def predict( self,
filename,
nb_autocontext=None,
mask=None,
debug=False,
return_all=False ):
"""
The prediction function must be defined outside of the main class in order to be used in joblib's Parallel.
"""
nb_labels = len(self.params['labels'])+1
if nb_autocontext is None:
nb_autocontext = len(glob(self.params['name'] + "_*"))
if self.params['predict_preprocessing_function'] is None:
img = irtk.imread( filename, dtype="float32" )
img = img.resample( pixelSize=self.params['resample'], interpolation='linear' ).rescale(0,1000)
else:
img = self.params['predict_preprocessing_function'](self, filename).copy()
if mask is None:
mask = irtk.ones( img.get_header(), dtype="uint8" ).as3D()
mask[img==0] = 0
else:
mask = mask.resample( pixelSize=self.params['resample'], interpolation='nearest' ).astype('uint8')
probas = predict_autocontext( self,
img,
mask,
nb_labels,
nb_autocontext,
debug=debug,
return_all=return_all )
return probas
def get_center_brain_detection(f,world=True):
input_mask = irtk.imread( f, force_neurological=False )
mask = irtk.ones( input_mask.get_header(), dtype='uint8' )
mask[input_mask == 2] = 0
x_min, y_min, z_min, x_max, y_max, z_max = map( float, mask.bbox() )
center = [ x_min + (x_max-x_min)/2,
y_min + (y_max-y_min)/2,
z_min + (z_max-z_min)/2 ]
if not world:
center = np.array(center,dtype='float64')[::-1]
return center
else:
center = input_mask.ImageToWorld(center)
return center
def predict(self,
filename,
ga,
nb_autocontext=None,
mask=None,
debug=False,
return_all=False):
nb_labels = len(self.labels) + 1
if nb_autocontext is None:
nb_autocontext = len(glob(self.params['name'] + "_*"))
img = irtk.imread(filename, dtype="float32")
img = img.resample(pixelSize=self.params['resample'],
interpolation='linear').rescale(0, 1000)
extra_layers = []
if mask is None:
mask = irtk.ones(img.get_header(), dtype="uint8")
mask[img == 0] = 0
else:
mask = mask.resample(pixelSize=self.params['resample'],
interpolation='nearest').astype('uint8')
metadata = None
probas = predict_autocontext(self,
img,
mask,
np.array(extra_layers, dtype="float32"),
metadata,
nb_labels,
ga,
nb_autocontext,
debug=debug,
return_all=return_all)
return probas
def predict( self,
filename,
ga,
nb_autocontext=None,
mask=None,
debug=False,
return_all=False ):
nb_labels = len(self.labels)+1
if nb_autocontext is None:
nb_autocontext = len(glob(self.params['name'] + "_*"))
img = irtk.imread( filename, dtype="float32" )
img = img.resample( pixelSize=self.params['resample'], interpolation='linear' ).rescale(0,1000)
extra_layers = []
if mask is None:
mask = irtk.ones( img.get_header(), dtype="uint8" )
mask[img==0] = 0
else:
mask = mask.resample( pixelSize=self.params['resample'], interpolation='nearest' ).astype('uint8')
metadata = None
probas = predict_autocontext( self,
img,
mask,
np.array( extra_layers, dtype="float32" ),
metadata,
nb_labels,
ga,
nb_autocontext,
debug=debug,
return_all=return_all )
return probas
def preprocess_training_data( patient_id,
img_folder,
seg_folder,
resample,
offline=False,
online=True):
if offline or online:
if ( offline
and os.path.exists( "offline_preprocessing/"+patient_id+"_img.nii.gz" )
and os.path.exists( "offline_preprocessing/"+patient_id+"_seg.nii.gz" ) ):
return
img = irtk.imread( img_folder + "/" + patient_id + ".nii.gz",
dtype='float32' )
seg = irtk.imread( seg_folder +"/"+patient_id+"_seg.nii.gz",
dtype="uint8" )
wall = nd.binary_dilation( seg,
morphology.ball(int(12.5*0.001/seg.header['pixelSize'][0])) )
wall = wall.astype('int')
points = np.transpose(np.nonzero(wall))[::4]
center,S,V = fit_ellipsoidPCA( points )
if V[0,0] < 0:
V *= -1
points = np.transpose(np.nonzero(wall))
projections = np.dot(points-center,V[0])
# valves
index = projections > (projections.max() - 40.0*0.001/seg.header['pixelSize'][0])
#print "VALVE size:",np.sum(index), projections.max(), 40.0*0.001/seg.header['pixelSize'][0]
wall[points[index,0],
points[index,1],
points[index,2]] = 2
#print "VALVE1", wall.max()
wall = irtk.Image(wall,seg.get_header())
img = img.resample( pixelSize=resample, interpolation='linear' ).rescale(0,1000)
seg = seg.transform(target=img,interpolation="nearest").astype('uint8')
wall = wall.transform(target=img,interpolation="nearest").astype('uint8')
wall[seg>0] = 0
seg[wall==1] = 2
seg[wall==2] = 3
#print "VALVE2", seg.max()
#irtk.imwrite("debug/"+patient_id+"_border.nii.gz",seg)
seg[img==0] = 255
if offline:
irtk.imwrite( "offline_preprocessing/"+patient_id+"_img.nii.gz", img )
irtk.imwrite( "offline_preprocessing/"+patient_id+"_seg.nii.gz", seg )
return
if not online:
img = irtk.imread( "offline_preprocessing/"+patient_id+"_img.nii.gz" )
seg = irtk.imread( "offline_preprocessing/"+patient_id+"_seg.nii.gz" )
mask = irtk.ones( img.get_header(), dtype='uint8' )
mask[img==0] = 0
return { 'patient_id': patient_id,
'img' : img,
'seg' : seg,
'mask' : mask }
irtk.imwrite( args.output + "/proba.nii.gz", proba )
irtk.imwrite( args.output + "/min_proba.nii.gz", min_proba )
seg = np.argmax( proba, axis=0)
seg = irtk.Image( seg,
img.get_header() ).astype('uint8')
irtk.imwrite( args.output + "/seg.nii.gz", seg )
# Regression
print "doing regression"
header = img.get_header()
header['dim'][3] = 3
offsets_heart = irtk.ones(header,dtype='float32') * np.finfo('float32').max
for l, forest, offsets in zip( [1],
[reg_heart],
[offsets_heart] ):
if args.theta < 360 and args.selective:
selection = np.logical_and( proba[l,
coords[:,0],
coords[:,1],
coords[:,2]]>0.5,
min_proba[l,
coords[:,0],
coords[:,1],
coords[:,2]]<0.5 )
else:
selection = proba[l,
def mask_image( file_img, file_mask, ga, r, neigh, output_dir ):
img = irtk.imread( file_img, dtype='float32' )
input_mask = irtk.imread( file_mask )
print "predicting..."
res = irtk.zeros( img.get_header(), dtype='float32' )
res2 = irtk.zeros( img.get_header(), dtype='float32' )
res3 = irtk.zeros( img.get_header(), dtype='float32' )
res4 = irtk.zeros( img.get_header(), dtype='uint8' )
mask = irtk.ones( input_mask.get_header(), dtype='uint8' )
mask[input_mask == 2] = 0
for z in xrange(img.shape[0]):
print z
YX = np.transpose( np.nonzero( mask[z] ) )
if YX.shape[0] == 0:
continue # this slice does not intersect the box
patches = extract_patches2D( img[z], r, YX )
patches = np.reshape( patches, (patches.shape[0],patches.shape[1]*patches.shape[2]) )
predictions = neigh.predict_proba(patches)[:,1]
res[z,YX[:,0],YX[:,1]] = predictions
x_min, y_min, z_min, x_max, y_max, z_max = mask.bbox()
proba = res[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1]
if args.mass:
BV = get_BV( args.ga )
box_volume = (z_max-z_min)*img.header['pixelSize'][2]*(y_max-y_min)*img.header['pixelSize'][1]*(x_max-x_min)*img.header['pixelSize'][0]
ratio = float(BV) / float(box_volume)
print "ratio", ratio
q0,q1 = mquantiles( proba.flatten(), prob=[0.5*(1.0-ratio),
1.0-0.5*ratio] )
print "threshold", q0,q1
#threshold = max(0.5,threshold)
# labels = res[z_min:z_max+1,
# y_min:y_max+1,
# x_min:x_max+1] > threshold
#res = 1 / (np.exp(-(res-threshold)/(res.max()-res.min())))
res[res<q0] = q0
res[res>q1] = q1
res -= res.min()
res /= res.max()
labels = res[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1] > 0.5
proba = res[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1]
cropped_img = img[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1]
if args.do_3D:
labels = irtk.crf( cropped_img,
labels,
proba,
l=args.l,
sigma=get_noiseXY(cropped_img),
sigmaZ=get_noiseZ(cropped_img) )
# elif args.do_patchZ:
# labels = irtk.crf_patchZ( cropped_img,
# labels,
# proba,
# l=10.0 )
# else:
# for z in xrange(z_min,z_max+1):
# labels[z] = irtk.crf( cropped_img[z],
# labels[z],
# proba[z],
# l=1.0 )
print "MAX LABEL:", labels.max()
irtk.imwrite(output_dir + "/bare_"+os.path.basename(file_img), labels )
tmp = irtk.zeros( img.get_header(), dtype='uint8' )
tmp[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1] = labels
( min_x_bare, min_y_bare, min_z_bare,
max_x_bare, max_y_bare, max_z_bare ) = tmp.bbox()
if not args.no_cleaning:
# clean by fitting ellipses enlarged of 10%
for z in xrange(labels.shape[0]):
edges = nd.morphological_gradient( labels[z] > 0,size=5 )
points = np.transpose(edges.nonzero())[:,::-1]
if len(points) == 0:
continue
points = np.array(map(lambda x:[x],points),dtype='int32')
ellipse = cv2.fitEllipse(points)
cv2.ellipse( labels[z], (ellipse[0],
(1.1*ellipse[1][0],1.1*ellipse[1][1]),
ellipse[2]) , 1, -1 )
irtk.imwrite(output_dir + "/seg_"+os.path.basename(file_img), labels )
irtk.imwrite(output_dir + "/res_"+os.path.basename(file_img), res )
# re-read the image in case we processed it
img = irtk.imread( file_img, dtype='float32' )
cropped_img = img[z_min:z_max+1,
y_min:y_max+1,
x_min:x_max+1]
cropped_img[labels==0] = -1
masked = cropped_img.bbox(crop=True)
irtk.imwrite(output_dir + "/masked_"+os.path.basename(file_img), masked )
# re-read the image in case we processed it
img = irtk.imread( file_img, dtype='float32' )
x0 = min_x_bare + (max_x_bare - min_x_bare) / 2
y0 = min_y_bare + (max_y_bare - min_y_bare) / 2
ofd = get_OFD(ga)/img.header['pixelSize'][0]
cropped_img = img[min_z_bare:max_z_bare+1,
max(0,int(round(y0-ofd/2))):min(img.shape[1],int(round(y0+ofd/2+1))),
max(0,int(round(x0-ofd/2))):min(img.shape[2],int(round(x0+ofd/2+1)))].copy()
irtk.imwrite(output_dir + "/very_large_"+os.path.basename(file_img),
cropped_img )
cropped_proba = res[min_z_bare:max_z_bare+1,
max(0,int(round(y0-ofd/2))):min(img.shape[1],int(round(y0+ofd/2+1))),
max(0,int(round(x0-ofd/2))):min(img.shape[2],int(round(x0+ofd/2+1)))].copy()
irtk.imwrite(output_dir + "/proba_"+os.path.basename(file_img),
cropped_proba )
def mask_image(file_img, file_mask, ga, r, neigh, output_dir):
img = irtk.imread(file_img, dtype='float32')
input_mask = irtk.imread(file_mask)
print "predicting..."
res = irtk.zeros(img.get_header(), dtype='float32')
res2 = irtk.zeros(img.get_header(), dtype='float32')
res3 = irtk.zeros(img.get_header(), dtype='float32')
res4 = irtk.zeros(img.get_header(), dtype='uint8')
mask = irtk.ones(input_mask.get_header(), dtype='uint8')
mask[input_mask == 2] = 0
for z in xrange(img.shape[0]):
print z
YX = np.transpose(np.nonzero(mask[z]))
if YX.shape[0] == 0:
continue # this slice does not intersect the box
patches = extract_patches2D(img[z], r, YX)
patches = np.reshape(
patches, (patches.shape[0], patches.shape[1] * patches.shape[2]))
predictions = neigh.predict_proba(patches)[:, 1]
res[z, YX[:, 0], YX[:, 1]] = predictions
x_min, y_min, z_min, x_max, y_max, z_max = mask.bbox()
proba = res[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1]
if args.mass:
BV = get_BV(args.ga)
box_volume = (z_max - z_min) * img.header['pixelSize'][2] * (
y_max - y_min) * img.header['pixelSize'][1] * (
x_max - x_min) * img.header['pixelSize'][0]
ratio = float(BV) / float(box_volume)
print "ratio", ratio
q0, q1 = mquantiles(proba.flatten(),
prob=[0.5 * (1.0 - ratio), 1.0 - 0.5 * ratio])
print "threshold", q0, q1
#threshold = max(0.5,threshold)
# labels = res[z_min:z_max+1,
# y_min:y_max+1,
# x_min:x_max+1] > threshold
#res = 1 / (np.exp(-(res-threshold)/(res.max()-res.min())))
res[res < q0] = q0
res[res > q1] = q1
res -= res.min()
res /= res.max()
labels = res[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1] > 0.5
proba = res[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1]
cropped_img = img[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1]
if args.do_3D:
labels = irtk.crf(cropped_img,
labels,
proba,
l=args.l,
sigma=get_noiseXY(cropped_img),
sigmaZ=get_noiseZ(cropped_img))
# elif args.do_patchZ:
# labels = irtk.crf_patchZ( cropped_img,
# labels,
# proba,
# l=10.0 )
# else:
# for z in xrange(z_min,z_max+1):
# labels[z] = irtk.crf( cropped_img[z],
# labels[z],
# proba[z],
# l=1.0 )
print "MAX LABEL:", labels.max()
irtk.imwrite(output_dir + "/bare_" + os.path.basename(file_img), labels)
tmp = irtk.zeros(img.get_header(), dtype='uint8')
tmp[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1] = labels
(min_x_bare, min_y_bare, min_z_bare, max_x_bare, max_y_bare,
max_z_bare) = tmp.bbox()
if not args.no_cleaning:
# clean by fitting ellipses enlarged of 10%
for z in xrange(labels.shape[0]):
edges = nd.morphological_gradient(labels[z] > 0, size=5)
points = np.transpose(edges.nonzero())[:, ::-1]
if len(points) == 0:
continue
points = np.array(map(lambda x: [x], points), dtype='int32')
ellipse = cv2.fitEllipse(points)
cv2.ellipse(
labels[z],
(ellipse[0],
(1.1 * ellipse[1][0], 1.1 * ellipse[1][1]), ellipse[2]), 1,
-1)
irtk.imwrite(output_dir + "/seg_" + os.path.basename(file_img), labels)
irtk.imwrite(output_dir + "/res_" + os.path.basename(file_img), res)
# re-read the image in case we processed it
img = irtk.imread(file_img, dtype='float32')
cropped_img = img[z_min:z_max + 1, y_min:y_max + 1, x_min:x_max + 1]
cropped_img[labels == 0] = -1
masked = cropped_img.bbox(crop=True)
irtk.imwrite(output_dir + "/masked_" + os.path.basename(file_img), masked)
# re-read the image in case we processed it
img = irtk.imread(file_img, dtype='float32')
x0 = min_x_bare + (max_x_bare - min_x_bare) / 2
y0 = min_y_bare + (max_y_bare - min_y_bare) / 2
ofd = get_OFD(ga) / img.header['pixelSize'][0]
cropped_img = img[min_z_bare:max_z_bare + 1,
max(0, int(round(y0 - ofd / 2))
):min(img.shape[1], int(round(y0 + ofd / 2 + 1))),
max(0, int(round(x0 - ofd / 2))
):min(img.shape[2], int(round(x0 + ofd / 2 +
1)))].copy()
irtk.imwrite(output_dir + "/very_large_" + os.path.basename(file_img),
cropped_img)
cropped_proba = res[min_z_bare:max_z_bare + 1,
max(0, int(round(y0 - ofd / 2))
):min(img.shape[1], int(round(y0 + ofd / 2 + 1))),
max(0, int(round(x0 - ofd / 2))
):min(img.shape[2], int(round(x0 + ofd / 2 +
1)))].copy()
irtk.imwrite(output_dir + "/proba_" + os.path.basename(file_img),
cropped_proba)
def preprocess_training_data(patient_id,
img_folder,
seg_folder,
resample,
offline=False,
online=True):
if offline or online:
if (offline and os.path.exists("offline_preprocessing/" + patient_id +
"_img.nii.gz")
and os.path.exists("offline_preprocessing/" + patient_id +
"_seg.nii.gz")):
return
img = irtk.imread(img_folder + "/" + patient_id + ".nii.gz",
dtype='float32')
seg = irtk.imread(seg_folder + "/" + patient_id + "_seg.nii.gz",
dtype="uint8")
wall = nd.binary_dilation(
seg,
morphology.ball(int(12.5 * 0.001 / seg.header['pixelSize'][0])))
wall = wall.astype('int')
points = np.transpose(np.nonzero(wall))[::4]
center, S, V = fit_ellipsoidPCA(points)
if V[0, 0] < 0:
V *= -1
points = np.transpose(np.nonzero(wall))
projections = np.dot(points - center, V[0])
# valves
index = projections > (projections.max() -
40.0 * 0.001 / seg.header['pixelSize'][0])
#print "VALVE size:",np.sum(index), projections.max(), 40.0*0.001/seg.header['pixelSize'][0]
wall[points[index, 0], points[index, 1], points[index, 2]] = 2
#print "VALVE1", wall.max()
wall = irtk.Image(wall, seg.get_header())
img = img.resample(pixelSize=resample,
interpolation='linear').rescale(0, 1000)
seg = seg.transform(target=img,
interpolation="nearest").astype('uint8')
wall = wall.transform(target=img,
interpolation="nearest").astype('uint8')
wall[seg > 0] = 0
seg[wall == 1] = 2
seg[wall == 2] = 3
#print "VALVE2", seg.max()
#irtk.imwrite("debug/"+patient_id+"_border.nii.gz",seg)
seg[img == 0] = 255
if offline:
irtk.imwrite("offline_preprocessing/" + patient_id + "_img.nii.gz",
img)
irtk.imwrite("offline_preprocessing/" + patient_id + "_seg.nii.gz",
seg)
return
if not online:
img = irtk.imread("offline_preprocessing/" + patient_id +
"_img.nii.gz")
seg = irtk.imread("offline_preprocessing/" + patient_id +
"_seg.nii.gz")
mask = irtk.ones(img.get_header(), dtype='uint8')
mask[img == 0] = 0
return {
'patient_id': patient_id,
'img': img,
'seg': seg,
'extra_layers': np.array([], dtype='float32'),
'metadata': None,
'mask': mask
}
proba = proba.transform(target=header)
if not args.score:
irtk.imwrite( args.output + "/proba.nii.gz", proba )
seg = np.argmax( proba, axis=0)
seg = irtk.Image( seg,
img.get_header() ).astype('uint8')
if not args.score:
irtk.imwrite( args.output + "/seg.nii.gz", seg )
# Regression
header = img.get_header()
header['dim'][3] = 3
offsets_heart = irtk.ones(header,dtype='float32') * np.finfo('float32').max
offsets_left_lung = irtk.ones(header,dtype='float32') * np.finfo('float32').max
offsets_right_lung = irtk.ones(header,dtype='float32') * np.finfo('float32').max
offsets_liver = irtk.ones(header,dtype='float32') * np.finfo('float32').max
for l, forest, offsets in zip( [1,2,3,4],#range(1,nb_labels),
[reg_left_lung,reg_right_lung,reg_heart,reg_liver],
[offsets_left_lung,offsets_right_lung,offsets_heart,offsets_liver] ):
selection = seg[coords[:,0],
coords[:,1],
coords[:,2]] == l
coords_reg = coords[selection].astype('int32').copy()
u2 = u[selection].copy()
v2 = v[selection].copy()
w2 = w[selection].copy()