Advice request for correcting illumination for subsequent quantification of fluorescence




My aim is to quantify the intensities of the protein stained in the red channel. The images are of cells stained with Hoescht (blue channel-D) to label nuclei and for a bone marker (red channel-R). The images are part of a high-throughout system whereby cells are grown on ‘microspots’ of various chemicals, and unfortunately some of these spots are auto-fluorescent, causing uneven background illumination that changes from one spot to the next.

1- I understand that I need CorrectIlluminationCalculate and CorrectIlluminationApply modules to get accurate quantification of the staining. Are there certain things to avoid when setting up these modules, i.e.: settings that would affect fluorescence quantification? E.g.: Does the re-scaling that automatically happens with these modules affect subsequent quantification?

2- I have been told that illumination correction modules should be done in a separate pipeline and then the corrected images loaded into the new pipeline for segmentation and measurement? Is this true? How would that be advantageous?

3- I plan to use “integrated” intensity for quantification of protein expression per cell (i.e fluorescence), then divide that by cell area to avoid any introduction of bias based on cell size. Is there a better/easier way of doing this, e.g.: median intensity?

OCN-Finalpipeline-MAmer.cpproj (1.3 MB)




  1. The help for both of those modules is also quite detailed. Those generally also assume though that you’re trying to correct is intensity of the whole image though, not intensity of a particular small regions. I wonder in that case if you’re not better off trying to define the spots (either by identifying them with IdentifyPrimaryObjects or by using DefineGrid), measure the intensity in those, and then mathematically remove it from the measurement of each cell afterward.

  2. We typically almost always do this because the kind of background correction we do for whole images involves averaging across all images from a plate a) averaging across all the images can take a long time and b) you want to only have to do it once, not everytime you want to try out a new set of segmentation parameters. Again, this doesn’t really apply in your case though.

  1. Integrated/Size == Mean, so you can save yourself that calculation by just using the mean. Whether mean or median is more appropriate really depends on your question though.

Hope that helps!


Thank you very much - This is very helpful! I am still a newbie, and I have been through the illumination correction tutorials on the website. They seem to give good results across my images, but the tutorials were unclear regarding what is okay and what to avoid when correcting illumination prior to fluorescence intensity quantification.
One more newbie question: Is the background subtraction module on ImageJ different to CellProfiler’s illumination correction modules?

Thanks a lot!