Challenging nuclei segmentation in tissues

cellprofiler
segmentation

#1

Hi - I need to segment nuclei in CP that are tightly packed and have inconsistent nuclear staining (due to protease treatment for FISH). The standard troubleshooting methods have been unsuccessful… I’m attaching a sample image. Any ideas? Thank you!


#2

Hello Magdastu !
Simply convert it to 32 bit binary since the red channel and blue channel are useless. (or you could split the channels and use only the green).
Hope this helped,
Bob


#3

Hello again,
If you want to increase the image considerably then go Process >Math>Macro and enter v=v/27*100 (or the value you wish for the 100)
Bob


#4

Hi Young one,
Here’s what I got. If it is good for you try out the above.
MAGc3e385adb6e501bd462dfd44e7da6cf6ece98002_1_500x500


#5

I do not think that is a “segmented” image.


#6

Thanks, Gabriel, for pointing this out. I wasn’t sure if I was missing something… Just to clarify for others that may have misunderstood my question, I am looking for “segmentation” of the nuclei from each other and the background. This effectively means defining them as primary objects correctly. Accurate segmentation has been difficult because most nuclei are touching and have nonuniform nuclear staining, hence are incorrectly assigned to the background or have a separating line run through them. I’m looking for a trick or a set of parameters for the segmentation that may help separate the nuclei more accurately. Thank you.


#7

Your request was clear, but perhaps you haven’t got many suggestions because it is a very challenging case. I do not think I could do it by hand either; the boundaries between nuclei are sometimes blurred and there are many nuclei overlapping (which in itself it poses a further problem of how to represent the result in a 2D image).
Can you prepare the material in a different way? (e.g. dissociate the tissue, so you end up with a monolayer of nuclei). If you come up with some solution, please post it here!


#8

Sorry folks, the image isn’t segmented. I was assuming your main problem was the dimmed image. I was leaving the segmentation to you.
I will try to segment it also.
Bob