Nucleus/Cytoplasm - Nucleolus/Cytoplasm ratios image analysis

Hello everyone,
I’m trying to calculate different parameters (see below) on AgNOR stained glass slides (at 100x mag). I tried to use the ImageJ plugin Intensity Ratio Nuclei Cytoplasm Tool" but as a kind forum user pointed out it is made for immunofluorescence and it will not provide any info on single cells.

Those are the parameters I will need to measure:

  • nuclear area of single cells
  • mean n of AgNORs per nucleus
  • mean area of AgNORs per nucleus
  • AgNOR area : nuclear area (ratio)

Can anybody provide any suggestions about which program (ImageJ, QuPath, others) should I use to do that? I am not very familiar with image analysis but eager to learn.

Thank you for your help.

Regards,

Sara

23154032-2.tif (1.0 MB) That’s an example of what my images look like at 8-bit

Another example of an original picture

Hello,

Your cells and staining are very difficult to observe. Do you think you could annotate on an image what is the nuclear area. In the title you mention Cytoplasm as well but you do not mention it afterwards.

Depending on what the annotations look like we can think of either ways of automating it with something like CellProfiler or QuPath. Do you have hundreds of images or just a few?

Best

Oli

2 Likes

Hello @oburri
Sorry I must have missed your message, thank you so much for

The cytoplasm is not actually stained with this technique, sorry, that was my mistake.
I have highlighted the cell nucleus (red) and the AgNORs (blue, in this case there are two). I have quite a few images (10 fields for each sample) but they will all look similar to this one (8-bit).

Thanks for your help!

Regards,

Sara

Thanks for the extra information.

Your task looks pretty complicated from a ‘single cell’ point of view.
As a quick example, which is something that was not really present in the image above, which of these manual segmentations makes the most sense to you?


Segmentation will be touch or will make the same kind of mistakes you see here. In cases like these, the experiment can answer “what is the global ratio?” based total nuclear signal area and AgNOR area, which can be found via threshold or k-means clustering.

Otherwise there is the possibility to try deep learning like cellpose but that’s way more contrived and I have very little experience so far. This is a result I got by just adding your 8-bit image.
With methods like these, it is likely you may have to annotate several images…

This implies that afterwards you would need to find another tool that could count or get the areas of the AgNORs.

I’m sorry I could not give a better answer.

2 Likes

Hello @oburri thanks so much for your help! Sorry I could not get back to you sooner. The second image is the one making sense to me. Which program can I try using to obtain something similar? Thank you

Sara