Edges of labels

When processing biological objects in image such as cells and nuclei, it may make sense to identify all pixels that lie on the surface of an object. This notebook demonstrates how to select pixels on the border of nuclei, just in case we would like to measure intensity in the nuclear envelope.

import pyclesperanto_prototype as cle
import numpy as np
from skimage.io import imread
import stackview

image = cle.asarray(imread("../../data/mitosis_mod.tif")[0:40,25:65])
image

cle._ image shape (40, 40) dtype float32 size 6.2 kB min 12.0 max 255.0

We then segment the nuclei.

label_image = cle.voronoi_otsu_labeling(image, spot_sigma=2, outline_sigma=1)
label_image

cle._ image shape (40, 40) dtype uint32 size 6.2 kB min 0.0 max 4.0

From the nuclei label image we can extract another label image which contains all pixels that are on the edge of the labels.

edge_label_image = cle.reduce_labels_to_label_edges(label_image)
edge_label_image

cle._ image shape (40, 40) dtype uint32 size 6.2 kB min 0.0 max 4.0

In case one wanted to measure in thicker areas along the borders, we could expand the borders.

thicker_edges = cle.dilate_labels(edge_label_image, radius=1)
thicker_edges

cle._ image shape (40, 40) dtype uint32 size 6.2 kB min 0.0 max 4.0

For visualization purposes we can also view the original image with the label borders on top.

stackview.imshow(image, continue_drawing=True)
stackview.imshow(edge_label_image, alpha=0.6, labels=True)

It is also possible to retrieve a 2-pixel thick boundary binary image of edges between labels and backgroaund.

cle.detect_label_edges(label_image)

cle._ image shape (40, 40) dtype uint8 size 1.6 kB min 0.0 max 1.0