Background subtraction and primary object identification

Hi Folks,

I am trying to develop a single-cell imaging assay that accurately quantifies the fluorescence intensities of cells in three channels and in order to do this I need to subtract the background fluorescence intensities from each channel. Thus, I was wondering how to accurately subtract background fluorescence from images containing objects and whether this should be performed before or after primary object identification? I have attached several representative images of cells expressing GFP, which are the images used for primary object identification, so if you have any recommendations for which thresholding and de-clumping methods work best that would be greatly appreciated also!



Typically, background correction is performed prior to object identification. If the heterogeneity is produced by the microscope hardware, you are best served by using the images from a plate to generate/estimate the illumination function using CorrectIlluminationCalculate (with the “All” option), and then correct each image in turn using CorrectIlluminationApply. If that is not possible, then you might be able to remove the background on a per-image basis, but you should check your controls to confirm that you are producing the expected results.

Also, one note: since your cells are touching, it is unlikely that using the GFP channel for primary object identification will work well. You are better off using a nuclei stain, if at all possible, as the primary objects then using the GFP channel to identify the cell boundaries from the nuclei.


Hi Mark,

Thanks for the quick reply and for all of the information! Based on your response, I am now wondering a) if I can perform flat-field correction and background subtraction sequentially and b) how to go about doing this? Since there is heterogeneity in illumination due to the imaging system would I first use the images (~1,000 per channel) from each plate to generate an illumination function using CorrectIlluminationCalculate (with the “regular” and “all” options) then flat-field correct each image using CorrectIlluminationApply (with the “divide” option)? Since I want to subtract the background fluorescence from these flat-field corrected images would I next use the CorrectIlluminationCalculate module (with the “background” and “all” options) to estimate the background fluorescence and finally subtract this from each image using CorrectIlluminationApply (with the “subtract” option)? Thanks again for your help and I look forward to hearing your thoughts!


Please see answer here: