Counting Double-labeled Cells (DAPI and Alexa 594)

Hi everyone,
thanks for all the help with Ilastik - now I’ve got some more questions but with CellProfiler (for an added twist). I’m trying to count the number of double labeled cells in many images: I have individual images for each channel (DAPI and Alexa 594). The essential part of this project is to count the number of RBPMS (c2) labeled cells, but given the difficulty I’ve had doing this, I’m trying to use DAPI staining (c1) as a seed to then identify the number of RBPMS positive cells.

Snap-1197_c1.tif (5.1 MB)
Snap-1197_c2.tif (5.0 MB)

I’ve also attaching my current pipeline DAPI RBPMS.cppipe (13.3 KB) , and am open to any suggestions - thank you all very much in advance!

Hi Philippe,

You haven’t mentioned the problem with what you have so far?
Rather than start from scratch, maybe try out the example CellProfiler give on their website
https://cellprofiler.org/examples/#cellparticle-counting-and-scoring-the-percentage-of-stained-objects
Hopefully you can then edit that instead of writing it all from the beginning.

Best wishes,
Cath

Hey Cath,
thanks for the link to the pipeline! I’m trying it out now. The problem that I’m having with my current one is that I’m getting numbers that are way too high for the cell count. I can segment the DAPI image very well, but I’m having more difficulty with the RBOPMS image, and I’m not really certain that I’m getting CellProfiler to count only the cells that are positively labeled for both DAPI and RBPMS.

Hi @Philippe_D_Onofrio,

I checked your images & pipeline, There are two things I suggest you,

  1. In the same pipeline you could use “Relate objects” to just identify the cells positive for RBPMS.
  2. You can use seperate primary object module to segment RBPMS.
    I have just edited your pipeline little with this, but you might hvae to optimise it with the parameters.
    Screenshot of relate objects & PFA pipeline,

    DAPI RBPMS1.cpproj (863.8 KB)
    Hope this helps.

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Thanks Lakshmi!
that’s good advice - I’m trying to optimize my pipeline. The problem I’m having now though is that CP doesn’t seem to recognize some of the dimmer DAPI stained nuclei

Any advice regarding this? I’ve tried changing the Threshold, but no luck so far.

Hi @Philippe_D_Onofrio,

You could try thresholding strategy as “Adaptive” instead of global & three class with foreground, since you have background, middle ground (cells with low) intensity & fore ground (bright cells). Also have to change the thresholding value accordingly.
Another option try rescale intensity module if it helps. Just sharing the screenshot of the primary object setting for your image set,

Hope this helps.

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Thanks Lakshmi,
these were great suggestions! I’ve tried to implement them and it looks like this really helped.

Hi @Philippe_D_Onofrio,

Great!! You are Welcome.

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Hey again,
I’m continuing to work on a CellProfiler pipeline to identify double labeled (DAPI and RBPMS) cells. For most of my images, my current pipeline works very well, but it seems to fall apart sometimes - I think this is due to variations in intensity across the image. I’ve attached some representative images that give the pipeline trouble Snap-1201_c2.tif (5.0 MB) Snap-1201_c1.tif (5.4 MB), Snap-1205_c2.tif (4.5 MB) Snap-1205_c1.tif (5.1 MB) Snap-1218_c2.tif (5.0 MB) Snap-1218_c1.tif (4.9 MB) . So far, I’ve tried to correct the illumination, but it hasn’t worked very well; I’ve also tried to use a smoothing filter, but also without great results. I’m posting my current pipeline DAPI RBPMS v5.cppipe (15.2 KB) and would appreciate any help - thanks!

Hi @Philippe_D_Onofrio,

I checked your problematic images & pipeline. Yes illumination is not working.
But I have tried smoothening alomng with the image math, it looks better. But slight dim the signal intensity, No worries about that since there is no background your segmentation is better & this is just to get the segmented boundary. you could take the measurements from the original channel. In case you need you could play around with the Smooth filter values.


You could use a similar approach for other channel as well.

PFA pipeline. Hope this helps.
DAPI RBPMS v5_LB.cpproj (864.2 KB)
Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Hi @Philippe_D_Onofrio,

I tried the illumination correction using different options, which doesn’t look bad. In fact no signal is lost in this way. I have added that too in the pipeline. you could choose the way you would like to use,

DAPI RBPMS v5_LB.cpproj (866.8 KB)

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

Hi Lakshmi,
thanks so much for your help! Do you think it’s better to use either the Illumination Correction pipeline OR the Smooth/Image Math pipeline - or should both functions be combined into one pipeline? I’m analyzing some images with both, but just wondering if you think they’d be useful combined.
Thanks!

Hi @Philippe_D_Onofrio,

Welcome. In your case you could use Illumination correction pipeline, if its works fine. I mean if you are able to segment the objects that you need from the illumination corrected image.
If you are not happy with the segmentation (or missing many objects), you could add smooth/Image math, but with lower smooth filter value. This is because you were missing slight signal in the smoothing & Image math step.
Hope this helps.

Regards,
Lakshmi
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Read more on our site.
Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging