Measure fluorescence in individual cells

Hello.

I have just started using CellProfiler and I want to do a very specific analysis. After a search in the forum I was able to find a pipeline that does a small part of what I wanted it to do and which I think is a good start point. I tried to combine it with some modules from a different pipeline to make the rest of the analysis I wanted, but there are still some problems that I can’t solve and other things I wanted it to do, but have no idea how. I would appreciate any tips on what modules might do what I want and what is currently wrong with what I am using.

So, ideally, I wanted the program to be able to identify the cells, and measure the fluorescence of each individual cell. Currently what it does is to identify the cells, but it considers the total of the image as one object and thus measures the intensity of all the cells there and not cell by cell, which is what I wanted it to do. Moreover, I keep getting the same error whenever I try to analyze several images (I send a print screen of the error in attachment). I can ignore the error and the analysis will proceed, but it is still annoying to have to click it every single time.
There are still some issues with cells detection, which I assume depend on a more appropriate tuning or the detection parameters. As you can see from the pictures, in some cases I lose a lot of information when identifying the cells.

To summarize, I would appreciate any help in knowing if it is possible to make the analysis of the fluorescence cell by cell instead of taking into account all the cells in one image and get those results in an excel file, while also getting back the pictures to make sure I didn’t lose too much information as in the case I show in appendix.

Thanks in advance for all the help!

Joana






yeast_cell_fluorescence.cp (5.88 KB)

This is an error that we have seen occur when the default input or output folders contains non-ASCII characters, e.g., non-English characters with accents over them such as listed here: extra.shu.ac.uk/emls/emlschar.html. If you can insure that the paths to your folders do not have such characters but are still getting this error, then let us know.

To figure out what the optimal settings should be, would you mind posting some sample representative images that your tentative pipeline is intended to be used with?

Cheers,
-Mark

This was exactly the problem. I changed the name of one of the folders and it is working fine now. Thanks!

Anyway, in the meanwhile, there was some improvement made on the pipeline, which is now doing more or less what I wanted it to do.

It was possible to improve object identification using RidlerCalvard adaptative instead of Background adaptative as the thresholding method for primary object identification, which was in the original pipeline I found for yeast. Do you also use these thresholding methods, or do you develop your own? Do you have any idea if this one is good for this job? I’ve only tested 3 images but it seemed ok. I’ve also managed to add modules to measure intensity, export it, etc. and it seems to be working!

I have another doubt that I couldn’t find mentioned in the manual. I can get intensity values as integrated or mean intensity. Integrated is not a good idea because it introduces a bias against smaller cells. However, I’m worried about using mean pixel intensity because of the edges. If the object boundaries are not defined exactly, won’t the program be averaging pixels from both the cell and the object edges/background? Maybe there’s another measurement that is more suitable, like mean intensity of the centre of the image or something? Is the median a good option for that?

Cheers,

Joana
yeast_DIC_pipe_intensity2.cp (5.93 KB)

We certainly use these thresholding methods as part of our daily work. As to which one is best, it strictly depends on your assay. If you find that one works consistently multiple images as compared to another, then go for it!

In general, I prefer to use the median intensity since it is statistically more robust to outliers than the mean.
You are correct that the quality of the object segmentation will affect the intensity measurements. This is why it is important to optimize the segmentation as much as you can before making measurements. If you are concerned about the influence of the edges, you could refine your segmentation to be more stringent, so that the object doesn’t extend beyond the edges as much.

Regards,
-Mark

Ok, new problem.

Although the pipeline works for some of the pictures, it doesn’t work for several of them (I send several examples in attachment). Not sure if it is because the fluorescence is low and it is unable to distinguish it from the background or not. Thoughts?

Thanks!
Joana









Hi Joana,

Since you didn’t attach a new pipeline, I went with the one you posted earlier. I removed some of the early modules, and went straight from LoadImages to IdentifyPrimaryObjects, and changed some of the settings:

  • Thresholding method: Otus Global, 3-class with the middle class set to foreground
  • Threshold correction factor: A less bit less than 1 (perhaps 0.95 or so)

Also, LoadImages should have “Individual images” under “File type to be loaded”, not “tif,tiff,flex,zvi movies”

Regards,
-Mark

I’m not sure how to open any of these attached pipelines. When I click the link, for example that under Joana’s signature, I’m brought to an empty window with the following address:
http://forum.image.sc//upload-forum-cellprofiler-org.s3-us-west-2.amazonaws.com/original/2X/f/f4be54702c4651fa63c5e0836c3ea501a8fec36b.cp

How can I download this pipeline to view its contents?

Thank you kindly!
Jackie

It appears that in moving the forum, the links to some of the file attachments were broken. We’re looking into and will get back to you when/if we’re able to fix it.
-Mark

Hi Jackie,
It looks like the broken links are working now. Give it a try if you’re still interested.
Best,
-Mark