Shape features#
In this notebook we will compare the impact of small manual modifications to labels/outlines on measured shape features.
import numpy as np
import stackview
import napari_skimage_regionprops as nsr
from skimage.morphology import disk
For demonstration purposes, we use a blank image and a circular region of interest.
image = np.zeros((50,50))
labels = np.zeros(image.shape, dtype=np.uint32)
labels[9:40, 9:40] = disk(15)
stackview.insight(labels)
|
|
We can print out the roundness and circularity of this single object like this:
stats = nsr.regionprops_table(image, labels, shape=True)
print("roundness: ", stats['roundness'][0])
print("circularity: ", stats['circularity'][0])
roundness: 0.9996522134693254
circularity: 0.9106696648110818
Manual modification#
When executing the next cell, a small user-interface will open that allows you to modify the object. Change the Eraser radius to 1 and hold the ALT-Key while clicking to remove individual pixels from the object.
stackview.annotate(image, labels, zoom_factor=4)
When you are done modifying the object, continue executing the next code cells.
stackview.insight(labels)
|
|
As you can see, small modifications to the outline can have huge impact on individual metrics.
stats = nsr.regionprops_table(image, labels, shape=True)
print("roundness: ", stats['roundness'][0])
print("circularity: ", stats['circularity'][0])
roundness: 0.8564505192042108
circularity: 0.13777164107308135
Exercise#
Modify the code above to also measure aspect ratio and solidity.