Count percent positivly staining cells

Image Analysis
#1

Hello friends,

I am trying to create a pipeline I can use to count the total number of staining and non-staining nuclei so I can find the percent of cells which are marked with my immunostain (MIB1 or KI-67). This is a task I do routinely manually and using an automated process would be advantageous. I have attached two example images. The example images are representative of what I do routinely. The cells which are marked/stain positively are the dark brown cells in the attached images. The nonreactive cells are those which are pale blue. I have also attached my attempt at creating a pipeline, but I know that this is above my head. I get a whole lot of errors along the way. Does the program have to crop my images? Can I not use any image size? I do appreciate your help with this.

Kindly,
Namocet
PercentPositive.cp (16 KB)

2.jpg1360×1024 516 KB


1.jpg1360×1024 529 KB

#2

Hi Namocet,

Thanks for trying CellProfiler. Hopefully this will help – please take a look at my attached pipeline, with these notes:

  • LoadImages – You only need a single instance to load ALL .jpg files.
    SIDE NOTE – avoid JPG files (more info) if possible and try and use something like PNG or TIF
  • You are correct – no need to use Crop here. I removed them.
  • For IdentifyPrimaryObjects to work (and many other modules), you need to input grayscale images. This was the primary error you had. Since you are using histological images, your best bet is to use UnmixColors
  • UnmixColors: I guessed at a ‘stain’ but you can use your knowledge or just empirically find one that best enhances the stained (and unstained) cells.
    SIDE NOTE – if you upgrade to our recent release version of CP, then you may notice a bug in UnmixColors that I just discovered. The workaround is to turn off the display window when running in Analyze Mode. But I will give you a version that will work with your version 11710.
  • EnhanceOrSuppressFeatures – these try and highlight the cells relative to background. They help some certainly, but only a little when the real test comes in the next module IDPrimary. These could be removed if you need the speed improvement.
  • IdentifyPrimaryObjects – I changed some of the parameters (thresholding method, threhold correction factor, and declumping settings) but these can also be tweaked by you.

Hope this helps!
David
PercentPositive_DL.cp (8.28 KB)