How can we eliminate unspecific overall high nuclear signal to be counted as foci?


#1

We have been trying to set up a pipeline that can identify and count 53Bp1 foci.
The pipeline that we have adjusted could identify most of the foci we want it to be counted. But we have a problem with cells that do not have clear foci but high intensity of the overall nuclear signal of the 53Bp1. In those cells, cell profiler recognises, especially the edges, as foci. How can we reduced this false counting?
As attached to this mail, you will see that we have an “enhanceor suppress feature” before IdentifyPimary Object. We observe that in problematic cells (no foci but high signal of 53BP1 foci), enhancements create a contrast and this then recognised as foci in the identify Primary Object. Thank you very much for your help!



#2

The problematic cell in the above image is the one at the top right…


#3

Hi there,
We might be able to provide better help if you can include your images and pipeline. It sounds like you have chosen a good strategy but it just might need some tweaking.

Worst case scenario, you can use MeasureObjectIntensity and then FilterObjects to delete cells with an integrated intensity above a certain amount.


#4

Hi Anne,

Thank you very much for your input. I tried to include the pipeline and images in my previous email, is it not visible by you? Maybe I attached wrongly, Please let me know,

Regards


#5

20170906-G1-53BP1-CYCA-GFP-RESCUE.cppipe (22.7 KB)

Here we go, thank you,
Ozlem


#6

There are 4 channels, red:53Bp1, far red:cyclinA, green:GFP, blue:DAPI.


#7

Dear Anne,

Have you had chance to look at our pipeline? I am looking forward to hearing from you,

Best,

Ozlem


#8

I hope someone from the team will be able to look soon. We’ve been swamped getting CellProfiler 3.0 ready to release!


#9

Hi,

Hmm…
I think you already did it the best way.

There’s little to improve here, with overshot /saturated signals like this. In general, if you hunt for tiny objects, the full dynamic range needs to be taken care of right at the time of image acquisition.

For these images, you probably can filter the bright and big objects at the edge of the nucleus by using FilterObjects with the restriction on size (after MeasureObjectSizeShape), intensity and child-parent distance (after RelateObjects)

Good luck.