Adipose tissue: cell surface area

Dear CellProfiler experts,

I am new to CellProfiler and i am impressed by all options this program offers. However, since so many options are available i don’t know where to start… :unamused: I checked out several examples (e.g. human HT29), and also searched this forum, but i could not get these to work for my question… I therefore would appreciate some assistance from the more knowledgeable people in setting up a pipeline.
I have a set of pictures obtained from adipose tissue from mice (HE staining). I would like to know the cell surface area of these cells, and the area frequency distribution. Plz see below for an example of my sections (also attached).

And to make the story complete here the graph i ultimately would like to prepare (panel A, B & C):


Taken from: http://www.jlr.org/cgi/content/full/43/6/986
Unfortunately these authors used a combination of commercial software (photoshop + set of special plugins) we cannot afford…

As said before, i don’t know how to get started so any hints to get me going are very much appreciated, especially how to properly process this color picture for analysis in CP, and how to actually determine the area of all cells.!

TIA,
Guido

Attached is a pipeline which finds the cells and measures their area, so you could create a plot such as the one you show. CellProfiler has a module called ‘DisplayHistogram’ that you can use to create a histogram if you’d like, or you can export the data into Excel.

-Kate
adiposePIPE.mat (1.14 KB)

Thanks very much Kate for providing a pipeline! It is much appreciated.
I’ll check it out and post my findings.

Regards,
Guido

Hi Kate and other experts,

I loaded the pipeline you (Kate) provided, and to understand all steps that take place I applied it to a randomly, selected file. The pipeline is finishing, however, I would appreciate if you could comment on the following (as you will see, i am a noob regarding digital imaging processing…):

My two main questions are these:

  • in step 2 of the pipeline (ColorToGray) the original image is separated in its three individual components (RGB). However, i don’t understand why in step 3 (CorrectIllumination_Calculate) (and all other subsequent steps) as input the Green color channel (OrigGreen) is used. Why is not chosen for ‘OrigGray’?
  • step 5 is the major step in the analyses, because here the cells are identified (IdentifyPrimAutomatic). I do understand all settings, but am I correct in concluding that ‘only’ circular cells are identified? I am asking because you have to set the ‘typical diameter of objects’. If i am correct, would it also be possible to identify cells with other shapes? In addition, i would appreciate if you could detail briefly on why thresholding is required (what is actually thresholded and compared to what?)

Also,

  • in step 6 the object Cells is measured. Could you actually select what is measured, or is this always a standard set of parameters?

I also added the modules ‘DisplayHistogram’ and ‘ExportToExcel’ to visualize resp. export the data. These two modules work fine.

Thanks again for your assistance!
Guido

[quote=“hooiveld”]

  • in step 2 of the pipeline (ColorToGray) the original image is separated in its three individual components (RGB). However, i don’t understand why in step 3 (CorrectIllumination_Calculate) (and all other subsequent steps) as input the Green color channel (OrigGreen) is used. Why is not chosen for ‘OrigGray’?[/quote]

The green channel was chosen because it gave the most favorable contrast out of the three color channels, in terms of high-lighting the “white” vs. “red” regions, for later cell identification. However, this may not always be the case if the color balance in your images changes.

Are you referring to the “Typical diameter…” setting? If so, then the answer is that objects of any shape are identified, but that the diameter here is actually the ‘equivalent diameter’, meaning the diameter of a circle with the same area as the object.

The thresholding is performed to identify the foreground areas (the Cells) from the background areas (the red regions). Some further adjustment of the settings in IdentifyPrimAutomatic may be required to optimize the detection.

No, this is merely a name to identify the objects. You may call them whatever you like, but remember that the objects must be referred to by that same name in later modules.

Hope this helps!
-Mark