Question on the identifySecondaryObject in the CellProfiler Pipeline (general question )

[arcolombo]

Hi. so my question is, 'how important is the 'identifySecondaryObject' / ImageMath module for the CP segmentation step using the bodenmiller pipeline? If i use nuclei as primary objects alone, will this lead to inferior phenotyping?'

So our tissue , some of the cytoplasm markers were inconsistent, so I was able to identify a segmentation mask using the nuclei primary objects, and I did not use the module 'identifySecondaryObjects/ImageMath', for summing the nuclear + cytoplasm probabilities in the probability map. note this was done because we don't have high confidence in our cytoplasm markers.

instead, i was able to identify a nuclei mask , and was able to identify cell objects using nuclei only.

my question is, how will ignoring the cytoplasm (and the image math) affect the downstream phenotyping?

Using histocat , the masks look quite excellent, so by visual inspection, it seems as though segmentation is quite good. I just don't know how this could affect phenotyping?

any suggestions will be greatly appreciated.

thank you!

AC

[nilseling]

Hi @arcolombo

That's a good point and we often observe that the nuclear mask potentially with a bit of pixel expansion leads to sufficient signal for phenotyping. You can test expanding the nuclear mask by one or two pixels and observe how that changes your results but overall your approach is valid.

[arcolombo]

Hi @nilseling thank you for the response. testing/comparison is a great idea. one concern i have is that using a strict nuclear segmentation, without border 1px expansion, do you think this will negatively affect the phenotyping such that CD4/CD8/CD3/ cell types will fail to be identified since these markers are mostly expressed on the cytoplasm? thanks again for the discussion

[nilseling]

I wouldn't worry too much about this as T cells are quite small and mainly consist of the nucleus anyway.

[nilseling]

Closing now