Positive cell detection per mmsq after QuPath classifier applied

New to using the QuPath classifier and have a possibly stupid question… With fast cell counts etc. QuPath outputs positive cells/mmsq. Some of my TMA cores have quite a bit of haemorrhage which has made getting an “average” value that works well across all more cores quite difficult (some cores quite pale and myxoid, other very cellular). Training a classifier has worked quite well to get around this problem. However I see I no longer get positive cells per sqmm as an output for the TMAs. I guess I could work it out in excel for each core afterwards. But there are thousands of TMA cores and I would prefer it to be outputted from the start. Is there anyway to do this? Thanks and apologies if the answer is obvious or has been asked somewhere before…

If you want to limit the area within the TMA cores you can use Simple Tissue Detection, which helps make density counts more accurate.

I use this Measurements script to calculate values for all existing detection classes. There are a few other scripts there that might help as well. Note that it applies the value to the annotation, but not to the TMA, you can then adjust it to target TMA cores or parents instead of annotations.

This thread and a couple of the other scripts within the above link should show how a similar script will work for TMA cores.

Also, when you created your classifier, did you call the classes Positive and Negative? I’m afraid that is required for the auto calculation and update.

The above was done using the Classify by Specific Feature option in the Classify menu, after cell counts were created.

If you choose any other names (other than the +1 2, etc for multi threshold), you would need to add the measurements yourself.


Thanks for that. I will take a look.

1 Like