Identifying apical cell area in epithelial cells

Dear CellProfilers,

I am trying to set up an automated way of measuring the apical area of drosophila pupal wing cells. An example of the type of picture I’d start with is shown below


The subapical adherens junctions between the cells are labelled with ECadherin-GFP. As you can see it is a very simple tissue with the two primary cell types (vein and ‘intervein’) easily distinguishable by apical cell area.

I tried to modify the ‘Tissue neighbours’ example pipeline but came up with 2 issues:

-   The 'overlay outlines' module would often put in junctions where no junction existed (i.e. through the middle of a cell). An example of mislabelled junctions is shown below:


As a result, many more cells were identified than were there, and average apical area was much smaller than it should have been.Is there a way of manually removing false junctions? Or is it just a case of changing the thresholding method?

  • Ideally I’d like a table of stats for the area of each individual cell measured in the picture, not just a summary of average cell area and standard deviation. Is there anyway of getting this?

I’m probably missing something obvious, but any tips would be very helpful.

Thanks,

Aidan

Hi Aidan,

The image indicates that the IdentifyPrimaryObjects settings need adjustment, probably tweaking the smooth filter size and/or maxima suppression distance to avoid mis-segmentation

The MeasureObjectSizeShape module obtains these measurements, just that the average/std is shown in the display window for clarity. ExportToSpreadsheet will allow you to export the per-cell table which will have the measurements for each cell; you should see both a per-image table and a per-object table as output.

Regards,
-Mark

Dear Mark,

Thanks a lot, that helped a great deal. I’m getting far less mis-segmentation on the images, and I also added the ‘display data on image’ module to show me the areas of specific cells.

I do have a couple of final queries. The first has to do with the area measurements themselves: I assume they represent the pixel area of each cell, but is there anyway of automatically converting them into micron measurements without doing it manually on excel? I know what to mutliply the pixel area by, but it would be nice if i could get the real measurements on to the image itself.

Secondly is there any way of colour coding each of the cells by their area (i.e. 200-300blue, 300-400purple etc), as a way of visualising trends in area differences?

If not, its not a big issue, as the new pipeline does everything crucial, and these are just presentation problems.

Thanks again for your help,

Aidan

You can use CalculateMath for this purpose. It can take a previously defined measurement as input and multiply by a constant. Select “None” as the operation and enter the conversion factor into “Multiply the above operand by”.

ClassifyObjects might be useful for this. Given the bins that you want the divide the cells into, it will output a color-coded image, show a histogram of the classification distribution and save measurements of the absolute number and percentages of the bins.

Cheers,
-Mark