Best way to de-clump nuclei?

Image Analysis
#1

Hello,

I’m very new to CellProfiler and am having trouble de-clumping nuclei. I’ve attached troublesome images and a current pipeline. Any way to get around this problem?

Thanks very much!
Maria-Louisa
TestLiveDay3_R_cp.zip (1.79 KB)



#2

Hi Maria-Louisa,

I’m attaching a revised version of your pipeline. You made a good start, especially with using the suppression in EnhanceOrSuppress; not many people make use of that feature.

The key revisions in IdentifyPrimaryObjects are the following:

  • Using 3-class thresholding and assigning the middle class to the foreground. This assists of getting more of the dim nuclei that would otherwise be missed with the regular 2-class Otsu.

  • Increasing the permissible object diameter

  • Using Laplacian of Gaussian which is often good for blobs which have a distinct intensity peak but are highly varied in overall brightness, which seems to be the case for your nuclei.

I also edited the regular expression that finds metadata in LoadImages. If you want to use it, help out the help for that setting.

Hope this helps get you started!
-Mark
TestLiveDay3_R_MAB.cp (4.14 KB)