Radiology Regret?

[GoPelicans]

I was also deciding between rads an anesthesia. At the end of pGY2 and I don't know if I made the better choice, but I don't think I would've been any happier in anesthesia. I think IR will be the thing that ultimately makes me most happy because I do want to do procedures, unfortunately 4 people in my class want to do IR and there are only two ESIR spots, which is uncharted territory for my program. So I have a lot of anxiety over my future right now.

---

Members don't see this ad.

 

[deleted746658]

You still never answered my question. Earlier, in a different thread, you claimed that there are way better specialties that are more insulated from encroachment/loss of jobs than radiology. I asked you which ones and you didn’t answer. So which ones?

IIRC, i said there are multiple fields that are more automation resistant.

let me list a few: anything procedural (GSurg and all its fellowships, CTS, urology, IR, anesthesia, etc)

general medicine (PCP and hospitalist)

all of the internal medicine subspecialties (rheum, A/I, cards, GI, renal, heme onc, etc)

misc: PMR, interventional pain, definitely EM, etc

in other words, almost every field is more automation resistant than rads

rads is about image pattern recognition. turns out, computers are really good at this

 

[Steve_Zissou]

IIRC, i said there are multiple fields that are more automation resistant.

let me list a few: anything procedural (GSurg and all its fellowships, CTS, urology, IR, anesthesia, etc)

general medicine (PCP and hospitalist)

all of the internal medicine subspecialties (rheum, A/I, cards, GI, renal, heme onc, etc)

misc: PMR, interventional pain, definitely EM, etc

in other words, almost every field is more automation resistant than rads

rads is about image pattern recognition. turns out, computers are really good at this

I guess I wasn’t looking for only automation resistant, but encroachment resistant. EM, FM, IM, etc are constantly being encroached on as well.

 

[Kuratz]

@cosine89

I’m not sure why you are so obsessed with AI in Rads. You bailed to go do IM, so not sure why you continue to troll here.

Interestingly our hospital purchased a very expensive, top of the line AI system for detecting sepsis. Anytime a patient’s lab and vitals profile hit certain parameters a screen pops up in the EMR suggesting we give the patient liters of fluid, start broad spectrum abx and take blood cultures. We are supposed to follow this 80% of the time, “It has unparalleled sensitivity!” They said. Do you feel threatened as a future hospitalist based on this?

Of course not! In practice the thing is beyond worthless and arguably does more harm than good. It suggests starting the sepsis bundle on post-op lap choles with expected leukocytosis and some baseline level of kidney disease. Giving them 2L of fluid and Vanc-Zosyn unnecessarily would obviously be bad medicine.

Some suits or even physicians sat through some sales pitch on why this program would improve outcomes and decrease length of stay or whatever. Perhaps the “evidence” was rock solid. But in practice the thing sucks. We ignore it entirely. This is what I envision Rads AI to be like for the foreseeable future. It’s going to be high sensitivity crap that no one trusts. Hell look at CAD for breast and it’s the same thing.

Once AI can put out a full, accurate report to clinicians on a complex MRI or CT and people actually trust it, the same with be for things like this Sepsis AI. Then you could argue midlevel + AI will be preferred over hospitalist as a cost-saving measure. I don’t think that will happen, but two can play the “sky is falling” game.

 

[medradthrowaway]

I guess I wasn’t looking for only automation resistant, but encroachment resistant. EM, FM, IM, etc are constantly being encroached on as well.

isnt IR also being encroached heavily? theres a thread here about IR PAs doing most of the drains and ports. Isn't there also a post about mid levels prelim reading studies? i dont think any field is totally safe from it. it just takes some lobbying and time.

 

[Dave1980]

isnt IR also being encroached heavily? theres a thread here about IR PAs doing most of the drains and ports. Isn't there also a post about mid levels prelim reading studies? i dont think any field is totally safe from it. it just takes some lobbying and time.

omg why are med students so worried about encroachment. there is more than enough radiology work (after covid subsides). just do what you like.

 

[medradthrowaway]

omg why are med students so worried about encroachment. there is more than enough radiology work (after covid subsides). just do what you like.

It's a counter argument. If "doing what you like" is always the answer then there wouldn't be any of these threads.

 

[Steve_Zissou]

It's a counter argument. If "doing what you like" is always the answer then there wouldn't be any of these threads.

It’s also a legitimate worry for a lot of us. “What is this field going to look like over the next 20 years” is a question I’ve asked myself with every field I’ve considered right alongside “will I hate my life if I choose to do this?”

 

[medradthrowaway]

It’s also a legitimate worry for a lot of us. “What is this field going to look like over the next 20 years” is a question I’ve asked myself with every field I’ve considered right alongside “will I hate my life if I choose to do this?”

Exactly. I know these attendings already have it set and they'll retire before any of this stuff affects them but from a student perspective, it's very important.

 

[Cognovi]

i think those who say that AI is not an imminent threat are naive and defensive

nearly every day or so there is a study like this one - https://pubs.rsna.org/doi/10.1148/radiol.2020201491 (here, an example of AI>>AI+human>>human). this tech is coming and it will transform the field. i have no doubt.

anesthesia much safer option. and if you liked it all along, go for it!

AI+human did better than AI alone. This model is not going to save any time.

 

[deleted746658]

AI+human did better than AI alone. This model is not going to save any time.

my point was there are proof of concept articles daily demonstrating what AI is capable of

the community of radiologists as a whole are entirely too dismissive of the threat of AI . "blah blah EKG.. blah blah CAD/ mammo"

the bottom line is ML/NN are completely different and far more advanced than old ecg / CAD tech.

we have evidence that AI is already non-inferior/ in some cases superior in narrow task use. that alone should cause pause when considering the field of radiology.

literally some of the brightest minds in the world backed by unlimited $$$ are going to work every day to put you out of a job.

this isn't anesthesia where MD's are competing with nurses. this is MD versus super intelligent self learning computer powered by google.

it's not a matter of "if" but "when"

i don't think its unreasonable to have AI in our workflow in 3-5 years. 10 years out , prelim AI reads (which would cause substantial decrease need for human rads). 15-20 years - ??.... but to anticipate a 30+ year career where you are a human interpreting scans. LOL. not a chance.

 

[clutch21]

my point was there are proof of concept articles daily demonstrating what AI is capable of

the community of radiologists as a whole are entirely too dismissive of the threat of AI . "blah blah EKG.. blah blah CAD/ mammo"

the bottom line is ML/NN are completely different and far more advanced than old ecg / CAD tech.

we have evidence that AI is already non-inferior/ in some cases superior in narrow task use. that alone should cause pause when considering the field of radiology.

literally some of the brightest minds in the world backed by unlimited $$$ are going to work every day to put you out of a job.

this isn't anesthesia where MD's are competing with nurses. this is MD versus super intelligent self learning computer powered by google.

it's not a matter of "if" but "when"

i don't think its unreasonable to have AI in our workflow in 3-5 years. 10 years out , prelim AI reads (which would cause substantial decrease need for human rads). 15-20 years - ??.... but to anticipate a 30+ year career where you are a human interpreting scans. LOL. not a chance.

It’s incredible that as a physician in a completely different field, you feel qualified enough to keep giving your opinion on the matter. It’s true that no one can say for certain what the future holds, but until you walk a mile in the shoe of a radiologist, it’s probably best to sit this one out.

Just a few years back there was an applicant who had a background/strong interest in AI who was applying to rads. Only a few months after starting R1 year, they quickly cooled on their optimistic outlook of AI.

Updated thoughts on AI/ML

I figured I'd give a little insight into my experience with what's happening in AI/ML and radiology. There's a great opinion piece from the New York Times that came out today. It echoes my thoughts and prompted me to write this post. The gestalt of the article is that deep learning has enabled...

forums.studentdoctor.net

 

[NDcienporciento100]

my point was there are proof of concept articles daily demonstrating what AI is capable of

the community of radiologists as a whole are entirely too dismissive of the threat of AI . "blah blah EKG.. blah blah CAD/ mammo"

the bottom line is ML/NN are completely different and far more advanced than old ecg / CAD tech.

we have evidence that AI is already non-inferior/ in some cases superior in narrow task use. that alone should cause pause when considering the field of radiology.

literally some of the brightest minds in the world backed by unlimited $$$ are going to work every day to put you out of a job.

this isn't anesthesia where MD's are competing with nurses. this is MD versus super intelligent self learning computer powered by google.

it's not a matter of "if" but "when"

i don't think its unreasonable to have AI in our workflow in 3-5 years. 10 years out , prelim AI reads (which would cause substantial decrease need for human rads). 15-20 years - ??.... but to anticipate a 30+ year career where you are a human interpreting scans. LOL. not a chance.

When am I going to be able to trust Google Translate English to Spanish?

 

[NDcienporciento100]

When am I going to be able to trust Google Translate English to Spanish? Because if they can’t do that that they can’t read a complex CT.

 

[SeisK]

my point was there are proof of concept articles daily demonstrating what AI is capable of

the community of radiologists as a whole are entirely too dismissive of the threat of AI . "blah blah EKG.. blah blah CAD/ mammo"

the bottom line is ML/NN are completely different and far more advanced than old ecg / CAD tech.

we have evidence that AI is already non-inferior/ in some cases superior in narrow task use. that alone should cause pause when considering the field of radiology.

literally some of the brightest minds in the world backed by unlimited $$$ are going to work every day to put you out of a job.

this isn't anesthesia where MD's are competing with nurses. this is MD versus super intelligent self learning computer powered by google.

it's not a matter of "if" but "when"

i don't think its unreasonable to have AI in our workflow in 3-5 years. 10 years out , prelim AI reads (which would cause substantial decrease need for human rads). 15-20 years - ??.... but to anticipate a 30+ year career where you are a human interpreting scans. LOL. not a chance.

Your single choice of article in defense of AI alone is better than AI+ radiologist came from a disease model where radiologists have no formal training, there is very little published to date in radiologic patterns, and therefore exceptionally difficult to reliably diagnose. The singular paper you’ve cited demonstrates only that AI is faster at picking up patterns than humans are.

You know, if you’re super disgruntled about IM, you can always apply rads again. Better late than never.

 

[DOFOSHO]

my point was there are proof of concept articles daily demonstrating what AI is capable of

the community of radiologists as a whole are entirely too dismissive of the threat of AI . "blah blah EKG.. blah blah CAD/ mammo"

the bottom line is ML/NN are completely different and far more advanced than old ecg / CAD tech.

we have evidence that AI is already non-inferior/ in some cases superior in narrow task use. that alone should cause pause when considering the field of radiology.

literally some of the brightest minds in the world backed by unlimited $$$ are going to work every day to put you out of a job.

this isn't anesthesia where MD's are competing with nurses. this is MD versus super intelligent self learning computer powered by google.

it's not a matter of "if" but "when"

i don't think its unreasonable to have AI in our workflow in 3-5 years. 10 years out , prelim AI reads (which would cause substantial decrease need for human rads). 15-20 years - ??.... but to anticipate a 30+ year career where you are a human interpreting scans. LOL. not a chance.

currently "AI" has worse specificity than a radiologist who doesn't even open the study. AI doesn't do anything but slow the radiologist down. AI is like that one referring doc who calls and asks "is that central fluid filled structure in the pelvis an abscess?" "no dude everyone has that"

 

[CyrusHabs7695]

Your single choice of article in defense of AI alone is better than AI+ radiologist came from a disease model where radiologists have no formal training, there is very little published to date in radiologic patterns, and therefore exceptionally difficult to reliably diagnose. The singular paper you’ve cited demonstrates only that AI is faster at picking up patterns than humans are.

You know, if you’re super disgruntled about IM, you can always apply rads again. Better late than never.

haha you can join the dark side, we won't tell anybody

 

[Taurus]

AI is overblown. I even consulted with a big tech company on their AI project. We are decades away if ever from being replaced by AI. Trust me, if radiologists can be successfully replaced by AI, absolutely no field in medicine, business, law, etc is safe. Performing radiology requires knowledge base which can be programmed but machines are very poor at recognizing less obvious image findings and synthesizing new findings and conclusions. It just takes one miss, one untimely death, and $20 million lawsuit to keep AI at bay. AI will supplement not replace radiologists. Much like an improved version of CAD, which isn't all that great to begin with.

I wrote this last year and nothing has changed.

Let's play devil's advocate and assume the worst. If AI is so good and powerful that you don't need radiologists anymore, the last thing I would worry about is my job as a radiologist. I would be more concerned about how do we prevent humans from becoming slaves to machine masters. That's the level of sophistication that AI would need to achieve to replace radiologists.

 

[deleted746658]

Your single choice of article in defense of AI alone is better than AI+ radiologist came from a disease model where radiologists have no formal training, there is very little published to date in radiologic patterns, and therefore exceptionally difficult to reliably diagnose. The singular paper you’ve cited demonstrates only that AI is faster at picking up patterns than humans are.

You know, if you’re super disgruntled about IM, you can always apply rads again. Better late than never.

Yes, AI is faster at picking up patterns. In fact, AI is damn good at identifying patterns. Radiology is pattern recognition. And it's all digital. That's convenient.

It wasn't my single choice. My point was there are studies such as the above cited that come out DAILY.

For example, just yesterday I came across the below -- AI distinguishing normal from abnormal CXR

Automated abnormality classification of chest radiographs using deep convolutional neural networks - npj Digital Medicine

As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology...

www.nature.com

You really don't think AI will be able to filter normal studies in the next 10-15 years? that would equate to a HUGE drop in # of studies and a massively reduced demand for radiologists.

right now, the studies are promising. think about in 5 years? how about 10? 15-30 years? that is all within current and future trainees working careers. i'm not trolling. i'm literally trying to be helpful to those on the fence between rads.

apply rads again? LOL. by that time it very well may be - "humans need not apply"

 

[NDcienporciento100]

Yes, AI is faster at picking up patterns. In fact, AI is damn good at identifying patterns. Radiology is pattern recognition. And it's all digital. That's convenient.

It wasn't my single choice. My point was there are studies such as the above cited that come out DAILY.

For example, just yesterday I came across the below -- AI distinguishing normal from abnormal CXR

Automated abnormality classification of chest radiographs using deep convolutional neural networks - npj Digital Medicine

As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology...

www.nature.com

You really don't think AI will be able to filter normal studies in the next 10-15 years? that would equate to a HUGE drop in # of studies and a massively reduced demand for radiologists.

right now, the studies are promising. think about in 5 years? how about 10? 15-30 years? that is all within current and future trainees working careers. i'm not trolling. i'm literally trying to be helpful to those on the fence between rads.

apply rads again? LOL. by that time it very well may be - "humans need not apply"

Who got to you and flipped you? Medicine attending you really respected or was it possible jealous medicine residents around you saying things, I’m just curious what tipped you from rads to hospitalist?

 

[deleted746658]

Who got to you and flipped you? Medicine attending you really respected or was it possible jealous medicine residents around you saying things, I’m just curious what tipped you from rads to hospitalist?

i wanted job security...

and done in 2 years, make 250k easily (more if u want), tremendous flexibility, option to sub-specialize is always there down the road

versus

grind for 5 more years getting paid 60k/year learning something completely new and worrying each year that someone is getting closer and closer to making my job obsolete.

yea, easy choice honestly if you can tolerate the work.

a job is a job.

 

[Steve_Zissou]

i wanted job security...

and done in 2 years, make 250k easily (more if u want), tremendous flexibility, option to sub-specialize is always there down the road

versus

grind for 5 more years getting paid 60k/year learning something completely new and worrying each year that someone is getting closer and closer to making my job obsolete.

yea, easy choice honestly if you can tolerate the work.

a job is a job.

Aren’t nurses much closer to taking your job than AI to Rad’s?

 

[SeisK]

Yes, AI is faster at picking up patterns. In fact, AI is damn good at identifying patterns. Radiology is pattern recognition. And it's all digital. That's convenient.

It wasn't my single choice. My point was there are studies such as the above cited that come out DAILY.

For example, just yesterday I came across the below -- AI distinguishing normal from abnormal CXR

Automated abnormality classification of chest radiographs using deep convolutional neural networks - npj Digital Medicine

As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology...

www.nature.com

You really don't think AI will be able to filter normal studies in the next 10-15 years? that would equate to a HUGE drop in # of studies and a massively reduced demand for radiologists.

right now, the studies are promising. think about in 5 years? how about 10? 15-30 years? that is all within current and future trainees working careers. i'm not trolling. i'm literally trying to be helpful to those on the fence between rads.

apply rads again? LOL. by that time it very well may be - "humans need not apply"

Yes, AI is faster at picking up patterns. In fact, AI is damn good at identifying patterns. Radiology is pattern recognition. And it's all digital. That's convenient.

It wasn't my single choice. My point was there are studies such as the above cited that come out DAILY.

For example, just yesterday I came across the below -- AI distinguishing normal from abnormal CXR

Automated abnormality classification of chest radiographs using deep convolutional neural networks - npj Digital Medicine

As one of the most ubiquitous diagnostic imaging tests in medical practice, chest radiography requires timely reporting of potential findings and diagnosis of diseases in the images. Automated, fast, and reliable detection of diseases based on chest radiography is a critical step in radiology...

www.nature.com

You really don't think AI will be able to filter normal studies in the next 10-15 years? that would equate to a HUGE drop in # of studies and a massively reduced demand for radiologists.

right now, the studies are promising. think about in 5 years? how about 10? 15-30 years? that is all within current and future trainees working careers. i'm not trolling. i'm literally trying to be helpful to those on the fence between rads.

apply rads again? LOL. by that time it very well may be - "humans need not apply"

I said the single choice of article defended AI > AI+rads. I’ve yet to see another study, cited by you or otherwise, defending that premise, and the first article you linked wasn’t a defense of an existential threat.

The NN in the study you currently cited, even if it was already packaged and ready to sell, would not be purchased by anyone. A ”normal” filter with any sensitivity or specificity below 100% (and they will always be that way) does not save any radiologist any amount of time, because a “normal” flagged abnormal that went unchecked by a radiologist would quickly hammer a lawsuit down.

I truly do not believe that even in 40 years AI will have replaced radiologists, for the same reason that something as rudimentary as autopilot hasn’t replaced a pilot. I am honestly beginning to think AI won’t even make radiologists much faster outside of tedium like counting pulm-nodules, it’ll just make them more accurate.

I’m sorry that your fear made you choose something that may not have been your number 1. Clearly the possibility that the AI train was overhyped is bothering you. I’m having a hard time imaging why another subspecialist would keep visiting this place.

 

[CyrusHabs7695]

cosine89 working his Saturday shift filling out a million charts while the rads guys are home chilling

 

[medradthrowaway]

I think the point alot of people are missing is that AI does not need to replace radiologists for the job market to take a sht. It just needs to be "good enough" to help improve efficiency. The analogy isn't to AI interpreting EKG (there's no money to be made there). The analogy is more akin to Walmart hiring 5 cashiers per store these days since we have self-check out lanes when it hired 20+ in the past to manually staff every lane. The self-check out lane only needs 1 staff to sort through and handle problemed transactions. I think this is a very realistic model in our lifetime.

Also, a "chilled" specialty isn't something to brag or showcase as it highlights opportunities for "inefficiencies." The MBAs will have a look at this and dig their corporate fingers through this in the future, I'm sure.

 

[SeisK]

I think the point alot of people are missing is that AI does not need to replace radiologists for the job market to take a sht. It just needs to be "good enough" to help improve efficiency. The analogy isn't to AI interpreting EKG (there's no money to be made there). The analogy is more akin to Walmart hiring 5 cashiers per store these days since we have self-check out lanes when it hired 20+ in the past to manually staff every lane. The self-check out lane only needs 1 staff to sort through and handle problemed transactions. I think this is a very realistic model in our lifetime.

Also, a "chilled" specialty isn't something to brag or showcase as it highlights opportunities for "inefficiencies." The MBAs will have a look at this and dig their corporate fingers through this in the future, I'm sure.

Im not necessarily saying I disagree, but what specific implementation of AI can you envision that would save a whole bunch of time?

 

[deleted746658]

Im not necessarily saying I disagree, but what specific implementation of AI can you envision that would save a whole bunch of time?

as mentioned above, filtering normal from abnormal studies.

if an AI application able to do that was deployed, it's effects on the job market would be catastrophic.

i'm just not sure how you can argue that

hence, why i said most in radiology are naive and overly defensive about AI in radiology and how it will effect the field.

 

[NDcienporciento100]

I think the point alot of people are missing is that AI does not need to replace radiologists for the job market to take a sht. It just needs to be "good enough" to help improve efficiency. The analogy isn't to AI interpreting EKG (there's no money to be made there). The analogy is more akin to Walmart hiring 5 cashiers per store these days since we have self-check out lanes when it hired 20+ in the past to manually staff every lane. The self-check out lane only needs 1 staff to sort through and handle problemed transactions. I think this is a very realistic model in our lifetime.

Also, a "chilled" specialty isn't something to brag or showcase as it highlights opportunities for "inefficiencies." The MBAs will have a look at this and dig their corporate fingers through this in the future, I'm sure.

You are comparing swiping bar codes to the complex intricacies of reading CT and MRI. And you are also misinterpreting what “chill speciality” means, does not mean intellectually easy what it means is know 5:30 am pre rounding, means no 2:00 am calls to renew Tylenol (unless you are doing IR). You are still studying hard.

 

[SeisK]

as mentioned above, filtering normal from abnormal studies.

if an AI application able to do that was deployed, it's effects on the job market would be catastrophic.

i'm just not sure how you can argue that

hence, why i said most in radiology are naive and overly defensive about AI in radiology and how it will effect the field.

Yeah m8 that post was for another person, you got your own higher up.

 

[medradthrowaway]

You are comparing swiping bar codes to the complex intricacies of reading CT and MRI. And you are also misinterpreting what “chill speciality” means, does not mean intellectually easy what it means is know 5:30 am pre rounding, means no 2:00 am calls to renew Tylenol (unless you are doing IR). You are still studying hard.

Honestly, radiology can be taught to almost anyone. It's all about repetition. Ultimately, the trend isn't going to be defined by those in the field but economics and capitalism. There is alot of money to be had here, and my wager is on something rather than nothing. Just like CRNAs have allowed anesthesia departments across the country to run multiple rooms with fewer anesthesiologists and increase RVUs dramatically, this too will happen across medicine in one form or another simply because there is a lot of monetary incentive.

International evaluation of an AI system for breast cancer screening - Nature

An artificial intelligence (AI) system performs as well as or better than radiologists at detecting breast cancer from mammograms, and using a combination of AI and human inputs could help to improve screening efficiency.

www.nature.com

 

[Tiger100]

as mentioned above, filtering normal from abnormal studies.

if an AI application able to do that was deployed, it's effects on the job market would be catastrophic.

i'm just not sure how you can argue that

hence, why i said most in radiology are naive and overly defensive about AI in radiology and how it will effect the field.

Unlike what you said, radiology as a field is very well aware of AI and its applications. Radiology societies and many academic departments have dedicated researches on AI, some are radiologists and have IT background. So saying that people in radiology are naive about AI is a false statement. You are trying to push your opinion by questioning other people's qualifications.


Now back to main discussion: There is no clear-cut line between normal and abnormal. Abnormal needs be defined. The clinical relevance of the findings is extremely important. Otherwise, many radiology exams are abnormal if you see them from just technical and not medical standpoint.

One of the hardest tasks in radiology is differentiating normal from abnormal in the clinical context. In other words, leaving alone the abnormalities that don't have significant clinical impact.

If AI becomes so good that it can differentiate normal from abnormal in the clinical context, the same AI can easily replace many other physicians, alone or with the help of an NP. If AI can do a nephrology or oncology consult for an NP, what's the point of doing a human consult?

 

[Tiger100]

Honestly, radiology can be taught to almost anyone. It's all about repetition. Ultimately, the trend isn't going to be defined by those in the field but economics and capitalism. There is alot of money to be had here, and my wager is on something rather than nothing. Just like CRNAs have allowed anesthesia departments across the country to run multiple rooms with fewer anesthesiologists and increase RVUs dramatically, this too will happen across medicine in one form or another simply because there is a lot of monetary incentive.

International evaluation of an AI system for breast cancer screening - Nature

An artificial intelligence (AI) system performs as well as or better than radiologists at detecting breast cancer from mammograms, and using a combination of AI and human inputs could help to improve screening efficiency.

www.nature.com

Let me correct you. Anything, can be taught to almost anyone.

 

[CyrusHabs7695]

Honestly, radiology can be taught to almost anyone. It's all about repetition. Ultimately, the trend isn't going to be defined by those in the field but economics and capitalism. There is alot of money to be had here, and my wager is on something rather than nothing. Just like CRNAs have allowed anesthesia departments across the country to run multiple rooms with fewer anesthesiologists and increase RVUs dramatically, this too will happen across medicine in one form or another simply because there is a lot of monetary incentive.

International evaluation of an AI system for breast cancer screening - Nature

An artificial intelligence (AI) system performs as well as or better than radiologists at detecting breast cancer from mammograms, and using a combination of AI and human inputs could help to improve screening efficiency.

www.nature.com

bro did u actually make a burner account just to rip into radiology people

 

[NDcienporciento100]

Honestly, radiology can be taught to almost anyone. It's all about repetition. Ultimately, the trend isn't going to be defined by those in the field but economics and capitalism. There is alot of money to be had here, and my wager is on something rather than nothing. Just like CRNAs have allowed anesthesia departments across the country to run multiple rooms with fewer anesthesiologists and increase RVUs dramatically, this too will happen across medicine in one form or another simply because there is a lot of monetary incentive.

International evaluation of an AI system for breast cancer screening - Nature

An artificial intelligence (AI) system performs as well as or better than radiologists at detecting breast cancer from mammograms, and using a combination of AI and human inputs could help to improve screening efficiency.

www.nature.com

When?
I predict that at some point in the future all doctors will be replaced by machines!
Notices how I put no end point so essentially I’ll never be wrong in my prediction.
I predict that radiologist will have no problem getting a job in my working life time 30ish years, now what say you?

 

[Steve_Zissou]

When?
I predict that at some point in the future all doctors will be replaced by machines!
Notices how I put no end point so essentially I’ll never be wrong in my prediction.
I predict that radiologist will have no problem getting a job in my working life time 30ish years, now what say you?

I would go even further and say for the next 50-100 years unless there is some massive breakthrough. Machine learning and AI are very complex things. I’m no expert but I’ve read “consciousness explained” by Daniel Dennett, which lays out pretty well a lot of the issues faced when it comes to developing AI and “deep learning.”

 

[NDcienporciento100]

I would go even further and say for the next 50-100 years unless there is some massive breakthrough. Machine learning and AI are very complex things. I’m no expert but I’ve read “consciousness explained” by Daniel Dennett, which lays out pretty well a lot of the issues faced when it comes to developing AI and “deep learning.”

I’m finishing my internship now and there is so much trash that’s gets said about radiologist, more then any other specialty and honestly I think that it is mostly jealousy.

 

[Steve_Zissou]

I’m finishing my internship now and there is so much trash that’s gets said about radiologist, more then any other specialty and honestly I think that it is mostly jealousy.

Im curious, what kind of stuff do people say about radiologists?

 

[NDcienporciento100]

Im curious, what kind of stuff do people say about radiologists?

Mostly Surgery residents talking crap about rads residents, over emphasizing one bad call from a 2nd year tads resident. My chief surgery resident honestly believes he reads an abdominal ct better then an average senior radiology resident, laughable. Oh and then when you talk about IR, the surgery teams “hates” my IR department, talking crap about them is one of there favorite past times.

 

[CidHighwind]

Damn I’m reading more and more of these threads and it’s making me really nervous about starting Rads next year. Maybe I should have ranked anesthesia programs before the rads ones. How likely is it to transfer out as a PGY1 if your program is advanced?

 

[Steve_Zissou]

Damn I’m reading more and more of these threads and it’s making me really nervous about starting Rads next year. Maybe I should have ranked anesthesia programs before the rads ones. How likely is it to transfer out as a PGY1 if your program is advanced?

Uff, I feel like anesthesia is gonna be the first or second one to go behind primary care.

 

[CidHighwind]

Uff, I feel like anesthesia is gonna be the first or second one to go behind primary care.

That’s what I figured as well. Even though I thought it would have been a more “fun” specialty, I worried about job security and the ability to work anywhere. TBH I usually don’t meet radiologist that “love” their actual work as much as their lifestyle/income, but seeing as how most hospitals still use 30 year old technology in beepers/pagers, even if the tech to replace radiologists came out tomorrow, would probably be another 10-20 years before we’d be out of a job.

 

[Steve_Zissou]

That’s what I figured as well. Even though I thought it would have been a more “fun” specialty, I worried about job security and the ability to work anywhere. TBH I usually don’t meet radiologist that “love” their actual work as much as their lifestyle/income, but seeing as how most hospitals still use 30 year old technology in beepers/pagers, even if the tech to replace radiologists came out tomorrow, would probably be another 10-20 years before we’d be out of a job.

Gotta remember man, it’s just a job. I’m 100% planning to live for my free time/lifestyle/income. I plan on being an outstanding radiologist and taking pride in my work, but that sure won’t be the thing that makes me happy.

 

[CidHighwind]

Gotta remember man, it’s just a job. I’m 100% planning to live for my free time/lifestyle/income. I plan on being an outstanding radiologist and taking pride in my work, but that sure won’t be the thing that makes me happy.

That’s how I felt. I realized that most surgeons that make it are truly passionate about the field. That said, I will still want to be good at my job, but am doing it so I have most of my weekends free and have the option of doing 7 on/14 off in the future if I wanted.

 

[seeingthings]

Damn I’m reading more and more of these threads and it’s making me really nervous about starting Rads next year. Maybe I should have ranked anesthesia programs before the rads ones. How likely is it to transfer out as a PGY1 if your program is advanced?

Reading the doom and gloom on these forums makes me feel the same way. I finish IM residency in a month and start rads on July 1st, but I am pretty miserable in IM. However after reading a lot of posts like these I’m wondering if I should have just stuck with it, maybe it’d be better with the money... or maybe not. The thing I try to remember is that most radiologists I’ve met are very happy. Most internists I’ve met are not.

 

[speedyxx626]

The arguments for AI by non-radiologists is absolutely RIDICULOUS. Honestly, I had some vague concerns about AI when I was in med school, but as soon as I stepped foot in a reading room and started interpreting studies on my own those fears went far out the window. There's no way in hell we're ever going to just trust AI to filter "normal" from abnormal studies, because how the hell would it ever do that? Everyone talks about AI identifying pulmonary nodules and pneumonia on plain films, etc, but how is that going to save us enough time to justify a computer reading through a study without a radiologist? A plain film takes a few seconds to not more than a couple minutes to be read by an experienced radiologist, so who the hell cares if AI can identify a pneumonia when a radiologist can do it just a fast. A radiologist is going to have to oversee AI anyways similar to how an attending overreads a resident because at the end of the day it's the radiologist's ass on the line. So where will time be saved? Not to mention having to look through all the false positive or missed findings that are going to happen.

Don't even get me started on cross sectional imaging. There are way too many normal anatomic variants, phases, sequences, subtleties, artifacts, incidentals, etc that a computer will not be able to read through anytime in the near future, if ever. Is AI going to look through a prior CT to tell us how the size or enhancement of a mass has changed over time? Is it going to compare findings on an CT to a prior MR or PETCT? How about finding that incidental renal mass on the last couple images of a chest CT (not super common, but I had two this week alone)?

I'm not saying AI isn't going to become part of radiology, because it probably will, but I think it'll mostly help us pick up more findings and not necessarily make us faster or more efficient. Hell, sometimes dictation software works so poorly that I would probably save more time if I typed my reports instead of dictating and correcting a bunch of errors.

 

[CidHighwind]

Reading the doom and gloom on these forums makes me feel the same way. I finish IM residency in a month and start rads on July 1st, but I am pretty miserable in IM. However after reading a lot of posts like these I’m wondering if I should have just stuck with it, maybe it’d be better with the money... or maybe not. The thing I try to remember is that most radiologists I’ve met are very happy. Most internists I’ve met are not.

True. Don’t meet many unhappy radiologists, dermatologists, ophthalmologists, or anesthesiologists IRL. Met a lot of unhappy peds, IM, and EM attendings surprisingly.

 

[NDcienporciento100]

You guys wishing you stuck it out in IM need to get it together, stop letting people on here get to you. They just hope they are right so they can justify a miserable note writing existence in which you are over worked, underpaid, and not appreciated. If you think procedure aspects of anesthesia are better then do IR, or if all you want to do is simple procedures (which is all An anesthesiologists gets to do) just do DR with procedural emphasis.

 

[GadRads]

Mostly Surgery residents talking crap about rads residents, over emphasizing one bad call from a 2nd year tads resident. My chief surgery resident honestly believes he reads an abdominal ct better then an average senior radiology resident, laughable. Oh and then when you talk about IR, the surgery teams “hates” my IR department, talking crap about them is one of there favorite past times.

When I was an R1, a senior surgery resident told me the same thing, claiming she did not need a radiologist's input for clinical decision making, but as a senior resident, I realize it's all nonsense. Surgeons can read some imaging, and many times, their knowledge of the patient, clinical exam and labs enhance their localization and interpretation of the findings. However their skills are limited and generally inferior to radiologists who, many at times, read imaging with limited clinical information. Surgeons also generally review imaging with tunnel vision. I have a lot of evidence to back this up.I have called the surgery team with critical findings which they missed. That said, subspecialty/niche surgeons, e.g. vascular/ENT will probably read vascular imaging or temporal bone CT better than a general rads, but the general radiologist will read everything else at a higher level. I helped a cardiac surgeon review a cardiac MRI and chest/cardiac CT before he took the patient to surgery, and I am not even a cardiovascular radiologist.

We also read imaging with complications and mishaps by surgeons and other physicians. We all make mistakes (this includes surgeons, no matter what they will tell you). It's all hubris, to be honest. Folks like to harp on radiology misses, because our mistakes and bad calls are documented and signed.

Most radiology residents will pit their CT skills against any surgeon anytime. It's ridiculous that senior radiology residents (who have a lot more experience reading imaging than surgery residents) still read imaging with some healthy trepidation, yet some arrogant surgery resident claims he is better. Just ridiculous.

Your chief resident, as he advances in his career, is free to ignore and not look at the radiology report. He will ultimately find himself in court.

 

[NDcienporciento100]

When I was an R1, a senior surgery resident told me the same thing, claiming she did not need a radiologist's input for clinical decision making, but as a senior resident, I realize it's all nonsense. Surgeons can read some imaging, and many times, their knowledge of the patient, clinical exam and labs enhance their localization and interpretation of the findings. However their skills are limited and generally inferior to radiologists who, many at times, read imaging with limited clinical information. Surgeons also generally review imaging with tunnel vision. I have a lot of evidence to back this up.I have called the surgery team with critical findings which they missed. That said, subspecialty/niche surgeons, e.g. vascular/ENT will probably read vascular imaging or temporal bone CT better than a general rads, but the general radiologist will read everything else at a higher level. I helped a cardiac surgeon review a cardiac MRI and chest/cardiac CT before he took the patient to surgery, and I am not even a cardiovascular radiologist.

We also read imaging with complications and mishaps by surgeons and other physicians. We all make mistakes (this includes surgeons, no matter what they will tell you). It's all hubris, to be honest. Folks like to harp on radiology misses, because our mistakes and bad calls are documented and signed.

Most radiology residents will pit their CT skills against any surgeon anytime. It's ridiculous that senior radiology residents (who have a lot more experience reading imaging than surgery residents) still read imaging with some healthy trepidation, yet some arrogant surgery resident claims he is better. Just ridiculous.

Your chief resident, as he advances in his career, is free to ignore and not look at the radiology report. He will ultimately find himself in court.

If they really did not need you they would not swarm around you every trauma pan scan.

 

[Mo991]

I'm a 3rd year radiology resident and just want to give my 2 cents
For the AI, I think it may impact the field but in a way we cant imagine. Separating the normal from the abnomal is a very hard task, one that humans find logical and easy.
For example, an AI can be trained enough to identify every item in an image (its hard but in theory can be done) but cant tell you what is worng in an image: it may spot the snowman and the shining sun but cant tell you that the snow would melt thereby something is wrong in the picture.
I think the current AI maximum ability is to generate pages and pages of thousands of questions and answers (like: aorta normal in size: yes, aorta atheroma: yes, SMI normal in siza: yes etc) and even then it can't handle comparaison with prior, post op status..
Radiology in particular is very vast, not only we deal with every part of the body, but we deal with life itself (iatrogenic injury, anatomic variants, every single pathology etc)
What is also important to know is that medicine is far from perfect (screening/treatment), therefore an AI might have a role in helping the radiologist making medicine one step.closer to perfect.
One last thing, 5 years ago they said that we should stop training radiologist because in 5 years (today) they would have no role

Here we are, with no proven app for Ai in radiology and not a single radiologist fired because of Ai.

(Sorry for the bad english, I am french educated)