Segmentation problem

Hey,
I am using CellProfiler to segment nuclei, bacteria and lysosome in images which have been obtained by infecting primary human macrophages with mycobacterium.
I am also interested in performing single cell analysis on these macrophages to know how many bacteria are colocalized with lysosome inside a single cell.
Now coming to the issues that I am facing:

  1. While segmenting lysosomes, CP fragments the lysosome into two or three parts which from naked eye we perceive as a single lysosome.
  2. While performing colocalization, I want to have cell as parents with two different children i.e. lysosome and bacteria and then do colocalization of these children for every parent.
    But I think this doesnot exist in cellprofiler so I ended up assigning lysosome and bacteria as parent-child. If there is a way to do the two children thing please let me know.
    I am attaching the pipeline that I am using.
    The images can be found here dropbox.com/sh/u39h82toajrw … cH498MNxwa
    The folder also contains Illumination correction images.
    Please let me know if you need anything else.
    Thanks
    Ashima

P.S. You guys have done a wonderful job making this software and that too open-source. Thanks a ton :smiley: :smiley:
Colocalization Pipeline.cppipe (23.6 KB)

Hi Ashima,

[quote]1. While segmenting lysosomes, CP fragments the lysosome into two or three parts which from naked eye we perceive as a single lysosome.
[/quote]

I would look at the setting in IdentifyPrimaryObjects for the Lysomsome object named “Automatically calculate minimum allowed distance …”. Set this to “No” and raise the default from 3 to, say, 5. You can also try adjusting the parameter right above this on (“smoothing filter” size) in the same manner.

Here’s a ‘before’ example with your settings, zoomed in:
cl.ly/image/3d3w3D333G2f

And here is the “After” I changed the minimum allowed distance from 3 to 5:
cl.ly/image/3t3c1o3d2o2G

Looks better to me! Since the threshold seems a little permissive (to my eye, and this is totally up to you), you might also change the Threshold Strategy from Automatic to say, MCT (which is the default method but now allows more customization), and then raise the Threshold Correction Factor to globally raise the threshold.

You can use the RelateObjects as you have it (relating lysosome and bacteria) and then use FilterObjects to create a new “Overlap” object, assuming you want objects that actually touch to be considered co-localized. FilterObjects should be set so that the Parent object from the previous RelateObjects has a minimum of 1 child object (not the 0 be default). Then you could use another RelateObjects with Parent=Cells and Child=NewOverlapObject. I think that will do what you want - does that make sense?

David

Dear David,
Thank You so much for your reply.
“you might also change the Threshold Strategy from Automatic to say, MCT (which is the default method but now allows more customization), and then raise the Threshold Correction Factor to globally raise the threshold.”

This threshold correction factor is not available with MCT i.e. Automatic in my version of CP. I checked for updates too but there are no updates too.

“You can use the RelateObjects as you have it (relating lysosome and bacteria) and then use FilterObjects to create a new “Overlap” object, assuming you want objects that actually touch to be considered co-localized. FilterObjects should be set so that the Parent object from the previous RelateObjects has a minimum of 1 child object (not the 0 be default). Then you could use another RelateObjects with Parent=Cells and Child=NewOverlapObject.”

With this method that I am losing out the information on Non-overlapping objects so as to say non-colocalized bacteria and lysosome.

Please correct me if I am wrong.

Thanks
Ashima

Hi

[quote]This threshold correction factor is not available with MCT i.e. Automatic in my version of CP. I checked for updates too but there are no updates too.
[/quote]

Ah, I meant to change from Automatic to Global (sorry that wasn’t explicit) to look like this:
cl.ly/image/2w210o3T3i3v
Then you will have more options to adjust.

[quote]With this method that I am losing out the information on Non-overlapping objects so as to say non-colocalized bacteria and lysosome.
[/quote]

True, though you could also add another FilterObjects with the opposite criteria to capture the non-overlapping objects (i.e. a maximum of 0 children).

Best,
David

Dear David,
I got it. Thank You so much for your help.

Cheers
Ashima