Segmentation help needed

Hi there,
I try to segment cells but I’m not satisfied with the result so far. Could you have a look at my example image and my settings and help me to improve the segmentation?
I’m not happy with the division of clumped objects. How can I prevent these small lines which go a little bit into some objects. And what about the holes?

Antje




Hi Antje,

The behavior of the Laplacian of Gaussian declumping is something we’ve noticed as well; we will look into it to examine the reason for the odd lines and the holes.

In the meantime, you can try two things:

  • A possible workaround is to use IdentifySecondary with these oddly-segmented primary objects as input, using exactly the same thresholding and upper/lower limit settings. I believe this will have the effect of filling in the holes, and closing off the lines into the interior.
  • This is more of a question: Did the intensity setting for declumping not work well? It appears as though some tweaking of the smoothing and maxima separation parameters might yield a good result…

Regards,
-Mark

Hi Mark,

thanks a lot for your answer. I’ve tried several solutions in the meantime but I’m still not satisfied. Your first proposal might work but I also would like to solve the problem with the nuclei which were splitted by mistake. And there are very faint nuclei present I’m loosing. I’ve got a better smoothing now but still it splits some nuclei in two (no matter which method I choose to distinguish and split clumped objects). I’m not sure what else, I can do…

Antje

Hi Antje,

At this point, I can’t access the problem without seeing an actual image and the pipeline you are currently using. Could you upload both of these items?
-Mark