|
|
|
|
|
|
by rksprst
5033 days ago
|
|
|
You're right physicians as a rule generally dislike software like that (partly because of their egos and distrust of software and partly because of job security). There are so many edge cases and so many examples physicians could give you of the AI/computer being wrong. A simple example is that any software will need input regarding the patient's history, his/her vitals. Right now heart rate data in ICUs is fed every second to storage systems. Let's say the this is a NICU and the baby's heart rate is suddenly elevated to extreme levels. The software would assume that something is wrong with the baby, when in reality a nurse is simply taking a blood sample and the baby started crying, being agitated by the nurse/needle. There are a lot of examples like this, it's a non-trivial problem to build use machine learning to take all of these factors into account to produce a diagnosis/treatment/intervention plan. I was recently at a medical/technology conference and the physicians in attendance were all actually in support of moving to using machine learning/AI (selection bias though, that was what the conference was about). They would rather have a machine with a 5% error rate than a human with 20% error rate. They believe that the role of physicians is going to change and become more of a management role as technology catches up. |
|
|