Measure amount and area of melanocytes and COS cells




I’ve been trying to create a pipeline to measure the mean cell surface area of melanocytes (see example picture). But since there is much variety in the cell phenotype, I have not succeeded. After the module MeasureAreaOccupied, I try to use the module IdentifyPrimAutomatic to identify the cells. However, if I set the diameter range to fit the range of all cells, larger cells are not recognized as one cell, but are rather identified as several smaller cells. I wonder how I could deal with this problem?
I also tried the FindEdge module, which draws a nice, thin, but somewhat broken line around the cells. I don’t know in which module I can further process this image. There is unfortunately no module like ‘filling the lines’, which can fill the area surrounded by line, and to use that to identify cells.

For another cell line, COS cells, I want to identify the percentage of small cells . How can I let Cellprofiler calculate the ratio between number identified small cells and total number of cells? In the Module CalculateRatio I could only calculate the ratio between measurements such as area, but not number of cells.

So to make this long story a bit shorter, I hope someone could help me creating a pipeline which can identify all the cells and to calculate the ratio of cell amount. Thanks!



These are both great questions…
To address your melanocyte segmentation problem, I would ‘turn off’ the declumping in IdentifyPrimAutomatic. To do this, you simply need to set one (or both) settings for ‘Method to distinguished clumped objects’ to ‘None.’ From your image, it looks like each cell is well-separated, so it should not be a problem to identify each cell individually.

For the cos cells, it looks like your background is a bit high, which might be a problem. However, if you are able to adjust the settings enough, you should be able to properly identify the small and large cells by using the IdentifyPrimAutomatic module. To determine the ratio of small to large cells, you can use the ClassifyObjects module to determine which cells are ‘small’ and which are ‘large’. Then, you can use the CalculateRatios module to calculate the ratio of small to large cells. For an example of how to use the ClassifyObjects module, please try the ‘Classified Colonies’ example from the CellProfiler Examples webpage.

Good luck!


Thanks a lot! I’ve been able to identify melanocytes in images in which they are single cells. Thanks also for the tip to use the module ClassifyObjects with the COS cells. However I haven’t been able to figure out how to process it further in CalculateRatios, I do not know how to load ‘small’ and ‘large’ cells into the CalculateRatio module since it asks for a category of measurements, for input.
Another whole different question, it is somehow possible now or in the future to measure the length or branches of axons on a neuron?


Ah, I see what you mean about the CalcuateRatios module. It might be easier for you to simply calculate the ratio post-CellProfiler (in Excel or MySQL).
Yes, we have plans to measure the length of nuerons, but we do not have anything just yet. We will keep you posted when it might be available.




again, thanks for all the help! Now I’m wondering if it is possible to identify and measure only colocalized cells. Since I now have double-transfected COS cells (as above), one transfected protein is colored red and the other is green, and we want to only identify and measure the cells which are both red and green. Is there a way to do this in CellProfiler? We’ve been trying to merge the pictures with Combine module, which works pretty well but the intensities in red and green picture are unfortunately not equal. And after this module we do not know how to proceed with the cells which are both red and green. It would be great if you could help me out on this, thanks a lot!



Hi Lili,

I would recommend using IdentifyPrimAuto on one of the color channels to identity the COS cells. Then you can apply MeasureObjectIntensity on the identified objects but using the other color channel as the image. So (for example), this will assign each detected red objects an intensity based on the green image.

At this point, you can use FilterByObjectMeasurement to remove those red objects which have a low intensity in the green channel (i.e, red cells which do not appear in the green image). You would need to specify a measurement to use, such as mean intensity (Intensity, measurement 2) and a minimum intensity value to exclude objects with.

The end result of this approach will be objects which are sufficiently bright in both the red and green channels, which you can further process as needed.

Hope this helps,