EMvision Research Notes, page-2540

i guess one thing I often wonder is that they may well have been faster in their development if they had trained and used a core team of actual human radiologists to read their images, rather than rely immediately on AI. The problem with AI is that there needs to be many many images included in the data package before the algorithm learns anything. Also going the full AI route might slow the integration and dissemination within the profession maybe.

I get that AI might be the future but I think it might slow the present.

But too late for that I guess.