Objective/unbiased identification of stains vectors for multiple stains

We need to differentiate between 3 stains for our analysis, therefore the Estimate stain vectors function does not work. I tried to calculate the intensity of the ROI but it only gives the mean value for each stain (see photo below)

I was wondering if there was a way to convert or determine the RGB values which were used to calculate the mean stain per pixel in an annotation from the compute intensity features function. We would ideally input the separate red, green and blue values into the Set stain vector window (see image below)

Alternatively, if this is not possible, is there a way to know the full color range (min/max RGB) for each stain (e.g. hematoxylin)

It is best to determine the vectors from single-stained (and background-corrected) slides.
What stain are you trying to analyse?

I suggest reading the section on Determining new vectors at:


You can attempt to separate the stains, and sometimes quite successfully, by changing the image type to Brightfield Other. This should not be taken to mean that any resulting intensity values are quantitative. However, it can be quite effective if you want to classify a certain cell as being primary blue, or purple, or whatever. More information on setting stain vectors here, check the YouTube link for details on setting individual color vectors from training annotations.

Basically, though, brightfield stain measurements are not terribly quantitative.
Whoops, realized I totally misread that. What I get for answering after getting back from the pub.
In case it helps, you can choose any one pixel, look at it’s RGB values in the lower right corner, and see exactly what color vector is calculated from that pixel.

Here I created a one pixel square, double clicked on the hematoxylin channel, and got the resulting numbers. There will never be a min/max RGB color value per stain, that isn’t how they work. A completely black 0,0,0 pixel will still have a certain percentage of hematoxylin based on the color vector. It’s more like a percentage, though that isn’t really a great description of a three dimensional measurement.

The problem with brightfield is that everything is interdependent. If you want three decent separate measurements, you use fluorescence.