Detection of multinucleated giant cells in QuPath

Giant cell.tif (3.2 MB)

Hello! I am completely new here and just started learning how to work with QuPath, I hope you can help me solve my problem. I uploaded a picture of mouse lung section stained for H&E. There are multiple multinucleate giant cells in this section, I chose a specific region to show you one of the cells of interest, the one visible in the center of the picture.

My goal is to automatically calculate all the mulinucleate cells in the section. Is there a way to teach QuPath to mark only the cells of certain size and with multiple nuclei?

Thank you very much for your help!
Kate

Hmm, if they are all smooth like that one in the center, you might be able to use the pixel classifier to create objects of about that size, but cell detection is not designed to work that way. One nucleus, one cell. Your analysis would probably be easier within QuPath using a specific marker for your cell of interest.

If the pixel classifier does not work, you may want to try Weka, Ilastik, or a deep learning model.

*If they are all approximately the same size, SLICs might work as well…

2 Likes

You could maybe try Stardist to find all cells and then an object classifier? If you set your pixel size correctly, it might think that whole cell is 1 round object. For this strategy, you need to either be comfortable with scripting or wait for QuPath version 0.3.0 (hopefully available soon).

2 Likes

Excellent point, I recall StarDist being used in a similarly creative way here:

2 Likes