Cell counting H&E stain

Hi there,

I’ve stained my mice tissue slides with H&E and am trying to quantify the nuclei and compare between different samples. I’ve tried a few pipelines and seemed to have some luck with a simple one. Unmix Colors - EnhancedOrSurpressFeatures - Identify Primary Objects. However it did not work for all my pictures, the ones that had less hematoxylin weren’t accurate, even when I added to Color to Gray.

I’m using CellProfiler 2.1.1 on a macbook…I’ve attached an example of a picture Im using.



Looking at your image, there is a large illumination artifact. The image is about 60% brighter in the center than the corners. I think the best thing you can do to have any hope of doing image analysis is to do illumination correction first. I have started a pipeline for you and you can see that the image is much “flatter” in intensity than the original. From there try your other modules as you have already done (EnhanceOrSuppressFeatures -> IDPrimObjs) and it may help.
Also, I think your colors might not correspond to the default H&E very well so you can try the Custom setting in UnmixColors and/or set the color deconvolution manually.

My other thought is to try ilastik (bundled with CellProfiler) as a first step to train various compartments.

Hope that helps!
DLpipe.cppipe (9.77 KB)