Part of the issue is that the doctor is on the hook if the tool fouls up. Consequently, having an extremely skeptical viewpoint on tools is perfectly rational. I can also tell you from discussions with doctors that one of the problems is that the intersection of GUI programmer, competent engineer, and medical domain knowledge is either a null set or a single person. (For example: EEG analysis seems to be a natural fi…
There is a bunch of ML/EEG devices on the market, it is a commodity technology by now, just google "fda machine learning eeg".