nastaví se jim hodnota liver
#data['segmentation'][data['segmentation'] == data['slab']['porta']] = data['slab']['liver']
#show3.show3(data['segmentation'])
import pdb; pdb.set_trace()
#data['segmentation'][vessels == 1] = data['slab']['porta']
#segm = data['segmentation']
#pyed = py3DSeedEditor.py3DSeedEditor(vessels)
#pyed.show()
print ("Select cut")
lab = vessel_cut.cut_editor_old(data)
pyed = py3DSeedEditor.py3DSeedEditor(lab )#, contour=segm)
pyed.show()
l1 = 1
l2 = 2
#import pdb; pdb.set_trace()
# dist se tady počítá od nul jenom v jedničkách
dist1 = scipy.ndimage.distance_transform_edt(lab != l1)
dist2 = scipy.ndimage.distance_transform_edt(lab != l2)
#segm = (dist1 < dist2) * (data['segmentation'] != data['slab']['none'])
segm = (((data['segmentation'] != 0) * (dist1 < dist2)).astype('int8') + (data['segmentation'] != 0).astype('int8'))
# vizualizace 1
# pyed = py3DSeedEditor.py3DSeedEditor(segm)
# pyed.show()
# import pdb; pdb.set_trace()
# pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'], contour=segm)
# pyed.show()
# import pdb; pdb.set_trace()
# vizualizace 2
linie = np.abs(dist1 - dist2) < 1
pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'], contour = data['segmentation']==data['slab']['liver'] ,seeds = linie)
pyed.show()
def interactivity(self):
"""
Interactive seed setting with 3d seed editor
"""
pyed = py3DSeedEditor.py3DSeedEditor(self.img,
mouse_button_map = self.editor_mouse_button_map )
pyed.show()
#scipy.io.savemat(args.outputfile,{'data':output})
#pyed.get_seed_val(1)
# control list of non zero values in seeds
#nzero_seeds_prev = ([],[])
nzero_seeds = pyed.seeds.nonzero()
pocitadlo = 0
opakovat = True
while opakovat:#(nzero_seeds_prev != nzero_seeds).all():
pocitadlo = pocitadlo + 1
self.voxels1 = pyed.get_seed_val(1)
self.voxels2 = pyed.get_seed_val(2)
self.seeds = pyed.seeds
self.make_gc()
# new pyeditor is created, seeds must be setted
pyed = py3DSeedEditor.py3DSeedEditor(self.img, seeds = self.seeds,
contour = self.segmentation,
mouse_button_map = self.editor_mouse_button_map
)
#pyed.seeds = self.seeds
pyed.show()
opakovat = not np.array_equal(pyed.seeds.nonzero() , nzero_seeds)
#opakovat =True #not all(opakovat)
nzero_seeds = pyed.seeds.nonzero()
def main():
#logger = logging.getLogger(__name__)
logger = logging.getLogger()
logger.setLevel(logging.WARNING)
ch = logging.StreamHandler()
logger.addHandler(ch)
#logger.debug('input params')
## read confguraton from file, use default values from OrganSegmentation
# input parser
parser = argparse.ArgumentParser(
description=' Show dicom data with overlay')
parser.add_argument('-i', '--inputdatapath',
default='',
help='path to data dir')
args_obj = parser.parse_args()
args = vars(args_obj)
#print args["arg"]
reader = datareader.DataReader()
data3d, metadata = reader.Get3DData(args['inputdatapath'], qt_app=None)
overlays = reader.GetOverlay()
overlay = np.zeros(data3d.shape, dtype=np.int8)
print "overlays ", overlays.keys()
for key in overlays:
overlay += overlays[key]
#import ipdb; ipdb.set_trace() # BREAKPOINT
pyed = py3DSeedEditor.py3DSeedEditor(data3d, contour=overlay)
pyed.show()
#print savestring
# import pdb; pdb.set_trace()
return
def showSegmentedData(self, data3d_thr, data3d_skel):
skan = SkeletonAnalyser(
data3d_skel,
volume_data=data3d_thr,
voxelsize_mm=self.metadata['voxelsize_mm'])
data3d_nodes_vis = skan.sklabel.copy()
# edges
data3d_nodes_vis[data3d_nodes_vis > 0] = 1
# nodes and terminals
data3d_nodes_vis[data3d_nodes_vis < 0] = 2
#pyed = seqt.QTSeedEditor(
# data3d,
# seeds=(data3d_nodes_vis).astype(np.int8),
# contours=data3d_thr.astype(np.int8)
#)
#app.exec_()
if not self.nogui:
pyed = se.py3DSeedEditor(
self.data3d,
seeds=(data3d_nodes_vis).astype(np.int8),
contour=data3d_thr.astype(np.int8)
)
pyed.show()
def simple_segmentation(self, data3d, voxelsize_mm):
simple_seg = np.zeros(data3d.shape )
#definice konvoluční masky
KONV_MASK = np.ones([10,10,10], float)
KONV_MASK = KONV_MASK/9.0;
#definice konvoluční funkce - param a - matice m*n kterou budeme přenásobovat konvoluční maticí, b - Konvoluční maska m*n
# nalezení kostí
simple_seg = data3d > BONES
#simple_seg[(simple_seg.shape[0]/5)*4:simple_seg.shape[0]] = 0
#nalzení páteře
spine_finder = scipy.ndimage.filters.convolve((simple_seg).astype(np.int), KONV_MASK)
#pyed = py3DSeedEditor.py3DSeedEditor(simple_seg)
#
simple_seg += ((spine_finder>25)*SPINE_ID)
pyed = py3DSeedEditor.py3DSeedEditor(simple_seg)
pyed.show()
return simple_seg
def get_seeds_using_prob_class1(self, data, class1, roi=None, dens_min=20,
dens_max=255,
thresholdType='percOfMaxDist', percT=0.5):
# calculates probability based on similarity of intensities
probs = tools.intensity_probability(data, std=10)
# py3DSeedEditor.py3DSeedEditor(data).show()
# py3DSeedEditor.py3DSeedEditor(probs).show()
# normalizing and calculating reciprocal values
# weights_ints = skexp.rescale_intensity(probs,
# in_range=(0,probs.max()), out_range=(1,0))
weights_ints = np.exp(-probs)
# py3DSeedEditor.py3DSeedEditor(weights_ints).show()
if roi is None:
roi = np.logical_and(data >= dens_min, data <= dens_max)
dist_data = np.where(class1 == 1, False, True)
dist_data *= roi > 0
# dists = distance_transform_edt(dist_data)
# py3DSeedEditor.py3DSeedEditor(dists).show()
# print 'dists max = %i' % dists.max()
# print 'dists min = %i' % dists.min()
# print 'weights_ints max = %.4f' % weights_ints.max()
# print 'weights_ints min = %.4f' % weights_ints.min()
# print 'probs max = %.4f' % probs.max()
# print 'probs min = %.4f' % probs.min()
# energy = dists * weights_ints
energy = weights_ints
# py3DSeedEditor.py3DSeedEditor(energy).show()
seeds = np.zeros(data.shape, dtype=np.bool)
if thresholdType == 'percOfMaxDist':
seeds = energy > (percT * energy.max())
elif thresholdType == 'mean':
seeds = energy > 2 * (energy[np.nonzero(energy)]).mean()
# TODO: tady je problem, ze energy je v intervalu <0.961, 1> - hrozne
# maly rozsah
print 'energy max = %.4f' % energy.max()
print 'energy min = %.4f' % energy.min()
print 'thresh = %.4f' % (percT * energy.max())
print seeds.min()
print seeds.max()
print 'seed perc = %.2f' % (
(energy > percT * energy.max()).sum()/np.float(energy.nbytes))
py3DSeedEditor.py3DSeedEditor(seeds).show()
# removing to small objects
min_size_of_seed_area = 60
print 'before removing: %i' % seeds.sum()
seeds = skimor.remove_small_objects(
seeds, min_size=min_size_of_seed_area, connectivity=1,
in_place=False)
print 'after removing: %i' % seeds.sum()
all_seeds = np.zeros(data.shape, dtype=np.int)
# allSeeds[np.nonzero(self.segmentation)] = 80
all_seeds[np.nonzero(class1)] = 1 # zdrava tkan
all_seeds[np.nonzero(seeds)] = 2 # outliers
# kvuli segmentaci pomoci random walkera se omezi obraz pouze na
# segmentovana jatra a cevy
all_seeds = np.where(roi == 0, -1, all_seeds)
py3DSeedEditor.py3DSeedEditor(all_seeds).show()
return all_seeds
def eval_all_from_dataset_metadata(inputdata, visualization=False):
"""
set metadata
"""
evaluation_all = {
'file1': [],
'file2': [],
'volume1_mm3': [],
'volume2_mm3': [],
'err1_mm3': [],
'err2_mm3': [],
'err1_percent': [],
'err2_percent': [],
'voe': [],
'vd': [],
'avgd': [],
'rmsd': [],
'processing_time': [],
'organ_interactivity_counter': [],
'maxd': []
}
for i in range(0, len(inputdata['data'])):
reader = datareader.DataReader()
data3d_a_path = os.path.join(inputdata['basedir'],
inputdata['data'][i]['sliverseg'])
data3d_a, metadata_a = reader.Get3DData(data3d_a_path)
print "inputdata ", inputdata['data'][i].keys()
try:
# if there is defined overlay
data3d_a = reader.GetOverlay()[inputdata['data'][i][
'overlay_number']]
logger.info('overlay loaded')
print 'overlay loaded'
except:
logger.debug("overlay not loaded")
pass
logger.debug('data A shape ' + str(data3d_a.shape))
data3d_b_path = os.path.join(inputdata['basedir'],
inputdata['data'][i]['ourseg'])
obj_b = misc.obj_from_file(data3d_b_path, filetype='pickle')
#data_b, metadata_b = reader.Get3DData(data3d_b_path)
data3d_b = qmisc.uncrop(obj_b['segmentation'],
obj_b['crinfo'], data3d_a.shape)
#import pdb; pdb.set_trace()
#data3d_a = (data3d_a > 1024).astype(np.int8)
data3d_a = (data3d_a > 0).astype(np.int8)
data3d_b = (data3d_b > 0).astype(np.int8)
if visualization:
pyed = py3DSeedEditor.py3DSeedEditor(data3d_a, # + (4 * data3d_b)
contour=data3d_b)
pyed.show()
evaluation_one = compare_volumes(data3d_a, data3d_b,
metadata_a['voxelsize_mm'])
evaluation_all['file1'].append(data3d_a_path)
evaluation_all['file2'].append(data3d_b_path)
evaluation_all['volume1_mm3'].append(evaluation_one['volume1_mm3'])
evaluation_all['volume2_mm3'].append(evaluation_one['volume2_mm3'])
evaluation_all['err1_mm3'].append(evaluation_one['err1_mm3'])
evaluation_all['err2_mm3'].append(evaluation_one['err2_mm3'])
evaluation_all['err1_percent'].append(evaluation_one['err1_percent'])
evaluation_all['err2_percent'].append(evaluation_one['err2_percent'])
evaluation_all['voe'].append(evaluation_one['voe'])
evaluation_all['vd'].append(evaluation_one['vd'])
evaluation_all['avgd'].append(evaluation_one['avgd'])
evaluation_all['rmsd'].append(evaluation_one['rmsd'])
evaluation_all['maxd'].append(evaluation_one['maxd'])
if 'processing_time' in obj_b.keys():
#this is only for compatibility with march2014 data
processing_time = obj_b['processing_time']
organ_interactivity_counter = obj_b['organ_interactivity_counter']
else:
processing_time = obj_b['processing_information']['organ_segmentation']['processing_time'] # noqa
organ_interactivity_counter = obj_b['processing_information']['organ_segmentation']['organ_interactivity_counter'] # noqa
evaluation_all['processing_time'].append(processing_time)
evaluation_all['organ_interactivity_counter'].append(
organ_interactivity_counter)
return evaluation_all
data['slab']['none'] = 0
data['slab']['liver'] = 1
data['slab']['porta'] = 2
#print np.max(output)
#import pdb; pdb.set_trace()
#data = {}
#data['data3d'] = oseg.data3d
#data['crinfo'] = oseg.crinfo
#data['segmentation'] = oseg.segmentation
data['segmentation'][output] = data['slab']['porta']
#data['slab'] = slab
pyed = py3DSeedEditor.py3DSeedEditor(data['data3d'], contour=data['segmentation']==data['slab']['porta'])
pyed.show()
#pyed = py3DSeedEditor.py3DSeedEditor(data['segmentation'])
#pyed.show()
# Uvolneni pameti
garbage.collect()
if args.outputfile == None:
savestring = raw_input ('Save output data? (y/n): ')
#sn = int(snstring)
if savestring in ['Y','y']:
pth, filename = os.path.split(os.path.normpath(args.inputfile))
startTime = datetime.now()
hr = TreeVolumeGenerator()
hr.importFromYaml(args.inputfile)
hr.voxelsize_mm = args.voxelsize
hr.shape = args.datashape
hr.generateTree()
logger.info("TimeUsed:"+str(datetime.now()-startTime))
volume_px = sum(sum(sum(hr.data3d)))
volume_mm3 = volume_px*(hr.voxelsize_mm[0]*hr.voxelsize_mm[1]*hr.voxelsize_mm[2])
logger.info("Volume px:"+str(volume_px))
logger.info("Volume mm3:"+str(volume_mm3))
#vizualizace
pyed = se.py3DSeedEditor(hr.data3d)
pyed.show()
# 3d vizualizace (debug)
if args.debug:
try:
from vtk_point_cloud import VtkPointCloud
pointCloud = VtkPointCloud(None,0,51,True)
for z in xrange(0,len(hr.data3d)):
for y in xrange(0,len(hr.data3d[z])):
for x in xrange(0,len(hr.data3d[z][y])):
if hr.data3d[z][y][x] == 1:
pointCloud.addPoint([x,y,z])
pointCloud.rander()
except:
logger.error("pointCloud 3d debug error: cant find vtk_point_cloud.py")
#ukládání do souboru
def vesselSegmentation(data, segmentation = -1, threshold = -1, voxelsize_mm = [1,1,1], inputSigma = -1,
dilationIterations = 0, dilationStructure = None, nObj = 10, biggestObjects = False,
seeds = None, interactivity = True, binaryClosingIterations = 2,
binaryOpeningIterations = 0, smartInitBinaryOperations = True,
returnThreshold = False, binaryOutput = True, returnUsedData = False):
"""
Vessel segmentation z jater.
Input:
data - CT (nebo MRI) 3D data
segmentation - zakladni oblast pro segmentaci, oznacena struktura se stejnymi rozmery jako "data",
kde je oznaceni (label) jako:
1 jatra,
-1 zajimava tkan (kosti, ...)
0 jinde
threshold - prah
voxelsize_mm - (vektor o hodnote 3) rozmery jednoho voxelu
inputSigma - pocatecni hodnota pro prahovani
dilationIterations - pocet operaci dilation nad zakladni oblasti pro segmentaci ("segmantation")
dilationStructure - struktura pro operaci dilation
nObj - oznacuje, kolik nejvetsich objektu se ma vyhledat - pokud je rovno 0 (nule), vraci cela data
biggestObjects - moznost, zda se maji vracet nejvetsi objekty nebo ne
seeds - moznost zadat pocatecni body segmentace na vstupu. Je to matice o rozmerech jako data. Vsude nuly, tam kde je oznaceni jsou jednicky.
interactivity - nastavi, zda ma nebo nema byt pouzit interaktivni mod upravy dat
binaryClosingIterations - vstupni binary closing operations
binaryOpeningIterations - vstupni binary opening operations
smartInitBinaryOperations - logicka hodnota pro smart volbu pocatecnich hodnot binarnich operaci (bin. uzavreni a bin. otevreni)
returnThreshold - jako druhy parametr funkce vrati posledni hodnotu prahu
binaryOutput - zda ma byt vystup vracen binarne nebo ne (binarnim vystupem se rozumi: cokoliv jineho nez hodnota 0 je hodnota 1)
returnUsedData - vrati pouzita data
Output:
filtrovana data
"""
dim = numpy.ndim(data)
logger.debug( 'Dimenze vstupnich dat: ' + str(dim))
if (dim < 2) or (dim > 3):
logger.debug( 'Nepodporovana dimenze dat!')
logger.debug( 'Ukonceni funkce!')
return None
if seeds == None:
logger.debug( 'Funkce spustena bez prioritnich objektu!')
if biggestObjects:
logger.debug( 'Funkce spustena s vracenim nejvetsich objektu => nebude mozne vybrat prioritni objekty!')
if ( nObj < 1 ) :
nObj = 1
if biggestObjects:
logger.debug( 'Vybrano objektu k vraceni: ' + str(nObj))
logger.debug('Pripravuji data...')
voxel = numpy.array(voxelsize_mm)
## Kalkulace objemove jednotky (voxel) (V = a*b*c).
voxel1 = voxel[0]#[0]
voxel2 = voxel[1]#[0]
voxel3 = voxel[2]#[0]
voxelV = voxel1 * voxel2 * voxel3
## number je zaokrohleny 2x nasobek objemove jednotky na 2 desetinna mista.
## number stanovi doporucenou horni hranici parametru gauss. filtru.
number = (numpy.round((2 * voxelV**(1.0/3.0)), 2))
## Operace dilatace (dilation) nad samotnymi jatry ("segmentation").
if(dilationIterations > 0.0):
segmentation = scipy.ndimage.binary_dilation(input = segmentation,
structure = dilationStructure, iterations = dilationIterations)
## Ziskani datove oblasti jater (bud pouze jater nebo i jejich okoli - zalezi,
## jakym zpusobem bylo nalozeno s operaci dilatace dat).
preparedData = (data * (segmentation == 1))#.astype(numpy.float)
logger.debug( 'Typ vstupnich dat: ' + str(preparedData.dtype))
# if preparedData.dtype != numpy.uint8:
# print 'Data nejsou typu numpy.uint8 => muze dojit k errorum'
if not numpy.can_cast(preparedData.dtype, numpy.float):
logger.debug( 'ERROR: (debug message) Data nejsou takoveho typu, aby se daly prevest na typ "numpy.float" => muze dojit k errorum')
logger.debug( 'Ukoncuji funkci!')
return None
if (preparedData == False).all():
logger.debug( 'ERROR: (debug message) Jsou spatna data nebo segmentacni matice: all is true == data is all false == bad segmentation matrix (if data matrix is ok)')
logger.debug( 'Ukoncuji funkci!')
return None
del(data)
del(segmentation)
## Nastaveni rozmazani a prahovani dat.
if(inputSigma == -1):
inputSigma = number
if(inputSigma > number):
inputSigma = number
seeds = None
if biggestObjects == False and seeds == None and interactivity == True and threshold==-1:
logger.debug(('Nyni si levym nebo pravym tlacitkem mysi (klepnutim nebo tazenim) oznacte specificke oblasti k vraceni.'))
import py3DSeedEditor
pyed = py3DSeedEditor.py3DSeedEditor(preparedData)
pyed.show()
seeds = pyed.seeds
## Zkontrolovat, jestli uzivatel neco vybral - nejaky item musi byt ruzny od nuly.
if (seeds != 0).any() == False:
seeds = None
logger.debug( 'Zadne seedy nezvoleny => nejsou prioritni objekty.')
else:
seeds = seeds.nonzero()#seeds * (seeds != 0) ## seeds je n-tice poli indexu nenulovych prvku => item krychle je == krychle[ seeds[0][x], seeds[1][x], seeds[2][x] ]
logger.debug( 'Seedu bez nul: ' + str(len(seeds[0])))
closing = binaryClosingIterations
opening = binaryOpeningIterations
if (smartInitBinaryOperations and interactivity):
if (seeds == None):
closing = 5
opening = 1
else:
closing = 2
opening = 0
## Samotne filtrovani.
uiT = uiThreshold.uiThreshold(preparedData, voxel, threshold,
interactivity, number, inputSigma, nObj, biggestObjects, closing,
opening, seeds)
output = uiT.run()
## Vypocet binarni matice.
if output == None:
logger.debug( 'Zadna data k vraceni! (output == None)')
elif binaryOutput:
output[output != 0] = 1
## Vraceni matice.
if returnThreshold:
if returnUsedData:
return preparedData, output, uiT.returnLastThreshold()
else:
return output, uiT.returnLastThreshold()
else:
if returnUsedData:
return preparedData, output
else:
return output
def visualization(self):
pyed = py3DSeedEditor.py3DSeedEditor(
self.segmentation == self.data['slab']['lesions'])
pyed.show()
def segmentation(data3d, segmentation, params):
""" Texture analysis by LBP algorithm.
data: CT (or MRI) 3D data
segmentation: labeled image with same size as data where label:
1 mean liver pixels,
-1 interesting tissuse (bones)
0 otherwise
"""
MAXPATTERNS = 10
slices = data3d.shape[2]
data3d = data3d.astype(np.int16)
params[2] = 0
data3d = data3d * segmentation;
data3d = sp.ndimage.filters.gaussian_filter(data3d,params*2.5)
pyed = py3DSeedEditor.py3DSeedEditor(data3d)
pyed.show()
lbp = lbpLib.loadLbpLibrary()
lbpReftmp = (ctypes.c_long * 256)()
lbpRef = np.zeros([MAXPATTERNS,256])
actualpatterns = 0
# vytvoreni referencnich LBP obrazu
for j in range(15,data3d.shape[0]-16):
if (actualpatterns == MAXPATTERNS) :
break
for k in range(15,data3d.shape[1]-16):
if (actualpatterns == MAXPATTERNS) :
break
if ((pyed.seeds[j,k,pyed.actual_slice] == 1)):
lbpReftmp = lbpLib.realTimeLbpImNp(lbp, data3d[j-15:j+16,k-15:k+16,pyed.actual_slice])
for z in range(256):
lbpRef[actualpatterns, z] = lbpReftmp[z]
print lbpRef[actualpatterns,]
actualpatterns += 1
# uprava velikosti dat
for i in range(20):
h2 = data3d.shape[0] - i*16
if (h2 <= 0):
h2 = i*16
break
for i in range(20):
w2 = data3d.shape[1] - i*16
if (w2 <= 0):
w2 = i*16
break
# obdelnikova data prevedeme na ctverec
if (w2 >= h2 ):
h2 = w2
else :
w2 = h2
numOfPartsX = (w2) / 16
numOfPartsY = (h2) / 16
minMaxData = np.zeros([h2,w2,data3d.shape[2]],dtype = np.float)
lbpSlice = np.zeros([h2,w2])
# hlavni cyklus vypocitavajici odezvy
for i in range(slices):
tempData = np.zeros([h2,w2],dtype = np.float)
tempData[0:data3d.shape[0],0:data3d.shape[1]] = data3d[:,:,i].astype(np.float)
lbpSlice = lbpLib.realTimeLbpIm2ImNp(lbp, tempData.astype(np.int16))
lbpRes = lbpLib.lbp2HistsNp(lbp, lbpSlice)
for j in range(lbpRes.shape[0]):
minmaxBestVal = 0
for k in range(MAXPATTERNS):
minmaxVal = minmax(lbpRes[j,:],lbpRef[k,:])
if minmaxVal > minmaxBestVal :
minmaxBestVal = minmaxVal
minMaxData[16*(j/numOfPartsX):16*(j/numOfPartsX)+16,16*(j % numOfPartsX):16*(j % numOfPartsX)+16,i] = minmaxBestVal
# zobrazeni vysledku
pyed2 = py3DSeedEditor.py3DSeedEditor(minMaxData)
pyed2.show()
return segmentation
