Identifying nuclei from low signal images

cellprofiler
identifyprimary

#1

Hi!

I have a set of images with very, very low nuclei signal and I would like to be able to identify the primary objects - nuclei - from them. These cells were stained with Hoescht. I am attaching an example. The signal is so low that I am not sure there is a way to detect the nuclei but any input from you would be appreciated.

Thanks for your help

Susana


#2

Hi Susana,

Your images didn’t attach. Can you try again or upload them elsewhere (Dropbox, Google Drive, etc) and link them here?

Otherwise, in general using the “CorrectIlluminationCalculate” and “CorrectIluminationApply” modules can often be helpful in identifying objects when there’s low signal-to-noise- read the module help and play with the various settings. You could also try the EnhanceOrSuppressFeatures module if your nuclei have a very typical size.

Good luck!


#3

Thanks! Sorry - thought it had attached. I think this time it did attach.

Susana

Dapi-Cam.tif.zip (6.6 MB)


#4

Here’s an example of what I meant by using the CorrectIllumination modules- the segmentation here is far from correct, but it’s included to show segmentation is possible. I’d play with maybe using Global vs Local thresholding, using EnhanceOrSuppressFeatures, etc.

All that to say, I think it’s worth it to try to improve your staining, because you’ll never get GREAT segmentation based on this. LowSignalNuclei.cppipe (6.5 KB)


#5

Hi Susana,

I think you can also try pixel classification, that method can do wonders on low signal to noise ratio like this one. The program Ilastik allows to easily start classifying pixels from background and Foreground. I used it on your image and in 5 minutes I had a probability map on which I could more easily use the Primary Objects module.



The pixel classification will give you a map of the probability for each pixel to belong to the foreground or background class that can be used as an input image by Cell Profiler. In the primary object module, use a manual threshold of 0.5, to select the pixels with the best probability to be foreground.


#6

Hi Lorraine!

Thanks so much for your feedback. I will look into itaskit. If i have
questions can I contact you?

Thanks again

susana


#7

Sure, you’re welcome !