Perinuclear Staining to Identify Nucleus?



I have some cells that contain a label with bright perinuclear staining. The staining occurs as a band of speckles surrounding the nucleus (the nucleus itself is as dark as background). I have been trying to use CellProfiler modules to select the nucleus. I can identify the speckles as primary objects, and I was hoping to use the UnifyObjects module to unify all the speckles and highlight one cell’s nucleus. So far it doesn’t look so promising, since UnifyObjects does not add any pixels. In addition, from your help section:

The space between the speckles is usually completely black, so I’m having trouble even unifying the speckles. Do you have any other suggestions for rigging up something that might work?



Hi Mark,

Would you be able to upload an example image, so we can take a look?


Yes, having an image to look at will help a lot.

While we are waiting, one suggestion that comes to mind is to smooth the perinuclear-stain image. With the right amount of smoothing, this will create a fuzzy ring around each nucleus that IdentifyPrimAutomatic should be able to work properly on. The edges of the nucleus identified by this method will not be perfectly accurate, but depending on your application this might be tolerable, and you can use ExpandOrShrink to adjust the edges slightly if needed.

Fundamentally, keep in mind that CellProfiler does not currently have any algorithms designed for finding ring shapes per se. But often, if the perinuclear staining can be smoothed to form a ring without gaps, you can use IdentifyPrimAutomatic (with the ‘fill holes’ option set to Yes).