Help with co-culture analysis

coculture

#1

Hi all,

I have been working on co-culture of two different cell types and trying to find a consistent way to quantify the percentage of each cell type in the co-culture. Nuclei staining usually turns out very nice so there is no issue quantifying them. However, both cell types exhibit quite similar-looking nuclei so it’s not possible to distinguish them based solely on the nuclei (green image). One cell type can be specially stained with collagen-I as shown in the red image. Mainly, I’m just looking for a reliable way to count these red cells, from which I could determine its percentage in the co-culture.

I’m completely new to Cell Profiler so I’m still playing around with all the settings and features of the software. However, if anyone can point me to some general direction on what I could try that would be very well appreciated. Is this achievable using solely CellProfiler or would ilastik also help?

Thank you!


#2

There is an example pipeline that might help you, on this page:
http://cellprofiler.org/examples/

“Cell/particle counting, and scoring the percentage of stained objects: CellProfiler is commonly used to count cells or other objects as well as percent-positives, by measuring the per-cell staining intensity. This pipeline shows how to do both of these tasks, and demonstrates how various modules may be used to accomplish the same result.”

Want to give it a shot and report back?

(and someday if you want to be very fancy, maybe you can assess whether there IS any feature of the nuclear stain that can distinguish the two cell types!)


#3

Hello,

I am completely new to CellProfiler and was wondering if anyone might be able to answer a few questions. The images I want to analyze are pap smears that contain both keratinocytes and neutrophil cells. I have been trying to work on a pipeline that can identify and count both types of cells, but I seem to have trouble having the program distinguish between keratinocyte nuclei and the neutrophils, since they are similar in size and intensity. Is there any way to improve the pipeline to be able to accurately distinguish between the two? I am attaching the pipeline I have been working, and would appreciate any help to modify it. I am also attaching an image that contains both keratinocytes and neutrophils as well as “control” images that contain mostly one or the other type of cell for reference.

Additionally, I am curious about how the CellProfiler Analyst works, and if this might be a useful tool for the project I am working on. If this is a valuable tool for this particular case, I would like to know how I can use the ExportData module to create a profile that can be used for CPA.

I would really appreciate any help, since I am so new to these programs!

Thanks,

Mika Caplan

Testing Cell Counting 9:26:17.cpproj (784.4 KB)


#4

Hello there,

What you’re aiming is actually more like a “high-level” classification, in which you’d combine many “low-level” features like pixel intensity, textural parameters etc… to distinguish neutrophils and keratinocytes.

Thinking in this way, CellProfiler only can provide to you the materials at low level, i.e. CellProfiler pipeline won’t be able to help you to immediately “segment” separately these 2 difficult cell types straight out-of-the-box.

So, I suggest you first just try to build a pipeline to segment correctly ALL cells in the pictures, and measure ALL features you could think of, like intensity, texture, size, shape, granularity, distribution etc…
You’ll then bring these materials into some form of machine learning to do the classification.

One of the ways is using CPA as you’ve mentioned, where you’d train a “Classifier”, i.e. hand-annotate which cell is which by drag-and-drop them into different bins. And hopefully one of the model can do the classification for you accurately.
Please have a look here

Hope that helps.


#5

Thanks for the quick response.

I’m wondering if you could help specify which modules would be necessary to segment all of the cells in these images, and which low-level features are best to use in this approach. Is there an example pipeline that has some of the features described? In order to use the CPA Classifier, would all cell types in the images need to be identified? If so, what is the best way for CellProfiler to identify everything captured in the image?

Thanks for your help,

Mika Caplan