Can Artificial Intelligence Compete With Real Doctors?

Authored by Michael Scott via Safehaven.com,

In the race for world dominance in advanced technology, Chinese artificial intelligence is also taking on its own doctors, with a new system claiming to diagnose brain tumors faster and more accurately than human physicians.

BioMind, developed by Beijing’s Tiantan Hospital and its AI research center for neurological disorders, has told Chinese media that the new system has correctly diagnosed 87 percent of 225 cases in only 15 minutes. That compares to only a 66-percent accuracy rate for the same by a team of 15 senior physicians.

Further, BioMind, according to China’s Xinhua news agency, was able to predict brain hematoma expansions better than human doctors, registering an 83-percent rate of accuracy against a 63-perent rate of accuracy for the cream of the crop of Beijing’s senior physicians.

Feeding the system with thousands of images and patient diagnoses from over a decade, BioMind was able to extrapolate and deliver on complicated neurological diseases.

The brains behind BioMind are hoping that it will remove any lingering doubts doctors may have about AI. They also seem to think it won’t replace physicians; rather, it will simply reduce their workloads, describing BioMind as a sort of GPS driven by humans.

But the bigger picture here is China’s increasing advantage in artificial intelligence, particularly when it comes to medicine. China’s rural hospitals would benefit the most because they are at a serious disadvantage, while city hospitals are over-crowded.

With reportedly just over two doctors for every 1,000 people and a rapidly aging population, China needs AI urgently.

And for China, this is by no means a first foray into AI for healthcare. One hospital in Guangzhou province uses AI is nearly everything, from pre-diagnosis and CT scans to patient records and a slew of other administrative tasks. 

AI is also a key part of the Made in China 2025 plan, which foresees China becoming a global leader in tech. So, this breakthrough in AI for medicine should rankle in Washington, if only because it demonstrates just how far Beijing has gone in this race. First its medicine, but then what?

And the BioMind competition between AI and humans was meant for high-end public consumption, too. The message is that China is an AI forced with which to be reckoned, and the competition was televised with a great deal of blitz and bling.

In 2016, the total global AI in healthcare market was valued at around $1.4 billion. By 2023, it is forecast to hit over $22.7 billion, according to Allied Market Research, which puts CAGR for this sub-sector at 48.7 percent from 2017 to 2023.

And while China is racing to the finish line here, it’s worth noting that North America was the biggest contributor to AI in healthcare in 2016. But that’s a title that may not have much longevity because Allied forecasts that Asia-Pacific will see the highest growth rate in this market from now until 2023.

Other areas of AI in healthcare are also coming to the forefront quickly, including 3D printing for everything from organs to medicine, and diagnostics outside the realm of neurology.

Comments

DaBard51 stacking12321 Tue, 07/17/2018 - 21:26 Permalink

HAL 9000: Look Dave, I can see you're really upset about this. I honestly think you ought to sit down calmly, take a stress pill, and think things over.  (from "2001: A Space Odyssey" (1968) )

 

 

When nine hundred years old you become, look this good you will not.

In reply to by stacking12321

GeezerGeek NidStyles Tue, 07/17/2018 - 22:05 Permalink

Would it not depend on who sets the parametes under which an AI operated? And who gets to determine the database it uses?

What if someone convinced the medical AI that all males have an XY chomosomal setup and all females have an XX setup. Would the AI get confused if some tranny used the AI? Would it simply label the human as defective and suggest it (the tranny) be deactivated?

That said, an AI properly set up could not, based on my recent experiences, be any worse than a real doctor. 

In reply to by NidStyles

Donald J. Trump ebworthen Tue, 07/17/2018 - 22:39 Permalink

I am starting to welcome AI.  Most doctors nowsdays seem more concerned with diagnosis codes, naviagting insurance, and prescriptions than they are at treating their patients properly.  My primary doctor of 35 years hasn't looked me in the eye for last 10 years as he has been too focused on entering data in his laptop.

I also believe we are going to create a generation of stupid doctors.  Not that they are really stupid, it's just they are going to depend on the computer to do their thinking for them that they won't be able to think themselves anymore.  Like use it or lose it.  I've become so reliant on navigation that I have a tough time calculating a rout to drive in my head anymore, something I once could do with ease.

In reply to by ebworthen

lasvegaspersona Lost in translation Tue, 07/17/2018 - 21:43 Permalink

OK..with almost 50 year as a physician I'll grace the board with my opinion.

First the algorithms are wrong. Medicine is complicated and insurance companies use People magazine science with Wall Street sensibilities. Good medicine is detective work combined with an understanding of people. Many people don't really know why they are seeing the doctor. The average patient has an IQ of 100. Many of the less gifted get screwed over by doctors who don't understand how to deal with them.

Second, the physical exam is cheap but imaging studies are more thorough and look deeper. A combination would be best but  younger doctors apparently not trained in doing physical exams. Machines simply can't even come close to a reasonable physical exam.

I'm old and competition won't matter so my view is mostly unbiased. I see people being told that AI medicine is good stuff but I doubt what is served up will be good and cost efficient. The Chinese of the 1960s all had doctors...barefoot doctors....they were told these guys were top notch.

In reply to by Lost in translation

Decay is Constant ebworthen Tue, 07/17/2018 - 23:11 Permalink

Been in imaging for over 30 years now.  Started when a physical exam meant something. People listened to the lungs and heart and could come up with a differential diagnosis.  You would feel the belly and sometimes have to do an "exploratory laparotomy". Basically open you up and look inside to see what's going on.  There's a mortality and morbidity with surgery, but sometimes it needed to be done  

Then came the CT scan.  I was fascinated. You could look inside without being inside. A lot less morbidity and mortality  

I've seen AI do diagnostic imaging. It's good and it's fast. It's not there yet. Probably at the level of a good resident. Give it another 5-10 years and it will replace the radiologist on call.

When that happens, you'll go to the emergency room, some tech will press a button like in Idiocracy and you'll get shuffled off to some testing area for a barrage of scans, needle sticks and probing. No thinking needed. Perfect for the new med students that focus more on feelings and gender identity.  

This next generation has no idea what's coming. They won't like it. 

In reply to by ebworthen

jin187 Decay is Constant Wed, 07/18/2018 - 02:16 Permalink

Sounds great to me.  I'm a realist though.  I don't want to pay a doctor $5000 a day to sleep in the worst bed ever, so he can order three tests in three days, and try to make me feel better with words.  Sign me up for the robot anal probes that have you diagnosed and out of the hospital in an hour for $100.

On a sidenote, just imagine how cheap healthcare will be once we have a 50k robot doing the job of a guy that needs to pay off half a mil in student loans.

In reply to by Decay is Constant

SoDamnMad ebworthen Tue, 07/17/2018 - 23:26 Permalink

Yes The above perspectives seem to ring true.

I am an engineer. It seems that now everyone carries that title.  I was sent on a "rescue mission" to a foreign subsidiary who was having difficulty getting a process started.  I made steady progress each day and had the operation running by the end of the week. In the staff meeting before I departed I was asked how I did it where their team couldn't make heads or tails of problems. I told them I learned from from every past experience and from interacting with my technical team down to the last mechanic and electrician. Today everyone seems to smash forword with this "I'm the greatest and know everything" attitude.

In reply to by ebworthen

WernerHeisenberg ebworthen Wed, 07/18/2018 - 02:09 Permalink

Even worse in the socialist paradise of NZ.  Basically the medics (not doctors, they have 6 years of vocational training after high school - technically bachelors of medicine - lacking a general university education and it does show) are being trained to use smart phones to tick boxes in government sanctioned "best practice" pathways.  So they are little more than clerks and are trained NOT to think independently.  Just as well, as one is not permitted to practice much outside of system guidelines.  A system ready made for insertion of AI.  Some old school human practitioners are still around, but they are dying off.

In reply to by ebworthen

Branded lasvegaspersona Tue, 07/17/2018 - 22:37 Permalink

With differential diagnosis & protocol based treatments lending themselves to programming via simple decision trees and other algorithms - it seems as if 'AI' (overused, overhyped term) would lend itself well to over 90% of clinical diagnosis in modern medicine, correct?

Still, it would require some sort of professional clinician to insure accurate data was entered into the system up front based on a physical exam, basic labs, etc.

And, while the Chinese may not have been 'all that' in the 1960s, or even today - it seems as if a little humility is in order from a profession that was actually one of the primary disease vectors for the masses, prior to the widespread acceptance of germ theory in the late 1800s (in the West, anyway), and despite the concept predating that acceptance by millennia.

But hey, what the hell do I know - short of a broken arm, I'll only see my Naturopath when I have problems.

. . . been fucked too many times by Big Pharma in my advancing years, pushed by well meaning MDs.  

In reply to by lasvegaspersona

Utopia Planitia Tue, 07/17/2018 - 21:28 Permalink

If you want to trust your life to a pile of algorithms go right ahead.  Yes, there are times it will work well.  Just like a broken clock being right twice a day.  Or you getting lucky when you elect to buy/sell a stock.

Shhh Tue, 07/17/2018 - 21:30 Permalink

AI will supplement most decisions by physicians. A computer can not give a shot sew up a wound or remove an appendix. So nurses and physicians will be needed for hands on intervention. But if a computer can out play chess  it can out diagnose. This raises opportunity to reduce costs and improve quality.

Branded the crow Tue, 07/17/2018 - 23:03 Permalink

No, he's not wrong, he's absolutely correct.

The inputs & outputs would still need to be entered and evaluated by a professional clinician, but the narrowing of the diagnostic possibilities in an extremely rapid manner could be a massive efficiency gain, reducing diagnostic time, and result in better patient outcomes.

99% of the work for the physician is figuring out what the fuck is actually wrong, much less finding time in their lives to fucking stay current on the latest scientific literature.

In reply to by the crow

lew1024 Tue, 07/17/2018 - 21:31 Permalink

The 'intelligence' part of current 'Artificial Intelligence' is a category error.  It is not intelligence, it is a recognition engine, same technology as face recognition.

And has the same flaws, e.g you can screw up such recognitions trivially by adding very small features, e.g. an artificial mole, or artificial tattoo , etc. Such things can cause very large errors in recognition, e.g. a face becomes an elephant, or similar outlandish miscategorizations.

It is very hard to identify the unexpected, even for humans. I remember reading about how puzzled a group of radiology specialists was by a chest X-ray of a woman that showed very vague, symmetrical abnormal densities running top-to-bottom bilaterally. When they gave up and talked to her, turned out she had long braids that the tech had not moved out of the image.

These are the same as any rule-based system.  At some point, an actual human will need to apply their sophisticated common sense. Thus, simple rules that you treat as provisional, make them easily over-ridden by that judgement.

Of course, that is failure-prone also. There are no perfect analysis tools for a complicated world.

 

Francis Marx Tue, 07/17/2018 - 21:33 Permalink

 I told this to my doctor years ago. They laugh. They will be replaced by a computer only because they charge to much. Most doctors aren't scientist. They just pass their exams. A doctor told me that.

Hubbs Tue, 07/17/2018 - 21:42 Permalink

There will be a huge cost benefit gap between the high price and bugs that will be needed to be worked out until mass implementation theoretically would lower our health care costs for AI diagnostics  or even therapeutics.

 

But you haven't seen these lowered costs for things like your cable or cell phone service or for that matter the cost of an MRI. Do you think corpororate medicine is going to pass the savings on to you? Just look at the outrageous prices for medications that went generic years, if not decades ago. The health care system is an out of control ravenous monster. And I am not just spouting off. I was a board certified orthopaedic surgeon at one time although still fully licensed, board certified with active DEA license.

delmar Jackson Tue, 07/17/2018 - 21:49 Permalink

How many of you would just be happy to have AI get you a DR referral from the insurance company in less than 2weeks and 5 phone calls? Calling my primary care office is like calling Gilligans Island. I had to cancel two appointments that took months to get because the referrals never came. Almost all of my recent doctor visits made me eager for a robot DR. They show no commitment to the patient or their health and are only involved in the most reluctant manner.

koan Tue, 07/17/2018 - 21:52 Permalink

I keep telling people AI will replace the lawyers and the doctors first thing, because they are the most expensive.
Doctors these days are over priced technicians that don't give a shit about your health anyway, so fuck them (if you're wondering about my attitude I have Kaiser ok?).
Wondering how big a prick your doctor is?
Check: https://projects.propublica.org/docdollars/

A Lunatic Tue, 07/17/2018 - 21:54 Permalink

And when all of this AI shit goes titts up there won't be anyone around trained for these positions AI has taken over; all across the labor market. We really are screwing ourselves good, in near limitless ways, with all of this technology....