AI threats? Please help an MS4 who wants to do rads but is terrified :(

[thecanadianIMG]

Hello AI, Goodbye Radiology as We Know It

A machine able to interpret diagnostic imaging studies better than radiologists has long been foreseen, yet its arrival comes almost as a surprise. We have underestimated the potential of AI to perform the kinds of work we do.

www.radiologybusiness.com

---

Members don't see this ad.

 

[Cognovi]

Can you put more effort into your posts

 

[Tiger100]

Won't replace radiologists. period.

There is also an article that says the world will end in 2030. Why do you want to go to residency at all?

 

[808s&heartbreak]

i do see where your concerns are coming from with regards to the exam you posted. Overall, I do not think it will replace radiologists. There are some exams which I think can be completely automated end-to-end without radiologist input. Screening mammography being one of them, and I give it 5-10 years before complete automation. The technology is here.

The exams which are at risk of automation are those of "perceptual" (e.g. is there a calcification or not, is there architectural distortion or not). The pathology in breast imaging is pretty narrow. Your one goal is to look for cancer vs no cancer. So in that regards I do think screening mammogram can be automated.

There are more than enough other modalities to keep radiologists busy, though. Overall I see AI as a plus.

 

[.Asterion]

resident's perspective on AI

how do radiology resident's feel about AI? excited? fearful? a combination of those emotions? also, do any of your training institutions offer any lectures on AI/ML? i try to follow developments in AI, and there are at least a few articles a week that kind of make me quiver as an MS4 entering...

forums.studentdoctor.net

I felt like the above was helpful.

Overall I was concerned a bit about AI when I decided on radiology last year but the more you actually look into AI you realize it's really good at narrow stuff. The people who say it will replace radiology are CS people with no experience with radiology or radiologists with no experience with CS.
The CS majors in radiology can give you the best insight into where things are. Among that group much more conservative estimates.
Also there's other posts of CS people that said it would be easy to replace radiologists then after their first year of radiology said they were mistaken and it will be way more difficult than previously thought.

I think it's reasonable to ask if your field is safe, anesthesia applicants look into CRNAs, EM into NP/PA, etc. Overall my take was AI is potentially our version of midlevels. Another thing is over the past 30 years CRNAs have taken over more than half the field of anesthesia but yet anesthesiologists are still in extreme demand making bank. There was a problem with the radiology job market in 2013-2015 because the combination of rapid expansion of residency spots and almost all rads about to retire all at once deciding not to retire (because of losing 30%+ of their savings 2009-2010). The net effect of surplus labor and complete lack of demand. That's why there was an issue. If AI gradually chips away at our work (like CRNAs) we will still have jobs and there's no issue. The most repetitive parts of our job will be automated first so it will just make our job more enjoyable.

There's no perfect field, only think we can say will 100% certainty is that in 30 years CS people will have jobs haha. I just figured I'd do what I love. Ai in radiology might even turn out to be like self-driving cars where in 2015-2016 they said it would be easy to have them on the road by 2020 and now they're saying it might be impossible (outside of good weather and predefined routes). Not to say it never will be there but driving is also WAY less complex than all the pathology that could possibly show up on imaging.

 

[Dave1980]

That article is so dumb:

"Until now, AI has outperformed radiologists only in limited aspects of a larger imaging study—counting and measuring lung nodules on CT, for example, or detecting pneumothoraxes on chest X-rays. But the system described in Nature is a major advancement: It’s the first system able to interpret an entire type of study better than radiologists"

it performed one function once again (finding a cancer). it is very misleading to say that it interpreted "an entire exam" where before it was only part of an exam.

I will be impressed when it can read a CT abdomen for HCC post ablation

 

[recycledpaper]

As I near the end of R1, and having heard a lot of talks and read papers on AI in radiology, I am comfortable that my job is safe, at least in my lifetime.

 

[deleted746658]

"in my lifetime"

this technology will improve at an exponential rate.

i'd be careful of using "my lifetime" - it's like saying "always" or "never"

 

[NecrotizingFasciitis]

I wondered the same thing for a minute (OMS-IV applying rads next season) until I watched this video. This dude is awesome; I follow him on Instagram too.

He basically says AI in rads is only going to help and that rads is one of the fields AI is synergistic with its' field in the sense that it will make life easier and not pose any threat to jobs.

 

[Prehealth1011]

I wondered the same thing for a minute (OMS-IV applying rads next season) until I watched this video. This dude is awesome; I follow him on Instagram too.

He basically says AI in rads is only going to help and that rads is one of the fields AI is synergistic with its' field in the sense that it will make life easier and not pose any threat to jobs.

Not a Rads applicant, but i would honestly be more worried about technology/nuisance that already exists in every other field (NP/PA) than AI. Rads seems "safer" to me than many other fields.

 

[odyssey2]

I'm more worried about midlevels in other fields than computers in this one.

 

[Epi]

There was this concern years ago when I was in residency, along with outsourcing night reads and eventually all work to Teleradiology companies. Concerns haven't materialized to a significant or career changing degree in 10+ years. IF anything I agree AI would be nice as an adjunct to radiologists work and help increase quality, would be a long time to actually replace a human being. Lots of radiology is not black or white, not just no fracture or fracture. There are clinicians that like to discuss cases with a live human being who knows medicine.
Volume of studies also keeps on increasing each year, sometimes as much as 5-10% per year in some modalities. I wouldn't mind if a computer helped me out and dictated a few cases for me. I could see compensation changing as a result of this, but lots of things in this current environment is putting pressure on compensation and workload.

 

[Kr#36]

There was this concern years ago when I was in residency, along with outsourcing night reads and eventually all work to Teleradiology companies. Concerns haven't materialized to a significant or career changing degree in 10+ years. IF anything I agree AI would be nice as an adjunct to radiologists work and help increase quality, would be a long time to actually replace a human being. Lots of radiology is not black or white, not just no fracture or fracture. There are clinicians that like to discuss cases with a live human being who knows medicine.
Volume of studies also keeps on increasing each year, sometimes as much as 5-10% per year in some modalities. I wouldn't mind if a computer helped me out and dictated a few cases for me. I could see compensation changing as a result of this, but lots of things in this current environment is putting pressure on compensation and workload.

How do you see compensation evolving over the next decade? Seems like there are competing forces between the increase in studies and the replacement of PP groups by large places that eventually drive down pay.