How does Porespy measure the combined pores that are not perfectly circular?

[takeruhukushima]

Hellow, @jgostick.

Previously, I calculated the pore size using porespy from the porous material I was researching as follows:

# Apply the porosimetry filter
im_filtered = porespy.filters.porosimetry(thresholded_image, sizes=sizes)

# Calculate the pore size distribution on the filtered image
pore_size_distribution = porespy.metrics.pore_size_distribution(im=im_filtered, log=False, bins=100, voxel_size=pixel_size_nm)

When I presented the results as part of a class assignment, a professor asked me, "How do you measure the combined pores that are not perfectly circular?"

To put it extremely, what kind of behavior does Porespy exhibit toward pore diameters that are somewhat awkward, resembling carbon sp hybrid orbitals viewed from the side?

I would appreciate your response when you have time.

[jgostick]

Hi @takeruhukushima
Pretty much ALL pores are not perfect spheres, so have various odd shapes like being ovoids or prisms. Drawing spheres is easy though, that is a quick and convenient way to approximate pore sizes. If you really want to get details you can do a watershed segmentation on the image and then analyze the shapes of each region individually. We have a bit of functionality for this in porespy. For instance, you can use snow partitioning to get the watershed image, then use the use regionprops_3d to get info about each region, like the sphericity for instance. Hope that helps.

[takeruhukushima]

Thank you for your response.

When there are pores resembling sp hybrid orbitals within an image, how does watershed segmentation divide them into two regions? I still don't understand it intuitively.

[jgostick]

The watershed segmentation is covered by a variety of internet articles, but is basically finds constrictions and uses them to define the boundary between cavities, aka pores.

[takeruhukushima]

Thank you. After implementing it, it worked fine.

[takeruhukushima]

I'll study the principles by reading that article. Thank you very much.