Standardization for Cell Detection

Image Analysis
#1

Hey QuPath community,

My pipeline has come a long way thanks to the community’s help but one thing I have overlooked was standardizing or preprocessing images before analysis.
I feel like the cell detection analysis tool does a very good job at detecting cells but I am wondering if there are any steps I should take before running the cell detection on my images.

For background, the images that I am running cell detection on are H&E stained slides where the nucleus shows up with a darker purple stain.

image
image834×570 270 KB

Here is the result with the parameters I have been using.

image
image1375×708 617 KB

Is there a better way to go about this?

Thanks in advance!

#2

I can attach a full image file in a minute if anyone would like to try it themselves!

#3
#4

You did not mention it, but probably the most important thing to do when using anything that relies on color deconvolution is to set the stain vectors correctly - which hopefully will be consistent across an entire project, but is not always. You may have to set them manually or figure out a script if your staining is inconsistent.
https://qupath.readthedocs.io/en/latest/docs/tutorials/separating_stains.html?highlight=stain%20vectors
And or a video

2 Likes