Pipeline Help - Cell counting



We are looking to utilize CellPrifiler as a tool to count cells bound to a surface. Trying to tweak the pipeline to give accurate cell counts across varying images has proven difficult. We are using a nucleus dye to stain and capture an image using a standard flourescent microscope. Any ideas on how to improve cell count accuracy? Below are two of our more difficult images.

Thanks in advance.



Here is what I was able to do first using a SmoothorEnhance module to EnhanceBrightRoundSpeckles (tophat filter) and then IdentifyPrimAutomatic, using MoG global thresholding.

Is this closer to what you were trying to acheive? What type of cells are these?

Let me know if you need any other help,


Also, it’s worth noting that the images don’t particularly look like a nucleus-specific stain…perhaps the sample preparation could be optimized to make the nuclei stand out better and reduce the cytoplasmic staining?