def find_nuclear_nucleoli(pixels: np.array):
"""
Detect nuclear in nucleoli stain images
:param pixels: np.array, nucleoli stain image
:return: label_nuclear_sort: np.array, grey scale labeled nuclear image
"""
bg_int = ana.get_bg_int([pixels])[0]
# nuclear detection
markers = np.zeros_like(pixels)
markers[pixels < (1.7 * bg_int)] = 1
markers[pixels > (3 * bg_int)] = 2
seg = random_walker(pixels, markers)
# nuclear binary mask
nuclear = np.zeros_like(pixels)
nuclear[seg == 2] = 1
# fill holes
nuclear_fill = ndimage.binary_fill_holes(nuclear)
# separate touching nuclei
label_nuclear = obj.label_watershed(nuclear_fill, 1)
# filter out:
# 1) nuclear size < 1500
# 2) nuclear touches boundary
label_nuclear_ft = clear_border(label_nuclear)
label_nuclear_ft = obj.label_remove_small(label_nuclear_ft, 1500)
label_nuclear_sort = obj.label_resort(label_nuclear_ft)
return label_nuclear_sort
def find_nuclear(pixels: np.array, local_thresholding_size: int, min_size: int,
max_size: int):
"""
Detect nuclear in nuclear stain image
:param pixels: np.array, nuclear stain image
:param local_thresholding_size: int, running size for local thresholding, varies for different magnification etc.
:param min_size: minimum allowable nuclear size
:param max_size: maximum allowable nuclear size
:return: nuclear: np.array, 0-and-1, nuclear mask with nuclei identified
label_nuclear: np.array, labeled image of nuclear mask
"""
# nuclear identification using local thresholding
local = threshold_local(pixels, local_thresholding_size)
# for Jose data under 60x objective, local_thresholding_size = 701
nuclei_local = pixels > local
for i in range(2):
nuclei_local = binary_erosion(nuclei_local)
for i in range(2):
nuclei_local = binary_dilation(nuclei_local)
nuclei_local = ndimage.binary_fill_holes(nuclei_local)
nuclei_local = remove_small(nuclei_local, min_size)
nuclei_local = remove_large(nuclei_local, max_size)
# for Jose data under 60x objective, min_size = 15000
# nuclear identification using background intensity
bg_int = ana.get_bg_int([pixels])[0]
nuclei_bg = pixels > 1.05 * bg_int
nuclei_bg = ndimage.binary_fill_holes(nuclei_bg)
nuclei_bg = binary_erosion(nuclei_bg)
nuclei_bg = binary_dilation(nuclei_bg)
nuclei_bg = remove_small(nuclei_bg, min_size)
nuclei_bg = remove_large(nuclei_bg, max_size)
# nuclear mask
nuclear = np.zeros_like(pixels)
nuclear[nuclei_local == 1] = 1
nuclear[nuclei_bg == 1] = 1
label_nuclear = label(nuclear)
return nuclear, label_nuclear
pointer_pd['num_ctrl_spots'] = [num_ctrl_spots] * len(pointer_pd)
# get raw intensities for bleach spots and control spots
pointer_pd['raw_int'] = ana.get_intensity(
label(bleach_spots, connectivity=1), pixels_tseries)
ctrl_pd['raw_int'] = ana.get_intensity(
label(ctrl_spots, connectivity=1), pixels_tseries)
ctrl_pd['pos'] = [pos] * num_ctrl_spots
# link ctrl spots with corresponding organelle
ctrl_pd['%s' % analyze_organelle] = obj.points_in_objects(
label_organelle, ctrl_pd['x'], ctrl_pd['y'])
print("### Image analysis: background correction ...")
# background intensity measurement
bg_int_tseries = ana.get_bg_int(pixels_tseries)
pointer_pd['bg_int'] = [bg_int_tseries] * len(pointer_pd)
# background intensity fitting
bg_fit = mat.fitting_linear(np.arange(0, len(bg_int_tseries), 1),
bg_int_tseries)
pointer_pd = dat.add_columns(
pointer_pd,
['bg_linear_fit', 'bg_linear_r2', 'bg_linear_a', 'bg_linear_b'],
[[bg_fit[0]] * len(pointer_pd), [bg_fit[1]] * len(pointer_pd),
[bg_fit[2]] * len(pointer_pd), [bg_fit[3]] * len(pointer_pd)])
# background correction
# use original measurement if fitting does not exist
if np.isnan(bg_fit[2]):
bg = bg_int_tseries
sg_tseries.append(sg)
if 1 in sg:
label_sg = label(sg, connectivity=1)
sg_props = regionprops_table(label_sg, pix, properties=('label', 'area', 'mean_intensity'))
sg_props1 = regionprops_table(label_sg, pix1, properties=('label', 'area', 'mean_intensity'))
sg_pd = pd.DataFrame(sg_props)
sg_pd1 = pd.DataFrame(sg_props1)
num_temp = len(sg_pd)
num.append(num_temp)
size_temp = np.mean(sg_pd['area'])
size.append(size_temp)
raw_int_G3BP1_temp = np.mean(sg_pd['mean_intensity'])
raw_int_G3BP1.append(raw_int_G3BP1_temp)
bg_G3BP1_temp = ana.get_bg_int([pix])[0]
bg_G3BP1.append(bg_G3BP1_temp)
raw_int_sample_temp = np.mean(sg_pd1['mean_intensity'])
raw_int_sample.append(raw_int_sample_temp)
bg_sample_temp = ana.get_bg_int([pix1])[0]
bg_sample.append(bg_sample_temp)
raw_int_G3BP1_full = raw_int_G3BP1_full + sg_pd['mean_intensity'].tolist()
raw_int_sample_full = raw_int_sample_full + sg_pd1['mean_intensity'].tolist()
x_frame = x_frame + [i] * len(sg_pd)
bg_G3BP1_full = bg_G3BP1_full + [bg_G3BP1_temp] * len(sg_pd)
bg_sample_full = bg_sample_full + [bg_sample_temp] * len(sg_pd1)
else:
num.append(0)
size.append(0)
raw_int_G3BP1.append(0)
# measure stress granule properties in both channel
sg_props = regionprops_table(label_sg,
pix,
properties=('label', 'area',
'mean_intensity'))
sg_props1 = regionprops_table(label_sg,
pix1,
properties=('label', 'area',
'mean_intensity'))
sg_pd = pd.DataFrame(sg_props)
sg_pd1 = pd.DataFrame(sg_props1)
# measure total intensity
fov_G3BP1 = regionprops(fov, pix)[0].mean_intensity
fov_sample = regionprops(fov, pix1)[0].mean_intensity
# measure background intensity
bg_G3BP1 = ana.get_bg_int([pix])[0]
bg_sample = ana.get_bg_int([pix1])[0]
# add information into list
pos_lst = pos_lst + [pos] * len(sg_pd)
well_lst = well_lst + [well] * len(sg_pd)
frame_lst = frame_lst + [frame + frame_offset] * len(sg_pd)
label_lst = label_lst + sg_pd['label'].tolist()
size_lst = size_lst + sg_pd['area'].tolist()
raw_int_G3BP1_lst = raw_int_G3BP1_lst + sg_pd['mean_intensity'].tolist(
)
bg_G3BP1_lst = bg_G3BP1_lst + [bg_G3BP1] * len(sg_pd)
fov_G3BP1_lst = fov_G3BP1_lst + [fov_G3BP1] * len(sg_pd)
raw_int_sample_lst = raw_int_sample_lst + sg_pd1[
'mean_intensity'].tolist()
bg_sample_lst = bg_sample_lst + [bg_sample] * len(sg_pd)
fov_sample_lst = fov_sample_lst + [fov_sample] * len(sg_pd)