Hello, and welcome to the NVIDIA AI Podcast.
I'm your host, Noah Kravitz. We talk a lot about technology, and data specifically, impacting all facets of modern life.
In this episode, we're going to look at data's role in the fight against one of the biggest threats to life as we know it, cancer.
We'll talk to Richard Wender, Chief Cancer Control Officer at the American Cancer Society, about the role of AI and machine learning in his work. and how technology is key to redefining how we look at and fight cancer.
Richard, thanks so much for joining the podcast.
I'm delighted to be here. How is the American Cancer Society using AI, machine learning, big data, all of this stuff in the battle against cancer?
Our mission is to lead the fight for a world without cancer.
You know, and each of those phrases matters. going to feel satisfied if we just help some of the people.
We're really committed to making sure that everyone benefits from the advances we've seen. every community in the United States and around the world.
And we are a global organization, particularly doing a lot of work in sub-Saharan Africa. where many people get cancer are never treated.
So, in reality, Noah, we're in a remarkable moment.
And I don't think there is a disease state that is more dependent on advances in technology to improve treatment than cancer.
And that's because just a few years ago we thought, oh, maybe there are 500 or 1,000 different types of cancer.
We know today. that there may be millions of different types of cancer, each defined not just by what organ they started in, or what they look like under a microscope, but we're now defining them by what their DNA sequence looks like, what their molecular signature is.
Even that we're going to learn is not sufficient.
It's going to be the cellular environment around it.
And then we're literally using that technology to develop tailored or targeted chemotherapeutic drugs or immunotherapy.
They go after that exact mutation abnormality or as close as we can to that exact mutation abnormality.
So just think about it. You have a patient who for years, you said, well, they have stage two. breast cancer and they have this hormone abnormality.
Well, you know, now when we get the molecular data, we're going to learn that they have a very specific very personal subtype of cancer.
And what you thought the treatment might have been is very different.
So literally occurring today, but within five or 10 years, every cancer treatment team in the United States and in high developed nations. and eventually around the world, will be completely dependent on feeding that patient's data into a large data set that can use machine learning to combine clinical information gathered from other patients in the past who had similar molecular signatures to spit out treatment options at the point of care at the time that you need it.
And then that, oncology team is going to be responsible for sharing that data back into these large data sets.
So there are sort of two facets of what you said that struck me.
One relates to sort of identifying the type of cancer.
And you mentioned that moving from a smaller set of definitions that are linked to, oh, it's pancreatic cancer, it's stomach cancer, it's breast cancer, moving to the molecular signature, which is something that... not to put words in your mouth, but technological advancements, I assume, have made that possible to... identify things on that level.
And then the second is now you've got many more definitions or types of cancer that you can identify, which opens up the possibilities for more types of data, more sortable, granular data feeding all of that data now into machine learning systems, AI systems, to help doctors diagnose, treat, and tailor treatments to individuals.
You did a really good job at summarizing that.
That's exactly right. It's really all I do in the podcast is listen and spit back what you said.
Well, you did a heck of a job. I'm impressed.
So that's exactly right. That, frankly, is just scratching the surface.
Think how independent we are on medical imaging today.
I was at a presentation right here at our conference earlier today about better harnessing of CT scans using machine learning. teaching them how to detect the texture and patterns of pancreas cancer is one of the most difficult cancers that we face to diagnose it more accurately, to find the stage of it more accurately, and to see if patients are amenable to surgery because if you find them too late, surgery is not an option.
That's going to be applied all over the country. just to take it even a step further.
Right now there's about 65 National Cancer Institute designated major cancer centers and about a thousand commissioned on cancer accredited hospitals.
They treat about 70% of the cancers, but there's a lot of variability even within that group.
That still has 30% of cancer patients in the United States treated at a facility that's not accredited specifically for cancer, often in rural areas.
We're not going to train enough oncologists who are going to live in each of those communities.
We're going to need to harness robotics.
We're going to need to harness distance learning.
We're going to need to harness technology. advanced telemedicine.
And some of it will require machine learning.
Some of it requires just really smart applications. of technologies we already know about, but solving them from the ground up.
What's the problem we're really trying to tackle and come up with the smartest solutions to do it?
It's very aspirational to think that a cancer patient, no matter where they live... could receive close to the highest quality care.
But if we don't set that aspirational goal, we have no hope of achieving it.
Well, your job title alone, Chief Cancer Control Officer, it's an aspirational title, so to speak.
It is. What's a day like for you or a project?
I mean, talking about cancer across the United States. a huge topic.
It was huge in my brain before listening to you explain how we've moved from you know, thousands to perhaps millions of types of cancer.
So how do you get started? What are all the things on your plate that you're Now, I don't want you to sell my job short because I lead our global team as well.
And fortunately, I have the privilege of working with an amazing group of people all around the country.
I can only imagine. And you're right. You did speak earlier about the global aspects of your work.
You know, here's the answer. About half of everyone developing cancer today has a cancer that might have been prevented By applying the things that are already proven and known.
So a great deal of... what my team tackles with our volunteers by the way around the country is putting the systems into place harnessing the systems to make sure that everyone benefits by what's already proven to work.
But again, just think a little bit about how technology can help us do that better.
Let's pick an example I'm involved with right now.
We've launched a few years ago, 2014, at the National Press Club, we launched a major campaign called 80% by 2018, 80 by 18. which is to make sure that 80% of eligible adults are screened for colorectal cancer by the end of this year, by the end of 2018.
About 150 organizations have achieved that 80% goal.
To back up, to be pragmatic, what does eligible adult mean?
Because maybe there are some listeners who should be screened.
Absolutely. Today, what it means. First is that you know your family history, you know your personal history.
So if you have inflammatory bowel disease, Crohn's disease, ulcers, colitis, if you have a personal history of polyps, or if you have a family history of someone who had colon polyps or colon cancer, you're not average risk.
And you may need to be starting screening as young as even in your 20s, but certainly by your 40s or 45.
So that's everyone's responsibility starting at a young age.
If you don't have those criteria, you're considered at average risk, not low risk, average risk.
80% of people who develop colorectal cancer have no special risks.
So that average risk screening starts at age 50, no later than 50.
We see too many people waiting until 52, 53.
Right. And colorectal cancer incidence is rising in these young people right now.
It's a really serious problem. We're looking at it very carefully.
So by 50. Everyone needs to be screened and some high-risk people before that.
So that's what we mean by the 80 by 18 goal is to get everybody up to date.
But guess how that gets done? it gets applied by applying evidence-based interventions that are proven to work.
And Every time we find a system that's hit 80%, 100% of the time.
And we ask them, how did they do it? We find the same elements.
We find a champion, we find a leader, We find a committed system that has applied not one, not two, not three, not four, but usually five or more evidence-based interventions.
So how does that happen in a systems approach?
Well, it depends on why's application in many cases, not all of them, of the electronic medical record data. that we've entered that data in a searchable field that the computer can recognize, the software can recognize. identifies the risk level of the patient, can look through natural language processing, because if you have to enter everything in a searchable field, it's not so easy.
But wouldn't it be wonderful if every EMR in the country could look through natural processing and say, oh, colonoscopy done in 2002. and capture that.
We're not there yet. That would be wonderful.
And so that when that patient walks into a office, immediately there's an alert comes out that that person is due for screening.
Now take it a step beyond that. not everyone comes to the clinician you can have machine learning do the entire process machine processing identify those at risk, do a automated generated phone reminder, and repeated phone reminders, and generate a letter.
It can all be done through technology application.
So how much of what stands between us now and us being able to take advantage of a system like that?
How much of that is technological and how much of that is whatever the right word is, political or whatever it might be around technology? notions of electronic health records and competing insurance providers and data portability and all of that stuff.
I don't know the exact percentage. Sure.
And you knew that when you asked the question.
Did I say percentage? You did, but it's okay.
That's quite all right. I knew exactly what you meant.
A lot of it. the lack of will. You know, a really organized will to say this is a critical, solvable problem and we need to solve it.
Because in reality, nothing that I just said – you know, I agree with you.
You kind of picked up when I was saying, you know, I said, gee, this sounds like things that should be pretty easy to do with, The NLP can be tough, but we have some good NLP out there.
Exactly. So what's standing in the way?
And what you correctly said is, gee, I have a feeling what's standing in the way is EMRs are trying to solve a different problem.
They're trying to make sure you have a note you can bill for.
And do we have enough financial incentive?
And are people really focusing on the right things?
So I agree with you. We do need... you know, to make sure that we're organizing healthcare delivery in a way that everyone feels the incentive to do what's right for an entire population, not just What in our kind of old payment models, which are based on how many treatments you gave.
Right. complex care generate the greatest surplus or profit for a particular entity So part of our job at the American cancer society is making sure that people do have those priorities straight.
I'll give you another good example though.
Um, I have a slide at a talk that I give and I said, Which technology is saving more lives, wearable fitness trackers or electronic cigarettes?
That's a provocative- That's an interesting one.
As soon as you said wearable fitness trackers, I said, okay, the answer is going to be B.
Well, good point, Noah. But that's interesting.
But I gave you an interesting B, didn't I?
You did. And, you know, what we know is that since electronic... And full disclosure, I'm wearing a fitness tracker now.
Got you. Throwing stones at my own glass house.
You know, and by the way, you know, certainly there are some individuals who've benefited from wearing fitness trackers, but you probably know the data because you knew to answer B.
That when you look at it at the population level, it's hard to show a big impact that's sustained on treatment change.
Look at what's happened with electronic cigarettes since they've been delivered.
You know, as soon as they came out, they became the preferred. nicotine delivery method for young people.
Many young people don't even think about smoking.
We would only... decide whether we're going to jewel or use an e-cigarette.
And smoking combustible tobacco is at an all-time low, both in young people and adults.
Look at Australia, which forbid e-cigarettes, their smoking rates are going up.
So now the problem is e-cigarettes aren't perfectly safe.
So there's a technology problem for someone to solve.
Can we find the right kind of nicotine delivery method? that is close to 100% safe, but it's appealing to potential smokers, it's cool enough.
And as strange as it sounds, we're probably preventing more death through different nicotine delivery than we are through fitness trackers today.
We just have to do it perfectly safely. Right.
Yeah. Our guest is Richard Wender. He's the chief cancer control officer at the American Cancer Society.
But don't be fooled by my own missteps in hosting here.
His work is global. And I'm wondering, we've been talking about broadly uses of starting with machine learning and big data. but kind of more broadly talking about all the different ways that technology can play a role in fighting cancer and hopefully eventually eradicating cancer from the planet.
Are there projects you have ongoing right now with the technology companies?
And you can, I'm not asking your name names necessarily, but I'm curious. what kinds of projects you have going on right now.
Sure. Let me just give you a few. And we're in discussions with a lot of technology companies.
One of the challenges... that we've tackled is that there are a lot of technology companies out there.
And figuring out the right partnerships.
And in a moment, I'd like to come back about the right business models, too, because that's a big deal for organizations like ours.
But here's some things we have going right now.
So we I have a wonderful app that is AI driven with an Israeli company called Belong.
Okay. and encourage people to download the app, play with it, take a look at it.
The app is called Belong? Belong. Okay. S-B-E-L-O-N-G.
And that's on Apple Store, Google Play? Yes, exactly.
Okay, perfect. And frankly, it's creating a connected community.
And with virtually no promotion whatsoever, in fact, ACS has done almost none, we've generated many, many thousands of members of this community.
We're able, they're connected with each other.
There's a lot of options for connecting with people with similar cancers.
Give you a great example. For years, we had a program.
We pioneered a program called Reach to Recovery.
It was kind of the first peer-to-peer program we ever had. where women who had had breast cancer often had mastectomies back then. were paired with other women who had mastectomies.
But imagine what we can do today. You know, now we can pair Not just women with breast cancer, but frankly, we can pair people with any kind of cancer.
How about if we pair them all the way down to their molecular subtype? and have people all over the world who are connected together, you know, through these kinds of communities.
We believe that's possible. And then in the community, the idea is to foster support networks.
Correct. And then, of course, the American Cancer Society is part of that dialogue, and we can connect with your cancer team.
There's lots of ways to use that. Let me give you a second one application.
And a big part of this partnership is with IBM and the IBM Health Corps, and that's our work in Sub-Saharan Africa.
So I'm not sure it's IBM that's developed this app.
So I apologize because I'm not sure who we did this with.
But we do have a pretty simple app where...
We did some real blocking and tackling. There's a wonderful organization called the National Comprehensive Cancer Network who writes treatment guidelines here in the United States.
But they're based on drugs available in the United States.
In sub-Saharan Africa, none of those drugs are available.
There's only generic drugs, and we're working to make those be more affordable.
They're hard to get even in Africa. Even those drugs are hard to get.
So we've convened a group of African oncologists, worked with the NCCN, and we've written treatment guidelines. by African oncologists tailored for African drugs that are available.
And now we've created an app. because here, if you have colon cancer, you'll see a colon cancer oncologist.
In Africa, you're lucky if you see an oncologist.
So that doctor can now take these guidelines that we've written and simply by asking, entering a series of questions about that patient, It will just take them immediately to the paragraph in the guideline that most closely applies to that Nothing fancy, I think, on the technology side, but really practical, really useful.
We are talking to an organization called Tempus, not done much yet, but they're a great example of a company's got big dreams. of bringing together big data, doing deep, deep molecular analysis of this data, and then Combining that with clinical information, just as I was saying at the start of the podcast, and being able to spit out those kind of treatment guidelines.
One of the roles we thought we might be able to particularly be able to play is just encouraging data sharing.
Sure. You know, the American Cancer Society is a big voice, big brand.
People trust us. We know that. And just saying this is one case where we can all learn together and it's okay to share data.
Right. And deep learning in AI is all about, the more data you have to feed in and the more quality data, the better the system works.
And it's interesting, one of the things that I've learned in doing the podcast and talking to people from... all sorts of industries, really, is the role of context.
And when they're training up their neural networks and they're the neural networks are learning and growing more intelligent, they're able to take context into account in certain situations.
And so thinking about what you were saying earlier about it's not just attached to the organ, it's a molecular signature, but then it also can have to do with the situation, the context.
And it's, whether you're talking about machine learning or something, It's technologically more simple, like an app that delivers text-based instructions.
That context is so key. i mean think about it i was talking just about the cellular context but let's talk about the social context We should be able to get to the point.
I think we are, by the way, getting there.
There's some work happening right now. where we also enter your insurance coverage.
This is a uniquely American solution to a uniquely American problem.
You know, where we enter in your insurance coverage or lack of insurance coverage.
We enter in cost data. We can generate some decision tools that are literacy adjusted and that allow people to make an informed choice, you know, say for much, much more money, there's a little bit more benefit.
You know, it's dicey, difficult, challenging things.
But I think value is a real hot word in healthcare right now.
You know, what are we getting for that dollar?
And when it's coming out of your own pocket, You know, it's very appropriate to do.
Earlier in the recording at the beginning of the conversation and then also before we hit record and we were just chatting before we sat down, you mentioned rural America a couple of times. and reaching the people who, and you're, talking about sub-saharan africa made me think of this as well reaching the people who it's not as easy for them to get to You know, a treatment facility and a credit facility, that sort of thing.
And you mentioned a few interesting ideas related to technology, autonomous vehicles, 3D printing, some other stuff.
Talk to us about that. What kinds of things, even if they're far flung?
I don't think they're far flung. In fact, I think they're going to happen.
They're not happening immediately, but I believe they will happen.
We are literally dwelling in the land of it will happen.
I'll confess, Noah, I can't remember how many years ago somebody came up to me and told me that, you know, there's something called the World Wide Web.
And everyone's going to own their own computer.
And I remember thinking, I have never spoken to a more misguided human being than that person.
I have learned to believe everything that my technology partners tell me because I know it's going to happen.
And obviously the ones I'm going to talk to about now, they're happening today.
It's solving the final. aspects to really take it into the real world in a reliable way.
So we know, by the way, improving access to high quality care is foundational for the American society.
That's a foundational platform. It makes sense.
We would not exist if it weren't for that platform.
Right. There's many other things we try to do, but we'll never depart from that one.
And here and around the world. So as a result, for years, we have tried to solve just the basic problems, advocating to make sure people have insurance coverage. providing patient navigation.
We have a 24 hour phone line, 1-800-ACS-2345, 24 hours a day. that provides basic information, good support, guidance about how to get through things.
This is who we are, our website, cancer.org, the most accessed. cancer website in the world so that's always been who we are but part of our mission has been to tackle the very practical problems of getting to therapy and having an affordable place to stay.
And the data are stunning and frightening.
Very substantial percentage of patients, 5% to 10%, depending on where you live, who do not complete care.
They don't initiate care in a timely fashion.
They don't complete care. And there's even a small percentage of patients, as many as 3% to 4% in certain parts of the country who never get treatment for their cancers at all.
Part of those barriers are financial and just overcoming the practical realities like driving, getting to treatment, having a place to stay.
For years, we've had a volunteer driving program called Road to Recovery.
We're now working with Lyft and Uber and all the other companies to try to find out different solutions for transportation.
But when I'm sitting looking about where autonomous vehicles are going, I say, this is the future.
We can't find volunteer drivers. in Wyoming who will drive 100 miles a day, five days a week for radiation therapy.
We can't find that. You know, our autonomous vehicle driver is real cool with doing that drive, doesn't bore them at all.
They're perfectly fine doing that. We have a Hope Lodge program, and we're building the first Hope Hostel, which is a more acceptable term, in Kenya right now because people are traveling hundreds of miles. have no place to stay, they stay outdoors.
I have videos of 3D printing of homes in 24 hours for $10,000.
You know, we can solve these kinds of lodging problems, and they don't have to be fancy.
No, sure. Very, very simple. But safe places for people to stay.
But if it's the difference between getting treatment or not.
Of course. And we know that in sub-Saharan Africa, people will make that trip once.
But if they have to come back, they often don't do it again.
So, boy, I must admit I get awfully excited. when I think about those very practical applications.
We love our volunteer drivers, believe me.
They're the greatest human beings. And it's an amazing experience for the patient.
Sure. So I don't mean to ever downplay that.
But there are just limits to how many volunteers we can do it.
So I see these solutions and I go, that's our future.
We need to be a part of that. So for folks listening who might want to join you and being a part of it, you mentioned the website, cancer.org.
Anywhere else they can go to find out more about What you're doing broadly, any of the specific programs?
Sure. I mentioned two things. Cancer.org and We love our website and our mobile version is pretty cool.
We just updated it. So if people are more comfortable looking at it on their phone, please do.
There's an amazing amount of cancer information as well as information about what we're doing.
But I'd also encourage people just to give us a call, 1-800-ACS-2345, 24 hours a day, 365 days a year.
These folks who are answering these calls know our organization well.
They can link people facing cancer up with specific services in their community.
That's, by the way, will be another great harnessing of technology because right now it's a lot of searching to get that data.
Right. The better and better we can make that, the easier.
The more that it can be available in the palm of people's hands, the easier.
So we're working on all of those issues.
But those are probably the best two places to go.
And then they're very happy to connect. you know, really motivated, skill-driven volunteer who wants to contribute to some of this work to the right person within the organization.
Great. So the AI network architect who's out there, who's been listening the whole time, just thinking, I have this idea for you, Richard.
They can just get connected to you, call the 800 number.
Absolutely. And ask for Dr. Richard Wender.
I promise it'll get to me. Perfect. Well, Dr. Wender, thank you so much for taking the time.
And I feel like it goes without saying, but I'll say it.
The work you're doing is of the utmost importance.
And on behalf of everybody I work with, all the best of luck to you.
And we just hope that tech folks can crank out the tools you need to keep doing the work you're doing.
Thanks, Noah. I'm very confident. that they can.
And it's a two-way street. We've got to be receptive and make sure that the opportunities are there to get them to contribute.
Absolutely.
