Counting all Stained Nuclei (Cells)

Hi everybody,

I have a small issue, I hope you can help me with. See the ImageJ-processed picture below. It has several stained (DAPI) cells in a 8-bit image. The staining efficiency is not that good because of some practical reasons.
Anyway, I want to count them. By eye it is clearly see-able, which parts are cells and which not. The problem is that with the threshold I cannot highlight the right parts. This is because in some parts the “background” or the space between two cells is brighter than in other parts the cells themselves. I wanted to use Threshold + watershed + Analyze Particles.

Do you have an idea how to deal with that?
Thanks in advance.
PS: This image is already processed: background is subtracted, brightness etc. is improved. Too high and too low values are excluded (with over/under and Huang). The high values came from dead cells wandering around.

20200331_100%-1_corrBackground-2.tif (5.6 MB)

And here is the original (background-corrected) image, if you prefer this one :wink:

Thanks.20200331_100%-1_corrBackground.tif (16,9 MB)

Hi everybody,

I wanna push this topic again, so maybe someone has a good idea?


Hi everybody,

my last try to push this topic again. Otherwise I have to think about another way to get an idea how to deal with that.

Thanks in advance.

I did take a quick look, but with the quality of the images, there were too many areas where the cells merged together. Also, it looked like the images provided were RGB, not fluorescent multichannel images. Except for the processed one, which seemed to be a merge of two of the RGB channels. I think you want better raw data, and yes, I noticed you mentioned practical reasons :slight_smile:

IIRC, and it’s been a few days, there might not have been any metadata in the TIFF either indicating pixel size.


thanks for your reply.
I know that the problem is the merged cells.
The picture is an RGB picture, but of fluorescent light (the microscope setting is like that).
I would love to have better raw data, believe me :wink:

Do you mean pixel size = a scale (bar)? I can provide that, but I do not know if that helps you with the processing?