you say AI might replace radiologists in 20 years, meaning you think it is possible, just not in the near future but I'm saying it will happen a lot sooner than you think, that is unless you have some intervention on the legislative side stopping AI from taking over, also whenever you ask a radiologist this question, they say you underestimate/ don't know what radiologists actually do, yet in the same breath hugely underestimate what AI is capable of, the thing is radiology is the perfect field for AI to thrive in, we can disagree on the time frame of when it'll happen, but everyone knows that it will happen.
I have been in practice in general radiology since completing my fellowship for 32 years. One thing that attracted me was the idea of being a doctor's doctor, a consultant to other physicians trying to solve problems. And sure enough doctors would come to the reading room all day and interactive problem solving would occur. Non radiologists would have a limited idea of how to view the images. Fast forward to now. The images are instantly available everywhere and a wide variety of physicians can interpret what's in front of them. The gap between what we know and what our physician customer's know has been reduced at least 80%. That erases much of the relative cognitive value that we can deliver. I have directly experienced how we have evolved from consultants into high paid factory workers who make measurements and generate reports. Modern radiologists and modern clinicians think that radiologists and clinicians are not supposed to interact, to the point where when a clinician rarely calls me to discuss a case the first phrase they utter is "I'm sorry to bother you...". I am just the report, that's it (not withstanding intervention). Over this period of time typical volume went from 13,000 studies per year to 22,000 and that's without considering the explosion in the number of images. Burnout is far more than the volume, it's the social isolation that comes with the IT revolution and the erasure of interaction and erasure of that higher consultative value. While EMR brings us an explosion in the quantity of patient information, the quality is greatly degraded, hence a smaller net gain. "Will AI replace us?" is a black and white question. AI will compliment what we do and increasingly free us up from the mundane factory work of clicking and measuring and generate reports all day, the factory work. It will in a certain way free us up and elevate us back up to the status of being consultants who are able to integrate the imaging into the entire patient situation. As AI evolves it will take over more and more of what we are used to doing in the current environment. I don't know exactly what radiology will look like, except to say it's going to look wildly different and in some way closer to the paradigm that I started out with. The thought leaders are going to need to redesign the training to return radiologist into being doctors who know how to integrate themselves back into the care team. That we need AI just to deal with the increasing avalanche of studies is a great point. It's hard to predict whether we will need more radiologist or less, just that it's going to be more of an expert niche integrated into the health care environment and less of a high paid factory worker. I no little of business but I recall a story about how the proliferation of ATM's was predicted to be the demise of bank branches. Instead the early impact was a proliferation of bank branches as resources were freed up to build out branches. So who knows. Sincerely ronald.garrell@gmail.com, Weston, MA.
If clicking a box that says “I am not a robot” on websites is all you need to stump AI, there’s a long way to go. Also who gets sued if there’s a miss? Years away in my opinion. Maybe for simple things at some point but nowhere near being able to do complex tumor follow up cases, etc.
@@Chris-ts9ok one of the main values of AI is to relief you of repetitive tasks of scroll, click, measure, scroll, click, measure so that you have time to do exactly what you say, evaluate the broader implications of complex situations. As an AI assistant you are feed to engage in the higher value and higher cognitive tasks.
Thank you! Please do share with those who think AI will replace us. I honestly can't wait to get a helping hand from AI to conquer the insane worklists.
ECG/EKG machines have been around for the longest time and we still dont take the machine diagnosis as the final answer but still wait for the cardiologist to read it. And thats just lines on a piece of paper. With AI its gonna take alot longer. Calculators/computers have not replaced mathmaticians or physicists
Nice presentation, Dr. Gupta. I was using an AI for helping me interpret mpMRI prostate. It was great. The company charges $25 per case fee to use it. Guess what? That is 50% of my reading fee, so the bean-counters said "No, you can't use it anymore, it's too expensive." All my 30 year career I have wanted a program to circle, measure, and characterize every potential pulmonary nodule. Seems simple enough, but nobody will implement it (liability fears=cost, I think). Maybe soon.
Thank you for this, I’m new to radiology and I love it but I hate these talks that are going around and so I really hope you’re right about everything 🥲🫶🏻
It is clear for every person with a bit of clear sense that AI will not REPLACE radiology , but more of a question is if the need of radiologist will decrease ! Cause if that happens, you will have more difficulties to find a place to work . Definitely for a radiologist who’s in work for over 20years this might not be a concern cause if this should happen he’ll be close to retire, but for young students who aren’t into the specialty yet this might be a thing they could think about .
hello Yasha Gupta, iam radiologist from indonesia, i really excited with AI , and i think its imposible replace rdiologist, it maybe some time in the future better at one or more things but for everything of course it cant replace ours, we better than it. it same like autonom car like tesla or maybe google autonom car, it can savely drive but it never ever get car licence, becuse it not human, so i think it same case, even it can diagnose accurately the deseases it not human so it never get that license. and for the your closing statement, i hope many radiologist have same oppinion like you, that AI can assist our work, it really helpfull. reduce our burnout...and help many thing. Please suport AI for this, and never afraid we will lost our job because it.
This analogy is exactly wrong. Self-driving cars have issues cause they have constantly make decisions in real-time within a highly non-standardized environment. This is absolutely not the case in medical diagnosis. You have time and there is sooo much standardized that is not influenced by sudden events you cant predict at all. What will happen is that ALL radiologists will work with what Machine-Learning-Systems are diagnosing and observe it and will step in if there is a case that is too complex or too rare (non-standardized). And the systems will get better and better and better and better... and will diagnose the standard stuff faster and cheaper. In the end the amount of observation and step in by human will go down and down and down. Of course most jobs wont be completely canceled by MachineLearning. But they will replace A LOOOOOOOT human in all possible white collar working areas. Especially where only brain function of human is needed. What is much harder is creating robots that have to do complex behaviour in real time. But brain work that is not time critical... will be taken over more and more by machines.
Radiology is not just CT /MRI Its also about Ultrasound/vascular and non vascular interventions/clinical meetings/ advice to others teams/ discussions and most importantly real life & past patients experiences
People who think AI is replacing radiology is because they do not understand what radiologist do, or maybe they do not understand how AI work, (or maybe is because of both reasons); very good video btw Dr. Gupta, I love your content!
you could have saved a lot of words, by just stating that AI in radiology, as in many fields, will remain an adjunct to diagnosis for the foreseeable future. And so radiology will adapt as AI improves and so will the medical field as a whole.
Very well said I see people saying this sort of nonsense that radiologists are about to be replaced by AI. I keep thinking all of these things I have commented as such on occasion as well. As far as mid level staff in radiology goes, they actually have a program for just that. And you can become a Registered Radiologist Assistant through the ARRT but it makes no sense to do it. I want to say it is a masters degree with a whole lot of extra CEU’s so you can bake less than $40,000 a year. So most people don’t bother, I had considered it until I realized it was nothing like being a PA and if you want to go down that type of route it is far better to switch to a PA.
Im a doctor and I see the power if Silicon,if software can differentiate iris patterns they can also differentiate Your xrays and CTs and make a ball park diagnosis to help radiologists fine tune their interpretation.Hand slills on the other hand will never be replaced
You might eat your words on that. We already have AI operated robots that can do dental procedures. There's an article released on it recently, you can google it
Respectfully, I have been in Radiology / Pathology information systems for the last 15 years. Consulted for medical AI firms as well. Generative AI and Moore's law is unfortunately very real. If you now look at procedural and image generation via AI today, the exponentially of AI is very apparent. Inference is at a completely different level than it was say 3 years ago. I am very neutral with regards to replacing Human Radiologists, BUT, the effectiveness and accuracy and speed of diagnosis across all modalities will come faster than you expect.
Blinders and cognitive dissonance. It’s hard to accept that one’s own field theyve worked so hard on is in jeopardy to AI. I get it. But it will happen. Not for like 20-30 years but it will for sure happen. #RemindMe in 30 years
I believe Professor Geoffrey Hinton's statement and especially the debate that came after it demonstrate how radiologists are undervalued by other professionals. Nobody wants to travel in a plane without a pilot for sure.
Hey yasha great video. A very nuanced and detailed analysis. I am just curious about something totally unrelated. You are a 2nd gen indian american. Maybe you could do a video with you and your dad who i know is a radiologist too and i am assuming 1st gen and talk about your feelings growing up with this identity and your dad on seeing you grow up to eventually pursue the same field as you. Its just a request. Sorry in advance if the request is too personal!
I understand it may not outright replace it. However, do you think the increased efficiency with AI may decrease the demand for radiologist making it much harder to get into? (More so 10 years in the future)
Yes I do, right now we are in a radiologist CRISIS where there are not enough radiologists to go around. A major shortage. AI can reduce the work burden on each radiologist which may reduce demand- but I only think this is a good thing for patient care. The way it is right now is not sustainable.
@@YashaGuptaMDwhy is there a crisis of not enough radiologists? At least for me, that seems like the most lucrative field of medicine to go into and it should attract more students.
AI could replace radiologists just as AI could eventually replace human beings, which I kind of think will happen but not for many many perhaps hundreds of years. So in short, I agree with you, for the current and next several generations, the risk of being replaced by AI is very very low. Furthermore, given that you currently can't even use AI to rub two sticks together as far as it relates to Radiology, should tell you we are way way far away from AI having a meaningful impact much less replace us alltogether.
Ai may also categorize, its not just yes/no questions, just so you know. For example it could go is there a tumor in this image and then IF yes, what type of tumor. One day it will be 'is there something abnormal in this image?"and then it'll tell you what. Doesnt mean it will replace radiologists but probably will get paid less and a lot less of them will be required 😶
AI will replace radiologists who don’t use AI. It’s all inside that sentence. We as radiologists faced already SEVERAL revolutions: contrast medium, CT, MRI, and PACS… and we are here to stay
In the evolving world of automation, AI is certain to alleviate the burden of repetitive, non-critical tasks, streamlining our daily workflows. However, it's essential to remember that there will always be room for individuals who can enhance and contribute value to these projects. Those who fear AI are either naive or ignorant. It’s here to stay and it’s here for better or worse. It’s the future. It is evolving, but so are we. The day AI completely replaces radiologists is the day when humans will prefer an AI doctor over a human doctor. Maybe that day will come someday. But are y’all really worried about that? 😂
@@fuzzfuzz4234 good question. This discussion was about the fear of AI replacing humans in this sector. If fear of AI replacing the radiologist is the reason not to join, that's ignorant.
Met doctors that seem completely clueless about nutrition. Also a doctor that didn’t diagnose a friends aunt lung cancer despite her going to him several times saying something is wrong. She ended up losing a lung.
I don't but you can. I made a video all about salary in radiology here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-PCF6cdQXJ4Y.htmlsi=4YgGDkTHTGAfkiLK
A good , experienced radiologist will do better than AI. I would rather AI read my films than a crappy radiologist. There are crappy doctors in every specialty.
Well ignorance is bliss. Diagnosing lesions is yesterday’s problem for AI. What we are developing is correlating EHR data with radiomics features, pathomics and genomics data and come up with prognosis and clinical decision making models.
She is in for a rude awakening, because it seems she dont know about AGI and how ai had already solved harder problems, radiology is really not even worth taking the job its so elementary
Sounds like you got no clue about what AI is. Universal lesion detection is already being implemented. One single AI model detecting multiple abnormalities
It is not just about making the most accurate diagnosis. If I’m diagnosed with cancer by the AI algorithm, and it may be 100 percent right, I will still run to a radiologist to cross check the scans before undergoing an invasive procedure like biopsy. If my biopsy comes out to be positive, i will not let chat gpt decide my chemo even if it spews the right algorithm. I will go to the best oncologist (preferably with years of experience under his belt) to discuss the chemo regimen. Similarly, disease patterns are also changing continually just like what happened in COVID. Just imagine COVID breaking out in a world with no radiologists and only AI. All the chest CT scans will be reported wrong. Bottom line: yes, AI will impact radiology job market, it will bring down compensation for sure but it will never replace them completely.
