Necrosis in H&E stain


I have some hematoxylin and eosin stained tissue samples that I would like to segment the necrotic regions. I have attached an example of the types of images I am working with and the necrotic region is roughly circled. I was hoping to get help writing a pipeline to automate this process.

Thank you for the help!


Hi Jocelyn,

From the image you posted, it is not visually clear to me exactly what features comprise the necrotic region. Without that knowledge, it is difficult to get the computer to do the same. Could you provide more details on what you looking for in the image? Also, do you happen to have a 3rd stain that delineates this region?


Hi Sorry for the wrong reply. The necrotic region is loosely bounded by the region drawn on the image. There is no stain for necrosis in these images. I am trying to differentiate the necrotic region by features such as a lack of nuclei and the tissue being patchy in the region. While I was able to find pipelines that segment both nuclei and tissue, I am not sure how to agregate these objects to characterize the larger region.

Hi Jocelyn,

Hmm, you might be able to filter the segmented tissue by using RelateObjects to establish a parent-child relationship between the segmented tissue objects and the nuclei objects, then using FilterObjects to discard tissue objects with a child count of < 1 (use a cutoff of 0.5). Then you can use ReassignObjectNumbers to reconnect the remaining objects that are touching each other.

There’s no guarantee that the above would work, but it might get you closer to producing an object set that represents a broad region with no nuclei.

Hope this helps!