Is Radiology a Dying Specialty?

[Piglet2020]

I want to prove to a nagging mother that radiologists wont become replaced by AI any time in the near future (at least in my generation).

Can someone post links to good sources that debunk claims that AI’s will somehow replace radiologists? Also she thinks physicians from India are gonna take over U.S. radiologist jobs (out-sourcing)... pretty sure only U.S. accredited MD physicians would have access to patient files though.

Edit: **Diagnostic Radiology**

---

Members don't see this ad.

 

[LizzyM]

If you haven't been admitted to medical school yet, why discuss this? You are, at the least, about 4 years away from interviewing for residency positions. At this point, just say, "Mom, you might be right. Time will tell." Don't give her a reason to nag you. You aren't going to be open a match day envelope until March 2022, so let things go for now.

 

[Piglet2020]

If you haven't been admitted to medical school yet, why discuss this? You are, at the least, about 4 years away from interviewing for residency positions. At this point, just say, "Mom, you might be right. Time will tell." Don't give her a reason to nag you. You aren't going to be open a match day envelope until March 2022, so let things go for now.

That’s true, but she’s so controlling that when I even suggest a specialty for which I might be interested in the future, she tries to shut it down.

Meh lesson of my life is just to not share any personal aspirations/goals with her in the future.

 

[Piglet2020]

I just want some honest opinions from radiologists out there about job prospects.

 

[WedgeDawg]

Tell her you want to be a dermatologic neurosurgeon who will work 30 hour weeks and have a starting salary of a quarter mil per year and only going up. Then, when reality sets in after you match into your specialty of choice, she can reevaluate her position.

 

[ndafife]

That’s true, but she’s so controlling that when I even suggest a specialty for which I might be interested in the future, she tries to shut it down.

Meh lesson of my life is just to not share any personal aspirations/goals with her in the future.

Well hopefully by the time you turn ~26 and start applying for residency you'll be able to make decisions for yourself

 

[Piglet2020]

Well hopefully by the time you turn ~26 and start applying for residency you'll be able to make decisions for yourself

That’s the dream

 

[CyrilFiggis]

You do both know that there are two different radiology specialties right?

 

[Med Ed]

I want to prove to a nagging mother that radiologists wont become replaced by AI any time in the near future (at least in my generation).

Can someone post links to good sources that debunk claims that AI’s will somehow replace radiologists? Also she thinks physicians from India are gonna take over U.S. radiologist jobs (out-sourcing)... pretty sure only U.S. accredited MD physicians would have access to patient files though.

Edit: **Diagnostic Radiology**

Outsourcing radiology has been talked about for about two decades. The basic problem is that no payer in the US is going to send money to unlicensed, non-US-trained docs in another country (e.g. India) for medical services. That's not going to change in the foreseeable future.

When it comes to AI, the tech companies can't even build a decent EMR system. It wasn't that long ago that everyone though personalized medicine was going to revolutionize care. Now we can actually sequence entire genomes rapidly and at low cost, but the revolution is yet to arrive. Meanwhile nobody saw the internet coming, even in 1994. About the only consistent feature of such prognostications is that they are consistently wrong.

 

[NotASerialKiller]

Don't tell her you want to be a pathologist, she'll start worrying about you being replaced by pigeons.

 

[sliceofbread136]

When we have an AI that can reliably read an ekg maybe then this will be something to worry about...

 

[21Rush12]

Tell her you want to be a psychiatrist and then ask her how she feels about that

A+

 

[kavorca]

I would be more worried about a T2 Skynet situation as opposed to them reading our MRs. I imagine the AI will be most useful in extracting more reads per minute from the humans.

 

[Planes2Doc]

AI will not replace radiologists. AI already does such an awesome job interpreting ECGs that you either require a cardiologist or emergency physician to read it over.

 

[Piglet2020]

Is precision as useful as we think? What will you do if your CT report says renal mass has a "0.8% chance of malignancy?" Sleep soundly? Remove the mass? Do follow-up imaging? Numbers are continuous but decision-making is dichotomous, and the final outcome is still singular. Is the human race prepared for all of this information?


The hallmark of intelligence is in reducing information to what is relevant. A dualism may emerge between artificial and real intelligence, where AI spits out information and radiologists contract information. Radiologists could be Sherlock Holmes to the untamed eagerness of Watson.


In the meantime, radiologists should ask which tasks need a medical degree. Surely, placing a caliper from one end of a lung nodule to the other end doesn't need 4 years of medical school and an internship. Then render unto AI what is AI's. Of all the people I spoke to for this article, Gregory Mogel, MD, a battle-hardened radiologist and chief of radiology at Central Valley Kaiser Permanente, said it best. "Any radiologist that can be replaced by a computer should be." Amen.

Medscape: Medscape Access

 

[LatentStudying]

Is precision as useful as we think? What will you do if your CT report says renal mass has a "0.8% chance of malignancy?" Sleep
In the meantime, radiologists should ask which tasks need a medical degree. Surely, placing a caliper from one end of a lung nodule to the other end doesn't need 4 years of medical school and an internship. Then render unto AI what is AI's. Of all the people I spoke to for this article, Gregory Mogel, MD, a battle-hardened radiologist and chief of radiology at Central Valley Kaiser Permanente, said it best. "Any radiologist that can be replaced by a computer should be." Amen.

\

But where do you draw the line? At what point do you say, "This is not for a model to decide"?

 

[calivianya]

\

But where do you draw the line? At what point do you say, "This is not for a model to decide"?

IMO, you just have to try it... and when AI fails, that's when someone has to admit to screwing up for letting AI direct treatment in the first place.

Happens all the time at my job with our protocols. We have very involved protocols related to glycemic managment, sepsis, etc. that require no physician oversight to implement interventions up to insulin drips, initiating vasopressors/inotropes, giving blood, etc. It never fails that some physicians goes, "Why the hell is XX receiving XX treatment?!" and the answer is always, "Because you ordered the protocol and that's what the protocol told us to do."

IMO, 99% of all protocols/AI driven medicine/whatever sucks. Every patient is different and using some sort of algorithm to predict outcomes just leads to lawsuits. It's almost always a bad idea to automate patient care.

 

[LatentStudying]

IMO, you just have to try it... and when AI fails, that's when someone has to admit to screwing up for letting AI direct treatment in the first place.

Happens all the time at my job with our protocols. We have very involved protocols related to glycemic managment, sepsis, etc. that require no physician oversight to implement interventions up to insulin drips, initiating vasopressors/inotropes, giving blood, etc. It never fails that some physicians goes, "Why the hell is XX receiving XX treatment?!" and the answer is always, "Because you ordered the protocol and that's what the protocol told us to do."

IMO, 99% of all protocols/AI driven medicine/whatever sucks. Every patient is different and using some sort of algorithm to predict outcomes just leads to lawsuits. It's almost always a bad idea to automate patient care.

Protocols are inherently different than the current models that use distributed representations. Protocols are a decision tree (often very shallow) that barely take into account the current information outside the current step. The advantage of the models used right now is that you can look at the decision, from the outcome all the way down to how it looked at the initial inputs (as well as past visits). The population needs to move away from thinking of these models as 'AI' and view it as learning a signal. So the question becomes, what is out there that has no mutual information with anything? The real bottleneck currently is the lack of data, and how awful we are at learning entirely from unsupervised data (think of how we learn a language as children).