H&E analysis

Hi,
I am trying to get the number of nuclei and ganglion cells from my images.
I tried few tricks I found on the forum but it is still not detecting 100% of the nuclei and cells are even more difficult.
I saw that you used MaskImage in few pipeline but I don’t understand why and how to append it into my pipeline.
Is it possible to crop the image without an ellipse or rectangle but instead a “manually clicked shape” ?
I attached my pipeline and an example of my pictures.
Thank you very much for your help.
Jeremie


HE_Pipeline.cp (17.6 KB)

Hi Jeremie,

I took a look at your pipeline thus far. The tricky thing with IHC images is that UnmixColors is not always reliable even with the stains selections in the module set to match what you actually used in the sample, because the sample preparation conditions are so often inconsistent even within the same assay. In this case, it seems that you may have better luck with splitting the RGB image into the component channels and working with them to identify the features that you want.

The explanation for the masking step is that thresholding by intensity for a certain feature is often confounded by regions of the image which are of a drastically different intensity than the regions of interest. In this case, the areas outside the tissue slice are white, which is very different than the colored tissue (nuclei, cells and all); conflating them together makes picking out certain features more difficult. Masking for the tissue explicitly helps in detecting only features you want, since the superfluous regions are not considered.

I’m attaching a pipeline which I think will will get you closer. However, I say again that you will probably need to check the performance of the pipeline across various slices and across multiple runs to make sure that it still works.

Regards,
-Mark
2013_04_11.cp (13.6 KB)