•	Ex 4.2: Threshold-based pixel classifier

4. Tissue segmentation and classification

4.1 Fully manual tissue segmentation

The most basic way to locate your objects is to create the polygons by yourself...

As you may expect, this approach is time-consuming, user dependend and becomes nearly impossible depending on what you need to find. This technique can be used whether you need to locate your objects of interest, or sub-areas within them.

4.1.1 Glomerulus-to-kidney ratio

Goals:

• The goal of this exercise is to process the glomerulus-to-kidney ratio for all images.
• To do so, you will need to:
  • Find the polygon corresponding to the whole kidney and give it the Kidney classification.
  • Find the set of polygons representing glomerulus and give them the Glomerulus classification.
  • Repeat these steps on two or three images.
  • Export the measurements to a TSV so you can process the ratio per image in a spread sheet software.
• After that, we will switch to some lungs images. They will demonstrate a case where manual segmentation has too many limitations to produce a usable result.

Required data:

Folder	Description	Location	License
Histology (PAS-HE-IHC)	H-DAB images of kidneys into which glomerulus are visible and lungs into which alveolus are present	DOI: 10.1038/s41467-023-40291-0	MIT

Note

Keep your project tidy for this exercise: we will use it as an input for another exercise!

A. Locate the kidney

• On these images, kidneys are solid circular blocks with no to little holes (so we will ignore them).
• Before starting, make sure that you have the Kidney and Glomerulus classes in your class list. If you miss them, you can generate them using the "+" button above the class list.
• To locate your kidneys, you can either use the ellipse tool or the wand tool.
• Like with any selection tool, you can push your active polygon from inside (left click) or from outside (alt+left-click).
• Once you are done, before going to the next image, don't forget to:
  • Assign the Kidney class to the annotation
  • Lock it with + or > "Lock".
  • Save the project.

Tip

Don't forget that the scale at which your wand tool works depends on your zoom level.

B. Locate the glomerulus

• What the glomerulus look like is indicated by the green blinking polygons in the first section of this exercise.
• To create polygons around the areas of interest, you have two options. You can either go with the wand tool or with the polygon tool.
• Don't forget to give the correct class to every new annotation you make.
• Once you selected every area of interest, take a look at your annotations list and:
  • on the first Glomerulus polygon of the list.
  • + on the last Glomerulus polygon.
  • Click on the next to "Select all" and "Delete".
  • In the menu showing up, go to "Edit multiple annotations" > "Merge selected".
• This should have merged all the polygons into one. Don't forget to lock the newly created polygon.
• You can now save the project (+) and go to the next image.

You should end up with polygons looking like that:

C. Export and process the ratio

• Make sure that you saved your project (+) before starting this step.
• If you click on any polygon from your annotations list, you should notice an Area px^2 property.
• In the top-bar menu, go to "Measure" > "Export measurements".
• In the list of images, transfer to the right column every image onto which you drew polygons.
• Choose a path for your TSV file. Keep it named measurements.tsv and leave it in the project's folder.
• The "Export type" indicates where the useful information is located. In our case, we want the areas, and we saw that they are present in annotations.

Warning

If you don't save your project after you made polygons on the last image, they won't be exported in the final results even if the polygons are visible.

D. In LibreOffice Calc

• If you open the produced TSV in your spread sheet software, you should notice that you have:
  • Two consecutive lines for each image (column "Image")
  • One with the Kidney class and the other with Glomurus (column "Classification")
  • The area for each one of them (column "Area px^2")
• If you want, to make the file clearer to read, you can remove useless columns (Object ID, Parent, ...)
• Create a new column that will contain the ratio (ex: "G to K ratio")
• Write the expression to divide the Glomurus area by the whole Kidney area and apply it to all images.
• Optionally, pass the formatting of the column to percentage ("Format" > "Number format" > "Percentage").

4.1.2 Locate lungs or crypts in large intestine

Goals:

• The overall goal here is to get an idea of the limitations that you will encounter if you stick to manual segmentation. Don't try to finish the tasks.
• The goal of this exercise is to:
  • Try to manually segment the lungs tissue present on an image.
  • Try to manually segment the crypts within some large intestine images.
• To do so, you will try to:
  • Find the polygon corresponding to the whole lung and give it the Lung classification.
  • Find the set of polygons representing crypts and give them the Crypt classification.

Required data:

Folder	Description	Location	License
Histology (PAS-HE-IHC)	H-DAB images of kidneys into which glomerulus are visible and lungs into which alveolus are present	DOI: 10.1038/s41467-023-40291-0	MIT

A. Lungs

• Create a new project.
• In the dataset, alongside the kidneys images, you should find both lungs and large intestine images.
• Import a couple of lung images into this new project.
• Lungs, contrary to kidneys, are very sparse. We cannot just make a whole circle around them.
• Try (without finishing, don't spend more than 20 seconds here) to use the wand tool to make a polygon representing the object.

B. Crypts in large intestine

• Still in the same project, import some large intestine images.
• The shape by itself is not as bothering as for lungs, but the crypts to be segmented are hard to catch with the wand. You would have to make a manual polygon around each one (again, don't spend more than 20 seconds to make polygons around crypts).

In the following exercises, we will cover more efficient ways to find our objects and to locate sub-structures within them.