Urgent help needed with cell profiler


I posted before regarding cell tracking and rescaling the intensity of my images. I am certain that the main issue is my original image as it is quite bad with too many cells in the field that I am not able to get any segmentation with other programs. However, I am able to get segmentation with cell profiler but the issue is now with tracking the cells as there is difference in number of objects identified between images of the same field of cells but obtained at different time points.
Please help as if this is impossible to analyse such image with any program then I can just focus on obtaining better images.

Thank you for the help.

Hi Durga,

I agree that as it stands now, using your current images for temporal object tracking would be quite difficult. However, I would be less worried about the number of cell per se, than the low image quality that seems to be the case here.

Based on my experience, I believe that achieving suitable object segmentation is the most challenging aspect of object tracking to overcome. Even if you have a lot of cells present, if they are well-separated and easily identifiable form one another, you can still make substantial progress. This is most often done by having a nuclei stain, or a suitable surrogate. In lieu of a nuclei stain, perhaps the dark space in the cytoplasmic stain can be used as a surrogate as you describe. However, there are two problems I see based on this image:

  • The high cell density and degree of confluence/overlapping will make differentiating even the cytoplasm difficult.
  • The presence of both the bright large blobs and small bright speckles will make the segmentation even harder.

I should note that these issues are general to object tracking in general, and are not specific to CellProfiler. I afraid that unless you plan to track the cells manually, the images right are just not suitable for a successful automated asssay.

Overall, I would concentrate substantial effort on optimizing the image quality to have the best chance at success with an automated solution, or even a semi-manual approach.


Hi Mark,

Thank you very much for your quick response. I have spent a week on trying to get some data from my bad images but needed an expert to confirm my original thoughts that I should go back to obtaining better imgaes with less confluent cells and it is not my lack of proper “communication” with the cell profiler software.

Thank you again for your help.