Integrated density of 3 channels

I have been analyzing integrated density of 2 channels (Red) and Green by converting the images to binary, and using image calculator that gives integrated density .However,is it possible to use image calculator in the same way for integrated density of 3 channels.,i.e. Red, Green, and Magenta (far red)?.


welcome to the forum.
Your approach might be flawed by a misunderstanding. If you binarize an image, the integrated density is not reflecting intensity from your sample you imaged. Binarization means that all pixels become either black (normally lowest value by convention, mostly 0) or white (normally the highest value by convention, mostly 255). So, it completely changes the original values in your image and render the intensity-based analysis or your image invalid.
In turn, if you use a binarized (or in other words thresholded image) to define regions of interest (ROIs) in which you measure pixel intensities (e.g. integrated density) originating from your original image, that would be an approach making sense in my opinion.
This principally does not have something to do with the image calculator. The image calculator makes pixel based math and if a resulting image from that tool serves for an intensity-based measurement needs to be evaluated in relation to your analysis.
To combine 3 images in one step in the image calculator is however not possible. One would need to do that step by step.
But, to give a more goal oriented answer, it would be helpful, if you could post an example image and describe a little bit in more detail, what the actual purpose of your measurement is and what you try to achieve.


Thank you so much for your prompt and succinct information. I forget to mention that I convert the stack image into binary (stack of individual channels, example; green, and red into binary mode). I have attached the stack image as an example. My goal is achieve co-localization of one of protein of interest with another (say how much of my protein is localized in neuronal or glial or synaptic receptors).In this particular instance, I am interested in understanding quantifying two of my proteins of interest that are possibly co localized in glia.
FF_image_FARRED.tif (1.4 MB)

Your attached image is not really a stack. Seems to be only one single channel. If you have trouble in attaching it here, you can also link to dropbox or similar.
Co-localization normally needs to be done in a different way than via binarization, but first looking at the images would be necessary to make more precise indications.
I will get back to you by tomorrow or after looking at your images

FF_image.tif (3.0 MB) .FF_image_DAPI.tif (1.1 MB) FF_image_FARRED.tif (1.4 MB) FF_image_GFAP_Red.tif (1.0 MB) FF_image_Green.tif (902.1 KB)

Here is the Stack of Overall channels with DAPI. There are also individual channels if its needed for your reference

Your images are saved as RGB color images. If this is just the case in the ones uploaded here and you have the originals really as a microscopic image stack, best in the original file format (such as Zeiss, Leica, Nikon, Olympus,… or any other original format) that would be optimal. Analysis from RGB directly will not be possible and might lead to inaccuracies or errors. Furthermore, the original files hold metadata which will be interesting and important to define e.g the actual resolution which is important to get an idea to which extent a co-localization analysis has a meaning.

The next step is your actual research question. Define exactly what you want to know. As explained in @Fabrice_Cordelieres video, it makes a huge difference if you just look for co-expression, co-occurence or correlation.

Next, in my opinion the magnification is way too low to make a meaningful co-localization study (at least on a per cell level)!
The magenta and green signal you refer to are quite diffuse. Diffuse and non-localized signals reduce the strength of a co-localization analysis to some extent.

Potentially, what might be helpful is the overlap of area between the different channels (and I guess that is what you referred to by “co-localization”). Then your binarization and image calculator approach would potentially help to some extend. But the diffuse nature of the signals will make it difficult to extract areas which one would consider being “correct” or “realistic” because there is no ground truth you could orient on.
Arbitrary thresholding will quickly make the areas too big or too small and you cannot really determine at which point it is “realistically” extracted. So, finally the overlapping areas you measure will have little meaning.
Additionally, be careful to not mix-up area measurements with integrated density. From a binarized image you cannot get integrated (or any other) pixel intensities or densities, since pixels are converted into black and white (lowest and highest intensity values available). There will be no link to your original intensities.
Finally, as mentioned before, the image calculator cannot compare 3 images among each other. This would need to be done e.g. by comparing green with magenta and the result of that comparison with finally red.

Don’t get me wrong, I don’t want to destroy your analysis idea. I just want elaborate that getting a meaningful outcome might be more difficult than just thresholding somehow the images and checking for pixel overlap. And here, it will be important to look at the question you and your group is trying to answer with the experiment.

The whole thing could be further deciphered, but I think first and foremost, important that you can specifically determine the exact question to be answered, then think about if the image material serves for such an analysis or if the imaging needs to be adapted and finally what kind of image processing-based analysis is necessary to answer your question.

Here some recommendations for watching and reading to get a deeper understanding of co-localization which might help to define the actual needs:

Very good video as introduction into co-localization analysis from @Fabrice_Cordelieres and NEUBIAS.

The basics for better understanding:

  1. Which Elements to Build Co-localization Workflows? From Metrology to Analysis
  2. A guided tour into subcellular colocalization analysis in light microscopy
  3. A practical guide to evaluating colocalization in biological microscopy

Optional deeper reading

  1. The 39 Steps: A Cautionary Tale of Quantitative 3-D Fluorescence Microscopy
  2. Image acquisition for colocalization using optical microscopy
  3. Quantifying colocalization by correlation: the Pearson correlation coefficient is superior to the Mander’s overlap coefficient
  4. Statistical tests for measures of colocalization in biological microscopy
  5. Automatic and Quantitative Measurement of Protein-Protein Colocalization in Live Cells
  6. Replicate‐based noise corrected correlation for accurate measurements of colocalization
  7. An automated method to quantify and visualize colocalized fluorescent signals
  8. A Syntaxin 1, Gαo, and N-Type Calcium Channel Complex at a Presynaptic Nerve Terminal: Analysis by Quantitative Immunocolocalization
  9. Dynamics of three-dimensional replication patterns during the S-phase, analysed by double labelling of DNA and confocal microscopy
  10. Partial colocalization of glucocorticoid and mineralocorticoid receptors indiscrete compartments in nuclei of rat hippocampus neurons
  11. Resolution, target density and labeling effects in colocalization studies – suppression of false positives by nanoscopy and modified algorithms

Thank you so much for elaborate discussion and technicalities. I appreciate your time and prompt response.

1 Like