nninteractive segmentation quirk

[Daniel Harrison]

I'm wondering if anyone has run into a similar issue to what I am having regarding the AI segmentation in 4.4.0 (nninteractive)? The addition of this to ITK-SNAP has been great and the segmentation is very nice. However, I have found that it can be very difficult to use this tool if you have multiple things of the same label on an image. For example, imagine you have a brain image that has multiple lesions in different parts of the brain, some of which have heterogenous signal, all of which you want to label with the same color/label #. You use the AI segmentation on the 1st lesion and do some feedback to make it a very nice, perfect segmentation. Then, you move on to the 2nd lesion elsewhere in the brain and do your initial scribble on the lesion, but not only does this segment this lesion, it also goes back and alters the segmentation of the 1st lesion based on how you segment the 2nd lesion. This can be highly problematic if you have an image with many, many things on it that you need to have the same label. The only workaround I can find is to select a different label color/# for every single lesion, but this is highly impractical.

Any other workaround people have figured out?