Masking cell aggregations


I am having some problems trying to mask some cell aggregations taken with phase microscopy.

I am trying to reduce the amount of cell segmentation so they can be counted as just one object rather than many. My current pipelines end up splitting the aggregates where the software ‘thinks’ there should be individual cells but in reality it’s impossible for any algorithm to count as they are too close together. Therefore I would be prefer to mask each aggregate as 1, or as close to 1 as possible!

So far I am trying the following pipeline…

Exp 34d.cpproj (417.3 KB)

The image should hopefully be saved onto the pipeline, but if its not loading properly please let me know and I can upload the image separately.

I am managing to capture some but not many of the aggregates. Also I have the option to disregard objects touching the sides checked so the final object identification looks better than it actually is…

I was hoping someone could offer some advice on whether this pipeline could be improved or whether it’s totally the wrong way to be doing this sort of thing?


Hi Clare,

You were close ! You could invert the phase first, then you do smoothing, and you got the expected result .

Bonus: when working with phase image, you can do an extra trick. After inverted the image, do another math: inverted image subtracts original phase image, it will further enhance your signal and darken the background.

Here’s a modified pipeline Exp 34d.cpproj (669.0 KB)



Thanks a lot for taking the time to help me! And for the trick with cleaning up my phase images. I will hopefully be able to do some fine tuning and have a good pipeline!