Counting objects (nuclei staining) with some background and unspecific staining


Dear all,

I really appreciate someone will help me out.

I have problems in counting the objects indicating the nuclei. The following is original image.

After I used the following setting in IdentifyPrimaryObject,

As could be seen from the images and table below, the counting was incorrect and the number of accepted objects in the output was too high

I am aware that I need to adjust the threshold, smoothing factors, and filters. But I am lost in how to start adjusting. For example, in the following image, the nucleus-like white spots in the green circles are the objects I want and the areas or lines in the yellow circles are the objects I don’t want to include in the counting.

I have the same problems in counting the nuclei with DAPI staining too. The following image is an example.

Many thanks for anyone who has the patience to finish reading my post!
Looking forward to any advice!

Best regards,


Hello there,

So I gave it a few try and think that the best way is to use CellProfiler in combination with ilastik.
Please follow this tutorial

Read more about the method.

In your particular case, you would need to train ilastik to recognize at least 3 labels: the nuclei, the basal lamina and the background.

A (super) quick test example:

Hope it helps.


Dear Minh,
That’s fantastic! Now I know where to look and how to improve. I am going to try it out.
Thanks again.