Intensity detection problems




I’m working on measuring the intensities of nucleoli. I’m doing fine identiying nuclei and nucleoli; you’ll notice a MaskObjects module in my pipeline designed to take out any odd particles that might have shown up as nucleoli that are not actually in nuclei. I’m aware that fluorescence varies naturally from coverslip to coverslip in a given experiment, even if all exposures are taken at the same length. To control for this natural variation, I compare nucleoli intensities to the intensities of surrounding nucleoplasm (which I call NENA in the pipeline for some reasons) with the idea that looking at these ratios should control for fluorescent variation since I’m just interested in changes in nucleolar intensity. However, when I look at these ratios, it seems as though CellProfiler does not tell me what my eyes tell me. For my “no damage” picture, it gives me a ratio of 1.08 for nucleoli:NENA, implying that the average pixel (I think?) intensity is greater in the nucleoli than in the nuclei. For the various other pictures I’ve included, either the ratio is again greater than 1 or only a bit below 1 when my eyes tell me it should be far below 1. I’m just not sure what is going on here. I’ve tried using Align to make sure my pictures were overlapping correctly, and found no difference in results. Do you have any suggestions?

I’m just sending the different channels for one experimental point, but I have more I can upload in separate responses if you wish. I’m also including the picture of the protein of interest; CellProfiler gives me a mean intensity ratio of .95 for mean nucleoli intensity to mean NENA (nucleoplasm) intensity. It looks like it should be less than that, right?

Thanks so much for your efforts on this program and website!

ku to send.mat (256 KB)


Hi Anne,

In your pipeline, you measure the intensity of both the MaskedNucleoli and NENA objects against the ku image. So the ratio of MaskedNucleoli to NENA is going to be a ratio of ku image intensities only. Is this what you want?

If so, then I see no reason why the ratio shouldn’t be around 1 since to my eyes, the objects in the ku image look fairly uniform in intensity, at least with respect to the background fluorescence. So the MaskedNucleoli and NENA objects will be roughly the same mean ku intensity, giving a ratio of approximately 1.



Hi Mark,

I want to measure the intensities the nucleoli on the ku image. Is that not what I’m doing? I use the other images just to identify either nuclei or nucleoli; ku is my protein of interest, and the only one with intensities of interest. I thought I was measuring only the part of the Ku image enclosed by either nuclei, nucleoli, or NENA.

Are you saying that because the nucleoli and NENA are both so much brighter than the background, their intensities will be approximately the same? I’m ok with the ratio of nucleoli to NENA being close to 1 for my no damage picture, but for the 100ug cis picture, I don’t think the objects have uniform intensity. The lesser bright areas of these objects correspond to nucleoli. When I run my 100ug cis picture with its matching nucleoli- and nuclei-identifying pictures, it still tells me that nucleoli:NENA is appoximately 1. Is this because the nucleoli are still much brighter than the background? I’m uploading the accompaniments for my 100ug cis picture.

Thanks again - sorry if I’m misunderstanding CellProfiler’s quantification of intensities!


]Hi Anne,

Thanks for clarifying. Yes, there is a more obvious difference in the 100ug picture. However, the ratios are ~0.8-0.9 in this case, which is lower than 1 but I assume not as much as you would have expected.

Attached is a picture of the intensities displayed using DisplayDataOnImage (nucleoli in green; NENA in blue), along with the object outlines; the background picture looks red due to a bug. It seems that you are correct; the nucleoli are still much brighter than the background, so the change in ratio is clearly present but not as drastic as you might expect.

My suspicion is that part of the issue is that several of the nuclei objects capture more of the background in the ku image than you would want, which is pulling the mean ku intensity of the NENA objects down.



Hi Mark,

Thanks! DisplayDataonImage is really showing me a clear picture of the situation.

You guys are the best!


Hi, mbray. Im a new user of cell profiler. I need to do something very similar to this (calculate nuclolus area, and the intesity of my interest protein. I tried to see your pipline, however, I have the last version of cell profiler and could not see it. I have a marker for nucleolus but it does not cover the nucolus compleatly, Im wondering whether there is a way to determine nucleolus area using DAPI intensity. Any help would be highly appreciated


Hi Miguel,

Can you make a new thread and post your pipeline so far and some images? That’ll make it easiest for us to help you out. Thanks.