Issues with Recognizing Abnormal Nuclei


I am trying to use Cell Profiler to measure the extent of deformation of the nucleus for different treatment conditions. I currently use the parameter of form factor. This parameter requires the identification of the nucleus in the uploaded images in DAPI channel. However, under some of my conditions, the nuclei become severely deformed such that segments of the same nucleus are mis-recognized as a discrete nucleus. See image below as an example. I have tried modifying the size parameter for the software to recognize the bigger nuclei as one object, without success.

Is there any way to increase the accuracy and the sensitivity of identifying the nuclei with more severe deformation? I would ultimately like to compare the extent of nuclear deformation across different conditions.


The IdentifyPrimaryObjects module generally tries to seperate touching nuclei based on either shape or intensity. It’s possible to disable this segmentation altogether and just consider any touching staining as a single object, but that may not be entire desirable. The question I would ask is how you tell the difference between touching and deformed nuclei by eye?

It might be possible to approach this with the Watershed module, perhaps after running some of the morphological operations modules to smooth out holes left by nuclei being deformed.