Help delineating complex ROI

Hello,

I’m trying to segment the plasma membrane vs. cytoplasm of fairly round cells. The cells typically have fluorescence enriched on the membrane, and also concentrated in the ER (which I’d like to exclude from the plasma membrane segment). Most of them are well-behaved and I can fairly easily get the membrane by identifying the nucleus as primary, identifying the cell as secondary using watershed, and then shrinking the cell by a fixed # of pixels to get a reasonable approximation of the inner membrane boundary. However, some of my cells have a more complex morphology - for example, in this image I’d like the membrane to be segmented as shown, but I can’t think of how to do it in CellProfiler. Algorithmically, I’d like to do something like start with the whole cells and then shrink each pixel inward by one unit until it detects an edge, kind of like a watershed but contracting instead of expanding. Is there a way to do this?

Thanks!

@DanP I can take a stab at it. Will you please upload the original image without any annotations?

Thanks you! Here are the images (they didn’t want to upload directly to this forum so I hosted them on Uploadfiles.io). c1 is my protein of interest (transiently transfected - only one of the two cells in the image is expressing it). c2 is my Hoescht-stained nuclei, where both cells are stained. My normal approach is to detect nuclei, identify cells as secondary, and then filter the cells and nuclei by intensity in the c1 channel to get rid of the nucleus in the untransfected cell. However, these cells are so round and sparse that I’ve been toying with the idea of just detecting them as primary objects directly.

Thanks again!

@DanP, please try out this pipeline. I’d be curious to hear your feedback on the outline of the “inner space” defined by the protein image. Unfortunately, the protein granule within the cell makes it difficult to find a circular segmentation. Am I correct that you are trying to quantify the amount of protein within the cell, while ignoring the protein signal in the membrane?

imagej-for-upload.zip (2.3 MB)