How do I filter out centrioles relative to centrosomes?

Hi, all! CellProfiler amateur over here. I have been playing around with CellProfiler to find another way to quantify the number of centrosomes and centrioles with the program rather than manually. I have three channels: DAPI to stain for nuclei, RFP to stain for centrosomes, and GFP to stain for centrioles.

For the most part, I have not had any significant issues in identifying the nuclei with CP. However, the original image was very dim and led to poor identification, thus I adjusted the brightness and contrast of the channel in Matlab and exported the image to CP to identify the objects (nuclei). There is just some minor discrepancies when identifying the nuclei. As for the RFP and GFP channels, I did not adjust the images in any way.

One issue I am having is finding a way to associate each centrosome with a nucleus in order to determine the number of centrosomes per each respective nucleus. I know how to relate the two objects with each other as child and parent and have selected to calculate the child-parent distances to help with filtration. Selecting to calculate the per-parent means for all child measurements did not appear to do anything. How would I proceed from here to report the number of centrosomes per nucleus?

Similarly, I want to subsequently filter out all of the centrioles that are associated with a centrosome in a similar fashion. When I try to filter out the centrioles, I either filter out none of the centrioles or almost all of them in the image. The GFP channel has a lot of background noise, so when I am identifying the objects by size, CP recognizes approximately 50,000 so-called “centrioles,” and I want to filter these out by the parent (centrosomes). How would I be able to do this?

Please feel free to reply with any feedback or tips! Thanks in advance for all the help!

Pipeline Attachment and Link to Google Drive folder with images:
CentrosomeTest.cpproj (1.0 MB)


Please take a look at the attached example where centrosomes are extracted from your data (from the newest version of CellProfiler). You can use the same approach for RFP which should be simpler since there is less background noise like you mentioned. For such sparse GFP and RFP the “Background” thresholding method appears to work okay. You can still improve the resolution by adjusting the threshold parameters. Also, “ExportToSpreadsheet” provides you the results from which you can extract the number of centrosomes/centrioles per cell (should appear in the last column).
Let us know if you need any more help for a head start on this project.

CentrosomesVC.cpproj (674.4 KB)