def segment(folder, params):
'''
Segment all layers in a folder
'''
#create folders for the output
filelib.make_folders([
params.inputfolder + '../segmented/masks/glomeruli/' + folder,
params.inputfolder + '../segmented/masks/kidney/' + folder,
params.inputfolder + '../log/' + folder
])
start = time.time()
#list all files in the folder
files = filelib.list_image_files(params.inputfolder + folder)
files.sort()
params.folder = folder
#segment all layers in parallel
boost.run_parallel(process=segment_layer,
files=files,
params=params,
procname='Segmentation of glomeruli')
# segment_layer(files[50], params)
t = pd.Series({
'Segmentation': time.time() - start,
'threads': params.max_threads
})
t.to_csv(params.inputfolder + '../log/' + folder + 'Segmentation.csv',
sep='\t')
def quantify(folder, params): ''' Quantify a stack ''' if not os.path.exists(params.inputfolder + '../statistics/' + folder[:-1] + '.csv'): #list files in the folder files = filelib.list_image_files(params.inputfolder + folder) files.sort() #create a folder for statistics filelib.make_folders([params.inputfolder + '../statistics/' + filelib.return_path(folder[:-1] + '.csv')]) #extract voxel size xsize, zsize = extract_zoom(folder) #compute volume of the kidney kidney_volume = 0 for i in range(len(files)): ind = np.load(params.inputfolder + '../segmented/masks/' + folder + files[i][:-4] + '.npy') kidney_volume = kidney_volume + len(ind[0]) stat = pd.DataFrame() stat['Kidney_volume'] = [kidney_volume*xsize**2*zsize] stat['Image_name'] = folder[:-1] stat.to_csv(params.inputfolder + '../statistics/' + folder[:-1] + '.csv', sep = '\t')
def segment(folder, params): ''' Segment all layers in a folder ''' #create folders for the output filelib.make_folders([params.inputfolder + '../segmented/outlines/' + folder, params.inputfolder + '../segmented/masks/' + folder]) #list all files in the folder files = filelib.list_image_files(params.inputfolder + folder) files.sort() ind = np.int_(np.arange(0, len(files), 10)) files = np.array(files)[ind] if not len(filelib.list_image_files(params.inputfolder + '../segmented/masks/' + folder)) == len(files): params.folder = folder #segment all layers in parallel boost.run_parallel(process = segment_layer, files = files, params = params, procname = 'Segmentation of glomeruli')
def quantify(folder, params):
'''
Quantify a stack
'''
start = time.time()
#list files in the folder
files = filelib.list_image_files(params.inputfolder + folder)
files.sort()
#create a folder for statistics
filelib.make_folders([
params.inputfolder + '../statistics/' +
filelib.return_path(folder[:-1] + '.csv')
])
#extract voxel size
xsize, zsize = extract_zoom(folder)
#compute volume of the kidney
kidney_volume = 0
for i in range(len(files)):
ind = np.load(params.inputfolder + '../segmented/masks/kidney/' +
folder + files[i][:-4] + '.npy')
kidney_volume = kidney_volume + len(ind[0])
#compute glomeruli characteristics
maxvol = 4. / 3 * np.pi * float(params.maxrad)**3 #maximal allowed volume
minvol = 4. / 3 * np.pi * float(params.minrad)**3 #minimal allowed volume
print 'load data', len(files)
ind0 = []
ind1 = []
ind2 = []
label = []
for i in range(len(files)):
ind = np.load(params.inputfolder + '../segmented/labels/glomeruli/' +
folder + files[i][:-4] + '.npy')
label.append(ind[0])
ind1.append(ind[1])
ind2.append(ind[2])
ind0.append(np.ones_like(ind[1]) * i)
ind0 = np.hstack(ind0)
ind1 = np.hstack(ind1)
ind2 = np.hstack(ind2)
label = np.hstack(label)
llist = np.unique(label)
volumes = ndimage.sum((label > 0) * 1., label, llist) * xsize**2 * zsize
ind = np.where((volumes < maxvol) & (volumes > minvol))
volumes = volumes[ind]
llist = llist[ind]
sizes = 2 * pow(3. / 4 * volumes / np.pi, 1. / 3)
stat = pd.DataFrame({'Label': llist, 'Diameter': sizes, 'Volume': volumes})
stat['Image_name'] = folder[:-1]
stat['Kidney_volume'] = kidney_volume * xsize**2 * zsize
stat.to_csv(params.inputfolder + '../statistics/' + folder[:-1] + '.csv',
sep='\t')
t = pd.Series({'Quantification': time.time() - start})
t.to_csv(params.inputfolder + '../log/' + folder + 'Quantification.csv',
sep='\t')
def label_cells(folder, params):
'''
label connected objects in a stack
'''
start = time.time()
#list files in the folder
files = filelib.list_image_files(params.inputfolder + folder)
files.sort()
#make output folder for labels
filelib.make_folders(
[params.inputfolder + '../segmented/labels/glomeruli/' + folder])
#initiate labelling
num = 0
shape = tifffile.imread(
params.inputfolder + folder + files[0][:-4] +
'.tif').shape #get image dimensions from the first image
mask = load_mask(params.inputfolder + '../segmented/masks/glomeruli/' +
folder + files[0][:-4] + '.npy',
shape) #load the first mask
l0 = np.zeros_like(mask)
labels = []
limsize = int(float(params.maxrad))
nfiles = list(np.array(files))
#label
for i in range(len(files)):
mask = load_mask(
params.inputfolder + '../segmented/masks/glomeruli/' + folder +
files[i][:-4] + '.npy', shape)
l, n = ndimage.label(mask)
mask = (mask > 0) * 1.
labels.append(np.zeros_like(l))
for j in range(n):
ind = np.where(l == j + 1)
ls = np.unique(l0[ind])
ls = ls[np.where(ls > 0)]
if len(ls) == 0:
num += 1
labels[-1][ind] = num
if len(ls) == 1:
labels[-1][ind] = ls[0]
if len(ls) > 1:
labels[-1][ind] = ls[0]
for k in range(1, len(ls)):
for s in range(len(labels)):
labels[s] = np.where(labels[s] == ls[k], ls[0],
labels[s])
l0 = labels[-1]
#if number of layers exceeds limit, save the first layer
if len(labels) > limsize:
mask = labels.pop(0)
fname = nfiles.pop(0)
ind = np.where(mask > 0)
lb = mask[ind]
np.save(
params.inputfolder + '../segmented/labels/glomeruli/' +
folder + fname[:-4] + '.npy', np.array([lb, ind[0], ind[1]]))
#save remaining layers
for i in range(len(labels)):
mask = labels.pop(0)
fname = nfiles.pop(0)
ind = np.where(mask > 0)
lb = mask[ind]
np.save(
params.inputfolder + '../segmented/labels/glomeruli/' + folder +
fname[:-4] + '.npy', np.array([lb, ind[0], ind[1]]))
t = pd.Series({'Labelling': time.time() - start})
t.to_csv(params.inputfolder + '../log/' + folder + 'Labelling.csv',
sep='\t')
def quantify(folder, params):
'''
Quantify a stack
'''
start = time.time()
#list files in the folder
files = filelib.list_image_files(params.inputfolder + folder)
files.sort()
#create a folder for statistics
filelib.make_folders([
params.inputfolder + '../statistics/' +
filelib.return_path(folder[:-1] + '.csv')
])
#extract voxel size
xsize, zsize = extract_zoom(folder)
print("xsize: " + str(xsize) + ", zsize: " + str(zsize))
#compute volume of the kidney
kidney_volume = 0
for i in range(len(files)):
ind = np.load(params.inputfolder + '../segmented/masks/kidney/' +
folder + files[i][:-4] + '.npy')
kidney_volume = kidney_volume + len(ind[0])
#compute glomeruli characteristics
maxvol = np.pi * float(params.maxrad)**2 #maximal allowed area
minvol = np.pi * float(params.minrad)**2 #minimal allowed area
print 'load data', len(files)
ind0 = []
ind1 = []
ind2 = []
label = []
for i in range(len(files)):
ind = np.load(params.inputfolder + '../segmented/labels/glomeruli/' +
folder + files[i][:-4] + '.npy')
label.append(ind[0])
ind1.append(ind[1])
ind2.append(ind[2])
ind0.append(np.ones_like(ind[1]) * i)
ind0 = np.hstack(ind0)
ind1 = np.hstack(ind1)
ind2 = np.hstack(ind2)
label = np.hstack(label)
llist = np.unique(label)
volumes = ndimage.sum(
(label > 0) * 1., label,
llist) * xsize * xsize # Modification: Calculate areas
ind = np.where((volumes < maxvol) & (volumes > minvol))
volumes = volumes[ind]
llist = llist[ind]
sizes = 2 * np.sqrt(volumes / np.pi)
stat = pd.DataFrame({'Label': llist, 'Diameter': sizes, 'Volume': volumes})
stat['Image_name'] = folder[:-1]
stat['Kidney_volume'] = kidney_volume * xsize**2
stat.to_csv(params.inputfolder + '../statistics/' + folder[:-1] + '.csv',
sep='\t')
t = pd.Series({'Quantification': time.time() - start})
t.to_csv(params.inputfolder + '../log/' + folder + 'Quantification.csv',
sep='\t')
shape = tifffile.imread(
params.inputfolder + folder + files[0][:-4] +
'.tif').shape # get image dimensions from the first image
for i in range(len(files)):
ind = np.load(params.inputfolder + '../segmented/labels/glomeruli/' +
folder + files[i][:-4] + '.npy')
lbl = ind[0]
mask = np.ones(len(lbl), np.bool)
mask[not lbl in llist] = 0
lbl[mask] = 0
lbl = np.reshape(lbl, shape)
skimage.io.imsave(
params.inputfolder + '../segmented/labels/glomeruli/' +
params.folder + files[i][:-4] + '.tif', lbl.astype(np.int))
def label_cells(folder, params):
'''
label connected objects in a stack
'''
start = time.time()
#list files in the folder
files = filelib.list_image_files(params.inputfolder + folder)
files.sort()
#make output folder for labels
filelib.make_folders(
[params.inputfolder + '../segmented/labels/glomeruli/' + folder])
#initiate labelling
num = 0
shape = tifffile.imread(
params.inputfolder + folder + files[0][:-4] +
'.tif').shape #get image dimensions from the first image
mask = load_mask(params.inputfolder + '../segmented/masks/glomeruli/' +
folder + files[0][:-4] + '.npy',
shape) #load the first mask
l0 = np.zeros_like(mask) # Label for last layer
labels = []
limsize = int(float(params.maxrad))
nfiles = list(np.array(files))
#label
for i in range(len(files)): # Go through each image (in order)
mask = load_mask(
params.inputfolder + '../segmented/masks/glomeruli/' + folder +
files[i][:-4] + '.npy', shape)
l, n = ndimage.label(mask) # Label the image
mask = (mask > 0) * 1.
labels.append(
np.zeros_like(l)
) # Create a new label image initialized with 0 and add it to the list
for j in range(n): # For each label:
ind = np.where(
l == j +
1) ## Get all indices where the current label is (j + 1)
ls = np.unique(
l0[ind]
) # Get all labels in last layer where current label is (j + 1)
ls = ls[np.where(ls > 0)] # Only consider non-background
if len(ls) == 0: # If there is no connection, create a new label
num += 1
labels[-1][ind] = num
if len(
ls
) == 1: # If there is a connection to 1 specific id, rename the current layer (labels[-1]) label
labels[-1][ind] = ls[0]
if len(
ls
) > 1: # If there are multiple connections, synchronize them
labels[-1][ind] = ls[
0] # Assign current label from last label (for the first)
for k in range(
1, len(ls)
): # Go through all last labels (except the first one)
for s in range(len(
labels)): # Go through the list of stored labels
labels[s] = np.where(
labels[s] == ls[k], ls[0],
labels[s]) #Replace all ls[k] with ls[0]
l0 = labels[-1]
#if number of layers exceeds limit, save the first layer
if len(labels) > limsize:
mask = labels.pop(0)
fname = nfiles.pop(0)
ind = np.where(mask > 0)
lb = mask[ind]
np.save(
params.inputfolder + '../segmented/labels/glomeruli/' +
folder + fname[:-4] + '.npy', np.array([lb, ind[0], ind[1]]))
#save remaining layers
for i in range(len(labels)):
mask = labels.pop(0)
fname = nfiles.pop(0)
ind = np.where(mask > 0)
lb = mask[ind]
np.save(
params.inputfolder + '../segmented/labels/glomeruli/' + folder +
fname[:-4] + '.npy', np.array([lb, ind[0], ind[1]]))
t = pd.Series({'Labelling': time.time() - start})
t.to_csv(params.inputfolder + '../log/' + folder + 'Labelling.csv',
sep='\t')