Help needed


#1

I am a new user of this powerful software. But I could not figure out how to get the number of positive cells and total cells of the Brdu staining picture.

Below is the link for the image:

pg.photos.yahoo.com/ph/talabia/d … pg&.src=ph

The dark brown is Brdu nuclear staining, blue one is hematoxylin counterstsing ansd also nuclear stain.

Thanks a lot!

Adrew


#2

Hi!
This is quite a challenging image. Most of our algorithms have been designed for fluorescence images. Still, I think you can possibly accomplish this project. You should start by using ColorToGray to split up the various colored channels in your image. You will then want to use an Invert module for each channel so that the nuclei are white(-ish) and the background is black(-ish). Then, you will need to see which channel clearly shows which type of cell (brown vs. blue). Perhaps combining a few channels together or subtracting one from another would actually be helpful; it’s hard to say.

Once you get grayscale images where your nuclei of interest are fairly white on a fairly dark background, you are ready to use IdentifyPrimAutomatic. You would probably use two of these modules: one module would count brown objects using whatever channel shows those nuclei best, and the other would count blue nuclei using whatever channel shows those best.

Hope this gets you started,
Anne


#3

Dear Anna,

All channels are correlated, why?

The proposed mixed image is quasi-Hematoxvlin and eosin (H&E).
This kind of stainning is more used to see the morphology.
I dont think is a good idea to use segmented (H&E) image to count the nucleus. And more any measurements related to (H&E) intensity is not suitable to conclude any biological fact about nucleus:

(H&E) is more related to the amount of protein, carbohydrate, lipid, and water, that filled the specific cell, stroma, epithelial and cytoplasm, for instance.

The stroma always stain pink as it have more protein, this is normal because it’s a muscle. In your image, stroma is scarce, may be the
biopsy is from a tumor, nucleus are invading the stroma …

Normally nuclei always stain blue not black, with at least a rim of dark blue at their edges.

The cytoplasm is more variable in its appearance:

Protein-rich cytoplasm stains dark pink.
Cytoplasm can stain rich purple, pink+bleu.
Cytoplasm that is mostly filled with carbohydrate, lipid,
or water will stain pale, close to white color.

On can assume that the information source is ergodic for any specific kind of cell. It implies that the information source has only one statistical structure for any specific cell. However, the data generated(image) from stained tissue is from a mixed source made up of a number of pure components that are each of homogeneous statistical structure.

This statistical structure can confuse blood with nucleus, for instance.

Most of cytoplasm will be detected as nucleus. Bias.

With my best regards.

F.


#4

Hello,
Thanks to Faysal for the input and the description of what is stained and appearing as the various colors.

I think the original question was simply to count the number of cells that are brown and the number that are not brown, which does not rely on any assumptions about the quantitative nature of what is being stained. Indeed, making assumptions about absolute protein levels based on intensity levels is very dangerous and not recommended. But a simple yes/no scoring of the brown BrdU stain is reasonable, assuming that the nuclei can all be identified properly.

A note I forgot to mention to the original poster - it may be necessary to obtain higher-resolution images, by the way, because I am worried you will not be able to accurately identify the nuclei at the current resolution.

Anne