Это видео недоступно.
Сожалеем об этом.
How AI is making it easier to diagnose disease | Pratik Shah
HTML-код
- Опубликовано: 3 авг 2024
- Today's AI algorithms require tens of thousands of expensive medical images to detect a patient's disease. What if we could drastically reduce the amount of data needed to train an AI, making diagnoses low-cost and more effective? TED Fellow Pratik Shah is working on a clever system to do just that. Using an unorthodox AI approach, Shah has developed a technology that requires as few as 50 images to develop a working algorithm -- and can even use photos taken on doctors' cell phones to provide a diagnosis. Learn more about how this new way to analyze medical information could lead to earlier detection of life-threatening illnesses and bring AI-assisted diagnosis to more health care settings worldwide.
Check out more TED Talks: www.ted.com
The TED Talks channel features the best talks and performances from the TED Conference, where the world's leading thinkers and doers give the talk of their lives in 18 minutes (or less). Look for talks on Technology, Entertainment and Design -- plus science, business, global issues, the arts and more.
Follow TED on Twitter: / tedtalks
Like TED on Facebook: / ted
Subscribe to our channel: / ted
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
I love his message. I'm a software engineer studying machine learning right now as a side project and a lot of people in the "popular" side of the industry, hence referred to as pop-ml, are obsessed with BIG data, it's a buzz word that I hear about 20 times a week. The message of LESS data and BETTER training.
So where is his work published? Has this or his technology been validated against current gold standard diagnostics in order to determine sensitivity and specificity?
He did not said that this is better than a standard gold but to show that there is an alternative on a Standard approach in AI technology.
Any idea of the sensitivity / specificity of their deep learning program at this point ?
Thank you - from a student paramedic on RUclips!
2OrdinaryGuys I’ve subscribed!
Chloe Jenkins wow thank you!!!
Thanks to research like this we can envision a near future where you can download a medical imaging app onto your smartphone and use its cameras to provide a rapid initial diagnostic capability, at home, that could alert you about a potentially serious disease while its in its early stages. Perhaps even in prodromal or presymptomatic phases of pathology. Imagine taking a selfie and your medical imaging app automatically screens it using a deep-learning neural network and informs you via push notification that you have melanoma developing on your face well before you or even a dermatologist could have identified it! Amazing! :)
Yes, you are spot on. That AI exists already. It is called VisualDx
Pratik, how does this really work? Did you make use of CNNs originally? But what did you do with the CNNs in order to intelligently dissect through the image and as it relates to the "color/pixel/geometry and rendering of the disease on the medical image"? This is pretty remarkable that you folks only used 50 images to train the neural network. And then used a regular camera and super-imposed the image taken over the trained image model that you have created. If I understand correctly. Remarkable! This could be a great module for the X-prize tricorder! And we could have all sorts of screening done at home for free and at a much much higher frequency and in a totally safe way.
Great! I was expecting this for very long..
very interesting talk.. I hope this makes a difference
"We had a problem where we needed 10 thousands images and thought, well what if we chopped one image up into 1 thousand images and then wed only need 10 images" this guys a genius
so now we can make our assumptions and extrapolate inferences from 10 images sliced 1000 ways instead of 10,000 regular images.. what?
AI, electrical ground and fuel delivery scanning with a camera drone should be able to pick up any problems in a machine for fixing, shouldn't be all that far away
That means some day it beats Dr.Gregory House to his diagnoses
Great presentation!
Do AI also helps in scanning or taking X-Ray of the chest through mobile phone
Maybe it will show you how to count change back
Brilliant!
I wait for the day that AI can not only diagnose disease but also cure them. Because there are currently loads of diseases that are easily diagnosable but are not yet curable even with early detection. Like Retinitis Pigmentosa.
Retinitis Pigmentosa does have an upcoming cure with gene therapy, according to clinical trial data
Anyone know if any of this work has been published? I've only see a paper on dental plaques
www.media.mit.edu/people/pratiks/overview/
www.pratiks.info/research/mit/
Not bad . Took me a sec to realize he was saying algorithms a few times
*I hope that this will help eliminate cancer.*
The Steadfast Duelist ...Don't listen to the quack snake-oil salesman "Awakened2Truth". They are a shill selling poisons by taking advantage of very sick and vulnerable people. Also, I also hope AI will be one many promising scientific and medical tools that will help us both better understand cancer pathology and treat it!
Does someone have a link to the paper? I can't find anything.
www.media.mit.edu/people/pratiks/overview/
www.pratiks.info/research/mit/
can’t understand how it works eventhough they dont have much data.
So, AI now will replace what the little work my brain does
Does anyone know about devices that can help stop a person from being electronically harassed?
brilliant
Very good
This is more like it TED
nice channel
The premise that costly imaging is used and expert physicians are used are on spot, but coming to a broad conclusion that imaging is used for diagnosis and is the limiting factor, makes the assertion not very fitting. AI can help in many ways, but not by over simplifying things, it can be done by understanding the process regarding the case in point properly and trying to solve it.
Ted where is the russian subtitles?
Ai will do what it is programmed to do, but if you program it to have emotions then starts the downhill.
Hi my name is connor, I'm the android sent by CyberLife
Clinic cua BS co nhan BS noi tru khong ? va co duoc cong nhan tai My khong ? neu co nhan thi nhan noi tru BS chuyen nghanh gi ?
I can see how this could make it easier for big pharma to sell even more drugs.
I saw this video from 3 months ago
engineers gone wild. great tech (maybe) but desperately needs some actual doctors helping them out. trials, evidence etc. just saying you will deploy to poor countries so its better than nothing is not good enough. you have to prove efficacy and that it meets standards of care. We dont throw experimental drugs at "third world" citizens because the unproven drugs are "better than nothing"
Just think of AI in 2030-35 . Lit🔥🔥🔥
Damn, you were my physics teacher..
ग्रेट वर्क फॉर युनिटी
Very good speaker but his tone was monotonous.
At least he isn't stood there just talking about himself and how brave he is battling nasty words and how oppressed he is despite being able to give a talk on TED
AI can be programmed to make deliberately wrong diagnosis.
Gotta listen to this at a speed of x0.75. Sigh.
Nicee
Those direct to use AI diagnosis or use AI diagnosis or accept AI diagnosis results will be regarded as murderers.
In before AI takes over the world
No, you're not.
All dislikes are from unsecure humans.
your sense of humor is horrible. Are you a primitive AI?
Why talk about something that does not exist!
www.media.mit.edu/people/pratiks/overview/
www.pratiks.info/research/mit/
med.ai diagnos.ai
1
2
3
Hahaha...
First like ❤️
A device that can detect disease, illnesses,cancers,disorders in one go a device that has a display the name of medication or fruit and natural remedy to cure, prevent, slow down or stop these illness diseases cancers in on go hopeful in coming years from now that may can do that and that tells you by blood tests in one go test whole body test in all in one test and may help future generations with kind regards? Im hopeing this information is helpful no lt harmful?. Im not feel 100% at the moment neither may comment next month or later this year 2022? A portable health device? With display screen?
Rehumanize the world dont dehumanize it!
People should not rely so heavily on AI it will be humanity's downfall
From AI's perspective, human is a disease that needs to be eradicated.
Precisely. We humans are born with sin thanks to adam and eve. Now if we create something in OUR image (AI) that will inherently be evil like us if it ever gets to that point. The bible talks in revelation about these things coming on the earth and Jesus prophesied of the last days too if he doesnt come back no flesh would be saved etc.. if your interested in the truth theres that story :P
Not interested in the bible.
@@mhtinla Yeah i figured. It is accurate and trustworthy though. AI is going to eradicate people like me first so when that happens you know lol
I think that believing we could fully understand what a true AI would be thinking is naive. Deep learning programs is another way of saying the computer learns from itself. So even the creators don't fully understand the primitive AI we have now. Just like you can't tell another persons motives just by looking at them, AI will probably think in a way that is beyond human comprehension. Now whether that is the end of the world, a utopia, or most likely somewhere in the middle, we won't know until it's to late to change course.
And how AI will be our doom
Ted ed
First comment
At last, your life has meaning!
What if AI turns evil? Who is to say it wont?
"What if AI turns evil? Who is to say it wont?"
I guess no AI will topple humans in that. If it happens, it will go against humans. By doing so, it will save thousands of species which are on the verge of extinction, thanks to human way of living. So, it will exterminate one to save thousands. Fair enough, my grandmother would say.
Yeah....now it's doctors turn ..to lose job b'cz of emerging AI...👍
Specialists are at more risk than nurses atm!
Humanity itself at risk! The future is war against the machines! People don't realize that they are giving their power away and losing their free will and human intelligence at the same time - Everyone will lose their jobs nurses and doctors alike !!!
Earth tech \ AI is so screwed when another Carrington event wipes out all the power like it did in the 1800's except back then we hard far less infrascructure than we do now!! Good times eh! I cant wait to see the zombies who are used to their phones being glued to their hand and eyeballs cringe in disbelief when it all eventuates - it will no doubt!
So you're thinking only about their jobs? And not the life saving benefits that AI is providing? 🙄
.. Ai.... ai will replace doctors in next decade, if don't agree, tell me the reason!
They will not replace but will assist. A.I. would take much more to understand the "Human Condition" and to deal personally with patients.
I don't agree about "next decade". Maybe in 100 years but probably it will be not enough. The reasons are associate to: -1 running time -2 missing of annotated causal clinical features -3 barriers of clinical decision support systems combined to electronic health records -4 medicine, at least for now, is not completely a science.
Ronak Patel Because of human nature. Doctors have gone thru many years of extensive training, delayed gratification, postponing their lives to study in exchange for the money and power their practice affords. They will fight before giving this up. The Medical Profession as a whole, is distrustful of new technology, slow to change their protocols, especially when there is no financial incentive, bc of the risk of lawsuits. And many patients are distrustful of machines. Possibly this invention to be used in poorer countries, where doctors don't like like kings, and prove itself there by overwhelming success. That is my hope. We need more low cost, effective inventions like this.
@@LordofSyn what if Google can make a assistant, that knows how do you like to get answer and how do like to get advice, and that will be better interaction than human does... Don't you think?
@@madarah8533 i trust numbers 99.92 vs 99.96 accuracy.... Which one would you prefer?