Definition of mean intensity and staining comparison in QuPath

Hello, I am trying to understand better the quantification process to report my data obtained using QuPath.
I have a bright-field image stained with Alcain blue and PAS. After separate both staining I used the intensity tool in Qupath to record the mean intensity of my annotations.
Now I have the results per image.
I am wondering,
1: Can I compare the intensity mean between one channel and the other?.
2: In each image, I have about 500 annotations with it corresponding mean intensity are in um2.
What would be the best way to represent the data when comparing two cohorts of patients?

Hope I was clear, I am sorry if this is repeated.

  1. No. It is tricky to even compare brightness of one channel between two images, as there isn’t a linear relationship between brightfield stains and any meaningful quantity. I think that your best option is to have someone familiar with the biology and the stain determine thresholds, and then quantify cell counts or areas that are higher or lower than that intensity of stain. @gabriel might go into more details or provide some references :slight_smile:

  2. Similar to 1, though I am not 100% clear on what you are saying. You are best off classifying the areas as positive or negative vs some threshold, and comparing the areas after thresholding. If there is sufficient variation in the areas, you probably want to use some sort of tiling, like SLICs and classification.
    QuPath Guide: Creating classified annotation areas using SLICs
    Though in your case, you might want to create some sort of high, medium, low staining threshold, and create sub-annotation areas based on that.

You need to know two things: whether the stain binds stoichiometrically, and if so, whether the the dye follows the Beer-Lambert law.
Some histochemicals stains might be stoichiometric, you need to do some research on this.
IHC is not, H&E is not.
On top, for IHC, DAB does not follow Beer-Lambert law.
Feulgen is stoichiometric for DNA contents.
Phalloidin is stoichiometric for actin (fluorescence applications)

Therefore you cannot use H&E e.g. to measure hyperploidy, regardless of the nucleus staining darker (hyperchromatic). For that you need to use e.g. Feulgen stain.

Colour separation is useful for segmentation. As you probably can tell now, quantifying stain darkness is not straightforward.


Neat, didn’t know that about Feulgen (*had never heard of either).

Thanks Gabriel for reply,
Actually both of the staining are stoichiometric for mucin. This colorant are usually used in solution to quantify mucin using spectrophotometers. However I am still doubting. I thought that by using this two staining methods it would be more straight forward. It looks as per the reply from Research_Associate it would safer to just use categories. I am new to Qupath and I’m struggling to use SLICs (even with the tutorials). I 'll se if can find some help on the analysis of Fuelgen stain.

1 Like

I’ll try to apply SLICs.

1 Like

Let me know if there is anything specific that could be improved in the tutorials, need feed back to make them better!

So here is a fun question. Even though each stain is stochiometric, are the unmixed channels stochiometric for pixels in which both stains are present when imaged on an RGB camera?

Great question!.
My first thought on this would be no. I would research this. However this reminds me RGB images in general are no good to quantify intensities.
Thanks for bearing with me.
I am working my way through your tutorial on SLIC. I’ll give comments as you requested.

1 Like