Is pixel classifying on positive cell detection possible?

Hi dear QuPath users,

My goal is: get a ratio of positive cells among tumor cells and immune cells.

I have Ki-67 stained slide and I want to know: In my tumor area of interest, how many of the tumor has Ki-67, and how many of the immune cells has Ki-67.

However, pixel classifier and positive cell detection produce mesurements independently, even applied to the same area.

Is there any ways to conjugate those mesurements?

Thanks you so much,

Ibrahim

Hi @ibrahim_uygun,

I think you might be interested in the Object classifier. It allows you to further classify your cells (and therefore has to be applied after cell detection). You can find details of how to use it here in the docs.

3 Likes

Thank you so much, it’s working as I want to, even tho i have some issıes with accuracy. (Bcuz if i train object classifier in H&E, it doesnt apply correctly to Ki-67 slides. And I also cant train it in Ki-67 bcuz I cant say which one is immune cell or tumor cell.:/)

Another quick question: In positive cell detection, we can only put threeshold for stain amount. Can we set the threeshold based on how big the cell is? ( bcuz in some cases it might be more helpful to differenciate tumor cells and immune cells)

Thank you.

I learned a way to set the threeshold as a cell area, or any measurement. (set cell intensity classification.) Thank you.