Connected component labeling on surfaces

Connected component labeling on surfaces#

This notebook demonstrates how to differentiate objects according to their connectivity.

import napari_process_points_and_surfaces as nppas
import pyclesperanto_prototype as cle
import napari_simpleitk_image_processing as nsitk
from skimage.data import cells3d
import stackview

We use a 3D image of nuclei…

image = cells3d()[:,1]
stackview.insight(image)

shape	(60, 256, 256)
dtype	uint16
size	7.5 MB
min	0
max	65535

… and segment the nuclei resulting in a 3D binary image.

labels = cle.voronoi_otsu_labeling(image, spot_sigma=9)
binary = cle.erode_labels(labels) > 0

stackview.insight(binary)

shape	(60, 256, 256)
dtype	uint8
size	3.8 MB
min	0
max	1

We convert this binary image into a surface dataset.

surface = nppas.all_labels_to_surface(binary)
surface

nppas.SurfaceTuple

origin (z/y/x)	[0. 0. 0.]
center of mass(z/y/x)	34.703,124.973,131.513
scale(z/y/x)	1.000,1.000,1.000
bounds (z/y/x)	16.500...59.000 0.000...255.000 0.000...255.000
average size	97.003
number of vertices	151354
number of faces	301006

By applying connected component labeling to the surface, we can identify vertices/faces that are connected and differentiate those which are not. The result is also a surface dataset where the vertex values correspond to the nth label these objects belong to. Thus, you can conclude from the maximum number of this surface that thereare 38 nuclei in this image.

surface_connected_components = nppas.connected_component_labeling(surface)
surface_connected_components.cmap = 'hsv'
surface_connected_components

nppas.SurfaceTuple

origin (z/y/x)	[0. 0. 0.]
center of mass(z/y/x)	34.703,124.973,131.513
scale(z/y/x)	1.000,1.000,1.000
bounds (z/y/x)	16.500...59.000 0.000...255.000 0.000...255.000
average size	97.003
number of vertices	151354
number of faces	301006
min	0
max	38