Average absolute intensity of Identified Objects


I need some of your expertise for analyzing GFP fluorescence of mammalian cells after CRISPR knockout of the stably expressed gfp gene. I am having some difficulties achieving this, and found great help from the solutions posted in similar topics, specifically to find the mean intensity of individual objects.

However, is this good to compare the intensities within one given image? Is the intensity given from a 0 to 1.0 scale an absolute intensity or is it a relative pixel intensity based on the image being processed? Given that the output of my KO is not an “on/off” type signal, I want to get the mean intensity of each individual object to be able to draw a histogram to compare the intensity distribution of WT vs. KO cells on ONE chart. I just want to make sure that the data I produce is not flawed.

I did my best to make this clear, let me know if my description was a bit fuzzy and I’ll try to clarify!

Thank you,


As long as you haven’t done any rescaling of the images outside of CellProfiler or inside it with the RescaleIntensity module*, the scaling is 0-1 based on the maximum pixel value possible in the image, not based on the maximum pixel value actually present; it’s absolute, not relative. Therefore it should be perfectly fine to compare between your WTs and KOs. Hope that helped!

(*= and if you’ve followed all other microscopy best practices such as using the same exposure time for both WT and KO, doing them on the same day, etc etc)

Hello @bcimini ,

Thank you very much for your answer! I did have a follow-up question however. While doing “IdentifyPrimaryObjects”, it usually automatically generates a threshold - this threshold is usually too high to detect my dim cells… I have not set any restriction on the intensity, so I do not understand why this happens. For example, for a given image, my threshold is automatically set at 0.188 when using an Adaptive Otsu threshold strategy/method… Should I be playing around with the threshold correction factor? If so, how should I play around with it? I have been pretty much guessing my way to the best images by trying random numbers and this is definitely not the way to troubleshoot!



I’d perhaps start by trying a different thresholding method- perhaps Otsu 3Class with the middle class set to the foreground. If that doesn’t work, then yes, adjusting the threshold correction factor is probably the best way to go- if you hover over your cells, you should be able to figure out how bright they are and therefore a ballpark figure for what the threshold correction factor should be (ie if your cells have a brightness of about ~0.09 in the image you mentioned above, try a threshold correction factor of roughly 0.5).