1) You may try:
First use ImageMath to combine or other manipulations of cytosol, nuclei and bright field images, then use the resulting ImageAfterMath for IdentifyPrimaryObjects. That way you can test which channel is best for making the dividing lines.
You may try to use cytosol channel for IdentifyPrimaryObjects, then IdentifySecondaryObjects with bright field images.
Or you may try using ilastik as explained here
2) In general, you can try a combination of FilterObjects, RelateObjects or ReassignObjectNumbers to combine objects in multiple ways.
Hope that helps.