Name: How to Make Medicine Personal Again
Uploaded: 2019-04-02
Duration: 10 min 19 s
Description: Putting the “person” back in personalized medicine. Leading physician Dr. Eric Topol shares the key takeaways from his new book.

Deep Medicine, your new book, and it is extraordinary. I'm gonna hold it as tightly as I can, so they can get a close shot on this, all right. The two takeaways that I have are, one, the one that might be expected, which is AI is gonna change everything in medicine. The second takeaway was much more surprising, which is you're optimistic about it. And your argument is that it will actually, the AI tools that we'll have, will give actually more time for doctor patient interaction, more time for humanity, more times for doctors to practice old type of medicine. Right, and that's really the key. I think The more short term story is, as you say, the augmentation of performance, the accuracy, speed, listening costs of care, but the bigger picture is if we can really work on giving this gift of time. The decompression, the liberation of keyboards, and getting rid of data clerk functions of doctors and clinicians And all the other things that, make their lives easier, at the same time giving patients their data, algorithmic support, and so out further decompressing the role of doctor. So all that effort can then get us to what we've been missing for decades now. Which is the true care, in healthcare. So I think you make a really compelling case for that in the book, but what's surprising about it is that the progression of medicine with more and more technology has given us, has taken us here where the average doctor visit is now eleven minutes and even that. Most of it is spent looking at a screen or filling out EHR, electronic health records. So why do we think it's gonna change this time? Well, actually that's my biggest fear [ UNKNOWN ], is that if we don't take an activist role and we just let administrators continue to squeeze which is put us to the peak, burn out and depression among clinicians in history. Right. If we don't know, and if we don't use this time to turn back inwards, Then we will have further squeeze and things could actually get worse. So it's now is the time to take this remarkable improvement that we can get, extract out of the world of deep learning with all these data which no human being could actually really assimilate all of it And now we can really rev up that accuracy and get rid of so many error which is a big problem. But that's not the, the real story, I mean the one that I think is, that is possible, that is [ UNKNOWN ] informative but will require tremendous amount of effort. It won't happen, the default mode will be things will get worse So, I mean, you talked a lot about the sort of augmenting of human intelligence here we've heard that phrase a lot augmented intelligence, but when you describe it you're saying things like radiologists will have, you know, ai deep learning algorithms will review. Thousands of radiological images and then flag ones that human will look at afterwards. There will be all sort of warning system will flag and then real human doctors will come in and make the evaluation. Yeah, I think the key point there is, we have to fess up about how bad things are now that is radiologists have a false negative rate of 32 %. Wow across all the different scan, that's a lot of missed findings like pulmonary nodules. gastric neurologist miss small ( diminutive polyps ?} all the time and that are just as pre cancerous as the bigger polyps so all these things can be improved by deep learning and machine vision and they already have been shown to so its just a matter of getting that into clinical validation and even today there has been new fda framework about that So we don't lock algorithms, but we continually take advantage of their autodidactic quality. So that's really encouraging. You know you write a lot of that. You talk about deep diet, deep mental health, deep, you know, cross the board. You looked at AI applications. What was the one that surprised you the most in your three years of research? Yeah, well, it's interesting. The deep diet thing was I mean actually did the experiment where, you know, having a right record in your phone everything you eat and drink your activity your sleep, your meditation, all this stuff plus. Your, your microbiome stool specimen, and then a glucose sensor for two weeks. And I did this output of a based on an algorithm of thousands of people that was put together by the Weizmann Institute, Aaron Siegel and colleagues in Israel. And it was a big surprise because all these foods that I would never go near where the ones I got the A ratings. And the ones that were things that I actually like got very poor rating. But you have a specific issue which is kidney stones. And so there was a diet, this wasn't just an exercise for you and Academic [ UNKNOWN ] when this was really affecting a very painful condition. Yeah we'll see that's point is why we have to have algorithms that dig in this multimodal input that has lots of information. So that particular work was done without knowledge of other pre existing conditions. Right. And so when that output came. Things that are high- oxilade food, which would be bad for somebody with kidney stones, wasn't even taken into account. But the point about all this is that we finally have the absolutely proof that there's no diet for all people, that we have highly individualized response. Our glucose responds differently in each of us. Our triglycerides in the blood respond differently to the exact same foods, the exact same amount at the exact same time. Now that we know that we are on the way, with the these deep learning machine learning tools to be able to fashion bespoke diet for to promote health we're not there yet, but we're getting there but it means get working on all the data that's available. I'd love to get a question here in the audience anybody. Have some. You bring up this issue and everybody so individualized. That raises the sore of challenge that everyone is end of one, right? It's very, very expensive to sort of process to progressive medicine the way we have Without something like AI, without some preliminary vetting. Well I think you have hand on something as really very important. We won't be able to make these germs like the individualized diets, like the ability to quantify a state of mind for people with depression and [ UNKNOWN ] health conditions. If it wasn't for these really extraordinary algorithm So the deep learning has insatiable appetite for date, whereas we humans have early sotety. So we have a situation now where we can finally take this multimodal data and work with it to understand things we never did before. We've got a question right up front. Yeah. Just identify. Hi [ UNKNOWN ] the guy with the early copy. Can you tell us what you're doing [ UNKNOWN ] using AI. I know you [ UNKNOWN ] what are specifically oing yourself with the [ UNKNOWN ] issue. I Couldn't with AI why you're working in your industry to Scripps yes API key detail about what you're specifically working. Right. Sure, well we have working in genomics, to actually use deep learning to rev up the interpretation of genomes. To find which variants in a person's genome are actually connected to health and medical and also with sensors so we are a center of excellence for biosensors with Nvidia. And we're using that to work with a time series analysis of things like heart rhythm, glucose and many other wearable sensors. You talked in the book about deep medicine really being three different deeps. One is deep phenoyping, then deep learning and then deep empathy, talk a little bit about this deep phenotyping idea. Right, so that's really getting all the different information. About a person that you can to define their uniqueness, their medical essence. So isn't just their genome sequence or their micro biome or the environmental that you can quantify features but also their sensor's physiology, their scans for their anatomy, all their electronic records. When you have all that data And the corpus of the medical literature, you can do some amazing things like a virtual medical coach to promote your health. So deep phenotyping, again, involves not just today terabytes of data, but it's continuously being updated. And if we have that data and we use deep learning to To process it that's where we can get to this state of deep empathy which is what is restoring that human connection that has been largely lost .>> You know academic researchers you've gotten some phenomenal reviews in the science science journal nature the lancet. So on academic side people have really upload this our medical doctors clinicians responding to this new thesis. Well I think it's good to have some circumspection. But what we need is some hope. Right now things are really not good and so many people are not just burnt out but are retiring early. It's very common for a physicians to tell their kids, don't go in the medicine. Mm- hm. I mean, it's not a good scene and it's also remember too that with burn out, there's a doubling at least of medical errors. So it's a vicious cycle that just fosters more burn out and depression. So we gotta break that up. And so the the people I speak to I ask well, do you have another idea? How are we gonna get out of this? And I haven't heard one yet. So it's reassuring that we have a path, and that if we work on it hard, we might get To get there. Dr. Eric Topol, his book is Deep Medicine and we will back with you in just a short minute break. Thanks. Thank you. [ APPLAUSE ] [ BLANK_AUDIO ]

How to Make Medicine Personal Again

Health

Putting the “person” back in personalized medicine. Leading physician Dr. Eric Topol shares the key takeaways from his new book.