Cell Clumping and Obvious Cells

Hello,

I am having some difficulties getting CP to recognize what I would consider to be fairly obvious cells. I am guessing that they are too close together (forming a clump that CP has a hard time separating), but cannot confirm this. As such, I am not entirely sure what to adjust for, since I do not know exactly what CP is having trouble with.

Here attached are a couple images (including a zoom with some of the cells that I am trying to identify) as well as the pipeline I am working with.
So far, the settings I am currently using produce the most accurate counts. I have tried using other declumping options (LoG) but they have not worked as well and for the sake of consistency across images in a same set, I would like to keep per-image analyses (i.e automatically counting different thresholds for each image) low.

Any suggestions as to why these cells are not being counted would be great.
Thank you for any help!
b431PIPE.cp (7.51 KB)





Hi,

None of the three images attached are the original raw images. Could you upload a sample of those images?

Also, your pipeline has an IdentifyObjectsManually module, but I don’t know what you are identifying. Can you clarify?

Regards,
-Mark

Hi Mark,

Thank you for taking the time for helping me out.

Here are the raw files (I should have thought of posting them earlier myself).
The IdentifyObjectsManually is there so I can define the object ROI, which is then used to mask the image.
The IdentifyPrimaryObjects then determines what is an object (i.e Cell) in the masked image.
There is probably a much simpler way to do this, I simply wanted to be able to specify what my ROI was so that not all the cells in the image were counted (I am only interested in a subset of the neuron population in these mouse brain sections).

Thanks,
Maxim



Hi Maxim,

I gave a try at improving the pipeline (attached) with the following steps:

  • Using IdentifyPrimary and IdentifySecondary to try to do an automatic tissue masking (which I’m assuming is what you were doing manually?). Basically, it tries to identify the nuclei and then simply expands them a bit.
  • The red stain regions are then identified by changing the thresholding and declumping methods in IdentifyPrimary.

Hope this points you in the right direction!
-Mark
2013_01_03.cp (10.2 KB)