Measuring mean intensity of fluorescence signal in channel composite and RGB format .tif images

Good day,

This is my first question in this forum, so I apologize in advance for everything I might have forgotten to include in this question!
I am currently analyzing my images, acquired with a Leica SP5 confocal microscope. I am staining a molecular structure in my cell line with a primary and secondary antibody (488 and DAPI) and then want to measure the mean intensity of the 488 fluorescence signal.
example.tif (695.1 KB)
When I export the images in the Leica software, I receive .tif images with rgb format.
I found that measuring mean intensity in the overlay image in rgb format produces quite different values than measuring the channel composite of the same image in the 488 channel.

Here, 1 is the left channel composite and 2 is the right rgb composite.
Can anyone explain why that is? It is the same image, just converted into channel composite with the options image>colour>make composite.
I use Fiji ImageJ 1.52p on my Windows 10 64-bit laptop.

Thank you in advance!

Hi @SabiFrank

I’m going to write a lot of text now but basically the TL:DR of it is what you are doing is not good practice for image analysis and it would be much better if you would measure the pixel intensity values of your original microscopy image. Please let me know if I don’t explain anything well enough.

When you convert to an RGB image, a pixel value is assigned to the red, green and blue components of the resultant image. In your case, probably what is happening is the intensity values in your DAPI channel are being assigned to the blue component, the 488 channel to the green component and the red component is full of zeros. When you measure an RGB image in FIJI and getting a mean intensity, the value you get is the sum of those three values divided by 3.

Then, when you are converting to composite you will have your three components split up, including an empty red channel. When you measure now, you will get a different result depending on if your channel slider is set to the red, green or blue channel. In your picture, I can see you are on the green channel. This measurement is there for the pixel intensity value assigned to the green component during the RGB conversion in LAF. If you moved the slider to the right you would get the blue channel measurement.

As an extra complicating factor, if your image was captured at something higher than 8-bit (quite likely I would say for most microscopes these days) then there is also scaling happen to make it 8-bit to get it into RGB. This whole conversion process is destructive, resulting in data loss and so it’s likely both of the measurements you are taking are inappropriate for image analysis.

I can see that you are using FIJI which automatically has Bio-Formats installed, this means you can open your .lif files without needing to export to .tif format at all. And this is what you should do really as it’s the raw data from your microscope. So I suggest you try that.

If you are having a problem with inputting LIF files into FIJI then please let us know what the problem is and we can try and help.


Thank you very much for your fast reply!
I was afraid that this is the case here and that is why I asked.
So if I want to measure the raw data images from my .lif file, do I just need to open them and measure with my rectangle, like I did before? Or do I need to process these images beforehand? I just need intensitiy values relative to the control measurement. I treat my cells with drugs, that change the intensity of the obtained signal.
Another problem that I got, is that I do not know how I should blind my data set if I use the .lif file. The images are named with the treatment concentration, but during the measurement, I must not know what kind of treatment I am measuring to not bring bias to my analysis. Do you know by chance, how I can systematically replace the names within a lif file with random names, so I do not know what kind of image I am analysing?

Thanks again!

This isn’t a FIJI thing, but QuPath is built off of ImageJ, and many of the same ImageJ functions can be called on your .lif images there. It also has a Preferences setting to blind images in the newer versions (0.2.0m2+, I think)

I haven’t really played with the downstream data export, but it might be an option. Have you tried opening the .lif files directly using the BioFormats Importer?

1 Like

I see, thank you very much for the tip!
Yes, I can open .lif files without problem. Is there maybe an option to save those lif files separately in a special format without data loss in channel composite format like when they are opened directly from within the .lif file? I have a script, that can blind images, but I need separate files for it to work properly.

I just looked it up and it seemes like my image properties are the follwing:" 1/2; 70.36x70.36 microns (1024x1024); 8-bit; 2MB

There is some variability to most file types, so I don’t know if it will work for you, but I was able to split the LIF file I have access to into many different files and save them as TIFFs with 3 channels. After the shown window, I had to Select all, and then got N images, which I could then cycle through with a script.

1 Like

That is perfect! So, when I save these single channel composites as .tif I will not have any loss of information? With this, I could work properly. Thank you so much!!

Again, double check the channels in the Image->Properties for the resulting TIF to be sure (C for channels should = the number of channels), but it worked correctly for my .lif files.

1 Like

I just did and it works fine. Thank you very much again for the competent and fast responses!

1 Like