Based on input you have. This would be like an enterprise software level program I guess and you would input history and then through trawling through data locally it can generate diseases and probability patient has each disease based on data inputted
Why doesn't something like this already exist? I am learning how to do differential diagnosis now and it seems use extremely rudimentary understanding of probability to diagnose things. You use clusters of symptoms and then use tests to eliminate stuff in the differential. It just seems like low hanging fruit that a program could do using tech we already have (I imagine LLMs will make it easier)
[link] [comments]