How long before >80% radiologists are replaced with a computer vision diagnostic system?

Medfag here. They will never be replaced - the AI will simply become a tool that radiologists will use to do their job.

You sound just as dumb as the people saying anesthetists will be replaced by machines. No, it's not that simple. The radiologist does far more than 'stare at xrays'.

> hurr durr medical databases and machine learning will replace X
> never happens, because X simply begins using it for their job and their job evolves a little bit but doesn't disappear

Okay, a little edit to my above post. They will EVENTUALLY be replaced in the way that all jobs of humanity will be, in like >100 years.

Sure, perhaps most of the radiologists job can be done by a computer eventually, but again, it's not that simple.

Something as 'simple' as an ECG analysis AI is still FAR from accurate.

>hurr durr my job is so complex and difficult xd

I said nothing about my particular job buddy

Im not your buddy, pal

Not soon enough. Given all the new research about how shitty humans are at shit like this I would much rather have my shit analyzed by a deep learning algorithm thank you very much.

Then you find out it was accidentally trained to recognize some irrelevant shit that happened to be correlated with the real symptoms in the training data.

>what is MRI
>what is kidney dialysis
>what is a defibrillator
Re-evaluate what you just said, and ask yourself what aren't humans willing to depend their lives on?

>oh no, it might make a mistake that humans already do!!

I suspect that it won't take very long, but it'll take longer than it should. Exactly because people like are the ones who will typically get to review papers that undermine their jobs.

I'm currently working on a paper in which various stages of coma are diagnosed based on resting-state EEG measurements, using machine leaning. Moreover, based on this diagnostic information I've developed a set of markers that also provide prognostic information. Using these markers I can predict, based on a single EEG measurement, with ~80% accuracy if someone is going to wake up from a coma, and importantly, what level of recovery can be expected. It's kind of a pain in the ass to have to deal with people who say shit like 'hurr durr but your machine might be wrong'. Fuck, it's demonstrably more accurate than any doctor could ever be by him/herself.

Until you can hold AI liable, never. Lawyers would prevent it.

Fucking democrat trial lawyers make so much from medical malpractice.

Usual misunderstand about automatozation. Automatization did not removed all the farmers, only 90% of them. Similarly it will not remove all radiologists, only a big chunk of them

>Machines that have critical-type tasks do not pass through hundreds of test before being put out in the field

The thing is, if the machine is trained and tested to work correctly, it just will.

People can be the brightest, but still make mistakes.

>nobody wants their lives depending on a dumb machine
> I am a radiologist

Having my life depend on a dumb radiologist seems even sketchier.

Can't be a day too soon, I have yet to meet a competent radiologist. As in competent enough to justify the amount of money they cash in for what they do.

they already have software that can determine what blood infection you have with a 90% accuracy and better than any doctor on the planet.

The reason why they dont use it is because there is general medication that can cure most and common blood infections.