What we need is a mandate for AI transformation of Radiology: Radiologists must be required to use AI every day on X% of scans, their productivity must double with the use of AI or they'll get fired, etc. To quote CEOs everywhere: AI is a transformative technology unlike any we've ever seen in our careers, and we must embrace it in a desperate FOMO way, anything else is unacceptable.
I wouldn't trust a non-radiologist to safely interpret the results of an AI model for radiology, no matter how well that model performs in benchmarks.
Similar to how a model that can do "PhD-level research" is of little use to me if I don't have my own PhD in the topic area it's researching for me, because how am I supposed to analyze a 20 page research report and figure out if it's credible or not?
Definitely an aggressive timeline but it seems like the biggest barrier to AI taking over radiology will be legal. Spending years training for a job which only continues to exist because of government fiat, which could change at any time, seems like a risky choice.
Every use of AI has its own problem of "person with 10 fingers" that AI image generation faces and can't seem to solve. For programmers, it's code that calls made up libraries and makes up language semantics. In prose, it's completely incoherent narratives that forget where they are going halfway through. For programmers it's making up case law and citations. Same for scientists, making up authorities and papers and results.
AI art is getting better but still it's very easy for me to quickly distinguish AI result from everything else, because I can visually inspect the artifacts and it's usually not very subtle.
I'm not a radiologist, but I would imagine AI is doing the same thing here, making up things that are cancer, missing things that aren't cancer, and it takes an expert to distinguish the false positives from true. So we're back at square one, except the expertise has shifted from interpreting the image to interpreting the image and also interpreting the AI.
All of the examples you gave (which I agree with, btw!), are generative AI, whereas I assume radiology would benefit more from the Machine Learning (ML), image in -> black-box ML decides whether it matches pattern -> verdict out, type of AI.
I suppose first of all, is that generally agreed? People aren't expecting a LLM to give a radiology opinion, the same as way that you can feed in a PDF or an image into ChatGPT and ask it something about it, are they?
I'm interested whether most people here have a higher opinion of ML than of the generative AIs, in terms of giving a reliably useful output. Or do a lot of you think that these also just create so much checking it would be easier to just have a human do the original work?
I think it's probably worth excluding self-driving from my above question, since that is a particularly difficult area to agree anything on.
As a doctor and full stack engineer, I would never go into radiology or seek further training in it. (obviously)
AI is going to augment radiologists first, and eventually, it will start to replace them. And existing radiologists will transition into stuff like interventional radiology or whatever new areas will come into the picture in the future.
I could see that as more radiology AI tools become available to non-radiologist medical providers, they might choose to leverage the quick feedback those provide and not wait for a radiologist to weight in, even if they could gain something from the radiologist. They could make a decision while the patient is still in the room with them.
> Three things explain this. First,... Second, attempts to give models more tasks have run into legal hurdles: regulators and medical insurers so far are reluctant to approve or cover fully autonomous radiology models. Third, even when they do diagnose accurately, models replace only a small share of a radiologist’s job. Human radiologists spend a minority of their time on diagnostics and the majority on other activities, like talking to patients and fellow clinicians.
Everything else besides the above in TFA is extraneous. Machine learning models could have absolute perfect performance at zero cost, and the above would make it so that radiologists are not going to be "replaced" by ML models anytime soon.
I find the radiologist use case an illuminating one for the adoption of AI across business today. My takeaway is that when the tools get better, radiologists aren't replaced, but take up other important tasks that sometimes become second nature when reads (unassisted) are the primary goal.
In particular, doctors appear to defer excessively to assistive AI tools in clinical settings in a way that they do not in lab settings. They did this even with much more primitive tools than we have today... The gap was largest when computer aids failed to recognize the malignancy itself; many doctors seemed to treat an absence of prompts as reassurance that a film was clean
Reminds me of the "slop" discussions happening right now. When the tools seem good, but aren't, we develop a reliance to false negatives, e.g. text that clearly "feels" written by a GPT model.
In May earlier this year, the New York Times had a similar article about AI not replacing radiologists:
https://archive.is/cw1Zt
It has similar insights, and good comments from doctors and from Hinton:
“It can augment, assist and quantify, but I am not in a place where I give up interpretive conclusions to the technology.”
“Five years from now, it will be malpractice not to use A.I.,” he said. “But it will be humans and A.I. working together.”
Dr. Hinton agrees. In retrospect, he believes he spoke too broadly in 2016, he said in an email. He didn’t make clear that he was speaking purely about image analysis, and was wrong on timing but not the direction, he added.
Oh yes it is. I have worked on projects where highly trained specialized doctors have helped train the models (or trained them themselves) to catch random very difficult to notice conditions via radiology. Some of these systems are deployed at different hospitals and medical facilities around the country. The radiologist still does there job, but some odd, random hard to notice conditions, AI is a literal life saver. For example, pancreas divisum, a abnormality in the way the pancreas ducts fail to fuse/etc can cause all kinds of insane issues. But its not something most people know about or look for. AI can pick that up in a second. It can then alert the radiologist of an abnormality and they can then verify. It's enhacing the capabilties of radiologists.
> Some of these systems are deployed at different hospitals and medical facilities around the country. The radiologist still does there job, but some odd, random hard to notice conditions, AI is a literal life saver
I would be very interested if you could provide specific examples.
This article is pretty good. My current work is transitioning CV models in a large, local hospital system to a more unified deployment system, and much of the content aligns with conversations we have with providers, operations, etc..
I think the part that says models will reduce time to complete tasks and allow providers to focus on other tasks is on point in particular. For one CV task, we’re only saving on average <30min of work per study, so it isn’t massive savings from a provider’s perspective. But scaled across the whole hospital, it’s huge savings
One thing people aren't talking about is liability.
At the end of the day if the radiologist makes an error the radiologist gets sued.
If AI replaces the radiologist then it is OpenAI or some other AI company that will get sued each and every time the AI model makes a mistake. No AI company wants to be on the hook for that.
So what will happen? Simple. AI will always remain just a tool to assist doctors. But there will always be a disclaimer attached to the output saying that ultimately the radiologist should use his or her judgement. And then the liability would remain with the human not the AI company.
Maybe AI will "replace" radiologists in very poor countries where people may not have had access to radiologists in the first place. In some places in the world it is cheap to get an xray but still can be expensive to pay someone to interpret it. But in the United States the fear of malpractice will mean radiologists never go away.
EDIT: I know the article mentions liability but it mentions it as just one reason among many. My contention is that liability will be the fundamental reason radiologists are never replaced regardless of how good the AI systems get. This applies to other specialities too.
So instead of having to train and employ radiologists, we will train and employ radiologists and pay for the AI inference. Excuse me, but how is this beneficial in any way? It's trivially more expensive, and the result has the same quality? And productivity is also the same?
When Tesla demoed (via video) self-driving in 2016 with a claim "The person in the driver’s seat is only there for legal reasons. He is not doing anything. The car is driving itself" and then when they unveiled Semi in 2017 - I tweeted out and honestly thought that trucking industry is changed forever and it doesn't make sense to be starting in trucking industry. It's almost end of 2025 and either nothing out of it or just a small part of it panned out.
I think we all have become hyper-optimistic on technology. We want this tech to work and we want it to change the world in some fundamental way, but either things are moving very slowly or not at all.
Things happen slowly, then all at once. Many people think ChatGPT appeared out of nowhere a couple of years ago. In reality it was steadily improving for 8 years. Before then, LLMs were being developed for Word2Vec. Before then, Yoshua Bengio and colleagues proposed the first neural probabilistic language model, and introducing distributed word representations (precursors to embeddings). Before then we had Statistical NLP took hold, with n-gram models, hidden Markov models, and later phrase-based machine translation. Before that we had work on natural language processing (NLP) which began with symbolic AI and rule-based systems (e.g., ELIZA, 1966).
These are all stepping stones, and eventually the technology is mature enough to productise. You would be shocked by how good Tesla FSD is right now. It can easily take you on a cross country trip with almost zero human interactions.
>"they struggle to replicate this performance in hospital conditions"
Are there systematic reasons why radiologists in hospitals are inaccurately assessing the AI's output? If the AI models are better than humans in testing novel data then, well, the thing that has changed in a hospital situation compared to the AI-Human testing environment is not the AI, it is the human, under less controlled constraints, additional pressures, workloads, etc. Perhaps the AI's aren't performing as poorly as thought. Perhaps this is why they performed better to begin with. Otherwise, production ML systems are generally not as highly regarded as these models when they perform as significantly below test data sets in production. Some is expected, but "struggle to replicate" implies more.
>"Most tools can only diagnose abnormalities that are common in training data"
Well yes, training on novel examples is one thing. Training on something categorically different is another thing all together. Also there are thresholds of detection. Detecting nothing, or with a a lower confidence, or unknown anomaly, false positive, etc. How much of the inaccuracy isn't wrong, but simply something that is amended or expanded upon when reviewed? Some details here would be useful.
I'm highly skeptical when generalized statements exclude directly relevant information to which an is referring. The few sources provided don't at all cover model accuracy, and the primary factor cited as problematic with AI review, lack of diversity in study composition for women, ethnic variation, children, links to a a meta study that was not at all related to the composition of models and their training data sets.
The article begins as what appears to be a criticism of AI accuracy with the thinness outlined above but then quickly moves on to a "but that's not what radiologists do anyway", and provides a categorical % breakdown of time spent where Personal/Meetings/Meals and some mixture of the others combine to form at least a third that could be categorized as "Time where the human isn't necessary if graphs are being interpreted by models."
I'm not saying there aren't points here, but overall, it simply sounds like the hand-wavy meandering of someone trying to gatekeep a profession whose services could be massively more utilized with more automation, and sure-- perhaps at even higher quality with more radiologists to boot-- but perfect is the enemy of the good etc. on that score, with enormous costs and delays in service in the meantime.
It's interesting to see people fighting so hard to preserve these jobs. Do people want to work that badly? If a magic wand can do everything radiologists can do, would we embrace it or invent reasons to occupy 40+ hours a week of time anyway? If a magic wand, might be on the horizon, shouldn't we all be fighting to find it and even finding ways to tweak our behaviors to maximize the amount of free time that could be generated?
While a lot of this rings true, I think the analysis is skewed towards academic radiology. In private practice, everything is optimized for throughput, so the idea that most rads spend less than half of their time reading studies i think is probably way off.
The only part of this article I believe is the legal and bureaucratic burdens part.
"Human radiologists spend a minority of their time on diagnostics and the majority on other activities, like talking to patients and fellow clinicians"
I've had the misfortune of dealing with a radiologist or two this year. They spent 10-20 minutes talking about the imaging and the results with me. What they said was very superficial and they didn't have answers to several of the questions I asked.
I went over the images and pathology reports with ChatGPT and it was much better informed, did have answers for my questions, and had additional questions I should have been asking. I've used ChatGPT's information on the rare occasions when doctors deign to speak with me and it's always been right. Me, repeating conclusions and observations ChatGPT made, to my doctors, has twice changed the course of my treatment this year, and the doctors have never said anything I've learned from ChatGPT is wrong. By contrast, my doctors are often wrong, forgetful, or mistaken. I trust ChatGPT way more than them.
Good image recognition models probably are much better than human radiologists already and certainly could be vastly better. One obstacle this post mentions - AI models "struggle to replicate this performance in hospital conditions", is purely a choice. If HMOs trained models on real data then this would no longer be the case, if it is now, which I doubt.
I think it's pretty clearly doctors, and their various bureaucratic and legal allies, defending their legal monopoly so they can provide worse and slower healthcare at higher prices, so they continue to make money, at the small cost of the sick getting worse and dying.
As the prediction of radiologists going dodo sprung out from improvements in image recognition, why don't we see premature and hysterically hyped predictions of psychiatrists being unemployed due to language models.
Their workday consists of conversations, questions and reading. Something LLMs more than excel at doing, tirelessly and in huge volumes.
And if radiologists are still the top bet due to image recognition being so much hotter, then why not add dermatologists to the extinction roster? They only ever look at regular light images, it should be a lower hanging fruit.
(I'm aware of the nuances that make automation of these work roles hard, I'm just trying to shine some light on the mystery of radiologists being perceived as the perennial easy target)
The best story I heard about machine learning and radiology was when folks were racing to try to detect COVID in lung X-rays.
As I recall, one group had fairly good success, but eventually someone figured out that their data set had images from a low-COVID hospital and a high-COVID hospital, and the lettering on the images used different fonts. The ML model was detecting the font, not the COVID.
If you're not at a university, try searching for "AI for radiographic COVID-19 detection selects shortcuts over signal" and you'll probably be able to find an open-access copy.
The thing with medical services is that there is never enough.
If you are rich and care about your health (especially as move past age 40), you probably have use for a physiotherapist, a nutritionist, therapist, regular blood analysis, comprehensive cardio screening, comprehensive cancer screening etc. Arguably, there is no limit to the amount of medical services that people could use if they were cheap and accessible enough.
Even if AI tools add 1-2% on the diagnostic side every year, it will take a very, very long time to catch up to demand.
72 comments
[ 2.5 ms ] story [ 65.6 ms ] threadSimilar to how a model that can do "PhD-level research" is of little use to me if I don't have my own PhD in the topic area it's researching for me, because how am I supposed to analyze a 20 page research report and figure out if it's credible or not?
If we had followed every AI evengelist sugestion the world would have collapsed.
AI art is getting better but still it's very easy for me to quickly distinguish AI result from everything else, because I can visually inspect the artifacts and it's usually not very subtle.
I'm not a radiologist, but I would imagine AI is doing the same thing here, making up things that are cancer, missing things that aren't cancer, and it takes an expert to distinguish the false positives from true. So we're back at square one, except the expertise has shifted from interpreting the image to interpreting the image and also interpreting the AI.
I suppose first of all, is that generally agreed? People aren't expecting a LLM to give a radiology opinion, the same as way that you can feed in a PDF or an image into ChatGPT and ask it something about it, are they?
I'm interested whether most people here have a higher opinion of ML than of the generative AIs, in terms of giving a reliably useful output. Or do a lot of you think that these also just create so much checking it would be easier to just have a human do the original work?
I think it's probably worth excluding self-driving from my above question, since that is a particularly difficult area to agree anything on.
AI is going to augment radiologists first, and eventually, it will start to replace them. And existing radiologists will transition into stuff like interventional radiology or whatever new areas will come into the picture in the future.
Everything else besides the above in TFA is extraneous. Machine learning models could have absolute perfect performance at zero cost, and the above would make it so that radiologists are not going to be "replaced" by ML models anytime soon.
It has similar insights, and good comments from doctors and from Hinton:
“It can augment, assist and quantify, but I am not in a place where I give up interpretive conclusions to the technology.”
“Five years from now, it will be malpractice not to use A.I.,” he said. “But it will be humans and A.I. working together.”
Dr. Hinton agrees. In retrospect, he believes he spoke too broadly in 2016, he said in an email. He didn’t make clear that he was speaking purely about image analysis, and was wrong on timing but not the direction, he added.
I would be very interested if you could provide specific examples.
I think the part that says models will reduce time to complete tasks and allow providers to focus on other tasks is on point in particular. For one CV task, we’re only saving on average <30min of work per study, so it isn’t massive savings from a provider’s perspective. But scaled across the whole hospital, it’s huge savings
At the end of the day if the radiologist makes an error the radiologist gets sued.
If AI replaces the radiologist then it is OpenAI or some other AI company that will get sued each and every time the AI model makes a mistake. No AI company wants to be on the hook for that.
So what will happen? Simple. AI will always remain just a tool to assist doctors. But there will always be a disclaimer attached to the output saying that ultimately the radiologist should use his or her judgement. And then the liability would remain with the human not the AI company.
Maybe AI will "replace" radiologists in very poor countries where people may not have had access to radiologists in the first place. In some places in the world it is cheap to get an xray but still can be expensive to pay someone to interpret it. But in the United States the fear of malpractice will mean radiologists never go away.
EDIT: I know the article mentions liability but it mentions it as just one reason among many. My contention is that liability will be the fundamental reason radiologists are never replaced regardless of how good the AI systems get. This applies to other specialities too.
I think we all have become hyper-optimistic on technology. We want this tech to work and we want it to change the world in some fundamental way, but either things are moving very slowly or not at all.
These are all stepping stones, and eventually the technology is mature enough to productise. You would be shocked by how good Tesla FSD is right now. It can easily take you on a cross country trip with almost zero human interactions.
Still true as work conditions are harsh, schedule as well, responsibilities and fines are high but payment is not.
https://www.outofpocket.health/p/why-radiology-ai-didnt-work...
Are there systematic reasons why radiologists in hospitals are inaccurately assessing the AI's output? If the AI models are better than humans in testing novel data then, well, the thing that has changed in a hospital situation compared to the AI-Human testing environment is not the AI, it is the human, under less controlled constraints, additional pressures, workloads, etc. Perhaps the AI's aren't performing as poorly as thought. Perhaps this is why they performed better to begin with. Otherwise, production ML systems are generally not as highly regarded as these models when they perform as significantly below test data sets in production. Some is expected, but "struggle to replicate" implies more.
>"Most tools can only diagnose abnormalities that are common in training data"
Well yes, training on novel examples is one thing. Training on something categorically different is another thing all together. Also there are thresholds of detection. Detecting nothing, or with a a lower confidence, or unknown anomaly, false positive, etc. How much of the inaccuracy isn't wrong, but simply something that is amended or expanded upon when reviewed? Some details here would be useful.
I'm highly skeptical when generalized statements exclude directly relevant information to which an is referring. The few sources provided don't at all cover model accuracy, and the primary factor cited as problematic with AI review, lack of diversity in study composition for women, ethnic variation, children, links to a a meta study that was not at all related to the composition of models and their training data sets.
The article begins as what appears to be a criticism of AI accuracy with the thinness outlined above but then quickly moves on to a "but that's not what radiologists do anyway", and provides a categorical % breakdown of time spent where Personal/Meetings/Meals and some mixture of the others combine to form at least a third that could be categorized as "Time where the human isn't necessary if graphs are being interpreted by models."
I'm not saying there aren't points here, but overall, it simply sounds like the hand-wavy meandering of someone trying to gatekeep a profession whose services could be massively more utilized with more automation, and sure-- perhaps at even higher quality with more radiologists to boot-- but perfect is the enemy of the good etc. on that score, with enormous costs and delays in service in the meantime.
"Human radiologists spend a minority of their time on diagnostics and the majority on other activities, like talking to patients and fellow clinicians"
I've had the misfortune of dealing with a radiologist or two this year. They spent 10-20 minutes talking about the imaging and the results with me. What they said was very superficial and they didn't have answers to several of the questions I asked.
I went over the images and pathology reports with ChatGPT and it was much better informed, did have answers for my questions, and had additional questions I should have been asking. I've used ChatGPT's information on the rare occasions when doctors deign to speak with me and it's always been right. Me, repeating conclusions and observations ChatGPT made, to my doctors, has twice changed the course of my treatment this year, and the doctors have never said anything I've learned from ChatGPT is wrong. By contrast, my doctors are often wrong, forgetful, or mistaken. I trust ChatGPT way more than them.
Good image recognition models probably are much better than human radiologists already and certainly could be vastly better. One obstacle this post mentions - AI models "struggle to replicate this performance in hospital conditions", is purely a choice. If HMOs trained models on real data then this would no longer be the case, if it is now, which I doubt.
I think it's pretty clearly doctors, and their various bureaucratic and legal allies, defending their legal monopoly so they can provide worse and slower healthcare at higher prices, so they continue to make money, at the small cost of the sick getting worse and dying.
Their workday consists of conversations, questions and reading. Something LLMs more than excel at doing, tirelessly and in huge volumes.
And if radiologists are still the top bet due to image recognition being so much hotter, then why not add dermatologists to the extinction roster? They only ever look at regular light images, it should be a lower hanging fruit.
(I'm aware of the nuances that make automation of these work roles hard, I'm just trying to shine some light on the mystery of radiologists being perceived as the perennial easy target)
As I recall, one group had fairly good success, but eventually someone figured out that their data set had images from a low-COVID hospital and a high-COVID hospital, and the lettering on the images used different fonts. The ML model was detecting the font, not the COVID.
[a bit of googling later...]
Here's a link to what I think was the debunking study: https://www.nature.com/articles/s42256-021-00338-7
If you're not at a university, try searching for "AI for radiographic COVID-19 detection selects shortcuts over signal" and you'll probably be able to find an open-access copy.
If you are rich and care about your health (especially as move past age 40), you probably have use for a physiotherapist, a nutritionist, therapist, regular blood analysis, comprehensive cardio screening, comprehensive cancer screening etc. Arguably, there is no limit to the amount of medical services that people could use if they were cheap and accessible enough.
Even if AI tools add 1-2% on the diagnostic side every year, it will take a very, very long time to catch up to demand.
Programming is the first job AI will replace. The rest come later.