AI in medicine

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.
Any field that heavily relies on objective data (imaging, histo, labs, etc) is going to be greatly impacted by AI.

The inertia in all of us compels us to anticipate only the negative impact. However, no one really knows how this is going to shape medicine.

For now, fields that deal with a high level of subjectivity (psych is the prime example) and heavily surgical/procedural ones (ie ortho) appear to be in a safe spot.
 
  • Like
Reactions: 1 user
Members don't see this ad :)
Maybe I’m naively optimistic, but I think this will be a lot worse for midlevels than for docs. Ultimately most people are going to want a human being to be the face of their healthcare, so the corporate overlords will likely need to keep some human “providers” around. I think the most likely near-future change is lots of midlevel layoffs as AI takes over their largely algorithm-based decision making.

Once AGI becomes a thing, idk, that’s singularity territory, and nothing is predictable after that.
 
Maybe I’m naively optimistic, but I think this will be a lot worse for midlevels than for docs. Ultimately most people are going to want a human being to be the face of their healthcare, so the corporate overlords will likely need to keep some human “providers” around. I think the most likely near-future change is lots of midlevel layoffs as AI takes over their largely algorithm-based decision making.

Once AGI becomes a thing, idk, that’s singularity territory, and nothing is predictable after that.


That's actually a pretty good point. We'll end up supervising computers rather than midlevels.

To the point about radiology...man....I feel like that would be the first domino to fall. It's ripe for algorithms to flourish in. No clinical history or context needed....just match that appendicitis r/o CT against 3 million positive and negative tests and it'll tell you yes or no.
 
Maybe I’m naively optimistic, but I think this will be a lot worse for midlevels than for docs. Ultimately most people are going to want a human being to be the face of their healthcare, so the corporate overlords will likely need to keep some human “providers” around. I think the most likely near-future change is lots of midlevel layoffs as AI takes over their largely algorithm-based decision making.

Once AGI becomes a thing, idk, that’s singularity territory, and nothing is predictable after that.

What's AGI?
 
What's AGI?
Artificial general intelligence. AI that is good at basically everything (unlike ChatGPT, which is just a language AI) and smarter than any human. Essentially gg to human effort being necessary. Lots of economic and philosophical concerns about if and when it comes about.
 
That's actually a pretty good point. We'll end up supervising computers rather than midlevels.

To the point about radiology...man....I feel like that would be the first domino to fall. It's ripe for algorithms to flourish in. No clinical history or context needed....just match that appendicitis r/o CT against 3 million positive and negative tests and it'll tell you yes or no.

They are already trying to do this. I think this an area of Rads that might have them concerned. They lurk here, maybe they can chime in.

As long as we don't train AI to say "could be atelectasis, consolidation, or random white lines on an xray, clinically correlate"...
 
Last edited:
  • Like
Reactions: 2 users
You sure about that?

Once I got an xray of an ankle and it said CT recommended

Got a CT. beginning of shift and I'm okay bedlocking. it says get MRI. At that point I consult ortho and get MRI

MRI says clinical correlation

I hope the robots can do better, but realistically I shouldn't have even gotten the xray
 
  • Like
  • Haha
Reactions: 5 users
Once I got an xray of an ankle and it said CT recommended

Got a CT. beginning of shift and I'm okay bedlocking. it says get MRI. At that point I consult ortho and get MRI

MRI says clinical correlation

I hope the robots can do better, but realistically I shouldn't have even gotten the xray
One (or many) bad apple(s) can spoil the bunch 🙁
 
They are already trying to do this. I think this an area of Rads that might have them concerned. They lurk here, maybe they can chime in.

As long as we don't train AI to say "could be atelectasis, consolidation, or random white lines on an xray, clinically correlate"...


Opinions vary widely about AI and rads. Some think its an imminent threat while others see it first being a supplementary tool. Others feel it will replace rads completely when AI is also replacing pilots, truck drivers, pharmacists, and tons of other jobs in other fields that don't have the same legal accountability.

My current group uses "primitive" AI for certain studies. It's great for flagging and bringing positive head bleeds, free air, and PE to one's attention but there are certainly false positives (eg. dural calcifications). Clearly can't dictate, compare to prior studies/other modalities, pick up incidentals, offer a differential, or notify clinicians.

Personally I am mid-career and my threshold for concern would be when AI is held to the same or greater standard to human rads from a medical legal perspective. I suppose my attitude is that anything is possible and really hard to predict.

Finally some in the field would welcome any substantial help from AI as we are drowning giving the shortage of rads and continued growth in imaging volume/complexity.
 
  • Like
Reactions: 1 user
Members don't see this ad :)
That's actually a pretty good point. We'll end up supervising computers rather than midlevels.

Oh, you mean I won't have to refuse to sign an order for bicarb pushes in a DKA with a pH of 7.27 because "the serum bicarb was low" or deal with an NP spam consulting surgery and GI for undifferentiated acute pancreatitis?

I mean... sure... I, for one, welcome our new AI overlords.
 
Last edited:
  • Like
Reactions: 1 user
Once I got an xray of an ankle and it said CT recommended

Got a CT. beginning of shift and I'm okay bedlocking. it says get MRI. At that point I consult ortho and get MRI

MRI says clinical correlation

I hope the robots can do better, but realistically I shouldn't have even gotten the xray
I don’t know man that sounds like a lot of billing generated. I’m sure our administrative hospital overlords would “hate” that result, especially if AI-based radiology created plenty of outputs like that.
 
Oh, you mean I won't have to refuse to sign an order for bicarb pushes in a DKA with a pH of 2.27 because "the serum bicarb was low" or deal with an NP spam consulting surgery and GI for undifferentiated acute pancreatitis?

I mean... sure... I, for one, welcome our new AI overlords.

Wait... what?
 
  • Like
Reactions: 1 users
I think the point at which AI takes over medicine is the point at which it takes over most jobs in the world and we can all go lay on the beach sipping martinis and there is no more need for human workers. Not worried. Although tbh the beaches would become full up.
 
When the AI Doctor starts saying things like, "Why am I working all the nights/weekends/holidays? It's because I'm a robot, right?," it has satisfied my medical Turing Test and is sapient :).
 
  • Like
Reactions: 4 users
I think people look at AI in medicine from a too "all or nothing" perspective. Physicians won't be replaced but the practice of medicine will likely dramatically change over the next decade or so in the same way it did when the internet became readily available. Think how different medicine was with paper charts and having to go to the library for a medical reference compared to the era of UpToDate, PubMed, and EMRs.
 
  • Like
Reactions: 1 users
I think people look at AI in medicine from a too "all or nothing" perspective. Physicians won't be replaced but the practice of medicine will likely dramatically change over the next decade or so in the same way it did when the internet became readily available. Think how different medicine was with paper charts and having to go to the library for a medical reference compared to the era of UpToDate, PubMed, and EMRs.
I’m a bit skeptical but mostly because the major reason for EMR adoption was not because it was more efficient or cost effective but instead the government said “you will adopt an EMR or we will cut your reimbursement by X% and that number will increase each year until you do” (IIRC it was part of PPACA)

I know of at least one local Peds group that held out and finally gave in within the past 5.

The problem with new technology is that people want to get paid a metric crapton for it. What good is eliminating a few MD/DO positions if you’re now having to support a bunch of tech developers to pay for it? It’s quite possible the numbers win out in favor of the AI tools but I don’t think it’s all that clear cut.
 
  • Like
Reactions: 1 user
Top