List of least in demand specialites?

[LebronManning]

---

Members don't see this ad.

Will CT Surgery fall into this category?

 

[LoGo]

The hype of AI taking over radiology is primarily driven by computer scientists and marketing departments; it's quite telling that almost no radiologist who actually understands the workflow thinks that this will come to pass anytime in this lifetime (in the distant future? Perhaps, but only at the point where many other careers have already been taken over). Non-radiologists somehow like to reduce the field to mere pattern-recognition and seem to be ignorant of the cognitive requirements of the field.

I mean, what do I care? I'm on the procedural side of things in IR, so I have no qualms about job security anyway. But having gone through residency, I know enough about the field that I realize that most of these computer scientists proclaiming the imminent demise of radiologists are blowing smoke. It's too bad that much of the laypeople are taken in by the hype.

I’d argue it’s not the computer scientists and engineers that perpetuate the hype - they know better than most that AI has its shortcomings and limitations and I haven’t seen them explicitly state AI will take over a branch of medicine. On the contrary, they seem to advocate its use as a clinical decision aid.

It is doctors in other disciplines who perpetuate this - JAMA editors are notorious for this, along with a few MD social media influencers. They spill the propaganda that AI is coming for everyone’s job to get articles read and picked up by media, when in fact 99.9% of doctors could not begin to explain what a neural network is actually doing.

There’s an old idea that knowing a little about something is worse than knowing nothing at all - it gives a false sense of thinking you know what the hell you’re talking about. This is rampant in AI right now, and has always been the case among physicians who think they know about other specialties because they did a 2 week clerkship in it 15 years ago.

 

[libertyyne]

I’ve never understood why people think computers will make radiology obsolete anytime in the foreseeable future. We’ve had computers reading ECG’s since the 1970’s, yet we still pay cardiologists to read the ECG’s.

because AI hasnt been reading ECGs.

The point is that even if computers augment radiologists at the current number of scans each radiologist will become 10-X times more efficient reducing demand of radiologists total number of radiologists by that amount.

Can there be a new imaging modality with explosion of required reads? Can there be more ordering of imaging per capita to offset efficiency gains? Unlikely to both of those questions.

 

[Sircalvin]

Interesting to read this as a radiologist 14 years into practice. Some thoughts on radiology: the folks who say AI will take this job in 10 years are unlikely to be correct. Yes, I am sure we can train computers to "see things", but can it then take the clinical information, old studies and ancillary findings to give a reasonable diagnosis and help formulate a treatment plan in a reasonable amount of time? Then call on critical results and discuss cases with clinicians? And do this across xray, CT, MR, US, nuclear medicine and mammography? Unlikely in my lifetime. People who offhandedly say AI will do it no problem have never actively worked in this field. It is not just seeing the abnormality, that's the tip of the iceberg. Not to mention the hands-on procedures that rads do now because hospitalists and primary docs and, sadly, even many surgeons don't like to stick needles in anything anymore, etc

But on the flip side, I say continue to bring all this speculation on. Always choose something you like. But perhaps now the better rad programs will be easier to get into because of all this angst, making other medical students and residents turn away from the field. And right now the job market is hot - we haven't been able to recruit for an open position for months. Yes, in around 2010-2014 it was bad. But not out of work, can't find anything bad. I mean, rads on a bad day make more than most physicians. I worked in teleradiology from home for a while and still made more than my brother in IM with all the issues he deals with.

Finally, people always seem to look at AI as an all or none. I hope it gets to the point of prescreening stuff that I am reading and can analyze in real time, help me focus on things that need to be looked at again. But I don't see it completely replacing what I do anytime soon. If it does, well then I'll just go into IT and help fix it when it breaks down 🙂

 

[Sircalvin]

You Sir, are the smartest guy in the room! and I know- I a dual- fellowship trained Neuro, and Cardiac radiologist with 35 years experience and what you say is spot on!!

 

[bearsfan95]

Interesting to read this as a radiologist 14 years into practice. Some thoughts on radiology: the folks who say AI will take this job in 10 years are unlikely to be correct. Yes, I am sure we can train computers to "see things", but can it then take the clinical information, old studies and ancillary findings to give a reasonable diagnosis and help formulate a treatment plan in a reasonable amount of time? Then call on critical results and discuss cases with clinicians? And do this across xray, CT, MR, US, nuclear medicine and mammography? Unlikely in my lifetime. People who offhandedly say AI will do it no problem have never actively worked in this field. It is not just seeing the abnormality, that's the tip of the iceberg. Not to mention the hands-on procedures that rads do now because hospitalists and primary docs and, sadly, even many surgeons don't like to stick needles in anything anymore, etc

But on the flip side, I say continue to bring all this speculation on. Always choose something you like. But perhaps now the better rad programs will be easier to get into because of all this angst, making other medical students and residents turn away from the field. And right now the job market is hot - we haven't been able to recruit for an open position for months. Yes, in around 2010-2014 it was bad. But not out of work, can't find anything bad. I mean, rads on a bad day make more than most physicians. I worked in teleradiology from home for a while and still made more than my brother in IM with all the issues he deals with.

Finally, people always seem to look at AI as an all or none. I hope it gets to the point of prescreening stuff that I am reading and can analyze in real time, help me focus on things that need to be looked at again. But I don't see it completely replacing what I do anytime soon. If it does, well then I'll just go into IT and help fix it when it breaks down 🙂

This nails it... I've made several architecturally different "image-reading" Convolutional Neural Networks and understand well ML capabilities. To "automate" a radiologist, and all their many clinical duties (which, as said above is beyond just spotting lesions) will be unlikely for a long, long time. The low-hanging fruit in radiology (and the art of diagnosis in general) is trained architectures/Neural Networks that help fill in holes due to cognitive biases and lapses of human interpretation -- as well as simple task automation, EMR's, etc. This will make all 'diagnosticians' better at their job, and this is good for our patients and the world. We want this change... We want Machine Learning to become a part of the medical field.

To automate the entirety of radiology would require a beyond enormous and comprehensive data set, covering everything from the most basic pathologies to the zebra's, represented by patients of all ages, shapes and sizes, encompassing PET, MRI, CT etc. (think an entirely new and different machine learning algorithm for every type of scan), that also has communicative abilities between other providers. This will take years and years to accomplish... So while never say never (regarding computers automating our professions... As one day this will happen to ALL of us), the "threat" of automation by Machine Learning alone should not scare one away from a career in radiology at this time... One should go in excited however to work increasingly with new, dynamic assistive technology that will help deliver better and better diagnostic care as time goes on