This is what I do for DAB quantification on slides with hematoxylin counterstain:

1. Download the free "Fiji" version of ImageJ from http://fiji.sc. All subsequent steps are performed in Fiji.

2. Open a DAB image.

3. Run Image > Color > Colour Deconvolution.

4. From the Vectors pull-down, choose "H DAB" as the stain (this assumes your images are correctly white-balanced, otherwise you have to define your own colors by choosing "From ROI" for the vector then drawing your own Regions Of Interest to define the stain colors).

5. Click OK in the Colour Deconvolution window, you will get three new images. The one with "Colour_2" in the title is the DAB image (Colour_1 is the hematoxylin image), you will quantify the Colour_2 image.

6. Run Analyze > Set Measurements and select "Mean gray value" and "Display label".

7. Select the Colour_2 image window.

8. Run Analyze > Measure (or press Ctrl-m); a "Results" window will pop up with the quantification in units of intensity.

9. You need to convert the intensity numbers in the Results window to Opitcal Density (OD) numbers with the following formula:

OD = log(max intensity/Mean intensity), where max intensity = 255 for 8-bit images.

This will quantify the average darkness of the image due to DAB signal. If you have areas of the image without tissue it will bias your results because it will bring the average OD down. If that is the case we need to discuss further. Also, you may need to use Integrated Density rather than Mean.

Good luck and let me know if you need anything else.

I've used CellProfiler for staining measurements: http://www.cellprofiler.org/. There are some very nice tutorials and "pipelines" for many different types of staining and tissues. Most of these methods follow what you describe above: thresholding followed by pixel intensity measurements. Unfortunately, it seems there is a bit more info for immunofluorescence rather than colourimetric IHC.

look at hierarchical normalized cuts (HNcuts) -- http://www.ncbi.nlm.nih.gov/pubmed/22180503

Try ImmunoRatio, which is a free web application (and also freely downloadable) at http://153.1.200.58:8080/immunoratio/

See the publication by Tuominen V et al. 2010

breast-cancer-research.com/content/12/4/R56

Tens of thousands of images analyzed worldwide. I have used it in clinical diagnostics since 2007.

Best regards,

Jorma Isola

SigmaScan is pretty good too, although ImageJ (which I use too) does pretty much the same things.

I have added some comments direct from the Image-J program about quantitative estimation of DAB products from immunostaining. As You can see is the conclusion using such methods is not useful Citate"Note: A very frequently asked question in the ImageJ mailing list relates to quantification of immunostain intensity (for example DAB intensity) to evaluate antigen expression. However, one needs to consider 2 main issues that prevent doing this in a quantitative manner:

1. Antigen-antibody reactions are not stoichiometric, so "darkness of stain" does not mean "amount of reaction products". In fact most histological stains are not stoichiometric (one exception is Feulgen stain det er jo ikke en antistof reaktion which is commonly used for DNA cytometry).

2. In particular DAB does not follow Beer-Lambert law. See, for example, CM van der Loos paper:

"The brown DAB reaction product is not a true absorber of light,

but a scatterer of light, and has a very broad, featureless spectrum.

This means that DAB does not follow the Beer-Lambert law, which describes

the linear relationship between the concentration of a compound and

its absorbance, or optical density. As a consequence,darkly stained

DAB has a different spectral shape than lightly stained DAB."

Further details are mentioned in this informative message posted by Al Floyd to the ImageJ mailing list.

Therefore, while colour deconvolution might be useful to segment immunostained structures or for image enhancement for colour blind observers, attempting to quantify DAB intensity using this plugin is not a good idea. end of Citat

I must say that I did not read the initial request properly (sorry)... I have used SigmaScan and ImageJ to count cells positively stained by Fluorescence or DAB, NOT too measure intensity of the staining, which I agree is not reliable and surely not quantitative.

Cheers,

Rv

Thank you all for your answers!

I guess the way to go will be to use a program such as ImageJ to select and measure the total area of stained cells or counting the stained particles in the samples. That way I can at least compare the proportion of each antigen to the other in the same area. I will have trouble comparing the quantity/intensity of staining between samples, it seems 0 vs 1 will have to do.

Of course I could always assess the staining subjectively...

I will go slighlty out of the topic. Has anyone tried using stereology method? Whichever way you do the intensity measurements or counting, if you are going to do it in an unbiased way, it will give you good results. Yes, I completely agree that counting is subjective, but counting in an unbiased way (by random and systematic section selection, random and systematic imaging and random and systematic counting) will always give you reliable results. I have seen some good stereology softwares for quatitative IHC with visiopharm (visiopharm.com). There are other softwares too which can do it. Good luck.

Is DAB an absolute essential for your studies?

Good question, for counting I prefer fluorescence, and also automatic counting using software allows usage of large or the whole section and not only "chosen" areas (unless the ones you use are the only relevant ones).

Cheers

^To Mr. Cebak. Regarding the DAB. Well not really, but all the stainings that have been done so far are DAB. I think they're the most common and routine IHC type staining that our pathology department does. Would you recommend some other staining?

^To Mr. Srinivasan. True, I guess there are means to minimize bias of manual assessment, I might look into doing that if the area measurements wont work.

Thank you for your answers.

Mr. Egilsson, Check out this paper. The author's use Integrated Density (a parameter of luminosity--though luminosity is somewhat of a misnomer) to measure DAB staining. Keep in mind that DAB staining does not follow the Beer-Lambert law that Mr. Albrectsen pointed out.

Article Endogenous Hedgehog Expression Contributes to Myocardial Isc...

I've been wondering a lot about this and reading this paper and some other comments regarding this it seems like quantification by measuring intensity on DAB stainings are not out of the question but provide a rough estimate. As has been mentioned it's because intensity of signal does not exactly reflect the concentration of antigen.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2326109/

"Having different spectra at different staining intensities of the chromogen also creates a potential problem with quantification of the immunostaining results. In this respect, the DAB chromogen is less suitable for quantification than, for example, LPR, which shows very similar spectra at high and low staining intensities (Figure 9), as is expected from a compound that follows the Beer-Lambert law.

The concerns about the DAB reaction product mentioned above certainly do not exclude DAB from a double staining application unmixed by spectral imaging. Its crisp reaction product is ideal for a good microscopic resolution, and dark brown staining intensities can be simple avoided by diluting the primary antibody of interest."

Mr Hugo AM Torres provides a convincing argument on the imageJ board:

"As G. Landini and Al Floyd mentioned, DAB is not in rigor a truly

absorptive material and you might have trouble publishing your data if

your method uses densitometry quantification based on the law of

Lambert-Beer.

It is very common, not withstanding, to find in situ hybridization

quantification of brains slices dipped into autoradiographic emulsions.

Silver granules, which likewise, do not satisfy Lambert-Beer criteria,

is quantified by densitometry in darkfield microscopy where light is

reflected!(A quick pubmed search will reveal)

Now, I don't want to justify an error with another one, the question is

whether there are more white pixels created by reflective silver spots

where there are molecules of interest. The rule that DAB can not be used

at all might not be a general rule and be more like a rule of thumb as

this can vary from antibody to antibody according to Watannabe et al(See

below). In case positive, although one can not estimate the true number

of molecules in situ on the basis of Lambert-Beer law, one would in

principle be able to compare rough relative amounts (e.g. a

semi-quantitative estimate of the proportion of molecules) in this

tissue site as opposed to another tissue site subjected to the same

conditions by photometric densitometry (granted exposure to the same

solutions, for the same amount of time, same illumination settings, lab

temperature, etc). "

http://imagej.1557.n6.nabble.com/regarding-IHC-PROBLEM-td3693036.html

Bypassing the issue of quantifying the staining density (which I agree is probably wrong) getting a measure of the surface covered with stained cells is probably OK. You already use Imagej, so why not use the colour deconvolution module written by gabriel landini? It si included in the Fiji distribution of ImageJ (Image menu-.Color-.color deconvolution) and is very good at separating dyes. follow this with a standard theshold and you're done.

A possibly slightly more quantitative signal is of course fluorescence. it does has it;'s problems as well of course, the main one being bleaching of the signal during observation. However if the image capture is performed correctly, you shou;d get a more quantitative readout.

If you are doing a broad IHC study, try using tissue micro array slides, compare the relative intensities.of the cores on the same slide. I enjoy scoring manually but put together with the above, your data will hold greater validity.

Your section thickness will have minimal variability as will your antibody exposure times/concentrations. We can make TMAs with cores close to slide label end so test sample can be floated onto the remainder of the slide which could help but the TMA and sample sections will probably differ in thickness.

Has any one tried Leica image analyser?

Hi Wafeay,

I have although it is quite a few years ago now and mostly use ImageJ these days. I remember it worked OK and as I had one of the more advanced versions the built in scripting language was very useful for creating routines for not-so-computer-savvy other people in the lab. However, it was rather expensive and I can't think of anything ImageJ can't do that Leica QWin can, while there's quite a bit for which the opposite holds true. Also the ImageJ community is very helpful for sorting out problems and fixing bugs. At the time there was nothing like that for QWin, but maybe there is now. If you use Leica microscopes it will probably be good for automatic control of the microscopes, but as I use Zeiss systems that's of no use to me.

Thanks Rob for your reply. I am new with image analyser use in immunohistochemistry. We will have a Leica software sooner. We have Olympus microscope and Leica camera.

It is possible to perform quantification of both DAB as well as the fluorescent staining by using the programmes like histoquest and tissuequest, which is provided by tissuegnostics.

Hi all. I'm very interested in this issue. Could someone explain me what really mean “the intensity of signal does not exactly reflect the concentration of antigen”? I ask this question because many IHC semiquantitative evaluations are performed by multiplying the score of percentage of positive cell ( often with a scoring scale with cut-off values) with the score for staining intensity (ei. 0 no stain, 1

weak, 2 moderate and score 3 strong staining). So that, it is reasonably evident that most IHC scoring systems take into account the intensity of DAB stained cells… Thanks

Daniel Kay, please note that there is evidence that subjective estimation of cell percentages and (especially) intensities is not reliable, even when performed by experts. Manual cell counting is fairly reliable, but if not I suggest using a computational approach as well, even one as simple as the following:

1. Use color deconvolution to separate the hematoxylin and DAB stains

2. In the blue channel use a gaussian filter followed by thresholding followed by nuclear segmentation to find the nuclei.

2b. Manually separate or remove any tissues in the image that are not of interest

3. Within each nuclei find the total and mean intensity

4. Quantify in terms of percentage of cells above a threshold and the density of intensity.

If you're less concerned about individual cells, then it's easier.

1. Follow step 1 and 2 of the previous, but do not segment the nuclei. Use this to create a mask image.

2. Apply the mask to the DAB channel and then measure the total intensity / total area and the area above a given threshold / total area.

I highly recommend APERIO...I used IA for over 6 years, easy to scan slides,

write algorithms, quantitate IHC tumors. top notch quality and service,,,

We have developed a method to quantify staining intensity. The software named TissueQuant that we developed based on this method can be used for quantification of a stained substance. The details can be found here: http://www.ncbi.nlm.nih.gov/pubmed/21924792

The optical intensity of DAB staining is not linearly related to the concentration of the compound: see http://www.mecourse.com/landinig/software/cdeconv/cdeconv.html and the links within.

I have found that imaging systems with even the most expensive scientific cameras are set up non-linearly about 50% of the time.  That means degrees of staining (staining densities) are not incremental.  Even if the gamma is at 1, it is no guarantee that consequent images will be linear.  Be sure to calibrate images against known optical density values from calibration slides.

Has any one tried olympus cellSens imaging software?

Has any one experience with Las AF lite of leica

I've used CellSens with a DP70. If the gamma is set at 1, this does not guarantee a linear image.  I found that it is linear at 0.8.  The other cameras I can't speak to.  If you really want to know whether you're getting a linear image, you can get a calibration slide from Datacolor (scientific.datacolor.com) along with software and you'll get the answer immediately; or, if you have the patience to take several pictures of each step wedge, use a Stouffer step wedge (stouffer.net) and calibrate optical density values to pixel values in your favorite quantitation software (of course, this assumes that you're getting true transmission values -- see Vlad's earlier posting). 

I haven't tried any of the Leica cameras, but hope to in the future...

I learned a lot from this conversation as well.

Dear Sara Mosaad,

First  some advice - do not use jpg images for  thresholding and analysis. JPG algorithm compresses color information , which introduces color artifacts that make the processing and analysis unreliable. Especially if you need to extract color information (as this is needed for the segmentation of your images).

Now on the point: You can try a bit simplified aproach at first:

The DAB areas in your pictures are brown, and brown color is actualy result of reduction of blue and to a less extend green in the recorded light. DAB scatters this spectra more than the red and the result is brown color. This means that the DAB stained areas will have high contrast (will be darker) in the blue and the I attach the result of a simple autothresholding in this channel and as you can see it segmented the DAB areas quite good . It isn't perfect as there are some nuclei which were selected too, but this can be further corrected by adjusting the threshold, or using local instead a global threshold and/or some smoothing before the actual thresholding.

Lastly be careful with analyzing DAB stained images. DAB doesn't absorb but rather scatter the light which makes some types of analysis (opticaal density based analyses) unreliable. Area measurement is ok, but not Optical density measurement.

Please read this very good article:

http://openwetware.org/wiki/Sean_Lauber:ImageJ_-_Threshold_Analysis

This is what I do for DAB quantification on slides with hematoxylin counterstain:

1. Download the free "Fiji" version of ImageJ from http://fiji.sc. All subsequent steps are performed in Fiji.

2. Open a DAB image.

3. Run Image > Color > Colour Deconvolution.

4. From the Vectors pull-down, choose "H DAB" as the stain (this assumes your images are correctly white-balanced, otherwise you have to define your own colors by choosing "From ROI" for the vector then drawing your own Regions Of Interest to define the stain colors).

5. Click OK in the Colour Deconvolution window, you will get three new images. The one with "Colour_2" in the title is the DAB image (Colour_1 is the hematoxylin image), you will quantify the Colour_2 image.

6. Run Analyze > Set Measurements and select "Mean gray value" and "Display label".

7. Select the Colour_2 image window.

8. Run Analyze > Measure (or press Ctrl-m); a "Results" window will pop up with the quantification in units of intensity.

9. You need to convert the intensity numbers in the Results window to Opitcal Density (OD) numbers with the following formula:

OD = log(max intensity/Mean intensity), where max intensity = 255 for 8-bit images.

This will quantify the average darkness of the image due to DAB signal. If you have areas of the image without tissue it will bias your results because it will bring the average OD down. If that is the case we need to discuss further. Also, you may need to use Integrated Density rather than Mean.

Good luck and let me know if you need anything else.

It seems that measuring numbers of photons emitted from the qualified fluorescent dyes using CellProfiler may be a better way for quantifying purpose since there are some insightful issues raised on using DAB.

Can someone comment whether quality of antigens (to be stained) in the cryosection is also compromised to the same extent as fixed antigens in the paraffin sections due to 4% PFA fixation?

Does anyone knows if it´s possible to count positive and negative nuclei with "Fiji" version of ImageJ ?

Dear Shehla Hridi; fiji has alot of plugin which is highly helpful.

Maybe this article will be interesting for some of the followers :

http://www.nature.com/articles/srep12096

(Also check the supplementary information.)

Dr. Mustafa, is your quantification method useful for PAS staining analysis?

Dear Dusan Racordon Glasinovich, I applied this method mainly for Immunohistochemical studies in my lab (on different antibodies markers).

But I will try for PAS and as soon as I can I will reply to your inquiry.

Dear Dr. Mustafa, I found your answer extremely helpful, thank you! I was just wondering if you could elaborate on what to do if there are areas of my image without tissue? Thanks again. 

Dear Dr. Mustafa, This is very helpful information. Thank you for sharing with us. 

Dear Dr. Mustafa,Thank you for sharing your resolution.  I am trying to quantify DAB deposition in a tissue and my pictures/ slides have some areas without any tissue.after deconvolution, I have already tried adjusting the threshold and inverting the LUT. what should I do?

I developed a simple method using ImageJ to quantifying the intensity of DAB. 

Here is the paper:

https://www.ihcworld.com/_books/Nguyen_Reciprocal%20Intensity.pdf

Here is the "How To" guide:

https://www.ihcworld.com/_books/Nguyen_Protocol_Reciprocal%20Intensity%20in%20Fiji.pdf

Dear Dr. Mustafa, how we can  number and count DAB stained CD 68 automatically?

Dear Dr Mustafa thank you very much for your answer, it was very helpful, but I wonder what to do if I want to determine the area percentage in immunohistochemical photoes. Thank you very much

Dear all,

Did someone find the method used by Dr. Mustafa published or written somewhere? I would like to take a look to a more formal SOP or perhaps to a paper including this method.

Thank you

Hi, came across this paper.. Probably it is useful for such application.. 

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0096801

 Am surprised nobody seems to be concerned about  Reidar Albrectsen contribution. Is it tru that we can not really quantify the intensity of the stains. Can I get a confirmation. And if so what about the various methods being described above by Mustafa and David Nguyen

Regarding your question Akinshipo Abdul Warith, my understanding is that although a precise quantification through stain intensity is not possible, it does give an estimate of the difference between samples, which if I understand this correctly, enables categorization of the results based on intervals (much like we do by manual subjective assessment).

hanks for the reply Mar, I would like to believe that the estimates given from the quantification is acceptable in scientific circles and among peers.

I think, you can use image J software only for counting. And determining the OD of stained slides are not really, due to verity of staining, pale to deep coloring.

I have already determine the  DAB color intensity with ImageJ . I chose the "DAB" in list of "colour deconvolution". Now I want determine intensity of flouresence and NBT in cells. Which item from "colour deconvolution" should I choose?

Dear Dr. Mustafa thank you for sharing. Update of the citations of your method will be really helpful.