Problems with defining GFP+ cells/ separating clumping cells

I just started working with your software and so far am very impressed. I’m not even scratching the surface but can still see the power of this tool.

My problems are probably quite basic to everyone here but even still have been frustrating me.

  1. I have two cell populations: GFP+ and GFP- cells. I am only interested in measuring the GFP positive cells therefore I would like to first identify nuclei (using DAPI) then correlate this to GFP expression (secondary objects). Not every nuclei corresponds to a GFP+ cell. However, when i do this the program identifies cells which clearly are not GFP+. Even when the background is completely black it “identifies” and outlines cells. Identifying the GFP+ cells as primary objects and nuclei as secondary doesn’t help as well.

  2. As you can see from my pictures, I have some clumped up cells that are relatively round and more intense in the centre. I would like to be able to distinguish these clumped up cells since I want to do a neighbour analysis. Right now my pipeline just completely ignores the nuclei problem and looks at GFP+ cells. I threshold using Background Adaptive: 1.3 and a Laplacian of Gaussian method to distinguish clumped objects, with minimal success.

My feeling is that using the nuclei would be easier to distinguish clumped cells, however, not every nuclei corresponds to a GFP+ cell.
Please, if anyone could help with my basic albeit trying questions I would be most grateful. Thanks.

ClusterPipeline.cp (5.67 KB)

okay, so just an update.
I first identified GFP cells as my primary object. From this I created a mask and masked the DAPI picture.
This new image I used to identify nuclei as another primary object.
Then I used the nuclei as a primary object on the GFP picture to help me distinguish clumped up cells.

I’ve attached my pipeline. It’s very crude and not always precise but works to a certain extent.
If anyone can help me streamline this or present a better alternative please help!
ClusterPipeline.cp (10.9 KB)

Your pipeline sounds good.

Do you still have issues?

You can set smoothing filter in idprim to 1 to help the declumping.

I have a more complex pipeline for declumping and determining appropiatly the limits but if you get good results this way it’s certainly more gently processing wise.

Hi Jonathan,

I’m attaching a pipeline that seems to do OK for what you want. As Feel noted, your approach seems to be the proper one: Detect the GFP+ regions first, mask the DAPI image, detect the nuclei and then re-detect the GFP+ cells. My main modifications are some changes to the choice of thresholding method for the GFP+ regions. However, this is based on my empirical estimation; you’ll have to check whether my choices are valid.

Here, the determination of clumping is performed using a FIlterObjects module to find cells with one neighbor or more. Provided that the GFP+ portion of the pipeline is OK, this approach should do what you want. You can adjust the neighborhood criterion as needed.

Hope this helps!
2010_05_31.cp (9.78 KB)