It's official: AI is more accurate than radiologists when it comes to reading mammograms

[Rads125125]

AI matches radiologist performance on mammo, DBT exams

A meta-analysis encompassing over 1.1 million screening breast imaging exams has concluded that artificial intelligence (AI) software performs equally to or better than radiologists, according to research published May 23 in Radiology.

www.auntminnie.com

Given how AI has reached this point in under a year after the release of ChatGPT, is it really realistic to say that radiology will still be a viable field in just 5 years? Maybe this wouldn't affect an attending breast radiologist in their 50s, but I would STRONGLY advise against premeds or even current med students applying to DR as a whole. Mammo is only the first domino to fall, the entire field is on extremely thin ice right now besides IR.

---

Members don't see this ad.

 

[NDcienporciento100]

That’s funny you say this. I just happened to be reading a mammogram. The AI says there is a mass in the left breast. Oh well I should not even bother to look at it. AI is more accurate than me anyway. I’ll just set her up for a biopsy. Ahh heck I’ll look for learning purposes! The AI is putting the mass star on top of a big ball of silicon which just so happens to be a breast implant! Who ever wrote this is clueless to the field of radiology. The odds of AI replacing radiologists in 5 years are about the same as humans colonizing Pluto.

 

[vm26]

AI matches radiologist performance on mammo, DBT exams

A meta-analysis encompassing over 1.1 million screening breast imaging exams has concluded that artificial intelligence (AI) software performs equally to or better than radiologists, according to research published May 23 in Radiology.

www.auntminnie.com

Given how AI has reached this point in under a year after the release of ChatGPT, is it really realistic to say that radiology will still be a viable field in just 5 years? Maybe this wouldn't affect an attending breast radiologist in their 50s, but I would STRONGLY advise against premeds or even current med students applying to DR as a whole. Mammo is only the first domino to fall, the entire field is on extremely thin ice right now besides IR.

Please tell me it can dictate studies for poly-trauma, strokes, dissection, ovarian torsion etc and negate the need for overnight rad coverage. Also needs to be able to compare to multiple prior studies (including other modalities), pick-up incidentals, notify clinicians of critical findings and discuss them in depth if needed, cross reference with EMR, and finally be held to the same legal standard to that of human rads. Forget about procedures, diagnostic breast imaging, and flouro.

With that said, who really knows. Tons of jobs in many fields that would be easier to replace with AI with way less liability. One of the main issues with imaging exams is they are not binary (eg. CTA is either positive/neg for P/E so no need to worry about that lung mass and liver mets).

 

[Anakinmemer]

Please tell me it can dictate studies for poly-trauma, strokes, dissection, ovarian torsion etc and negate the need for overnight rad coverage. Also needs to be able to compare to multiple prior studies (including other modalities), pick-up incidentals, notify clinicians of critical findings and discuss them in depth if needed, cross reference with EMR, and finally be held to the same legal standard to that of human rads. Forget about procedures, diagnostic breast imaging, and flouro.

With that said, who really knows. Tons of jobs in many fields that would be easier to replace with AI with way less liability. One of the main issues with imaging exams is they are not binary (eg. CTA is either positive/neg for P/E so no need to worry about that lung mass and liver mets).

These are all great points for current radiologists but aspiring medical students or premeds? Idk. I'm in medical school and feel uneasy saying confidently that none of what you mentioned will happen in the next FORTY years I practice.

Some things will never be replaced, but we don't need the current amount of radiologists to just do procedures. I'm scared the job market might go the way of rad onc where the top dogs managing the AI still make bank but there aren't a lot of jobs to go around. (in a 40 year timeline, not imminently)

Still applying radiology though because I'll take any excuse to retire early...

 

[NDcienporciento100]

These are all great points for current radiologists but aspiring medical students or premeds? Idk. I'm in medical school and feel uneasy saying confidently that none of what you mentioned will happen in the next FORTY years I practice.

Some things will never be replaced, but we don't need the current amount of radiologists to just do procedures. I'm scared the job market might go the way of rad onc where the top dogs managing the AI still make bank but there aren't a lot of jobs to go around. (in a 40 year timeline, not imminently)

Still applying radiology though because I'll take any excuse to retire early...

Fields being replaced actively as we speak. Family medicine, right now as I type some manager is haveing a conversation with someone saying “why do we need to pay a physician 180-250 to be our family practice doctor in rural Kansas when we can pay an NP half that”. Anesthesia…. Talk about a field that has ****** themselves. There are hospitals and not a small number that don’t even have an Anesthesiologist at the hospital. Same scenario, why pay double when I can get the “same thing” for half.
Right now, a board eligible general radiologist can make 3000k/day minimum to locums for an 8 hour day. Not even realistically speaking, like easy! Interventional radiologist can make a minimum of 4k/day doing locums and the same thing with Neuro, Breast etc. AI is really good with ischemic stroke, the vessel either stops or it doesn’t. AI does not do well with borderline situations. People have been writing paper’s saying that AI will replace radiologists for 20 years and has failed to really even make a big impact as of yet in my opinion.

 

[bronx43]

That’s funny you say this. I just happened to be reading a mammogram. The AI says there is a mass in the left breast. Oh well I should not even bother to look at it. AI is more accurate than me anyway. I’ll just set her up for a biopsy. Ahh heck I’ll look for learning purposes! The AI is putting the mass star on top of a big ball of silicon which just so happens to be a breast implant! Who ever wrote this is clueless to the field of radiology. The odds of AI replacing radiologists in 5 years are about the same as humans colonizing Pluto.

Not sure how you can be so sure about this. The medical system has already shown that it doesn’t really care about outcomes - only that the medicolegal costs don’t overtake the cost savings. If the system finds that it can get away with AI interpretation of certain “low risk” cases, and just be able to swallow the tail risk of lawsuits then it will implement AI. And it won’t likely be broad replacement of rads. Certain imaging studies will go first then others will follow. Admin will love it. Clinicians won’t really care. Public won’t care or won’t even know.

 

[VA Hopeful Dr]

Fields being replaced actively as we speak. Family medicine, right now as I type some manager is haveing a conversation with someone saying “why do we need to pay a physician 180-250 to be our family practice doctor in rural Kansas when we can pay an NP half that”. Anesthesia…. Talk about a field that has ****** themselves. There are hospitals and not a small number that don’t even have an Anesthesiologist at the hospital. Same scenario, why pay double when I can get the “same thing” for half.
Right now, a board eligible general radiologist can make 3000k/day minimum to locums for an 8 hour day. Not even realistically speaking, like easy! Interventional radiologist can make a minimum of 4k/day doing locums and the same thing with Neuro, Breast etc. AI is really good with ischemic stroke, the vessel either stops or it doesn’t. AI does not do well with borderline situations. People have been writing paper’s saying that AI will replace radiologists for 20 years and has failed to really even make a big impact as of yet in my opinion.

You do realize that FM has the best job market and one of (if not the) highest demand in all of medicine, right? And we haven't really seen the big boomer retirement wave yet (around 40% of practicing FPs are boomers).

With the increasing importance of quality-based care and focus on cost-effective care, midlevels aren't the hot commodity for primary care that they were.

As a fun side note, averaged across the year as an FP I bring in just shy of $2000/day and I'm middle of the pack for the doctors in my office.

 

[NDcienporciento100]

You do realize that FM has the best job market and one of (if not the) highest demand in all of medicine, right? And we haven't really seen the big boomer retirement wave yet (around 40% of practicing FPs are boomers).

With the increasing importance of quality-based care and focus on cost-effective care, midlevels aren't the hot commodity for primary care that they were.

As a fun side note, averaged across the year as an FP I bring in just shy of $2000/day and I'm middle of the pack for the doctors in my office.

I didn’t mean to offend you. I think what FM doctors do is amazing work.

 

[NDcienporciento100]

Not sure how you can be so sure about this. The medical system has already shown that it doesn’t really care about outcomes - only that the medicolegal costs don’t overtake the cost savings. If the system finds that it can get away with AI interpretation of certain “low risk” cases, and just be able to swallow the tail risk of lawsuits then it will implement AI. And it won’t likely be broad replacement of rads. Certain imaging studies will go first then others will follow. Admin will love it. Clinicians won’t really care. Public won’t care or won’t even know.

Alright everyone, you heard it from him/her first. Set your clock! In 5 years put down your dictaphone down because ChatGPT is going to be doing your job.

 

[bronx43]

Alright everyone, you heard it from him/her first. Set your clock! In 5 years put down your dictaphone down because ChatGPT is going to be doing your job.

Show me where I said that. I’m just pointing out that your certainty regarding the issue is unfounded.

 

[OnlyScans420]

AI matches radiologist performance on mammo, DBT exams

A meta-analysis encompassing over 1.1 million screening breast imaging exams has concluded that artificial intelligence (AI) software performs equally to or better than radiologists, according to research published May 23 in Radiology.

www.auntminnie.com

Given how AI has reached this point in under a year after the release of ChatGPT, is it really realistic to say that radiology will still be a viable field in just 5 years? Maybe this wouldn't affect an attending breast radiologist in their 50s, but I would STRONGLY advise against premeds or even current med students applying to DR as a whole. Mammo is only the first domino to fall, the entire field is on extremely thin ice right now besides IR.

I swear, all these posts are just med students hyper focused on compensation and job security that are trying to convince themselves to chose a different specialty.

Just do a different specialty and join everyone else who regrets not doing rads(and can't transfer anymore because it's only getting more competitive). No need to go through the song and dance of posting ****ty articles on an internet forum as cope.

 

[NDcienporciento100]

Show me where I said that. I’m just pointing out that your certainty regarding the issue is unfounded.

I mean “unfounded”? I don’t know about that. Literally millions of people today call a FNP or NP their primary care provider. It wasn’t like that 20 years ago. You have to know that. “unfounded”? I disagree. Salaries I’ll defer to you since you are practicing physician in in the field.

 

[VA Hopeful Dr]

I didn’t mean to offend you. I think what FM doctors do is amazing work.

Not offended at all, just clarifying things.

This isn't meant to be critical of you in any way. I think medicine, especially for those of us in primary care, is more fragmented than it used to be. For example, I haven't set foot in a hospital professionally since residency. Don't misread, I love that hospitalists exist so I don't have to. But, this means I don't regularly talk to (and never in person) doctors in other specialties. This leads to many of us having no idea what's going on in the rest of medicine. If it weren't for SDN (and the fact that I regularly go into other specialties' forums here), I would have no clue about changes in other areas since residency which is 10 years ago now.

So I don't expect anyone outside of primary care to know what's going on in primary care.

 

[RadOncDoc21]

AI is real just like the internet was real. We all need to learn how to adapt and use it to enhance our efficiency and knowledge. We can either embrace it or ignore it but it doesn’t change the fact that it’s real… just go check out the 200 billion bounce in Nividia’s market cap this week.

 

[colbgw02]

These are all great points for current radiologists but aspiring medical students or premeds? Idk. I'm in medical school and feel uneasy saying confidently that none of what you mentioned will happen in the next FORTY years I practice.

Some things will never be replaced, but we don't need the current amount of radiologists to just do procedures. I'm scared the job market might go the way of rad onc where the top dogs managing the AI still make bank but there aren't a lot of jobs to go around. (in a 40 year timeline, not imminently)

Still applying radiology though because I'll take any excuse to retire early...

It's smart of you to think of a timeline beyond what you see everyday while in training. It seems like so many of us only focus on the next step or two, which--in this context--means overly weighting the lives and careers of the residents and academic attendings when picking a specialty.

I wasn't around forty years ago, but many of my residency attendings trained in the 1980s and early 1990s, so I got glimpses into what radiology was like back then. MRI was still a novel technology and not super available. CT scanners were single-slice and shut down for the day circa 4 pm. Residents would have to stay late to hang the next day's films. PACS and voice-recognition software was virtually unheard of, and turnaround times were measured in days rather than minutes.

It's hard for me to imagine any medical student in 1983 could have envisioned what radiology would look like in 2023 with any degree of accuracy. I don't know how any of us could even dare to guess what it will look like in 2063.

I'm skeptical about the claims that human radiologists won't exist in x, y, or z number of years. I also don't know enough about it to be sure that AI won't have a significant impact on our field, but that's not the point. The point is that no one should choose radiology, or any specialty really, thinking that it's static. Accept the fact that you'll be practicing differently in 5 years than you are today, never mind 40. Embrace the suc...I mean, change.

 

[doctalaughs]

I don’t see AI replacing radiology or any field in medicine. If AI can’t do my landscaping or replace my cleaning lady or the burger flippers yet, no way it can replace doctors.

The being said I could see AI assisting many doctors with concomitant expectations that we “do more volume”

If that happens radiologists may be expected to read twice as many studies with AI doing the prelim reads for 80% of situations. Whether that increases or decreases job prospects and pay has more to do with outside factors (lobbying, laws, training supply, private equity trends etc than the AI itself).

 

[bronx43]

I mean “unfounded”? I don’t know about that. Literally millions of people today call a FNP or NP their primary care provider. It wasn’t like that 20 years ago. You have to know that. “unfounded”? I disagree. Salaries I’ll defer to you since you are practicing physician in in the field.

It hasn’t really reduced salaries but it probably has had some effect on job availability in desirable locations.
I’m not totally sure what the Np argument is here. If anything I would argue the presence of midlevels is argument for potential adoption of AI, since it was a unilateral decision by the “system” onto its unwilling participants. It didn’t matter to the admin that midlevels are much worse in terms of diagnosis and treatment, but they were cheap enough and “good enough” for their profit seeking motives.

 

[NDcienporciento100]

It hasn’t really reduced salaries but it probably has had some effect on job availability in desirable locations.
I’m not totally sure what the Np argument is here. If anything I would argue the presence of midlevels is argument for potential adoption of AI, since it was a unilateral decision by the “system” onto its unwilling participants. It didn’t matter to the admin that midlevels are much worse in terms of diagnosis and treatment, but they were cheap enough and “good enough” for their profit seeking motives.

The argument with regards to mid levels is simple. Their presence will continue to grow and the fields that will be affected the most will be primary care.

When I was a medical student I remember and internist working in a large family practice clinic having a conversation with an administrator in front of me and she said to the administrator I’m the only doctor taking new patients at this clinic and the administrator “no” so and so is taking new patients. So and so was an NP. and that’s how administration views primary care physicians in large clinics, the same as NPs and PAs. I don’t agree with it but it is a fact. I’m glad you’re still doing well for yourselves.

 

[SeisK]

Still no clinical trial, still no heterogenous datasets. Still no comparison to an AI+ radiologist arm. Still no address of real AI weaknesses that are artificially smoothed out with retrospective meta-analyses.

Still no acknowledgement of the pathologies to which AI is weak to. Still no generalizability of this result to thousands of specific pathologies. Still no integration into the patients’ specific clinical contexts. Still no succinct and concise narrative reports.

I’ve never said certain tasks won’t be replaced. What I’m saying is the task replacement will be so slow and piecemeal no radiologist will feel the difference anyway.

This is all med students trying to undermine their competition’s spirit to make their match easier, or people who didn’t go into rads doing the best to overcome their cognitive dissonance with choosing a different, worse field.

The only thing I know for certain is that 5 years is a goddamn HILARIOUS timeline.

 

[NDcienporciento100]

Demand for radiologists keeps rising! More studies being done now then ever before. As I write there is a recruiter out there begging some 5th year resident to take a 1 week on two week off 500k/year job doing nights.

 

[SeisK]

If the system finds that it can get away with AI interpretation of certain “low risk” cases, and just be able to swallow the tail risk of lawsuits then it will implement AI.

This is far and away not the only financial consideration given to AI implementation, and not even the most important one.

 

[bronx43]

This is far and away not the only financial consideration given to AI implementation, and not even the most important one.

What is the most important financial consideration?

 

[Anakinmemer]

It's smart of you to think of a timeline beyond what you see everyday while in training. It seems like so many of us only focus on the next step or two, which--in this context--means overly weighting the lives and careers of the residents and academic attendings when picking a specialty.

I wasn't around forty years ago, but many of my residency attendings trained in the 1980s and early 1990s, so I got glimpses into what radiology was like back then. MRI was still a novel technology and not super available. CT scanners were single-slice and shut down for the day circa 4 pm. Residents would have to stay late to hang the next day's films. PACS and voice-recognition software was virtually unheard of, and turnaround times were measured in days rather than minutes.

It's hard for me to imagine any medical student in 1983 could have envisioned what radiology would look like in 2023 with any degree of accuracy. I don't know how any of us could even dare to guess what it will look like in 2063.

I'm skeptical about the claims that human radiologists won't exist in x, y, or z number of years. I also don't know enough about it to be sure that AI won't have a significant impact on our field, but that's not the point. The point is that no one should choose radiology, or any specialty really, thinking that it's static. Accept the fact that you'll be practicing differently in 5 years than you are today, never mind 40. Embrace the suc...I mean, change.

This was really reassuring and I'm definitely still excited to be applying radiology and being on the forefront of technological evolution.

Kind of hoping there is a dip in competition this year due to AI fear because the total applications have been rising more than any other specialty over the past 3 years...

 

[SeisK]

What is the most important financial consideration?

Medicare reimburses physician services highly, and are already dragging their feet on approving reimbursements for things that are demonstrably life saving in clinical trials. Once an AI replaces a clinician’s task, medicare will approve reimbursement for it, but is guaranteed to ratchet it to the floor. No hospital is going to rake in cash from AI for long.

 

[bronx43]

Medicare reimburses physician services highly, and are already dragging their feet on approving reimbursements for things that are demonstrably life saving in clinical trials. Once an AI replaces a clinician’s task, medicare will approve reimbursement for it, but is guaranteed to ratchet it to the floor. No hospital is going to rake in cash from AI for long.

Sure, if AI is touted as a complete and immediate replacement for a physician. But, it may not go down in such a way. It may be a situation where AI is initially utilized to make physicians "more efficient" but a physician still has to sign off on a particular read or clinical decision. Medicare will likely keep reimbursement status quo, since they can't just take an axe to reimbursement simply for efficiency gains. As this "efficiency" increases, then the job market will become ever tighter and it will allow for greater and greater arbitrage potential for hospitals. The end game may be a small team of clinicians/radiologists with software support signing off on huge caseloads. These physicians will have to endure the unsavory byproducts of such a system, because the alternative is unemployment.

 

[SeisK]

Sure, if AI is touted as a complete and immediate replacement for a physician. But, it may not go down in such a way. It may be a situation where AI is initially utilized to make physicians "more efficient" but a physician still has to sign off on a particular read or clinical decision. Medicare will likely keep reimbursement status quo, since they can't just take an axe to reimbursement simply for efficiency gains. As this "efficiency" increases, then the job market will become ever tighter and it will allow for greater and greater arbitrage potential for hospitals. The end game may be a small team of clinicians/radiologists with software support signing off on huge caseloads. These physicians will have to endure the unsavory byproducts of such a system, because the alternative is unemployment.

Spoken like someone who isn’t a radiologist, and doesn’t really know what they’re talking about.

efficiency gains aren’t really real outside of cherry picked retrospective studies. AI slows me the f down. Pretty much everyone else says the same thing.

I really am hoping for the first radiologist who leans on an AI read for a PE, anticoagulation started leading to cerebral hemorrhage, and that guy gets sued for the false positive read. Id love to be the one to testify against them as an expert witness.

 

[NDcienporciento100]

Look, here is the reality. You can’t even get two radiologists to agree and certainly not everyone will agree with a machine in the next 5 years.
Go on google translate right now and look at the inaccuracy. It’s really bad! Like 75% accurate in the best of scenarios. And there is literally millions of people using it everyday and probably >1000 people working on that app and it’s simple ******* English to Spanish translation! How is it that inaccurate?! Bringing it back to the issue. How the **** is a machine going to give me a concise read on a disaster abdomen read when 1000+ of the great minds at google can’t even figure how to translate English to Spanish for a simple breakfast order at a McDonald’s in Spain?! Do you see my point? It’s crazy to think these jumps are going to be made overnight.

 

[NDcienporciento100]

I’m serious! someone who is friends with someone who has said AI is taking over radiology. Ask them why google can’t get a simple English to Spanish translation correct. We all know that person. Ask them. And then ask how a machine is going to read my disaster abdomen study with 20 priors.

 

[Ho0v-man]

This has been floated on Reddit by a couple of people too. Some even claiming to be residents or know residents trying to flee radiology because the end is nigh.

Turns out they were med students who hadn’t applied yet trying to scare people away from rads lol. Gee, I wonder who this person is?

 

[Cognovi]

Let's return to the point made by the OP, which is that "AI has reached this point in under a year after the release of ChatGPT".

The linked article is a meta-analysis of studies that predate ChatGPT.

ChatGPT has nothing to do with the study.

ChatGPT is a chatbot that produces text from text input.

 

[CallMeBigJeff]

I said this on a forum 3 years ago when I was an uncertain M3 and I'll say it now. If you're a med student and AI makes you uneasy, realize that you're making a bet on a nebulous fear. I would rather retrain in the doubtful event I get replaced than NOT go into my preferred specialty and be ABSOLUTELY MISERABLE if radiologists are still practicing 40 years down the line, and I let anonymous people on the internet and an article on business.com scare me out of a fulfulling career. Rising into R2 year right now and love what I do day in and day out. Also no way an AI is performing/interpreting the VCUG I just did or interpreting the poorly done retroperitoneal US on a squirming 1 year old.

 

[nycitygas]

Fields being replaced actively as we speak. Family medicine, right now as I type some manager is haveing a conversation with someone saying “why do we need to pay a physician 180-250 to be our family practice doctor in rural Kansas when we can pay an NP half that”. Anesthesia…. Talk about a field that has ****** themselves. There are hospitals and not a small number that don’t even have an Anesthesiologist at the hospital. Same scenario, why pay double when I can get the “same thing” for half.
Right now, a board eligible general radiologist can make 3000k/day minimum to locums for an 8 hour day. Not even realistically speaking, like easy! Interventional radiologist can make a minimum of 4k/day doing locums and the same thing with Neuro, Breast etc. AI is really good with ischemic stroke, the vessel either stops or it doesn’t. AI does not do well with borderline situations. People have been writing paper’s saying that AI will replace radiologists for 20 years and has failed to really even make a big impact as of yet in my opinion.

FYI The Anesthesia job market is the hottest it has been in decades

 

[qxrt]

It's pretty telling that speculation over AI taking over radiology is always propagated by non-radiologists who somehow know the actual practice and future course of radiology better than radiologists. I have yet to personally meet a radiologist who actually thinks AI is going to be taking over substantial portions of radiology practice in the foreseeable future.

I mean, you don't often see most other specialists being told by others not in their field what the future of their specialty is. And yet, radiologists are constantly told by non-radiologists including software engineers, businessmen, non-radiologist physicians, med students, etc. what the future of radiology holds. No one actually seems to believe the radiologist what the near future of their own specialty might realistically be. Isn't that weird?

Not all of us radiologists are blinded by self-interest and prevented from seeing what AI can and can't do. It's just not realistic anytime in the near future. And yet, here we are constantly facing the non-stop flow of non-radiologists telling us what will happen to our specialty and that we're wrong.

It doesn't really matter what people believe. Either AI can or can't read radiology exams, and simply believing that it soon will be able to doesn't get the scans read or the work done.

 

[NDcienporciento100]

It's pretty telling that speculation over AI taking over radiology is always propagated by non-radiologists who somehow know the actual practice and future course of radiology better than radiologists. I have yet to personally meet a radiologist who actually thinks AI is going to be taking over substantial portions of radiology practice in the foreseeable future.

I mean, you don't often see most other specialists being told by others not in their field what the future of their specialty is. And yet, radiologists are constantly told by non-radiologists including software engineers, businessmen, non-radiologist physicians, med students, etc. what the future of radiology holds. No one actually seems to believe the radiologist what the near future of their own specialty might realistically be. Isn't that weird?

Not all of us radiologists are blinded by self-interest and prevented from seeing what AI can and can't do. It's just not realistic anytime in the near future. And yet, here we are constantly facing the non-stop flow of non-radiologists telling us what will happen to our specialty and that we're wrong.

It doesn't really matter what people believe. Either AI can or can't read radiology exams, and simply believing that it soon will be able to doesn't get the scans read or the work done.

I think because what we do is mostly computer based people just naturally think that it will be easy for a computer to do. What they don’t realize is most of what other (non procedural) fields do is computer based. How often have you seen a full work up done on a patient by a provider who actually never saw the patient? How did they implant that work up? They looked at the computer saw the vitals, labs chief complaint etc. The bigger risk to physicians is mid level providers. I guarantee anyone reading this that is a trainee has to compete with mid levels at some level. That may have been for minor procedures or a multitude of other things. Where I train and many other places nurse anesthetist tap out the anesthesia residents. Early on I had PAs pissed at me for taking there little minor procedures. I believe that Will negatively affect specialty’s far before AI affects radiology.

 

[Hypothalamus]

It's been over 10 years since my chairman told me AI would do my job within 5 years. Since then, it has had 0 impact. He scared many students out of radiology, who would otherwise have been entering the hottest radiology job market in a couple decades.

Med students for the most part don't know what radiology is; even most physicians don't. Surgeons and radiation oncologists know what we do. Oncologists are familiar as well. Our value is in problem solving and judgement on difficult cases, where teams are deciding on whether or not to change chemotherapy, or operate. I don't think AI can do that for many years, and I don't think an AI company will ever assume liability for it. In fact, I think it'll be many years before AI can read an image.

On the other hand, AI can always be sold to policymakers and hospital administrators, who themselves know little about imaging and patient care, as a suitable replacement for radiologists. Who knows?

It'll be many, many years before our jobs are computerized, if ever. But it is amusing to see AI back as a bogeyman over a decade later when my chairman was scaring everyone about it.

I'd worry more about the next down cycle in the radiology job market.

 

[NDcienporciento100]

It's been over 10 years since my chairman told me AI would do my job within 5 years. Since then, it has had 0 impact. He scared many students out of radiology, who would otherwise have been entering the hottest radiology job market in a couple decades.

Med students for the most part don't know what radiology is; even most physicians don't. Surgeons and radiation oncologists know what we do. Oncologists are familiar as well. Our value is in problem solving and judgement on difficult cases, where teams are deciding on whether or not to change chemotherapy, or operate. I don't think AI can do that for many years, and I don't think an AI company will ever assume liability for it. In fact, I think it'll be many years before AI can read an image.

On the other hand, AI can always be sold to policymakers and hospital administrators, who themselves know little about imaging and patient care, as a suitable replacement for radiologists. Who knows?

It'll be many, many years before our jobs are computerized, if ever. But it is amusing to see AI back as a bogeyman over a decade later when my chairman was scaring everyone about it.

I'd worry more about the next down cycle in the radiology job market.

One of my attending’s who trained going almost 20 years ago said they were saying the same thing then.

Look at breast AI. Anyone reading that has read breast knows that the breast AI system is complete garbage. Most attending’s I know forget to look at it even though to bill for it you technically have to. But make no mistake about it, its useless garbage. It was pushed and sold but the people who developed it who managed to get the pink ribbon wearing breast cancer warriors to believe its the greatest thing ever and it was going to save lives who then brow beat politicians to put in a system so that it could be billed for.

 

[Sawback3]

It's pretty telling that speculation over AI taking over radiology is always propagated by non-radiologists who somehow know the actual practice and future course of radiology better than radiologists.

The same thing has happened in pathology over the last few years, with many pathologists believing the hype and getting freaked out. Most pathology departments aren’t even digital yet, which would be a prerequisite. Is the AI gonna go down to the gross room, ink the margins, and cut up the organ for me?

 

[Cognovi]

The same thing has happened in pathology over the last few years, with many pathologists believing the hype and getting freaked out. Most pathology departments aren’t even digital yet, which would be a prerequisite. Is the AI gonna go down to the gross room, ink the margins, and cut up the organ for me?

A PA can do that, no worries.

 

[VivaElBirdos]

It's been over 10 years since my chairman told me AI would do my job within 5 years. Since then, it has had 0 impact. He scared many students out of radiology, who would otherwise have been entering the hottest radiology job market in a couple decades.

Med students for the most part don't know what radiology is; even most physicians don't. Surgeons and radiation oncologists know what we do. Oncologists are familiar as well. Our value is in problem solving and judgement on difficult cases, where teams are deciding on whether or not to change chemotherapy, or operate. I don't think AI can do that for many years, and I don't think an AI company will ever assume liability for it. In fact, I think it'll be many years before AI can read an image.

On the other hand, AI can always be sold to policymakers and hospital administrators, who themselves know little about imaging and patient care, as a suitable replacement for radiologists. Who knows?

It'll be many, many years before our jobs are computerized, if ever. But it is amusing to see AI back as a bogeyman over a decade later when my chairman was scaring everyone about it.

I'd worry more about the next down cycle in the radiology job market.

Which radiology subspecialty (besides nucs/breast) is consulted most by oncologists? Body?

 

[RadOncDoc21]

Which radiology subspecialty (besides nucs/breast) is consulted most by oncologists? Body?

Cancer can occur in any cell, so if there are cells, there can be cancer.

 

[NDcienporciento100]

Which radiology subspecialty (besides nucs/breast) is consulted most by oncologists? Body?

Honestly probably IR. In fact definitely IR. Other than the brain if it’s a mass that need’s biopsy we can pretty much do it. Big pelvic mass needs diversion call IR. Panc head mass with obstruction and ERCp failed call IR. Patient needs mediport for chemo, call IR unless your a surgeon that is short on rvus then you will go fumble your way through a mediport. Unresectable liver mass call IR for Y90/tace/ablation. Renal cell carcinoma call ir for ablation. Lung mass in poor surgical candidate that failed radiation call IR for ablation. Painful spine Mets call ir fkr osteocool/Kyphoplasty. The list goes. You were probably talk about diagnostic only. Probably body. MSK the least ( if you work at mallinckrodt don’t get defensive talking about how you have seen 10 telangiectatic osteosarcomas this week. Some of us work in the real world. No offense Dr Jennings, keep treating this unicameral bone cyst. Your doing good work).

 

[Cognovi]

Which radiology subspecialty (besides nucs/breast) is consulted most by oncologists? Body?

Hard to say. Cancer imaging and participating in tumor board is a central part of every radiology subspecialty.