Varity in images due to background


I´m trying to identify stained nuclei (HTX) and also are of positively stained CD68 in our tissue samples. Though it seems that the images varies due to different background staining and therefore are hard to quantify. I have tried to change the intensity and diameter, though the problem is that when the settings fits one Picture another Picture is detected with a lot of background. Is there any possibility of detecting all of the Pictures in spite of different backgrounds?

I have attached some example Pictures and also the pipline i have used.
CD68.cpproj (100 KB)


Yes you may be able to automate this, howver histological image analysis is difficult due to the variation in staining. But the main problem with your pipeline is that you are using a “Manual” thresholding method. You definitely want to use one of the other settings, like “Automatic” or Global or Adaptive.

We would need to see the original images to be sure (without the outlines), but I would guess that you could try in both IdentifyPrimary Objects, Global-> Otsu-> Three classes-> (Assign Middle intensity class to) Background. Then adjust the Threshold Correction Factor for one image and test on a number of images.



And thanks for your replay!

I have understood that the settings for manual threshold is not to prefer, though I have never managed to get the settings right for the global. May I bother you again when attaching the raw images and see if you may find settings for the global threshold?
Best regards!

Sorry we lost track of this post. Did you still want help?