Cytoplasm looks like patchwork



Good evening,

I have a pipeline that works quite well at defining the nuclei and the cytoplasm in order to quantify the levels of transcription factors in the two areas.

Even though most pictures are quite comparable in terms of background and are accurately interpreted by the pipeline (example in file A), my most recent data contains some files that confuses the pipeline leading it to draw huge cytoplasm. The picture eventually looks like patchwork!

My IdentifySecondaryObjects module is Global threshold strategy and Otsu. I have tried the three classes (example in file B) and the two classes (example in file C) and it’s not getting better. Increasing the threshold to the point where the cytoplasm is accurately interpreted results in too stringent delineation.

Would you have any pointers on how to deal with this issue?

Many thanks in advance for your support!



Do I understand correctly that your pipeline could identify the objects well with A but not so well in B and C?

Can you try any of the following suggestions (or both):

  1. Use CorrectIlluminationCalculate and then CorrectIlluminationApply for all the images, to “normalize” the discrepancy of brightness between images.
  2. Use EnhanceOrSuppressFeatures to enhance the appearance of the objects, then try to identify enhanced objects. Here’s a screenshot

Hope that helps.


You can also try one or both of these:
-Using the Distance-B method in IdentifySecondary instead of Propagate or Watershed; that will allow you to set a maximum number of pixels the cell body can extend out from the nucleus even if the threshold isn’t set quite right
-Look at the threshold values chosen for images that work well and images that work poorly; if they are substantially different (ie images that look nice have a threshold of 0.2 and ones that have large cell areas have a threshold of 0.05), you can try to use the threshold minimum option in IdentifySecondary to require that the threshold be at least a certain number (say 0.1 or 0.2 in the examples I just gave)


Thank you very much! I will give it a try!