The quick answer is that:
(1) CellProfiler expects that foreground objects be brighter than background. For fluorescently labeled objects, that is no problem, but for histological sections or any traditionally stained bright-field samples, you’d first have to transform the images (try the UnmixColors module!).
(2) Bright-field image analysis is hard! If it is at all viable, add a fluorescent marker, e.g. DAPI. The added benefit is that you will be able to get a much simpler count of all cells (by counting separated DAPI-marked nuclei, and then count the stained cells too. As it is, I find it hard to distinguish between adjacent brown cells, but even harder is counting the unstained cells!
I did create pipeline to segment the stained cells (attached). Even this is difficult as it is, and I don’t think that counting the unstained cells will be that viable without a change in staining (nuclear, or possibly a better and more continuous membrane stain), though feel free to try and post your pipeline attempt!
DL_attempt_stained.cppipe (5.6 KB)