Using EnhanceEdges in CellProfiler to identify nuclei

Hi everyone.
I have a question about nuclear segmentation. I’m using CellProfiler modules, EnhanceEdges and IdentifyPrimaryObjects to accurately identify the nuclei in my image (10x objective, tissue sample). I’m attaching the CellProfiler pipeline I’m using and an example of what the image looks like after the EnhanceEdges module.
My issue is with the IdentifyPrimaryObjects module. I find the EnhanceEdges module is much nicer, cleaner and accurate in how it is identifying the cells. So when I execute the IdentifyPrimaryObjects module, the nuclear segmentation becomes wonky. Is it possible to just identify the nuclei using the EnhanceEdges module somehow? I want to be able to measure other signals in the cell (AF 488, AF 647, etc.) as well after nuclear segmentation.
Thanks!


→ An example of what the image looks like after EnhanceEdges

tissue_pipeline2.cpproj (126.0 KB)
→ my pipeline

1 Like

Hi, I think you will for sure need the IdentifyPrimaryObjects module, which expects the objects of interest to be the bright pixels in your image. So, I would expect that you have to ImageMath>Invert the output of the EnhanceEdges module before feeding it into the IdentifyPrimaryObjects module. Does that help?
(Please, of course, ignore my suggestion in case you are doing this already, as I did not download your pipeline).

Hi @vivi,

As @Christian_Tischer commented, IdentifyPrimaryObjects identifies objects with bright centers and dark edges. If you’re not able to get a good segmentation on the original image with this module, you can make adjustments to your images to try to increase the brightness of the center of the objects and decrease the intensity at the edges of the objects. For example, you could try using ImageMath to subtract the Edges from your original image – that may help you separate adjacent nuclei.

This CellProfiler introductory workshop has an excellent overview of the principles of object detection and thresholding, which may be helpful for you as you troubleshoot. It’s available on the Center for Open Bioimage Analysis YouTube Channel.

Hope that helps!
Pearl