Adipocyte H&E Cell Profiler Pipeline

I am trying to analyze adipocyte size from some mouse fat depots using H&E sections…

I would like some help with a pipeline because there is little cytoplasm and often the adipocyte cell membrane’s “break” during the histological prep, causing CellProfilier to count 2 or more cells as a single cell. I have attached our pipeline and 4 representative images of the H&E sections.

Thank you for you help

Chris








AdipocyteSize.cp (5.13 KB)

Hi Chris,

I see that “Male EWAT HFD 12 Week #1.1” is listed twice, once as an original and another as a copy. Did you mean to upload a different image? Also, is it possible to get versions of the images without scale bars?

Thanks!
-Mark

Hi Mark,

Thank you for taking a look at the pipeline and images. Yes that was a copy of the same file, my mistake.

Please find attached some different images without a scale bar.

Chris








Hi,

I have attached images taken with the fluorescence setting as this sometimes helps the contrast due to high auto fluorescence. If I can supply better images or samples, please do not hesitate to ask.

Thank you

Chris








Hi Chris,

These flourescence images definitely help; they are much easier to work with than transmitted light images. However, generally speaking, JPG is not the best format for image analysis (see here); would you be able to provide these images as TIF or PNG?

Regards,
-Mark

Hi Mark,

Please find attached the PNG files. The Tiffs were >10Mb each but I can send a server link if you would prefer the TIFFs.

How is the pipeline going?

Thanks

Chris








A question for clarifycation: What sort of readout do you want here? A per-image measurement (e.g., average cell area), a per-cell measurement (i.e, a list of each cells’ area), or something else?
-Mark

Hi,

I am looking for a per cell measurement so the output will include a list of all the adipocytes measured. We will have multiple images of the same fat depot.

Our aim is to take many images of adipose tissue from different fat depots and multiple mice to come up with a measure of cell size increase (or decrease) in response to obesity. As the fat depots are large it would be great to count larger areas and lots of cells. Currently we count around 100 cells by circling adipocytes manually using image J. Hopefully we can input lots of adipose images and have a distribution of adipocyte size per depot and per mouse.

Does cell profiler measure pixels? If this is the case then we will need to convert the number of pixels (or area) back into adipocyte size/area (um/um2).

Thanks

Chris

Hi Chris,

Attached is a pipeline that should get you started. I’ve annotated the pipeline in the module notes (at the top of the module settings panel) so you can see my thought process for the workflow.

Note that at several places in the pipeline, I mention that if the image resolution/magnification changes from that used in the current images, some size-dependent settings will need to be modified accordingly.

Regards,
-Mark
2012_04_22_Flu.cp (13.9 KB)

Hi Mark,

Thank you, the pipeline is looking much better running the 4 png images. I will gather more images and run the pipeline.

Have a great weekend

Chris

I’ll refer you to the FAQ on this issue. The pipeline I uploaded incorporates the FAQ’s response.
-Mark

Hi Mark,

We are testing out the pipeline on some more images. Unfortunately, some of the images are not as good as the ones we set up the pipeline with as there is more vasculature and broken membranes…

Is there a way to label the cells on the output image so that I can go back and remove cells that are obviously wrong by eye e.g. broken membranes or very large cells drawn from the vasculature? This extra quality control would really help the accuracy.

Thanks

Chris

You might want to try inserting the EditObjectsManually module in the pipeline if you want to remove such objects as the pipeline is running. Alternately, you can use DisplayDataOnImage to place the cell number on the image if you want to do this afterwards. The cell number is referenced as “Number” (category) and “Object_Number” (measurement) for the Adipocyte objects. Lastly, you can remove very large cells by adjusting the upper diameter limit in the 2nd IdenitfyPrimaryObjects module.

Cheers,
-Mark

Hi there,

I am in need of some help…

I am trying to quantify adipoctye size and number from H&E stained samples…

I have been playing around with the flu(2) pipeline that was posted to help others…

Using this pipeline I am still omitting about 50% of my cells…

Any help would be greatly appreciated!

Thank you
Britt

Hi Britt,

In order to help, you would need to post some example images that you are using in your assay, so we can try out the pipeline on them.

Regards,
-Mark

Please find attached the slightly modified cell profiler pipeline we use for calculating adipocyte size from H&E sections.
Adipocyte Pipeline numbered Rodeheffer Lab .cp (14.2 KB)

Hi cdchurch,
Thanks for contributing the pipeline. Hopefully, bcoats06 finds it helpful.
-Mark

Hi there,

Sorry to revive this thread after so long. I’ve been using the pipeline that cdchurch posted and it works really well. However, I’m having a small problem with a part of the pipeline.

The software does a really good job of predicting where the adipocyte membranes have been broken and filling them in. But it’s been overzealous and filling in large regions of empty space with a predicted membrane where there aren’t any adipocytes at all.

Does anyone have any suggestions on how to adjust the pipeline to make it more selective in where it “completes” broken membranes?

-Alex

Hi Alex,

Can you please send your pipeline and the images where you see that occur and we can take a look at it? It will make it master to find the solution.

Thanks,
Vas

Hi Cdchurch,
Have you by any chance modified your pipeline to make it compatible with the 3.0 version? I am totally new to CellProfiler and haven’t really started understanding the way it works. Any help with working pipelines is appreciated!
Regards,
Nicolas