How to measure the background fluorescence of a fluorescence (cells) image?

How to measure the background fluorescence of a fluorescence (cells) image?
I have a fluorescence image with cells. I can calculate the intensity of the cell but I guess its not accurate as I need to substract the background fluorescence. How can I do that? Your assistance is appreciated.

Hi @taufiq_bge

You can measure any selection (including background areas) by defining an area (see magenta rectange below) then running [ Analyze > Measure ].

If Mean doesn’t appear in the results, you may need to select the outputs in the [ Analyze > Set Measurements... ] dialog.

You can then subtract this from the entire image by running [ Process > Math > Subtract ] and entering the mean background value. Remember to remove any selections in your image before doing this otherwise the Subtract will run only within the selection.

Hope that helps!

2 Likes

Hello Dave,
Thank you for the explanation. I did it but after doing it all the background turned black. The mean of the background is 1625 and the mean of fluorescence cell is 1930. This is a z-stack image.

Did i make any mistake here. Thanks for your help.

If you don’t have enough real signal above the background, that is what I would expect to see. Without sharing your original images, it is going to be hard to tell, though. Thus the recommendations that were made in your original post :slight_smile:

1 Like

Hi, Thank you for your reply. I have attached the raw image without z-stack. So can I get the exact value of the fluorescence? As if I subtract 1600 then I see only black image no bright cells. Also what is the difference between [Process > subtract background] and [ Process > Math > Subtract ].
Thank you.

Maybe I am missing something but I do not see the raw image. If it is too large, you may need to host it elsewhere and provide a link.

Math->subtract will subtract one single value from each pixel. It is highly controlled and will result in an easily known change.
The other subtract background option, as it states, is a rolling ball, so the amount of background subtracted will vary across an image. This can be useful if the level of background staining is not constant across the image. It also requires knowing the approximate size of the objects you are looking for, and that those objects be of fairly consistent size (“or smaller”). Otherwise you can end up subtracting them out.

1 Like

Hi, Thanks for your reply. Here is the original image. image1.tif (10.6 MB)
I am confused that the flourescence and background intensity difference are very close. So does it mean background noise? And, what is the subtraction procedure in this case.

Thank you.

Yes, your image has high background. The background in most areas is around 1600, while the signal at it’s strongest is around 1700, with some of what I assume is real signal being as low as 1650. That makes your signal to noise ratio around 1.03.
In this case it would probably be more accurate to say you have low signal to background, rather than signal to noise, but the result is nearly the same - it is hard to pick out the positive signal.

A post discussing both SNR and SBR, but in cellprofiler. Signal to noise Ratio and Signal to Background Ratio in fluorescent images

After subtracting 1600 from your image, this is what it looks like, which shows a decent amount of speckling still in the background along with a few objects.

You may have better luck with the Background subtraction option if you have a good grasp on the size of the objects you are looking for. Alternatively, if you have no way of improving the image quality, you may want to look into other options like Noise2Void.

1 Like

Thank you for your details explanation. As I know the size of the cell (radius about 0.5 micrometers). So in that case can I use Process> subtract background. If possible then how can I select the rolling ball radius in pixel. Thanks for your time. Its really worth learning from you.
Thank you.

Side note, signal to noise is different from signal to background. You may have noise issues, but it is hard to tell since your background is very high relative to your signal.

The pixel size needs to be known or taken from the metadata of the image. Raw image files like CZI, LIF, etc should include pixel size metadata. If you are on a custom built system, you need to figure out and set that yourself.
Once you know it, you can set metadata using Image->Properties
https://imagej.nih.gov/ij/docs/menus/image.html#properties

The rolling ball radius will be in pixels, regardless, so you could also check Preview and test out which size works for your experiment. Just don’t limit yourself to testing it on a single image.

1 Like