Cell Size Measurements

[rolo0024 (Zack)]

Hello,

First and foremost, I am very new to microscopy, and even more so with CellProfiler, so I apologize if this question is basic. I am working with someone in my lab on measuring the area of satellite cells, and I’m running into some problems with the analysis picking up too much of the background. I have gone through the CellProfiler tutorials, and have done some other ones, but I have yet to figure out how to get my pipeline to quit picking up so much noise.

The images I have taken were done using brightfield microscopy, and I think therein lies the problem. The background of the image and the cells are similar on the grayscale, and thresholding is almost impossible. My question then is, is there a module that will allow me to better differentiate between background and foreground? Or would it be better if we just retake our images with fluorescence so we can separate the two?

As this is my first post, I am unable to attach any files, but they will follow in a subsequent comment.

Thank you,
Zack

[rolo0024 (Zack)]

Here is one of my images, and my current pipeline:
Snap-1165.tif (2.8 MB)
SC Pipeline.cppipe (7.8 KB)

Thanks for any help or advice,
Zack

[pearl-ryder]

Hi @rolo0024,

Thanks for your message and congrats on getting into microscopy and image analysis. I think it’s a fun world to explore!

I think you are exactly right that segmenting these cells will be challenging in CellProfiler due to the intensity values not being very different. Check out the two posts below, which offer you some options – some things to try inside CellProfiler, and if that doesn’t work, how to train an ilastik pixel classifier to help identify and measure your cells in CellProfiler. Good luck and hope this helps!

Pearl

Counting cell nuclei and positive marker expression with strong substrate autofluorescence

Hi @Francesco_P, I have a couple of ideas that might help. Your pipeline uses the Robust Background method for segmentation, which works best for a mostly black background (not the higher intensity background present in your sample images). An Otsu or Minimum Cross Entropy method may work better. Keep in mind that IdentifyPrimaryObjects works best at detecting objects with bright centers and dim edges In general, I think a major challenge for your thresholding is that the autofluorescence of …

Measuring the nucleus to cytoplasm ratio of a cell in an image with significant background noise

Hi @KirstenSteinke, I took a look at your pipeline and think that it will be difficult for CellProfiler to segment these nuclei and cells using classical image segmentation techniques. If you take a look at the pixel intensity values in the grayscale image that you create in the first module, you can see that background intensity values (~ 0.5) are essentially the same as the intensity values within your cells (~ 0.4 - 0.6). You can see these intensities by hovering your mouse cursor over the i…

[rolo0024 (Zack)]

Thank you @pearl-ryder! I will look into these and post a comment on where things are at. I appreciate the advice!

Zack

[rolo0024 (Zack)]

Hi @pearl-ryder,

I watched the tutorial you recommended in those previous questions/topics, and it really helped distinguish foreground from background. I was able to replicate all of the steps with my images. However, when I try to test it, I get the following message:

If I click ‘Continue Processing’ it brings me out of test mode. I’m wondering if it has something to do with the Metadata section? I used pretty much the same settings from the tutorial, but I had to convert my original images from .tiff to .png so the NamesAndTypes rules would work. Do you have any thoughts, or do you think I should post this as a new topic?

Thanks again,
Zack

[pearl-ryder]

Hi Zack,

That does seem to be an error with the import of your images in NamesAndTypes or the Metadata module. Perhaps you can share a test set of images (original image + exported probabilities from ilastik) and your pipeline? With that, our team can take a look and see if we can figure out the error.

Best,
Pearl

[DStirling (David Stirling)]

This looks like something caused by the image set matching system from Metadata/NamesAndTypes. If you could upload a copy of the pipeline which triggers this that’d be really helpful.

You should get a more helpful explanation than it’s provided, so there’s a bug in there somewhere.

[rolo0024 (Zack)]

Hi Pearl and David,

Thanks for getting back to me. I have attached one of my images along with the probabilities, and the pipeline I am using. For some reason I was getting an error when I tried to upload the original image in .png format, so attached it as a .tif file, but it will need to be converted in order to get the NamesAndTypes rules to work. Please let me know if I can send along anything else that will be helpful.

Cheers,
Zack

Snap-1165.tif (2.8 MB)
Snap-1165_Probabilities_0.tiff (1.2 MB)
Snap-1165_Probabilities_1.tiff (1.2 MB)
skmSC.cppipe (9.4 KB)

[pearl-ryder]

Hi @rolo0024,

I’m not able to reproduce the error you describe using these images. I converted the image to a .png file using ImageJ and then loaded it into CellProfiler without any error messages.

The first module then ran without any problems:

Did you make sure to click “Update” in the Metadata and NamesAndTypes modules? Sometimes errors arise when the image sets haven’t been fully specified.

Also, another way you can use NamesAndTypes to identify your original image is to specify that the file “Does Not” contain “Probabilities”. This allows you to use the original .tif file format (screenshot here taken from a different project w/ a similar workflow):

Hope this helps.

[rolo0024 (Zack)]

Hi @pearl-ryder,

Thank you for all your help! I was able to successfully analyze my images today with the information you provided. I did run into a couple more Javascript errors, but I was able to work those out by converting my original .tif files to .png files in ImageJ.

Thanks again,
Zack