Odds radiology will be defunct in ~5-10 years due to automation

[907914]

I would keep a close eye on what AI starts handling in the non-medical field.

Most of what you are describing could also be said of someone working in a call center, and I doubt those jobs survive the end of the decade.

AI can also become pretty good at putting people in boxes (not literally - for now), this is why Facebook is a leading company in AI, so categorizing patients as more or less likely to follow a medical regimen is well within the realm of possibility. This could probably be done reasonably now with the right dataset.

People are also getting more and more tolerant of asking AI questions, even letting it drive their car.

There was a strong, human component when a trained travel agent sat down with a customer too.

People are fickle and will give up human physicians in a second if it means lower cost of reasonable quality care, especially if it becomes more available and convenient.

So gotta do a procedures based specialty for that 50 year job security

---

Members don't see this ad.

 

[Lawpy]

 

[907914]

I’m not sure there’s 50 year job security in anything.

I mean, prostitution, soldier, and healer are probably the three oldest professions.

 

[CheGuevaraMD]

I mean, prostitution, soldier, and healer are probably the three oldest professions.

Pornography and drones have already dramatically reduced the demand for the first two...

 

[907914]

Pornography and drones have already dramatically reduced the demand for the first two...

But the careers still Perceiver. If prostitution can survive pornography then I am sure radiology can survive AI.

 

[Moko]

Where’s the human compassion though?!

Just have a tissue compartment that pops open whenever tears are detected, and a separate compartment with a hand that gives high-fives whenever a patient adheres to their fluid restriction. Simple.

In all seriousness though, all this talk about AI replacing doctors is overblown. AI will eventually serve as a powerful tool for delivering better patient care, but it's got a long ways to go. Computers can't even reliably diagnose straightforward arrhythmia on EKG, nor can they reliably differentiate between septic and non-septic patients who may just happen to be wiggling in their bed (I'm on to you Sepsis Alert!). Several years ago, I was part of a research project to 'teach' a computer to identify malignant cells on path based on staining (something the human eye can do relatively easily), the company had difficulty even reliably identifying the cells' nuclei (there's probably / hopefully been more progress on this front).

Given the nuances with imaging and the limits to pattern recognition, I would be shocked if AI can achieve this any time soon with the degree of accuracy that clinical medicine demands. The need for doctors to double-check these algorithms will exist for the foreseeable future, though I am excited to see how AI can eventually revolutionize the way we practice medicine.

 

[Gurby]

I was doing a Step 1 practice question earlier. Patient is presenting with some numbness and weakness of the thigh. It gave you a single slice of an axial CT scan and said "what's wrong with this patient?" You were supposed to notice the retroperitoneal hematoma, and realize that the femoral nerve - lying between the iliacus and psoas muscle - would be susceptible to compression here and could cause these symptoms.

It seems like AI is soon going to be able to do a decent job of answering discrete questions like "is there a head bleed?" But I think we're still light years away from "Watson, interpret this CT scan".

 

[907914]

I was doing a Step 1 practice question earlier. Patient is presenting with some numbness and weakness of the thigh. It gave you a single slice of an axial CT scan and said "what's wrong with this patient?" You were supposed to notice the retroperitoneal hematoma, and realize that the femoral nerve - lying between the iliacus and psoas muscle - would be susceptible to compression here and could cause these symptoms.

It seems like AI is soon going to be able to do a decent job of answering discrete questions like "is there a head bleed?" But I think we're still light years away from "Watson, interpret this CT scan".

And I don't think anyone is arguing that they will be able to interpret in a significant clinical context. But "Is there a bleed?" "Is there a fluid buildup?" "Is there a microtumor?"

 

[Dave1980]

It's only a bad question to ask because it indicates ignorance for the field. It's not insulting to the radiologist.

 

[Princeton Medical Student]

It's only a bad question to ask because it indicates ignorance for the field. It's not insulting to the radiologist.

It also indicates an ignorance in overestimating computers and computer scientists.

We still aren't amazing at analyzing ECG results which at the end of the day are a bunch of 2D plots. Radiology won't go anywhere for a long time.

 

[coolkul]

Not an expert on this topic, but I saw some work being done in image analysis through a large hospital research group and the tech is reaching a point where it can diagnose problems better than physicians. Radiology as a whole will not disappear, but it is somewhat likely the number of radiologists needed will decrease over the years simply because the diagnostic part of the job can be simplified and reduced. Imagine having 1 radiologist managing an AI based diagnostic tool and communicating with other doctors instead of 3-4 radiologists in a clinic.

It concerns me that radiologists are dismissing AI completely without thinking about the industry as a whole. AI will not replace radiologists, but it can definitely reduce the number of radiologists a hospital will have to hire. Considering the direction medicine is heading with cost cutting and increasing corporate salaries, someone will have to take a hit.

However, who knows how long it will take for this disruption to occur. All we know is that AI/machine learning has shown an enormous amount of progress in the last two decades and future improvement will be even more rapid.

 

[Maxamillion12]

Playing devils advocate here. If anything, I think primary care has the greatest risk of going by the wayside. Imagine just plugging your symptoms and family history into an app and being spit out several diagnoses with a recommended specialist to see. If you really believe that radiology will go by the wayside, then you can't be hypocritical and say PC won't go out of favor. No one like that conversation.

 

[RNthenDoc]

Playing devils advocate here. If anything, I think primary care has the greatest risk of going by the wayside. Imagine just plugging your symptoms and family history into an app and being spit out several diagnoses with a recommended specialist to see. If you really believe that radiology will go by the wayside, then you can't be hypocritical and say PC won't go out of favor. No one like that conversation.

This.
As things go toward algorithms for every Dx it seems less crazy that a piece of software could manage a medicine admission or do an f/u with a diabetic.

 

[CyrusHabs7695]

:troll:

 

[CyrusHabs7695]

Playing devils advocate here. If anything, I think primary care has the greatest risk of going by the wayside. Imagine just plugging your symptoms and family history into an app and being spit out several diagnoses with a recommended specialist to see. If you really believe that radiology will go by the wayside, then you can't be hypocritical and say PC won't go out of favor. No one like that conversation.

Yeah except every person with a headache will have brain cancer and anybody with intestinal pain will think they have Crohns

 

[RNthenDoc]

Idk man the painting industry is already DONE FOR! Look at this bad boy. Imagine it doing a colonoscopy.

 

[Princeton Medical Student]

Playing devils advocate here. If anything, I think primary care has the greatest risk of going by the wayside. Imagine just plugging your symptoms and family history into an app and being spit out several diagnoses with a recommended specialist to see. If you really believe that radiology will go by the wayside, then you can't be hypocritical and say PC won't go out of favor. No one like that conversation.

That's seriously not happening in the next 40-50 years and this is coming from someone that will graduate with a CS degree. It's just not that easy. We aren't even that good at reading ECGs.