Osteoclast segmentation


I would like to segment large multinucleated cells (osteoclast) and reject mononuclear cells. I have two channels, one to characterize the nuclei and the second the cytoskeleton (tubulin). Similarly, I would like to measure the peripheral density of this tubulin in the osteoclast. Attached a crop file as example. Thanks for your help.


The osteoclast has tricky brightness that is lower than the “noisy” surrounded mononuclear cells and merely brighter than background.
If you look for precise segmentation, please follow this lead https://blog.cellprofiler.org/2017/01/19/cellprofiler-ilastik-superpowered-segmentation/

You can try use the following CellProfiler-only pipeline, however you may not obtain perfect segmentation:

  • Use module “EnhanceOrSuppressFeatures” to neglect small mononuclear cells.

    Note: you may have to fine-tune the size of small objects to-be-neglect, which is around 20-45.

  • Then do a series of ImageMath to improve the signal of osteoclast

  • Use IdentifyPrimaryObject to identify osteoclast region.

  • Then shrink primary objects for about 30 pixels

  • Use IdentifyTertiaryObjects to define a “peripheral”.

I attach here a primitive pipeline, please modify to get better result.
Osteoclast.cpproj (641.6 KB)