Cell detection without nuclear dye

Image Analysis
#1

Hi Team,

I have many confocal images in lsm format. The images have cells which have been stained with three dyes.
Two dyes are for two different proteins which translocate from cytoplasm to nucleus and thats what the goal is to track their translocation pattern. However, 3rd dye was used to mark the nuclear boundary NOT by staining nucleus BUT by mitochondrial dye which simply outlines the nucleus. So, all three dyes are cytoplasmic, two of them will show the translocation pattern.

How can we work with these images as boundary detection is first step to move-on?
Plz guide me.
Thanks
Mridul KK

#2

Hi,

It seems to me that whatever stain you are choosing to use to delineate the nuclear boundary, two things need to be true:

  • There needs to be good contrast between the perinculear region stained with the mitochrondial dye and the interior of the nucleus. Any sort of thresholding or image processing procedures will require a reliable difference in intensity between these two areas.

  • The perinuculear stain needs to enclose the entire nucleus perimeter, for the purposes of properly delineating these regions.

If this is the case, you would suggest looking at this thread. The experimental situation is very different, but the basic assay is the same: using a circular image feature to identity an enclosing object of interest. This may help you get started.

Regards,
-Mark

#3

Hi Mark,

Thanks for your reply and pointer to the thread. I assumed the similar thing to work with and I did tried with the pipeline given in the mentioned thread however, I didn’t get acceptable result.

So, I thought of giving you three sample images each representing one channel. Just briefly again,
Green: tagged protein will translocate from cyto to nucleus after stimulation.
Red: tagged protein will translocate from cyto to nucleus after stimulation.
Blue: tagged protein is basically a peri-nuclear one and will NOT translocate to nucleus and should show the boundary.
The nuclei itself is not stained with anything.

AIM: need to calculate the translocation ratios of two tagged proteins: green & red.

Kindly help me.
Thanks





#4

Hi Mark,

Thanks for your reply and pointer to the thread. I assumed the similar thing to work with and I did tried with the pipeline given in the mentioned thread however, I didn’t get acceptable result.

So, I thought of giving you three sample images each representing one channel. Just briefly again,
Green: tagged protein will translocate from cyto to nucleus after stimulation.
Red: tagged protein will translocate from cyto to nucleus after stimulation.
Blue: tagged protein is basically a peri-nuclear one and will NOT translocate to nucleus and should show the boundary.
The nuclei itself is not stained with anything.

AIM: need to calculate the translocation ratios of two tagged proteins: green & red.

Kindly help me.
Thanks

blue - Ch3-T3 - C3 Z1 T1.jpeg1024×1024 222 KB


red- Ch3-T2 - C2 Z1 T1.jpeg1024×1024 185 KB


green - Ch2-T1 - C1 Z1 T1.jpeg1024×1024 155 KB

#5

Hi Mark,

Did you get any chance to look at the images that I sent you?
Thanks
Mridul KK

#6

Hi cellprofiler team,

Im glad to receive constant help from you for any problems, i came across.
But this time, I’m not sure whether Mark is available currently to look into and help me. Can anyone of you suggest a pipeline/strategy to solve the problem mentioned here.
I have tried all sorts of possibilities to work it out but couldn’t get any.
Please guide me.
Thanks

Mridul KK

#7

Hi,

I’m attaching a pipeline that attempts to detect the cells and the nuclei in the blue channel. A lot of tweaking will be required to capture both the nuclei and the cells, but this should get you started.

Even though your intent in the blue channel to establish a clear boundary of the nuclear area, it is difficult to do so from these images. I can make out the nucleus in the middle cell but I would be hard pressed to do so in the upper and lower cell by eye. If a human is unable to do it, the software is unlikely to do the same. For the best accuracy, I think you really do need a nuclear stain. Is it really impossible to find one that will not perturb the biology response you are interested in?

Regards,
-Mark
2010_07_21.cp (5.8 KB)

#8

Thanks Mark.
I’m getting started with your given pipeline.

we are using RNA as a stimulating agent which means any dna binding dye will show florescence in both cyto & nuclear regions, making it eventually meaningless for the study we are doing. Thats the reason, we need something which basically performs a peri-nuclear appearance.

will using illumination correction or enhaceEdges or enhanceSupressFeatures can help in this direction?

Mridul KK

#9

One idea that comes to mind is the fact that you are using confocal images. Perhaps you could take a minimum projection of all the slices in a stack to better highlight the nuclear region. Since the region enclosed by the nucleus will have a lower intensity than the cytoplasm, you might be able to capture that area across several slices rather than a single slice where the nucleus may or may not be distinct. CellProfiler has a MakeProjection module which can take the pixel-wise minimum of a series of images as input and can be saved on the last cycle using SaveImages.

Regards,
-Mark