"AI equal with human experts in medical diagnosis based on images" - Took er jerbs moment?

[Doctoscope]

---

Members don't see this ad.

Don't know if this has been posted here but,

AI equal with human experts in medical diagnosis, study finds

Research suggests AI able to interpret medical images using deep learning algorithm

www.theguardian.com

The article says deep learning systems correctly detected disease state 87% of the time, compared with 86% for healthcare professionals, and correctly gave all-clear 93% of the time, compared with 91% for human experts.

The article does acknowledge that it's based on a small sample size, so it could be just clickbait.

Is AI/ML truly capable of automatizing medicine? I feel like there were so many fields that people were confident were "machine-proof," and yet here we are. Maybe 30-40 years down the line there will be some Japanese anime robot that's capable of performing brain surgery while simultaneously dancing around. Or is this a nothingburger?

 

[907914]

This thread will be:

1/2 people saying it will never happen
1/2 people saying radiology will be non-existent in 10 years.

You’re all wrong. You’re all right. No one really knows. Probably gonna be somewhere in between.

 

[piii]

can't expect premeds to understand what a radiologist does, image interpretation is a small part of the job

 

[Cornfed101]

Most likely it will just make physicians more efficient and it will help them make better, more informed decisions. I don't think there is anything to worry about. I would be much more concerned (and excited) about your uber driver losing his job to an automated car. We still have pilots flying planes even though most of their job is done by the plane's autopilot. Robots are really bad at empathy and relating to people, which is a very important job of a physician.

 

[Matthew9Thirtyfive]

These studies are usually not as impressive as the headlines make them out to be. You have to look at the methods. Occasionally you’ll see a study like this and they will have the AI test on the same database of images it learned on, while the physicians were blind to the cases.

Also, FWIW, I spoke with a few radiologists this week about AI and they were very excited about it. Their points were basically:

1. AI will only make radiologists more efficient and free them up to do the cooler things and the more difficult cases

2. People said MRI would be the end of radiologists because it was going to show pathology so clearly that anyone could read it, and now it’s one of radiology’s biggest fields because no one has any idea what they’re looking at for the most part which is why rads residency is 5 years long.

3. If you think rads are going to be out of a job, go down to the reading room and then tell them they’re going to be out of work.

 

[AnatomyGrey12]

 

[scoopdaboop]

From the article itself:


Writing in the Lancet Digital Health, Denniston, Liu and colleagues reported how they focused on research papers published since 2012 – a pivotal year for deep learning.

An initial search turned up more than 20,000 relevant studies. However, only 14 studies – all based on human disease – reported good quality data, tested the deep learning system with images from a separate dataset to the one used to train it, and showed the same images to human experts.
The team pooled the most promising results from within each of the 14 studies to reveal that deep learning systems correctly detected a disease state 87% of the time – compared with 86% for healthcare professionals – and correctly gave the all-clear 93% of the time, compared with 91% for human experts.

However, the healthcare professionals in these scenarios were not given additional patient information they would have in the real world which could steer their diagnosis.

Now imagine this:

HOSPITAL BUYS (WHATEVER AMOUNT OF MONEY) AI MACHINE THAT CAN RECOGNIZE HUMAN DISEASE 1-2 PERCENT BETTER THAN THE HUMAN COUNTERPART WHO HAD NO ADDITIONAL INFORMATION. Yeah this isn't happening. It makes no sense. Why replace a physician who is just as capable, likely cheaper (considering all the IT support that is likely needed for AI), and most importantly, can talk to patients, answer their questions, and DO PROCEDURES. There is no way a physician is being replaced, just based on the fact that procedures are intricate, and can go awry at anytime and needs quick thinking to be able to act on the complication etc. What machine is gonna do a central line in the ER on a morbidly obese patient???

 

[Princeton Medical Student]

As someone studying computer science, no, not any time soon.

 

[Doctoscope]

It’s always fun to fear monger and irrationally speculate though, right? I mean if you can’t do that what’s the fun in life?

 

[Cornfed101]

It’s always fun to fear monger and irrationally speculate though, right? I mean if you can’t do that what’s the fun in life?

You could always take a cruise on the ConspiraSea. Great article from popular mechanics. Andrew Wakefield was one of the main people there!