Louis Vaickus

This problem would be very amenable to a self-supervised approach like CLAM (clam.mahmoodlab.org/). This process takes as input whole slide images and quantitative outcomes (e.g. malignant vs benign), "learns" to tell the difference, and shows you why it thinks that #CytoChat

I think one of the biggest problems with automating these tasks is the widely variable quality of IHC. In a fully digital ecosystem, one could envision automatic quality checks and perhaps even stain adjustment through style transfer techniques like GAN's #CytoChat

I can't tell you how many consults I get with estimates that turn out to be way off #CytoChat

Couldn't agree more. What I always say to my colleagues is that AI won't take your job, it'll take the worst parts of your job #CytoChat