Trouble identifying objects



Cell Profiler Crew -
I’m a new cell profiler user that would like to identify objects and retrieve data from the objects. I am trying to identify microbial colonies growing on agar. I have been using IdentifyPrimAutomatic and EdgeFind, but can’t seem to identify the objects. I have had mild success from EdgeFind, but no success from IdentifyPrimAutomatic. I have been manipulating ApplyThresh to get a better contrast between objects and the background, but this hasn’t seemed to help.
I also closely followed your speckles example, without success.

My typical Pipeline:

  1. Load
  2. Apply threshold - make colonies darker
  3. Invert
  4. apply theshold - make background darker (Essentially I have binarized the image)
  5. identifyPrimAutomatic

Do you have any suggestions?
Here is the picture I am trying to analyze.
Graduate Student
Dartmouth College


Hi Lucas,

My suggestion is to try the following:
(1) LoadImages
(2) ColorToGray: Split the image up into its red, blue and green components.
(3) Invert: Invert the red image, since it seems to have the best contrast of the three channels
(4) SmoothOrEnhance: This is the key module. Use “Enhance bright round speckles (tophat)”. Set the object width value to be as large or larger as the largest colony that you’re interested in, in pixels (for your image, I set it to 75).
(5) Crop: Get rid of the border, if possible.
(6) RescaleImage: Rescale the inverted red image result to be between 0 and 1
(7) IdentifyPrimAutomatic: Otsu Global seems to be OK, but you’ll need to adjust the size exclusion settings. Also, you probably won’t want declumping so set both “Method to distinguish clumped objects” and “Method of draw dividing lines” to “None”

Hope this helps!


See also the “Classified Colonies” example pipeline at
This seems very similar to your application. Best of all, there is a tutorial to walk you through adjusting it to your own images!