Mesure color intensity of DAB image

I just wanted to confirm if all the steps are correct?
And if there are no intermediate steps between step (3) and (4) ?
On the photo: color 2 = I will directly do my measurements and I will take the Mean gray value.?? or I have to make the marking black? and how?
Thank you

Not really you need an image with neutral background (that is corrected) yours has a blue background so the colour separation will be incorrect.

Also concerning is why you want to measure the DAB intensity. Did you read the plugin warning message?:
Warning: This plugin is not suitable to quantify the intensity of immunostained slides because immunostains are not stoichiometric"

You can only use colour deconvolution in DAB stain for segmenting, not for quantifying amount of antigen.

2 Likes

ok thank you for your reply but i need more explanation please!!

  1. So how do I do to correct my images (image with neutral background)?
  2. Yes we use colour decovolution in DAB stain for segmenting … SO why then we segment the photos if we do not use for the quantification
    Can I have more informations how do I quantify/mesure the intensity of antigen

Thank you so much for your help

  1. Try the “a priori method” for brightfield correction:
    http://imagejdocu.tudor.lu/doku.php?id=howto:working:how_to_correct_background_illumination_in_brightfield_microscopy

  2. Antigens do not have “intensity”. They are there or not. There are also problems in finding how many copies of an antigen there are. Computing the concentration of a substance based on the Beer-Lambert law https://en.wikipedia.org/wiki/Beer-Lambert_law
    works for stoichiometric stains (like, for example, the Feulgen reaction for DNA) but not for non-stoichiometric ones for two reasons (one is obviously the non-stoichiometric nature of the Ag-Ab reaction and the amplification steps, the other is that DAB precipitates scatter light). See the “note” here:
    http://www.mecourse.com/landinig/software/cdeconv/cdeconv.html

which will also lead you to this other post:
https://list.nih.gov/cgi-bin/wa.exe?A2=ind0902&L=IMAGEJ&P=R18412

Some people ignore all this, do meaningless statistics on “DAB intensity” to claim that there is a statistical difference in expression of this or that and then get surprised when they get papers rejected.

5 Likes

Hi, according to your suggestion, how should we do quantification analysis of DAB staining, like IHC. Thanks a lot.

Hi, once you understand how IHC works, you will realise that it is not a good idea to attempt measuring “expression” of an antigent based on the intensity of the stain, so I would not do it.

So I wonder positive cell counting maybe a better choice to quantify the results. But it seems difficult if the expression in cytoplasma. So I’m quite confused. Do you have any suggestions?

IHC is more like “yes or no” test but “how dark” depends on too many variables.
Counting cells is less bad. For cytoplasmic stain, you can identify first the nuclei and then use them to define regions around then and test inside if there is any stain or not in that region.

Hi @11112 @Nass_Bella,

I think you may find a couple of lessons from this online course interesting in order to perform your analysis:

https://courses.edx.org/courses/course-v1:EPFLx+IPA4LSx+3T2019/course/

  • 7.2.1 Color & Color Models
  • 7.2.2 Quantifying DAB

In brief, as @gabriel pointed, there is no relationship between the quantity of DAB and the darkness, so quantifying the intensity is not a valid analysis. To overcome this, you can segment both the DAB and the tissue area of your interest. Then you can obtain the DAB-positive area fraction.

If you want to perform a cell counting (e.g., DAB+ vs DAB-), you might be interested in using QuPath.

Hope it helps

2 Likes

In brief, as @gabriel pointed, there is no relationship between the quantity of DAB and the darkness, so quantifying the intensity is not a valid analysis.

What I am saying is that there is no quantitative relationship between the darkness of DAB staining and the amount of antigen (i.e. the “expression”).

1 Like

Oh, yes, sorry about that

Dear Pau,

Thanks for your help. I can segment the colors by color deconvolution and got DAB positive area. But my question is if the protein is expressed in cytoplasma, how to evaluate it, it seems that cell counting is not suitable. By the way, I’m learning Qupath to auto count positive cell.

Best regards,
Huang Jumin

Pau Carrillo-Barberà via Image.sc Forum <imagej@discoursemail.com> 於 2020年5月18日 週一 上午12:26寫道:

If you are using QuPath, it really depends on the cell type and staining. Sometimes it is best to treat the whole cell, cytoplasm+nucleus as a “nucleus” using OD Sum to detect nuclei, and then use the level of DAB in the “nucleus” (which is now the whole cell) to decide which objects are positive. Alternatively, if there is a strong independent nuclear signal (the DAB doesn’t cover it up), you might be able to achieve more accuracy by finding areas that are DAB positive within the cell.

1 Like

Thanks for your suggestion. I got it. And I’m wondering whether we can calculate the %Area of DAB by doing color deconvolution.

Best regards,
Huang Jumin

MicroscopyRA (old Qupath forum) via Image.sc Forum <imagej@discoursemail.com> 於 2020年5月18日 週一 下午10:23寫道:

I generally do that by keeping the “expected spot size” at 1, min spot size at 1, and using the

Estimated spot count/Cell Area*100

as a new measurement. Note that this does exclude tiny speckles of DAB, but usually for this type of measurement I have found that to be a good thing, as I am looking for large contiguous regions of stain, not ISH spots.

It does require scripting, but there are plenty of examples of that floating around. Collected a few here.
https://gist.github.com/Svidro
See the Making measurements section.

Can we quantify analysis by %Area after color deconvolution with DAB segment? Thx!

If you only want to do percent area, without cell counts, that is even easier (once you have set your color deconvolution correctly) using the Thresholding tool.
https://qupath.readthedocs.io/en/latest/docs/tutorials/thresholding.html
The options for thresholding a brightfield image now include the color deconvolution channels. So you can either get percentage information within an annotated area (which you could also create with another thresholder. For example a tissue detection thresholder followed by a DAB thresholder), or create objects based on a threshold and do further processing with those.

Color deconvolution:

1 Like

Many thanks! %Area seems much easier, but I’m not sure whether it can be accepted by experts or magazine.