Identifying closely packed nuclei

Hi, I am an undergraduate student using this software for my honors project.

I have some problems in identifying closely packed nuclei of stem cells. I have played around with the settings in primary object identification, have also increased the resolution of the picture. I understand that the software might not be able to identify closely packed nuclei. But I am trying my best to identify as much nuclei as I can to extract information from the different colors. Am currently using otsu global on three class settings (foreground), have changed the nuclei settings accordingly to the image as well.

Have attached the pipeline I am using and 2 images too

Thanks in advance!


Quantifying DAPI.cp (5.43 KB)

Hi,

One issue that comes to mind is one with image acquisition: your images are somewhat saturated (i.e., many pixels are at the max intensity), which reduces the contrast and makes object detection more difficult. So you may want to check your exposure settings.

For images such as yours, in which there are many overlapping objects of various intensities, you may want to consider using Laplacian of Gaussian as the declumping method; basically, it’s a blob-finding algorithm. The downside is that it tends to be fairly sensitive to the setting choices, and what they should be is not intuitive. I’m attaching a pipeline which is my attempt to tune the settings to your images; more adjustment will certainly be needed.

Regards,
-Mark
2012_02_08.cp (4.05 KB)

Hi Mark,

Thank you for the reply. Have tried the settings that you have recommended for the nuclei identification. Works much better.

I have another question which would require your assistance as well. I would like to identify the cell outline (Secondary object identification). However, I only have the stainings of the outline of the cell via antibody staining of E-Cadherin. The outlines are not very accurate under the settings that I am using? I am using otsu global, propagation and 3 class thresholding and foreground for middle intensity.

Attached are the files

Thanks again

Yee Siang


Secondary object identification.cp (5.4 KB)

Hi Yee,

Part of the reason that the secondary identification isn’t as good is that the cell staining is fairly dim in places. This means that the initial thresholding only captures some of the foreground and the propagation method doesn’t expand from the nuclei as much as it should. To remedy this, you could try adjusting the threshold correction factor downwards. I tried 0.9 and the results seemed to improve.

Regards,
-Mark

[quote=“mbray”]Hi Yee,

Part of the reason that the secondary identification isn’t as good is that the cell staining is fairly dim in places. This means that the initial thresholding only captures some of the foreground and the propagation method doesn’t expand from the nuclei as much as it should. To remedy this, you could try adjusting the threshold correction factor downwards. I tried 0.9 and the results seemed to improve.

Regards,
-Mark[/quote]

Hi Mark,

Thank you for the prompt reply and help.

Yee Siang