Diffrentiating individual cells

I have this H&E staining of human breast cancer, and I want to differentiate between individual cells (nuclei and cytoplasmic).
Do you think it’s possible? If so, what do you think is the best way to do it? (maybe machine learning?) I remember someone presented this option in the last NEUBIAS conference.

IDC_37-3.tif (16.9 MB)
IDC_37-5.tif (16.9 MB)


Hello @Oshrat,

There is an article that tackles histological images with ilastik an cellprofiler: https://blog.cellprofiler.org/2017/01/30/be-a-histology-hero-with-cellprofiler/
Do you think this could be a viable approach?


I am not really sure what you mean here, but if you are just looking to classify the nuclei vs the cytoplasm, you can do that with the pixel classifier as shown here:

With the Positive Pixel detection command in Analyze->Region detection
or with small SLICs (Analyze/Region detection/Create tiles) which you can then add measurements to and classify.
If you want to detect cells as objects, you would use the Analyze->Cell Analysis->Cell detection
This is somewhat difficult, though, as you have two very different cell types with the smooth muscle tissue. You may be better off first classifying each type of tissue and then running different cell detections on each.

I didn’t take too much time with it, but you can see the fairly strong difference in Htx stain between the smooth muscle nuclei and the rest of the cells was problematic, as was the amount of background Htx staining in the sample.