Associate cytoplasmic stain with nuclear mask?

Hello Everyone,

This is a long post, so I will try to organize it to make it easier to read. I am relatively new to confocal image analysis, but I have learned a lot through ImageJ tutorials. However I cannot seem to find the best way to associate cytoplasmic staining with nuclear masks.

My data: I have FFPE embedded tissue, stained with DAPI and various nuclear and cytoplasmic markers. I have tiled 40x confocal images (currently in 2D). I can make a nuclear mask to quantitate nuclear signals, however I am having trouble finding the best way to quantitate membrane markers, and keep their association with nuclear mask/nuclear ROIs. *** Beyond DAPI, none of the other markers will be positive in all cells. Weka seems like a popular tool (I haven’t really tried it yet), but not sure if this would be the best route for my data?

Ideal Analysis: I want to keep the X/Y coordinates of nuclei and associate positive cytoplasmic staining with corresponding nuclei. I don’t necessarily care to compare the intensity of cells/signal for this analysis, I just need the cytoplasmic values to be associated with the nuclear mask/ROIs and spatial coordinates (I can also choose the threshold for positivity in downstream analysis). Is there a way to analyze the outer perimeter of a nuclear mask within an x # pixel border?

Previous Attempts:

  1. Dilating the nuclear mask is not good enough because too many nuclei become merged and inseparable with watershed.
  2. I have tried JACoP using nuclear mask with cytoplasmic mask, but I think this uses center based strategies which don’t seem to work for for my data either.
  3. Vornoi segmentation based on nuclei doesn’t work properly and always outputs a very strange segmentation.

Any tips/suggestions/strategies/plugins/key words to read up on are much appreciated! I keep going around in circles of ideas. Thanks in advance!!!


Accurate segmentation of cytoplasmic staining is a difficult problem even in a perfect 2D plane, and even at 40x confocal, there is enough thickness that you can get some overlap between cells. If you can get good nuclear segmentation, though, it should be possible to build off of or use that.
Now that you have made a first post, you should be able to share a demo image and indicate what kind of problems you are having. Google Drive can allow you to host images up to around 15GB if size is a problem.

I suspect there is, but I usually use QuPath for this since it handles that by default.

Thank you for the tip, QuPath looks like a great software!

1 Like