Future of DR - Impact of Technology on Job Market

[Axcella]

I'm a 3rd year with a background in machine learning. I had my heart on doing DR at an academic institution and applying my ML background to DR. Last week the chair of the surgery department at my institution strongly encouraged me not to seek a DR residency. He claimed that DR would be a dead field (due to AI/outsourcing) within ten years. He felt strongly about this but could not provide me with any information about why he thought this was the time scale. I have no doubt that new tech will change DR more drastically than others, but where is this 10-year number expiration date coming from? Why not 20 or 30? Where is a good place to find real information about the direction of the field?

My sense has been that practicing radiologists think computers will never be able to do most of what they do and those outside the field think that we should stop training radiologists now. I have not yet found convincing evidence for either of these positions.

---

Members don't see this ad.

 

[Gadofosveset]

My surgery preceptor told me the same thing... 10 years. That was literally 11 years ago. Not sure why doctors make pronouncements about specialties that they don’t really know much about.

 

[Axcella]

I had no idea this sort of thing had been going on for so long. Seems like there is big time uncertainty on this point.

 

[Neopolymath]

It's a silly circular discussion. Your stereotypical obnoxious and critically outspoken surgeon, ER doc or noctor can't find clear everyday value in a human radiologist because they feel that the reports don't tell them anything they don't know from examining the patient most of the time. That's because they keep ordering ****ing imaging for things that they should be diagnosing clinically. The real value of radiologists would be seen more clearly if imaging was mostly ordered when it was actually needed and appropriate. In fact I would venture to say that no one would even be talking about outsourcing, AI, machine learning, etc if there was not an overabundance of purely stupid exams for radiologists to read each day. The truth is that all the critical findings justify each of the seemingly normal/low value reports because that's how medicine in practiced in America.

Basically, don't listen to random people talk about radiology that don't understand what an expert consultation in radiology provides to the system. It's a waste of energy. The same people say dumb things about anesthesiologists as well.

 

[sloh]

If anything, the improvement in technology and imaging has created MORE work for radiologists. The only people genuinely concerned about AI taking over radiology are non-radiologists.

 

[NDcienporciento100]

I had no idea this sort of thing had been going on for so long. Seems like there is big time uncertainty on this point.

That’s because surgeons don’t have a clue! Just like I don’t know the best way to go about doing a partial colectomy. The difference is I don’t pretend to be and expert in all things medicine were is they do.

 

[yungspleen]

My surgery preceptor told me the same thing... 10 years. That was literally 11 years ago. Not sure why doctors make pronouncements about specialties that they don’t really know much about.

Because they read a vox article about machine learning one time and want to sound smart/interesting

 

[Cognovi]

My surgery preceptor told me the same thing... 10 years. That was literally 11 years ago. Not sure why doctors make pronouncements about specialties that they don’t really know much about.

Your surgery preceptor must have been ahead of the curve if they mentioned AI 11 years ago. AlexNet didn't win ImageNet until 2012, which lead to the recent boom in interest in convolutional neural networks.

Certainly at that time, people were talking about how teleradiology technology would lead to off-shore outsourcing and loss of jobs. A decade later, we see that domestic teleradiology has flourished and enabled many radiologists to live where they want to live rather than the middle of nowhere.

 

[Gadofosveset]

Right - the inevitable demise at that time was from “outsourcing to India”. He also had the fixed belief that surgery would soon own IR.

In the 90s it was HillaryCare.

Every decade something new.

Your surgery preceptor must have been ahead of the curve if they mentioned AI 11 years ago. AlexNet didn't win ImageNet until 2012, which lead to the recent boom in interest in convolutional neural networks.

Certainly at that time, people were talking about how teleradiology technology would lead to off-shore outsourcing and loss of jobs. A decade later, we see that domestic teleradiology has flourished and enabled many radiologists to live where they want to live rather than the middle of nowhere.

 

[NDcienporciento100]

Right - the inevitable demise at that time was from “outsourcing to India”. He also had the fixed belief that surgery would soon own IR.

In the 90s it was HillaryCare.

Every decade something new.

Surgery owning IR is most laughable offense on this thread! The other day I saw a surgeon put in a chest into a patient with a chronically elevated diaphragm thinking it was a pneumo and asked us to asses it. half of the time they are asking us to drain abscesses that are not even abscesses. Carotid sticks for central lines. You name a procedure and I can tell you a horrible story about how surgeon tried it screwed it up lol.

 

[Cognovi]

Right - the inevitable demise at that time was from “outsourcing to India”. He also had the fixed belief that surgery would soon own IR.

In the 90s it was HillaryCare.

Every decade something new.

Was anything in HillaryCare specific to radiology or was it across all specialties, like Medicare for All?

 

[Gadofosveset]

Was anything in HillaryCare specific to radiology or was it across all specialties, like Medicare for All?

Not specific to radiology — the theory was that specialists were gonna get rocked in favor of primary care spectrum. The threat was grave enough to depress radiology residency applications for a few years: end of radiology, yada, &c. Similar know-it-alls at work generation after generation. Then the helical CT renaissance... so far, residency applicants always seem to be a step behind the wave, which is unfortunate.

 

[Mh0311]

^well fortunate for those of us who do their research and are applying

 

[2brads]

Look. If we cant even figure out how to keep “self driving” cars from smashing into things, we are very far from AI reading radiology reports. Sure there will be incremental steps that will make rads lives easier like pointing out lung nodules, subtle pneumonia, brain bleeds, fractures. But it’s going to provide a heat map of probability that the rad can take into account. It’s not going to say “transverse fracture of the mid femoral shaft with apex lateral angulation “. Don’t let the specter of AI dissuade you from a great career.

 

[Cognovi]

Not specific to radiology — the theory was that specialists were gonna get rocked in favor of primary care spectrum. The threat was grave enough to depress radiology residency applications for a few years: end of radiology, yada, &c. Similar know-it-alls at work generation after generation. Then the helical CT renaissance... so far, residency applicants always seem to be a step behind the wave, which is unfortunate.

The reality is that the now-final Medicare rules to give more weight to E&M codes is going to rock specialists in favor of the primary care spectrum. Radiology is going to see a 10% cut.

 

[NDcienporciento100]

The reality is that the now-final Medicare rules to give more weight to E&M codes is going to rock specialists in favor of the primary care spectrum. Radiology is going to see a 10% cut.

Is that what you’re banking on?

 

[2brads]

The reality is that the now-final Medicare rules to give more weight to E&M codes is going to rock specialists in favor of the primary care spectrum. Radiology is going to see a 10% cut.

If you’re not going to apply to rads because of this one thing, you shouldn’t be a radiologist anyways.

 

[Cognovi]

Is that what you’re banking on?

How can one bank on that?

 

[NDcienporciento100]

How can one bank on that?

What speciality are you going into?

 

[Cognovi]

What speciality are you going into?

I'm a radiology resident.

 

[Gadofosveset]

The reality is that the now-final Medicare rules to give more weight to E&M codes is going to rock specialists in favor of the primary care spectrum. Radiology is going to see a 10% cut.

Yup -- guess it's time to jump ship for FM & IM -- and I've been told this is what a lot of med students did in the mid 90s.

I think one would feel this new change to a greater or lesser degree depending on one's practice type. If one is say, an academic neuroradiologist at, say, MGH, then I imagine one would probably be more insulated from an acute change. Big systems already engage in a lot of cost sharing among the different medical specialties. That's why I always found the radiologist "parasite" claim from the ED amusing... in the few big systems I know, cost sharing effectively meant that radiology was indirectly funding the ED. Hey -- it's all a team effort anyway.

 

[scoopdaboop]

Yup -- guess it's time to jump ship for FM & IM -- and I've been told this is what a lot of med students did in the mid 90s.

I think one would feel this new change to a greater or lesser degree depending on one's practice type. If one is say, an academic neuroradiologist at, say, MGH, then I imagine one would probably be more insulated from an acute change. Big systems already engage in a lot of cost sharing among the different medical specialties. That's why I always found the radiologist "parasite" claim from the ED amusing... in the few big systems I know, cost sharing effectively meant that radiology was indirectly funding the ED. Hey -- it's all a team effort anyway.

But then next year, radiology will see a 2 percent increase... etc. etc. It'll shift and get lower over time, but I doubt radiology salaries will ever fall so low that the average is less than 450k MGMA.

 

[Radology]

You should listen to the chair of surgery’s input on radiology as much as your should listen to the chair of radiology’s input on anesthesiology.

the concerns (particularly outsourcing) are non issues for the foreseeable future

there is literally One legitimate threat to radiology — our self destructive practice of selling our premier practices to Private equity firms for a quick buck.

 

[90sportsfan]

There have been so many posts on this topic (the question pops up weekly at times). The answer is always the same. Radiologists will not be replaced. If anything, in our lifetime the new technologies would be a diagnostic adjunct that would improve efficiency (assuming it can be perfected). I am not going to fall for changing my desired specialty (Radiology) based on trying to project on future technology. There are still critical reads and the reports have to be "signed." Would a non-trained physician be willing to slap their name on a report that's been generated by AI?

 

[recycledpaper]

I can't wait for all the shiny AI algorithms to be released for clinical use. It's going to make my life so much more better and efficient.

 

[g3tb0mbed]

I can't wait for all the shiny AI algorithms to be released for clinical use. It's going to make my life so much more better and efficient.

I see this being said a lot, but doesn't more efficient = less radiologists needed = worse job market/salary?

 

[Jamiu22]

I see this being said a lot, but doesn't more efficient = less radiologists needed = worse job market/salary?

There continues to be exponential need for radiology services, even as we have become more efficient over the past 20yrs. I dont see this dynamic changing with the advent of artificial intelligence. Clinicians no longer rely solely on physical exam/labs, and their patients are demanding increased sensitivity, perfection, and in some cases, increased specificity. Progressively increased aging population is also a challenge to an ever increasing work load burden i.e. increased imaging utility in the ED workup, cancer staging etc...

Overall, there is a lot of dynamics to medicine, pretty fluid if I might add, so anyone who tells you they have all the answers is lying to you. In my humble opinion, radiology will continue being a hidden gem to a lot of med students who dare to dream of the darkness - a road to Nirvana (lots of vaca, high pay, mentally stimulating, not having to deal the BS aspects of clinical medicine).

 

[NDcienporciento100]

Heck ya to not having to deal with the bs stuff of medicine!

 

[recycledpaper]

I see this being said a lot, but doesn't more efficient = less radiologists needed = worse job market/salary?

Yes and no. It will be a gradual process. Right now, the direction radiology is headed is not sustainable. Something has to change.. volume is not going down, so we need to be more efficient just to survive and maintain our sanity.

It is possible that with more efficient radiologists, you can have one radiologist in place of two for a shift. Then that would cancel out the efficiency you gain from AI tools. However, the scenario is dynamic. All I know is, we need AI tools to improve our workflow. Some hospital systems and groups are already employing them.

 

[Listeningprince]

People will always make projections based on their experience. Sometimes these projections ignore factors outside of their personal experience. I would not loose sleep over it. I doubt AI will replace the DR. instead, I see it being used to make the system more efficient. Remember when everyone thought smart EKG machines will replace the cardiologists input in that regard. Never happened. Instead it made the cardiologists job easier.

 

[Steve_Zissou]

What are peoples thoughts on residency programs that are beginning to integrate AI development training? UW seems to have recently added a "deep learning" pathway which seems really interesting.

 

[fun8stuff]

Slightly off topic. I had a case last week where the surgeon had placed the chest tube into the liver! Last year the case of the year was a chest tube that a surgeon placed into the right ventricle !!
...

Anyway, I’d be more worried about AI if i was a non-procedural clinician. In my area they’re already staffing urgent cares and primary care practices with NP’s and PA’s.... as well as certain surgery days with all CRNAs. It would be much easier to replace these types of docs with midlevels augmented with AI where midlevel’s role is mostly data entry and patient communication. There isn’t much high level thinking needed for general primary and urgent care.

 

[Qwerty122]

What are peoples thoughts on residency programs that are beginning to integrate AI development training? UW seems to have recently added a "deep learning" pathway which seems really interesting.

I think it's an important and necessary step that will eventually become a requirement for most radiology programs, just as how things like PACS and MRI were new technologies in the past that now all radiologists must be comfortable with. Radiologists don't have to code and create AI programs, but at least should be able to converse intelligently on AI and machine learning models with data scientists and programmers. Looking at the most recent RSNA conference, clinical integration of AI especially into radiology is starting to pick up speed, with the "top" institutions like UCSF and BWH leading the way. (Data science pathway prepares radiology residents for machine learning). I think our generation of radiologists are the ones who need to be aware of this change and take the initiative to facilitate this shift in clinical practice setting correctly. It's important for us to be able to separate the "hype" from actual clinical relevance (especially in this current boom of AI research) in order to address the impending future problems (high workload, aging population, retiring radiologists, etc) that other posters above have been talking about.

 

[Cognovi]

Looking at the most recent RSNA conference, clinical integration of AI especially into radiology is starting to pick up speed, with the "top" institutions like UCSF and BWH leading the way. (Data science pathway prepares radiology residents for machine learning).

Why is "top" in quotation marks?

 

[Qwerty122]

Just to signify that it's hard to accurately rank and differentiate radiology residency programs, since there are a lot of factors in play, with each factor having a different weight to each student, so "top" as in the widely accepted top programs, nothing too definitive and just going off doximity rankings

 

[Mh0311]

Maybe it’s just me but I think AI is going to make us even busier, imagine having to double check all the AI reports on top of your own

 

[Qwerty122]

That can be a definite concern too especially at the early stages of implementation, but progress on algorithms that can rule out diseases and identify negative findings with a high degree of accuracy might help reduce the more mundane aspects of radiology where you are just ruling out findings and are reading a normal scan. I also think there is great utility for algorithms that can tell you a probable finding, and tell you how confident it is on its analysis, then triage it up or down your workflow accordingly. UCSF is implementing this to some extent. I've also heard that the process of dictating and writing a report can be significantly reduced. Eliot Siegel (of Maryland, who created the first PACS radiology department in the world) is heavily involved in this and claims that radiologists spend 15% of their time actually reading studies, and that can be improved up to 70-80% if more efficient report-writing and small quality of life changes due to AI are correctly implemented. He could have taken these stats out of his butt, but he has some credibility, especially as a technology innovator in radiology.