Cell segmentation w/ cytoplasmic stain—identifying cytoplasmic regions

I have cells where the nuclei are marked with DAPI, and the cytoplasmic regions are marked with an anti-enolase stain. My goal is to identify all of the individual cytoplasmic regions as separate primary objects.

Some cells are clustered, while others are isolated. The isolated cells are segmented properly, with the cytoplasm highlighted and the nucleus present as an empty circle.

However, when there are multiple adjacent cells, the segmentation fails—my pipeline combines half of one cell’s cytoplasm, and half of another cell’s cytoplasm, into a single cell. The nuclei are wrongly identified as empty space at the edge of a cell, and are used for segmentation. I’m new to CellProfiler. How do I fix this? Thanks in advance. I have attached my basic pipeline and some example images.

Test_Pipeline_2_S3W4.cpproj (649.2 KB)

image

Here is the original image, containing a blue and red channel. Blue is the nucleic DAPI stain and red is the cytoplasmic stain.

image

Here is the image with the blue channel subtracted using ImageMath.

image

Here is the attempted segmentation.

If your cells are mononuclear, we typically recommend you make nuclei your primary objects, cell bodies as secondary objects, and cytoplasms as tertiary objects; because cells that are clumpy ARE indeed hard to break up in a segmentation (CellProfiler or any other program), it makes it more likely to succeed.

If there is a reason why you can’t follow this approach and must have cytoplasms be primary objects, can you explain why? It is definitely harder, but it may be possible (if tricky).

2 Likes

Thanks for the help! There was really no need to make the cytoplasms primary objects. I’d just been sticking to primary objects up to that point, and I felt comfortable using them, but having them as tertiary objects was also fine. Anyway, the segmentation is a lot better now.