The issue is that you don’t actually have a clue what a radiologist does and how complex it is. I can show you three different MSK radiologists and they would have 3 different levels of sensitivity for calling, say, rotator cuff tendinosis vs tear, and they would each have different ideas of what makes the impression to help guide clinician decision making. There are so many variables, and having someone on the other end to talk to a clinician about what their findings mean/don’t mean and having that radiologist give his or her opinion is invaluable. The whole process is much more complex than you realize, and people seem to think it’s just look at pictures -> spit out findings, but it’s far more difficult than that and AI is unlikely to be able to recreate/replace/significantly reduce load for a very long time if physicians have to go back and read the AI prelim reports anyways.