Trouble with Primary Object on neuron dapi staining

Dear CellProfiler users and team,

I’m a new cellprofiler user. My goal is to do a colocalization between two RNA fish done on mouse brain section. Hera is a weTransfer link to see the images. I have already done a crop with ImageJ to only keep my region of interest.
WeTransfer-Images

So I have three channels:

  • Blue for dapi staining
  • Red and Green for the two FISH

My idea was first to identify primary object with the dapi staining and secondary objects with the FISH.

But I’m still stuck with the primary object segmentation. As you can see on the dapi image, cells can be clumpy and the dapi staining is not uniform inside a cell. So I first apply a smooth on the image to have a more uniform dapi:


I think that the smooth works pretty well (tell me if I’m wrong!).
After that, I tried to identify primary objects and with all the parameters that I tried, my best solution is presented here:

As you can see, I have a lot of clumpy cells that I cannot separate.

If someone can help me to optimize that, I would be super happy !
Thanks

  • To my eye, that smoothing might be too broad and contributing to your issues- I would try a smaller sigma, like 2-3
  • You could also try a different threshold method, since the one you’re currently using seems PRETTY good but also seems to be including some of the gaps between nuclei (though that could also be due to your “Fill holes in identified objects setting”, which you could play with setting to either “After declumping only” or “None”).
  • Finally, you can change the size of the smoothing filter for declumping and/or the distance between local maxima, which right now are both set to automatic- you can read the help for both of those settings in CellProfiler itself, and/or check out our video tutorial on segmentation, which discusses those settings and how to tune them (the section on splitting objects starts at about the 21-minute mark, with deeper discussion of those settings starting at around 28 minutes).

Thanks a lot ! Will play with these parameters !