What specialties can be least easily done by computers?

[deleted103644]

Again, by that point we'll have all time records for unemployment rates. You dont think there will be a massive uproar?

Nope. Just look at how things went down with other automated industries.
Also look at anesthesia with CRNAs.

The senior people will benefit from the increased productivity, even though it comes at the expense of employment opportunities for younger physicians.

Similar stuff is happening in law - there are now tools for automated discovery, etc, that can displace roles formerly performed by associates.

---

Members don't see this ad.

 

[druggeek]

Nope. Just look at how things went down with other automated industries.
Also look at anesthesia with CRNAs.

The senior people will benefit from the increased productivity, even though it comes at the expense of employment opportunities for younger physicians.

Similar stuff is happening in law - there are now tools for automated discovery, etc, that can displace roles formerly performed by associates.

By the time we have technology to replace a majority of physicians (not surgeons yet) the baby boomers will be dead. Surgeon replacement would be a far next level.

But anyway, you missed the point. There is such a big demand for family doctors right now, they dont want to hire nurses! they want to hire doctors. I live in Toronto area where every single thing in every field is saturated, yet a family doctor can make 200k first year out of residency with average clinical skills and average financial skills. This demand isnt going anywhere. Come back to me when there's a trend showing massive unemployment for physicians due to mid-levels....Then you can talk about robots.

What happened to machines replacing pharmacists? The technology is there and is actually used! Yet the job market is unaffected by this, only techs are reduced in numbers.

But anyway my central point has been... by the point physicians are replaced by computers/robots, other fields will be long replaced. That will = massive unemployment. I think that's easy to understand....

 

[deleted103644]

That's more due to technophobia in medicine and inertia than anything else.

The number of hospitals without full EMR is still shocking. There's no good excuse.

The technology will be there very soon, and healthcare is ripe for someone from the technology sector to b**** slap our business models. Inertia and red tape can only hold it back so long.

And I still don't care about Canada.

 

[druggeek]

That's more due to technophobia in medicine and inertia than anything else.

The number of hospitals without full EMR is still shocking. There's no good excuse.

The technology will be there very soon, and healthcare is ripe for someone from the technology sector to b**** slap our business models. Inertia and red tape can only hold it back so long.

And I still don't care about Canada.

Well considering our job markets are very similar, I think my example was a great one.

You didnt actually address what I said lol... I've been saying that there's a crapload of jobs that robots/computers can easily do and replace in the near future. Those would be gone first before people trust computers and robots to be in charge of their life. Then unemployment rates would soar and people would be pissed cause everyone would be broke.

 

[deleted103644]

Well considering our job markets are very similar, I think my example was a great one.

You didnt actually address what I said lol... I've been saying that there's a crapload of jobs that robots/computers can easily do and replace in the near future. Those would be gone first before people trust computers and robots to be in charge of their life. Then unemployment rates would soar and people would be pissed cause everyone would be broke.

Oh I'm not denying other jobs will disappear.

I'm just not planning on playing an active role in the creative destruction of those jobs.

 

[SchroedingrsCat]

What happened to machines replacing pharmacists? The technology is there and is actually used! Yet the job market is unaffected by this, only techs are reduced in numbers.

Might wanna talk to some recent PharmD grads.

 

[druggeek]

Might wanna talk to some recent PharmD grads.

I'm aware of the major saturation in pharmacy, but this has nothing to do at all with technology. It's basically too many pharmacists per population. You can only have so many pharmacies in one region anyway...each pharmacy can only hire so many pharmacists... so obviously if the supply far exceeds the demand, theres gonna be a problem.

 

[Perrotfish]

Computers are starting to take over tons of previously high-skill jobs (ex. pharmacy, law cases) and I see much diagnosis work in medical fields is becoming automized as well with IM physicians basically becoming machine technicians. What fields will be good to prevent your job getting taken by a computer?

Better engineers, physicians, economists, and science fiction writers than post on this board have tried to figure out the answer to this question. I don't think youre going to find a clear answer. My guess:

Short term: Specialties that focus on analyzing inputs off a computer screen are at the highest risk. Anesthesilogy gets simplified to the point where it can be done exclusively by CRNAs, who are in turn replaced by nurses, who are finally replaced by techs with a high school degree plus six months training. Diagnostic radiologists are replaced by diagnostic radiology programs. Both are gone fast.

Mid term: Generally people, extrapolating on existing trends, guess that the diagnostic specialties diappear before the the procedural ones. The logic is that we have seen an exponential increase in the pattern recognition capacity of computers, but robototics has been a relatively static field. On the other hand I think we might be about to reach the end of Moore's law, while robotics seems like its on the brink of a renaissance. Sooner or later computers will take over both specialties (like towards the end of your life) but I bet that the mechanical issues of robotics falls before the issues of deciphering emotional, human communication in diagnostic medicine. Surgery goes, then Medicine and Peds.

Long term: Implicit in your question is the assumption that human ability is a static benchmark that automated ability is measured against. I'm guessing about 50 years from now human ability is going to start advancing and diversifying, following its own Moore's law. in 100 year the question of humans vs machines is going to be irrelevant: mechanical and biological systems are going to be largely integrated.

So my advice: avoid diagnostic radiology and otherwise hope for the best.

 

[newtype]

First to go: radiology
Last to go: surgical

 

[druggeek]

Better engineers, physicians, economists, and science fiction writers than post on this board have tried to figure out the answer to this question. I don't think youre going to find a clear answer. My guess:

Short term: Specialties that focus on analyzing inputs off a computer screen are at the highest risk. Anesthesilogy gets simplified to the point where it can be done exclusively by CRNAs, who are in turn replaced by nurses, who are finally replaced by techs with a high school degree plus six months training. Diagnostic radiologists are replaced by diagnostic radiology programs. Both are gone fast.

Mid term: Generally people, extrapolating on existing trends, guess that the diagnostic specialties diappear before the the procedural ones. The logic is that we have seen an exponential increase in the pattern recognition capacity of computers, but robototics has been a relatively static field. On the other hand I think we might be about to reach the end of Moore's law, while robotics seems like its on the brink of a renaissance. Sooner or later computers will take over both specialties (like towards the end of your life) but I bet that the mechanical issues of robotics falls before the issues of deciphering emotional, human communication in diagnostic medicine. Surgery goes, then Medicine and Peds.

Long term: Implicit in your question is the assumption that human ability is a static benchmark that automated ability is measured against. I'm guessing about 50 years from now human ability is going to start advancing and diversifying, following its own Moore's law. in 100 year the question of humans vs machines is going to be irrelevant: mechanical and biological systems are going to be largely integrated.

So my advice: avoid diagnostic radiology and otherwise hope for the best.

So what do you do for complicated surgeries where many complications can result and a high level of education is needed ? (anesthesia)
Basically you're saying there will be a large human tolerance for errors made by machines.....I seriously doubt that.

People like to predict things like this, but we were supposed to be on Mars since 10 years ago and have flying cars as well... what happened to all that?

 

[Perrotfish]

People like to predict things like this, but we were supposed to be on Mars since 10 years ago and have flying cars as well... what happened to all that?

Its always amazing to me how people can be so unimpressed by the wonders we're surrounded with. You can use a wireless device to ask a a computer program any question in the world and it will answer you. You snack on genetically engineered foods,play with children fertilized in test tubes and rased in incubators, and if your adventerous you can even buy a cloned pet (though they're still rare). All of this was science fictiona single generation ago, and not even good science fiction. The all knowing computer was a plot device for when star trek couldn't think of a beter way to get on with the exposition, and now we just call it Google. But you're unimpressed,, because we have 'only' landed robots on Mars rather than a colony. So far.

And we do have flying cars. If your car flies they call it a plane. Its got wheels you can drive it on the ground, and it flies. I've never understood whatelse people wanted here.

 

[SchroedingrsCat]

Its always amazing to me how people can be so unimpressed by the wonders we're surrounded with. You can use a wireless device to ask a a computer program any question in the world and it will answer you. You snack on genetically engineered foods,play with children fertilized in test tubes and rased in incubators, and if your adventerous you can even buy a cloned pet (though they're still rare). All of this was science fictiona single generation ago, and not even good science fiction. The all knowing computer was a plot device for when star trek couldn't think of a beter way to get on with the exposition, and now we just call it Google. But you're unimpressed,, because we have 'only' landed robots on Mars rather than a colony. So far.

And we do have flying cars. If your car flies they call it a plane. Its got wheels you can drive it on the ground, and it flies. I've never understood whatelse people wanted here.

And 2pac just performed as a ****ing hologram. This sci-fi future we're living in is pretty scary tbh.

 

[jejun]

As someone who studied computer science... my two cents:

Diagnostic Radiology WILL NOT be the first to go. We can barely do face recognition. EKG analysis is far from perfect (and it's a LINE, not a CT/MRI). Taking a complicated 3d image and analyzing it in a classification algorithm is not trivial. It's extremely difficult

Alternatively... having someone check off a list of symptoms, in conjunction with vitals and a few numerical lab values is much easier to classify. Check out the diagnostic algorithm Watson... it's already being done . So yeah: I would say ER could be automated to some extent but we'll always need physical exam. Things that are not acute that simply need a decision of: admit/not-admit could be handled by a computer since the decision and inputs are pretty simple. Also, social issues are also probably pretty hard to automize and we all know that they make up a significant portion of medicine.

At the end, I don't think this would be bad for physicians. Doctors would spend less time on "trivial cases" and more time on the 10-20% of things that don't follow the pattern. They would probably also have more time to talk to patients about social issues that often get missed due to time constraints.

 

[deleted103644]

As someone who studied computer science... my two cents:

Diagnostic Radiology WILL NOT be the first to go. We can barely do face recognition. EKG analysis is far from perfect (and it's a LINE, not a CT/MRI). Taking a complicated 3d image and analyzing it in a classification algorithm is not trivial. It's extremely difficult

Alternatively... having someone check off a list of symptoms, in conjunction with vitals and a few numerical lab values is much easier to classify. Check out the diagnostic algorithm Watson... it's already being done . So yeah: I would say ER could be automated to some extent but we'll always need physical exam. Things that are not acute that simply need a decision of: admit/not-admit could be handled by a computer since the decision and inputs are pretty simple. Also, social issues are also probably pretty hard to automize and we all know that they make up a significant portion of medicine.

At the end, I don't think this would be bad for physicians. Doctors would spend less time on "trivial cases" and more time on the 10-20% of things that don't follow the pattern. They would probably also have more time to talk to patients about social issues that often get missed due to time constraints.

Exactly. I think there's some cool stuff on the horizon for image recognition, but it's still a tough nut to crack. (So the guy saying radiology will be the first to go doesn't know what he's talking about...)

The first thing that should happen is automation of basically everything in the MGH handbook. All of those algorithms could just be built into a decent EMR, giving you automatic suggestions for diagnoses/tests/treatments according to current guidelines. The next step after that would be to use something like Watson to update those guidelines automatically as the literature progresses (and maybe some automatic QC and hypothesis testing on the data available in the hospital EMR).

I really like physiology, but never took anesthesiology seriously as a choice for me because I do think that will be one of the first specialties to get automated almost out of existence. It will go long before radiology (at least as far as anything that can be handled by a CRNA is concerned).

 

[docshop12]

As someone who studied computer science... my two cents:

Diagnostic Radiology WILL NOT be the first to go. We can barely do face recognition. EKG analysis is far from perfect (and it's a LINE, not a CT/MRI). Taking a complicated 3d image and analyzing it in a classification algorithm is not trivial. It's extremely difficult

Alternatively... having someone check off a list of symptoms, in conjunction with vitals and a few numerical lab values is much easier to classify. Check out the diagnostic algorithm Watson... it's already being done . So yeah: I would say ER could be automated to some extent but we'll always need physical exam. Things that are not acute that simply need a decision of: admit/not-admit could be handled by a computer since the decision and inputs are pretty simple. Also, social issues are also probably pretty hard to automize and we all know that they make up a significant portion of medicine.

At the end, I don't think this would be bad for physicians. Doctors would spend less time on "trivial cases" and more time on the 10-20% of things that don't follow the pattern. They would probably also have more time to talk to patients about social issues that often get missed due to time constraints.

I agree about rads and EKGs. But what you describe (someone checking off a list of sxs) and a computer looking at lab values is already in place... If it is ever automated to any level 200 yrs from now the first to go would be the LNP. ER docs are going nowhere.

I'm not sure what level med student you are but once you see those checklists and then you actually go and talk to the patient you'll understand the limits things like that have... i.e. patients don't even understand what the sxs mean therefore skewing and making formulating a diff dx more difficult and patients will often exaggerate sxs or list every single sx they've ever had making the list basically useless since that wouldn't be reflected on the checklist. We also have systems that could easily incorporate all abnormal lab values. But you neglect that "abnormal" is a range and levels at either extreme could be significant. Also interpretation of lab values is beyond difficult to program into a computer just based on the shear number of possibilities. And also patients rarely present like they do in the textbook and patients nowadays come in so early for any problem they are having it makes it far more difficult to quickly get a dx - both of which a computer would not be able to interpret.

As a result what you describe won't ever be automated. Never. We clearly have the technology to do that now in any industrialized country and it is done exclusively in a total of ZERO hospitals or clinics.

 

[jejun]

Hey finishing up my third year. Some quick comments (I find this topic pretty interesting).

The patients wouldn't be entering or filling out a checklist. It would be carried out by an MD/Nurse/Medical Student. The point is, after it is done the majority of the workup would be complete in terms of further labs, scans, and admit/not-admit would be formulated dynamically when enough results came back.

Speaking of abnormal lab values, computers would not look at them as binary entities (Low/Normal/High) the way many physicians do. You They would probably look at the degree to which they are elevated based on it's the statistical training of the algorithm. This is something computer actually do very well. You're right, not everything is clinically significant. In today's ER though, an elevated lab often leads to expensive workups without and justification... this is a mistake these algorithms would be far less likely to make.

 

[docshop12]

Hey finishing up my third year. Some quick comments (I find this topic pretty interesting).

The patients wouldn't be entering or filling out a checklist. It would be carried out by an MD/Nurse/Medical Student. The point is, after it is done the majority of the workup would be complete in terms of further labs, scans, and admit/not-admit would be formulated dynamically when enough results came back.

Speaking of abnormal lab values, computers would not look at them as binary entities (Low/Normal/High) the way many physicians do. You They would probably look at the degree to which they are elevated based on it's the statistical training of the algorithm. This is something computer actually do very well. You're right, not everything is clinically significant. In today's ER though, an elevated lab often leads to expensive workups without and justification... this is a mistake these algorithms would be far less likely to make.

well here's the thing... who's going to develop these programs and algorithms? I'd imagine you'd need a group of highly trained and respected physicians who are also extremely adept at computer programming or can explain how to do it to a programmer. You'd then have to show its effectiveness being far superior to the current system with a lower cost. Next you'd need to get it adopted. Aside from the difficult task of completing the first phase ALL of this stuff would cost a very large amount of money. Who's going to fund it? Where's the profit in it since there would have to be a profit for everyone - the company developing the programs, programmers, docs, hospitals, etc? You think government funds would do it hahahaha? These of course are all rhetorical questions...

so aside from me thinking a lot of it is not very useful, the logistics of getting anything like this done are so extreme and there are so many barriers I don't think it'll be happening any time soon if ever.

 

[impact2d]

One of my primary reasons for becoming a physician is a surefire demand for my skills in the upcoming zombie apocalypse.

That being said, while there might be a grind towards computer diagnostics, look at unmanned aircraft. There are still fighter pilots, (or pilots supervising the drones). Physicians will always be needed.

Is it sad that this is secretly my reason for wanting to become a doctor lol

 

[druggeek]

Its always amazing to me how people can be so unimpressed by the wonders we're surrounded with. You can use a wireless device to ask a a computer program any question in the world and it will answer you. You snack on genetically engineered foods,play with children fertilized in test tubes and rased in incubators, and if your adventerous you can even buy a cloned pet (though they're still rare). All of this was science fictiona single generation ago, and not even good science fiction. The all knowing computer was a plot device for when star trek couldn't think of a beter way to get on with the exposition, and now we just call it Google. But you're unimpressed,, because we have 'only' landed robots on Mars rather than a colony. So far.

And we do have flying cars. If your car flies they call it a plane. Its got wheels you can drive it on the ground, and it flies. I've never understood whatelse people wanted here.

I'm just saying 20 years ago... the predictions made for >2010 were far more "sci fi" than the things we have now. Sure we could all be replaced by robots/computers, but I'm saying it most likely wont happen in our working lives.

As someone who studied computer science... my two cents:

Diagnostic Radiology WILL NOT be the first to go. We can barely do face recognition. EKG analysis is far from perfect (and it's a LINE, not a CT/MRI). Taking a complicated 3d image and analyzing it in a classification algorithm is not trivial. It's extremely difficult

Alternatively... having someone check off a list of symptoms, in conjunction with vitals and a few numerical lab values is much easier to classify. Check out the diagnostic algorithm Watson... it's already being done . So yeah: I would say ER could be automated to some extent but we'll always need physical exam. Things that are not acute that simply need a decision of: admit/not-admit could be handled by a computer since the decision and inputs are pretty simple. Also, social issues are also probably pretty hard to automize and we all know that they make up a significant portion of medicine.

At the end, I don't think this would be bad for physicians. Doctors would spend less time on "trivial cases" and more time on the 10-20% of things that don't follow the pattern. They would probably also have more time to talk to patients about social issues that often get missed due to time constraints.

I dont know why we should be concerned about doctors being replaced by machines when pharmacists have yet to be. The technology is there, the machines are there... but they're still where they should be.

My issue with Watson, is how do you distinguish something with a 62% chance and lets say.. 48% chance? how do you use visual judgement (ex. person with intense pain vs. someone with extreme pain)? etc etc

 

[druggeek]

Exactly. I think there's some cool stuff on the horizon for image recognition, but it's still a tough nut to crack. (So the guy saying radiology will be the first to go doesn't know what he's talking about...)

The first thing that should happen is automation of basically everything in the MGH handbook. All of those algorithms could just be built into a decent EMR, giving you automatic suggestions for diagnoses/tests/treatments according to current guidelines. The next step after that would be to use something like Watson to update those guidelines automatically as the literature progresses (and maybe some automatic QC and hypothesis testing on the data available in the hospital EMR).

I really like physiology, but never took anesthesiology seriously as a choice for me because I do think that will be one of the first specialties to get automated almost out of existence. It will go long before radiology (at least as far as anything that can be handled by a CRNA is concerned).

so for many extremely complicated cases where many complications can result, you hand it over to the CRNA or the computer?

The average poor person may be okay with it, and maybe the average person who doesnt know **** about anything... but intelligent/richer people more than likely wouldnt be.

 

[CatFactorial]

I'm just saying 20 years ago... the predictions made for >2010 were far more "sci fi" than the things we have now. Sure we could all be replaced by robots/computers, but I'm saying it most likely wont happen in our working lives.


I dont know why we should be concerned about doctors being replaced by machines when pharmacists have yet to be. The technology is there, the machines are there... but they're still where they should be.

My issue with Watson, is how do you distinguish something with a 62% chance and lets say.. 48% chance? how do you use visual judgement (ex. person with intense pain vs. someone with extreme pain)? etc etc

Tsunade was able to determine that Rock Lee had a 57% chance of surviving his surgery after opening the sixth gate. But she used chakra.

 

[Acoustic]

Does anyone else think it's funny that this thread popped up at the same time as the one about midlevels takin' errrr jobs? It seems to me that any Watson based program would basically be another midlevel, good at finding the patterns but horrible at interpreting the ones that don't fit. At any rate, they both need physician supervision and if anything will allow the physician to make more money.

That is, unless we completely lose the political war. That's probably the bigger threat to our futures, making sure we have a voice in what's coming and educating the public about our perspectives.

 

[Oxer45]

Its far easier for two horny teens to make a baby that's more intillegent than any supercomputer. So if I'm not worried about a human replacing me, why should I be be worried about "artificial intelligence" replacing me. If a nurse, who can interact with patients, take history, draw labs, follow diagnostic/treatment guidelines cant replace me, how can a computer? Can you mention anything that AI is supposed to do that a nurse cant already do?

 

[Whiskeypunch]

If there is one lesson from finance/econ/history of business it is that the future is difficult to predict. But it is fun to speculate.

I don't see robots replacing surgeons until every other job in America is replaced.

Even just the da Vinci robot is 1.5-2million fixed cost plus like 1800-2200 in supplies per surgery. Add on top of that expensive software... That's really not cost effective.

Plus you've got the difficult task of integrating visual input, disparate anatomy, critical thinking, etc.

It makes way more sense for something like radiology to go (as mentioned before) because all of the data is in a computer. There's no need for expensive robotics, just the software and a computer. This is something that could actually happen in our lifetimes I think. I mean just think about it, it really is a job that should be done by a computer. The computer could use every image ever produced as a learning tool.

I could see something like what IBM is working on with Watson allowing mid-levels to take over primary care. They input the h+p into the computer and their diagnosis + treatment plan and the computer spits back things to look for, or recommendations, or tells the mid-level to call a specialist.

All of this is of course going to depend on liability.

 

[Law2Doc]

First to go: radiology...

again, I worked tangentially with some if the CAD stuff many many many many years ago as an undergrad. There has been minimal to no progress or improvement on this since then, i guess because they haven't figured out how humans process the images given multiple normal variations, which was also our problem back when. We used to put special sensors on radiologists eyes to see what they looked at, but we got very little insight. Looks like the technology stalled out since then, because the CAD programs I've seen lately are awful and borderline useless. All they really do is highlight densities for the reader to consider, which doesn't really add much. So I'd say radiology is probably last to go because there's a science to the pattern recognition that has the programmers baffled. Path probably presents similar issues. Based on what Watson has been shown to do, IM/FM, neuro and probably EM are the easiest to make a dent in. Anesthesiology can be significantly helped as well. Surgery will take a lot longer, but very routine things could conceivably be taken over. So no, I'd say rads is going to be late if ever, and the more cerebral differential diagnosis fields will be first to go. The old Eliza program actually was lauded by folks in psych decades ago, so a more sophisticated version probably could make inroads in that field as well.

 

[Law2Doc]

... Can you mention anything that AI is supposed to do that a nurse cant already do?

work 24 hours a day, never draw a salary, never complain, never have an off day, never have HR (discrimination/harassment) issues...

 

[Law2Doc]

...
It makes way more sense for something like radiology to go (as mentioned before) because all of the data is in a computer. There's no need for expensive robotics, just the software and a computer. This is something that could actually happen in our lifetimes I think. I mean just think about it, it really is a job that should be done by a computer. The computer could use every image ever produced as a learning tool....

the fact that they've tried for many many years (decades) and made negligible progress is pretty good evidence that this will actually be one of the last bastions to fall. Anything which requires significant visual processing and a huge range of normal has been shown to really not be currently technologically feasible, and maybe never will be. Doesn't matter that the information is digitized, if you can't figure out how it should be processed, it's sort if like trying to pick out the grains of sand that don't fit on a mile long stretch of beach.

 

[shadowfox87]

Who's worried? I want to help develop it.

Amen, as a engineer, I can't ever look down on technology. I look forward to any advances in medicine. Being in med school, I can see so many opportunities to automate. Anything systematic can be programmed. Also, image recognition is not so easy because of anatomical variation. We have been able to do it on standard patients, but patients are far from standard most of the time. This variation is pretty difficult to program. The same goes for surgery. I have worked with robots similar to Da Vinci and haptic technology. Programming works given certain constants, when those constants become variables, a human has to interject to set extra parameters. Humans are still needed. I believe technology will make the job of the doctor easier such that there will be less memorization and more room for creative thinking. This is what I'm looking forward too.

 

[badasshairday]

My issue with Watson, is how do you distinguish something with a 62% chance and lets say.. 48% chance? how do you use visual judgement (ex. person with intense pain vs. someone with extreme pain)? etc etc

Watson will never have gestalt an experience physician has.

 

[badasshairday]

again, I worked tangentially with some if the CAD stuff many many many many years ago as an undergrad. There has been minimal to no progress or improvement on this since then, i guess because they haven't figured out how humans process the images given multiple normal variations, which was also our problem back when. We used to put special sensors on radiologists eyes to see what they looked at, but we got very little insight. Looks like the technology stalled out since then, because the CAD programs I've seen lately are awful and borderline useless. All they really do is highlight densities for the reader to consider, which doesn't really add much. So I'd say radiology is probably last to go because there's a science to the pattern recognition that has the programmers baffled. Path probably presents similar issues. Based on what Watson has been shown to do, IM/FM, neuro and probably EM are the easiest to make a dent in. Anesthesiology can be significantly helped as well. Surgery will take a lot longer, but very routine things could conceivably be taken over. So no, I'd say rads is going to be late if ever, and the more cerebral differential diagnosis fields will be first to go. The old Eliza program actually was lauded by folks in psych decades ago, so a more sophisticated version probably could make inroads in that field as well.

Agreed. On the interview trail, a prominent western academic program showed us some of there research. They were using what you described, a special sensor on radiologists eyes and comparing it to a special sensor on a radiology residents eyes and comparing how each one looks at a study. They basically are using this technology now to figure out why mistakes are made and what are the differences between a novice trainee in how they read the image compared to an experienced board certified individual. It was actually quite interesting to see the prelim data. But overall I agree with your salient points.

Again look at the EKG machine. This is actually something that is numerical/graph/math driven compared to something like pattern recognition which is not an exact science at all. As we all know, the EKG machine has a ton of kinks in it.

 

[Law2Doc]

Agreed. On the interview trail, a prominent western academic program showed us some of there research. They were using what you described, a special sensor on radiologists eyes and comparing it to a special sensor on a radiology residents eyes and comparing how each one looks at a study. They basically are using this technology now to figure out why mistakes are made and what are the differences between a novice trainee in how they read the image compared to an experienced board certified individual. It was actually quite interesting to see the prelim data. But overall I agree with your salient points. ....

yeah that would seem like much more interesting and promising research if we weren't already doing the exact same study so many many many years ago. I'm sure it looks impressive to applicants who don't realize that its a rehash, but it's yesteryears news. And not very fruitful at that.

 

[Shnurek]

yeah that would seem like much more interesting and promising research if we weren't already doing the exact same study so many many many years ago. I'm sure it looks impressive to applicants who don't realize that its a rehash, but it's yesteryears news. And not very fruitful at that.

got a link?

 

[Law2Doc]

got a link?

Not without outing myself on here.

 

[45408]

You really think that?

Money talks. 10 years from now, if the US were to fall in to dire financial trouble, I doubt they would be so desperate about making sure patients could sue their radiologist, especially if any randomized studies came back to show equivalent effectiveness.

Just realize, anything can happen.

Yet, I don't think any of us are in trouble as far as job security.

I do. I don't think the American public could fathom having no one to blame for something.

You're not even a medical student.

Have you ever sutured anything? (I hope not a patient, but you could have tried pigs' feet or orange peels.)

I'm not going into a surgical specialty, but I've sutured plenty of wounds/incisions and honored my surgery rotation.

Suturing would be difficult to automate, but far from impossible. The hardest part would be wound approximation. (And obviously the needles would be different.)

I'm pretty sure we'll have cold fusion and flying cars before we have robots suturing our wounds. Approximating skin is one thing, and it's not that hard to do an okay job (which is why they let me do it as a med student). It's sewing the many other things in surgery that is much more complicated, like an AV fistula or sewing in a dural patch.

Have you seen/read about the DaVinci surgical machine...blows my mind. Still requires a surgeon to operate, but blows my mind all the same.

It's highly useful in the right setting, but it is simply an extension of your own movements with some dampening and clutching.

 

[45408]

Its always amazing to me how people can be so unimpressed by the wonders we're surrounded with. You can use a wireless device to ask a a computer program any question in the world and it will answer you. You snack on genetically engineered foods,play with children fertilized in test tubes and rased in incubators, and if your adventerous you can even buy a cloned pet (though they're still rare). All of this was science fictiona single generation ago, and not even good science fiction. The all knowing computer was a plot device for when star trek couldn't think of a beter way to get on with the exposition, and now we just call it Google. But you're unimpressed,, because we have 'only' landed robots on Mars rather than a colony. So far.

And we do have flying cars. If your car flies they call it a plane. Its got wheels you can drive it on the ground, and it flies. I've never understood whatelse people wanted here.

[YOUTUBE]8r1CZTLk-Gk[/YOUTUBE]

 

[badasshairday]

Had to bump this.

http://www.foxnews.com/scitech/2012...r-robot-prostitutes-by-2050/?intcmp=obnetwork

 

[sportsperson]

Had to bump this.

http://www.foxnews.com/scitech/2012...r-robot-prostitutes-by-2050/?intcmp=obnetwork

Except I I would think most "clients" want to have sex with a real human. Only people with blowup doll fetishes would be into this type of thing.

 

[The kitchen sink]

The specialty most in danger of losing to computers is anesthesiology. Computers may not be able to administer the gases correctly, but they damn sure can complete a crossword puzzle.

they took er jerbssss

HAHAHAHA well done

 

[Substance]

again, I worked tangentially with some if the CAD stuff many many many many years ago as an undergrad. There has been minimal to no progress or improvement on this since then, i guess because they haven't figured out how humans process the images given multiple normal variations, which was also our problem back when. We used to put special sensors on radiologists eyes to see what they looked at, but we got very little insight. Looks like the technology stalled out since then, because the CAD programs I've seen lately are awful and borderline useless. All they really do is highlight densities for the reader to consider, which doesn't really add much. So I'd say radiology is probably last to go because there's a science to the pattern recognition that has the programmers baffled. Path probably presents similar issues. Based on what Watson has been shown to do, IM/FM, neuro and probably EM are the easiest to make a dent in. Anesthesiology can be significantly helped as well. Surgery will take a lot longer, but very routine things could conceivably be taken over. So no, I'd say rads is going to be late if ever, and the more cerebral differential diagnosis fields will be first to go. The old Eliza program actually was lauded by folks in psych decades ago, so a more sophisticated version probably could make inroads in that field as well.

Pathology is the first to go.

1. The images are 2D and don't require reformatting into 3D like radiological images.

2. Most diagnoses are going to be based on either in-vivo imaging or molecular genetics/proteomics, both of which make slide analysis obsolete.

 

[surftheiop]

Not that anyone really is, but I wouldn't be too concerned with this sort of thing. If computers/AI actually get to the point to replace any medical field, then that means there was probably a big enough breakthrough that they could replace all the fields, so there isn't much point worrying about it.

But if we have to conjecture, I would say diagnostic radiology would be the first go just because thats the only field where currently all the data available to the clinician is already collected by a machine and the computer already has every single data point that the radiologist reading the image has.

Not to say that our image processing software is anywhere close to being able to do this, but the "table is set" more in radiology than any other field seeing as the problem is now solely a data processing problem at this point. While in other fields its a data collecting and data processing problem (as well as robotic problem for anything procedural).

 

[deleted103644]

Pathology is the first to go.

1. The images are 2D and don't require reformatting into 3D like radiological images.

3D images are actually easier for computers to process in some ways than 2D projections.

We're hard wired to convert 2D data into 3D, but that's actually tough to program.

2. Most diagnoses are going to be based on either in-vivo imaging or molecular genetics/proteomics, both of which make slide analysis obsolete.

Agreed, although reagents will be cost-prohibitive for awhile.

 

[cowme]

I remember coming across an article last year entitled "Microsoft software to "replace" radiologists." I freaked out. I'm thinking: just my luck, days before matching in the specialty, here is some groundbreaking technology to replace the field I have selected. It was called "innereye," and it was a multimillion dollar project that spanned nearly five years. I was freaking out.

Then I actually read the article. This "revolutionary" new software that was ready to take my job could do one thing: point to 8 human organs on a CT scan.

That's right people. The world's brightest young minds, working with unlimited funds, half a decade, and one of the leaders in software development came up with an with a program that does something that I could teach a 10 year old to do in 1 hour.

Radiology may one day be replaced by computers. But that is a long, long, loooong way away.