Selectively count identify primary object

Hello everyone,

What kind of module do I have to add into he pipline to get rid of the dividing cells when counting the nucleus?

Canyou please suggest me on the pipeline?

href="/uploads/cellprofiler/original/2X/e/e0fe1b403f4d8bec49115a1cacc7d3055a60850a.cpproj">question-on cut off shape.cpproj (98.6 KB)

Thank you,
Ning

Hi,

Probably not perfect, you can try to add:

  • MeasureObjectSizeShape
  • FiltleObjects >> Limit >> object with FormFactor less than 0.75 and MinorAxis length less than 30 or so.

For better result, you may consider to use CellProfiler Analyst to train machine learning to do so for you.

Bests,

You could also try a texture measure for the FilterObjects module if the suggestions above are not specific enough.

If you want extremely high accuracy you could use the supervised machine learning strategy in CellProfiler Analyst called “Classifier”, BUT that would require adjusting the CellProfiler pipeline to measure a bunch of features and export the data in the right format so that CellProfiler Analyst can do the learning. I think you may get quite nice accuracy with the simpler approach Minh described.

Tell us how it turns out!

Thank you very much, Minh and Anne, it can differentiate between the dividing cells out form the non-dividing cell by uisng form FormFactor module only. I range very narrow values which indicates the circle shape.

I try to figure out the way to count the cells (FilteredNuclei) which containing >5 dots (Children) and calculate as percentage. From my pipeine, where should I change the parameter to get the %positive?

I have tried to open the cellprofiler analyst but it keep looking for the property file which I don’t quite get it yet.

17F25_foci count-positive cell.cppipe (16.9 KB)

If you want to use CellProfiler Analyst you’ll have to have your output in a database and have a properties file- the ExportToDatabase module will help you make both.

If you want to count the nuclei only containing >5 foci, you could add a FilterObjects module after your RelateObjects module to filter your FilteredNuclei based on the Children_AF647_Count (you could call the resulting objects something like “PositiveNuclei”). You could then just look at the Image spreadsheet to get the number of Count_FilteredNuclei and Count_PostiveNuclei in each image; you can also calculate the % of positive using the CalculateMath module- take the Image_Count_PositiveNuclei (optionally multiply that operand by 100), and divide by Image_Count_FilteredNuclei.

Does that make sense?