Identifying secondary objects

Hi Mark,

I think I posted this in the wrong place. So I copied in here.

I labeled my cells with DAPI and phalloidin to identify nuclei and outline of individual cells, respectively. Although I can clearly see single cell outlines in my fluorescent images, “identify secondary objects” module doesn’t follow true outlines but mostly find non-specific particles.
What am I doing wrong? Can you help me?


P.S.: I attached my DAPI and phalloidin stained images. Could you help me finding individual cell outlines?


IdentifySecondaryObjects didn’t work as expected for two reasons I think:

  • IdentifyPrimaryObjects needs some adjustment to detect the nuclei better since they are rather clumpy and image saturation seems to be an issue with this example.

  • In IdentifySecondary, using Progagte as the method may not be as appropriate for your image as it could be. I would suggest the Watershed-Gradient might be better since the cell edges are bright. Also, every nucleus idenfied receives a secondary cell body regardless of strength of stain. So the uniformly-stained region to the left and right edges will also be used to find cells, but since there is little in the way of intensity cues to guide identification, the cells found will probably be less meaningful. If you intend to restrict cell identification to the middle bright area, you will need to mask it out first.

I’m attaching a preliminary pipeline, but as mentioned above, it will require additional tweaking. I hope this gets you started!
2011_03_02.cp (5.07 KB)

Thanks for your help, Mark.
Watershed-gradient works better than propagation does. Before hand, I optimized nuclei identification by manually identifying area of interest and using entropy function, by the way. However, cell boundary doesn’t match up correctly yet. In order to identify individual nucleus and cytoplasm with high accuracy, what is a better way of sample preparation to acquire good images for CellProfiler? Should I label cytoplasm with weak cell boundaries signals? Or should I use cell boundary specific markers to minimize cytoplasmic signals? Or either works equally?

Let me hear your experience. Thanks again.

Hypothetically, either can work: Propagation and Watershed-Gradient usually works better as long as there is a strong intensity difference between the cytoplasm and boundary (brighter or dimmer), whereas Watershed-Image should work well with strong cytoplasm/weak boundary. However, the bulk of our assays fall into the strong cytoplasm/weak boundary category, and we find Propagation usually works pretty well, provided your cells aren’t severely overlapping.

I should mention that your images seemed somewhat saturated, so be careful during image acquisition, as saturation can abolish the intensity differences needed for any of these methods to work effectively.