Confocal analysis with Fiji

Hi everyone,

How could I analyze with the same way, regarding thresholds and brightness/contrast, the intensity between different samples in the same experiment? I compare different conditions for both red and green images in order to find the yellow colocalization (puncta). How can I write a script for that so the colocalization analysis which is what I am looking for will be quicker for the big dataset I have?

Moreover, is it ok to manipulate the images firstly in photoshop and then proceed for the colocalization with image j? My problem also is that regarding the changes I am doing with the threshold tool, I am taking different numbers regarding the yellow spots that refer to the colocalization of my red and green image :confused:

Thank you for your time!

You asked:
Moreover, is it ok to manipulate the images firstly in photoshop and then proceed for the colocalization with image j?

No, it is not ok. That is essentially manipulating the results. Such results will not be reproducible and will prevent reaching any sort of generalisation.
One good start to try and resolve this is to think why your images change in brightness and contrast and try to control that before image capture.
Are samples standardised? Were the settings in the confocal changed across images/samples?

Hello gabriel,

Thank you very much for your message.
My experiment is this:

  1. I take pictures for my whole experiment (number of samples I have) with the same microscope settings) in order to compare them in the end.
  2. I have a ctrl case and a treatment
  3. I have neuronal samples, so I take pictures: red (protein no. 1), green (protein no.2), blue (this is my MAP2, dendrite marker) and turquoise (nuclei, that I do not care for now).
    I have to find and measure the colocalization of the yellow color generated from red and green.

How could I do this? Is it ok though to fix the brightnes/contrast in image j before performing colocalization? Should I keep this changes the same for every picture? Unfortunately, not all my pictures are the same regarding the intensity of the green, red or blue even in one case (i.e control), maybe because the coverslip is not homogeneously stained? But why? Is it because I do the merging using specific stacks (i.e z3) and I do not compare z-stacks?

Could you please tell me the very first basic changes you would do before proceedding to color threshold for yellow and colocalization?

Thank you for your time!

@nou3lla I encourage you to read this page:

And use quantitative, statistical measurements of colocalization, rather than qualitative red-green color merge images.

1 Like

For colocalisation applications, you can as well have a look at the following plugin:
http://punias.free.fr/ImageJ/colocalization-finder.html
which has more “on the fly” features

First, I can recommend to have a read on the following 2 very good publications to get into co-localization analysis:

  1. Bolte and Cordelières

  2. Dunn et al.

Second, brightness and/or contrast changes should never be done before any measurements related to pixel intensities, since they change the values (as @gabriel stated already). The only manipulation which needs to be done is a background subtraction if there is a considerable amount of background (because this will give you false-postive colocalization indications). Here, Dunn et al. describes an efficient method. Additionally, noise should be as low as possible (e.g. due to line/frame averaging at the microscope, if possible)

Third,…

… does that mean you take Z-stacks and then do a Z-Projection (combination of Z-slices into a single image)? This could not be used as an input in a co-localization analysis. Rather take the full Z-stack of the 2 channels you want to analyze and start the analysis on them (after background subtraction) by using either JACoP (explanations in the Bolte and Cordelières publication) or Coloc2.

If you have small puncta you are interesteed in the object-based co-localization analysis in JACoP might be a good option and gives you an easy to interpret readout. Also the Manders quotients might be worth looking at. As @ctrueden stated, better use those (statistical) tools because looking for yellow in a combination of a red and a green channel is only in very obvious cases giving a hint on a colocalizing situation.

Once you are familiar with the principles stated in the Bolte and Cordelières paper the Colocalization finder mentioned by @philippe.carl will also be very helpful.

2 Likes

I have updated the Colocalization Analysis page of imagej.net to link to more colocalization plugins, including Coloc 2, JACoP and @philippe.carl’s fork of Colocalization Finder. See here:

@philippe.carl Please feel welcome to edit that section describing the plugin in more detail—or even better, create a Colocalization Finder page on the wiki with as much detail as you like, and link to that.