The Future of AI in Medicine - A conversation with Dr. Sameer Antani

Show Notes

Transcript

00:00:03:17 - 00:00:29:15
Yamila El-Khayat
I'm librarian Yamila El-Khayat, and this is NNLM Discovery. A podcast from the network of the National Library of Medicine. This podcast series explores how the NNLM is engaging with communities to provide access to trusted information for the purpose of improving the public's health. Today's episode is “The Future of A.I. in Medicine.” This episode makes me think about when we were kids and robots seemed so unreal.

00:00:29:15 - 00:00:48:11
Yamila El-Khayat
Or should I say TV’ish? I remember watching “The Jetsons” and thinking, this is way out of this world. Remember Rosie, the maid that would go around and sweep for the whole family? Now we have the robot vacuum that goes around the house and cleans for us. I didn't think I'd experience this in my lifetime. Griff Partington produced our story.

00:00:48:12 - 00:00:49:04
Yamila El-Khayat
Hey, Griff.

00:00:49:07 - 00:00:59:19
Griff Partington
Hi, Yamila. Today, we're featuring a researcher at the National Library of Medicine, Dr. Sameer Antani. Honestly, I didn't realize that the National Library of Medicine even had researchers.

00:00:59:20 - 00:01:00:19
Yamila El-Khayat
Oh, come on, Griff.

00:01:02:05 - 00:01:08:21
Griff Partington
Well, when I think of libraries and librarians, I think of them collecting books and papers and data from other researchers.

00:01:08:21 - 00:01:12:01
Yamila El-Khayat
And the typical checking in and checking out of books.

00:01:12:13 - 00:01:17:09
Griff Partington
Yeah, I didn't want to go there. I didn't want to offend our host librarian here.

00:01:17:23 - 00:01:21:08
Yamila El-Khayat
Offense?! I will take great offense to that Griff.

00:01:22:23 - 00:01:37:01
Griff Partington
Well, moving forward. Today, we're going to start with Dr. Sameer Antani's job title, which basically sums up the whole story. So easiest question first, if you wouldn't mind repeating your name and your role at the National Library of Medicine.

00:01:37:01 - 00:01:45:19
Dr. Sameer Antani
Sameer Antani, Principal Investigator at the National Library of Medicine, doing research in medical imaging, machine learning, and A.I..

00:01:46:09 - 00:02:00:11
Griff Partington
So he's a Principal Investigator, meaning he's a leading researcher at NLM. That's cool. We all get that. He's doing research and medical imaging. Yamila, can you describe some of the types of medical images The National Library of Medicine is cataloging and collecting.

00:02:00:15 - 00:02:14:16
Yamila El-Khayat
Absolutely. So there's all kinds of images that we got at the National Library of Medicine. We catalog X-rays, CT scans, radiology, all these images to help students, doctors and researchers learn more and to advance the future of medicine.

00:02:14:20 - 00:02:20:14
Griff Partington
And he's also doing research in machine learning and A.I.. Do you know anything about A.I. or machine learning?

00:02:20:22 - 00:02:25:00
Yamila El-Khayat
A.I. Griff, I've got an Alexa. Does that count?

00:02:26:12 - 00:02:38:05
Griff Partington
An Alexa is an example of A.I. or artificial intelligence. Artificial intelligence is the capability of a computer system to mimic human cognitive functions, such as learning and problem solving.

00:02:38:07 - 00:02:42:02
Yamila El-Khayat
Check you out, Griff. Please tell me you did research at a library.

00:02:43:06 - 00:03:24:18
Griff Partington
Yeah, I did do my research for this story. I had to keep up with Sameer. So imagine, the robot vacuum at home. An example of A.I. would be if the robot vacuum could understand what’s dirt and what isn’t dirt in a home while cleaning. And now the vacuum uses this information to predict where dirt may be in the future, in totally new areas. A.I. is an application of Machine Learning. Machine Learning would be the vacuum learning the layout of room. It's the process of using mathematical algorithms to help a computer learn without direct instruction. So it's like the computers going to school. The vacuum is learning and improving on its own based on experience.

00:03:24:18 - 00:03:33:01
Yamila El-Khayat
At least the computer is not going to like nag back at you like, Oh, I got to go to school like kids do, right? So it's kind of like the real deal.

00:03:34:03 - 00:03:37:10
Griff Partington
You know, I should have asked Sameer that question. That's a good one.

00:03:39:14 - 00:03:58:02
Griff Partington
He does have a few kids, so I'm sure it's easier to teach the computer with his algorithm than teaching his kids. So back to Sameer's story. Sameer is from India. He's from a family of doctors and engineers, and he grew up studying computer science engineering. Sameer, when did you become interested in artificial intelligence?

00:03:58:16 - 00:04:23:22
Dr. Sameer Antani
Machines and robots always excited me as a child. Having something move in the physical sense itself is exciting. We're getting something that's inanimate to behave like we do. We move, but we want it move on its own and with purpose. Now I take that idea and translate it into my passion, which is computers. The motion for me is to make computers improve biomedical research and clinical care.

00:04:24:09 - 00:04:31:22
Griff Partington
Why is the National Library of Medicine studying A.I., and how did you, a computer scientist and engineer, become part of this institution?

00:04:32:07 - 00:05:12:11
Dr. Sameer Antani
I'm at a library and I'm not a librarian. I'm a researcher. I have my Ph.D. and I'm an investigator at the National Library of Medicine. How does this all fit in? So NLM, for those who may not know, is one of the oldest institutes at that comprise the National Institutes of Health. And we go back to John Shaw Billings in the Civil War and collecting knowledge from the field of how the wounded were being treated and then making a repository of information so that the clinicians of the age learned how to advance their medical careers.

00:05:13:08 - 00:05:54:01
Dr. Sameer Antani
Fast forward 150 years, and we are in the age of computers. And the idea that a library of the future might have data that was visual and then look for patterns in these collections of visual data. So they were looking for postdoctoral fellows who might want to pursue a research idea in that realm. I had come from a world of video processing and image analysis and computer vision, which is trying to understand the content of pictures and the idea of similarity, visual similarity, which is a very simple concept when you think about it.

00:05:54:11 - 00:06:20:23
Dr. Sameer Antani
If I showed you two shoes, two pairs of shoes, you could tell them apart fairly easily. You can if I if I show you two kind of cars that are easy to set apart, but two medical images are not so easily separated because, one, they look very similar. All chest X-rays, for example, look alike. All tissue images tend to look alike. And the differences are at the disease level and they're very subtle.

00:06:21:15 - 00:06:30:02
Dr. Sameer Antani
And so looking for these minor patterns and then expressing them mathematically as as features, as we call them.

00:06:30:20 - 00:06:52:23
Griff Partington
So Yamila, Sameer is looking for patterns or features in the thousands of medical images that are gathered from around the globe at NLM. Yamila, I'm going to put you on the spot. I'm going to show you three chest x-rays. One has bacterial pneumonia, one has viral pneumonia from COVID, and the third has viral pneumonia without COVID.

00:06:53:09 - 00:06:54:23
Griff Partington
Can you tell the difference between the three?

00:06:55:04 - 00:07:02:00
Yamila El-Khayat
Listen, Griff, when I said I was a librarian, I meant it. There's a reason why these people go to medical school.

00:07:03:04 - 00:07:05:19
Griff Partington
I totally agree. I can’t tell a difference.

00:07:05:23 - 00:07:14:20
Yamila El-Khayat
What can I say, just looking at these three images, looks like lungs and looks super cloudy to me. That's about all I could see is black and white.

00:07:14:20 - 00:07:37:00
Griff Partington
I wish everybody else could see it. They look identical. But the reason I showed you chest x-rays is because this is one of the type of medical images that Sameer specializes in. He actually published a paper recently and he developed an A.I. algorithm that allowed a computer to tell the difference between the three. He was able to tell the difference between a COVID positive x-ray and a COVID negative x-ray.

00:07:37:07 - 00:07:40:23
Yamila El-Khayat
I don't know how that makes me feel knowing that a robot’s going to know more than me.

00:07:41:17 - 00:07:54:01
Griff Partington
Well, this isn't about robots knowing more than you. We're going to get into that more in a second. Let's jump back into our story where Sameer explains how pattern recognition works and how this can be applied in the field of medicine.

00:07:54:13 - 00:08:22:13
Dr. Sameer Antani
So pattern recognition is a family of techniques that looks for characteristics, pixel characteristics, inside an image and learns to recognize those objects. And then researchers like me spend time converting those visuals into numerical risk predictors that could be used by clinicians then to guide them, providing best therapies, best treatments, so that the best guidance can be given to a patient.

00:08:22:20 - 00:09:00:15
Dr. Sameer Antani
People have often asked me what is the value of my work? I often say that imagine your a clinician and you are presented with a patient with their images, perhaps an MRI, and you have to make a decision on how much risk the patient has for developing some disease. You can either look to yourself or based on your exposure, or you can look to guidance from a machine which has seen hundreds and thousands and millions of examples.

00:09:01:01 - 00:09:27:11
Dr. Sameer Antani
Every clinician have seen a certain number of patients in their clinical training, they perhaps have spent more time at hospitals or clinical centers and exposed to certain population, and they become very adept at that population. Machines, on the other hand, could be trained on data that is from different parts of the world, different ethnicities, different age groups, so that there's an improved care giving and therefore better expectation on treatment and care.

00:09:28:05 - 00:09:47:04
Griff Partington
I love how modest researchers are sometimes Sameer speaks so casually about his research, but I hope what he said is meaningful. He's building the A.I. tools that will help allow clinicians to make more informed decisions in the future based upon data collected around the globe. Yamila, are you a sci-fi fan, like Star Trek?

00:09:47:09 - 00:09:49:13
Yamila El-Khayat
Not at all. I'm not a Trekkie.

00:09:49:23 - 00:09:58:01
Griff Partington
Well, I hope what Sameer says next is meaningful to you. Here's how Sameer sees us using A.I. systems in the future.

00:09:59:00 - 00:10:21:19
Dr. Sameer Antani
I would like to see a future where all data about a patient is pulled in and guidance is given, much like the Star Trek and Babylon five and other TV shows that we grew up with, where the computers worked in sync with the human experts. Supplying other information that could be used to make an advanced, high-level decision.

00:10:22:04 - 00:11:05:00
Dr. Sameer Antani
That's a vision I see for the machine for machines that can augment human intelligence in the future. A machine can guide you if on the knowledge that it has gathered from a larger population of patients so that they can make a better informed decision for their patients. So providing guidance to the clinician carries a very high degree of responsibility for the algorithm, and the algorithm designer and consequently the researcher like me, who is at the early stages of trying to build something that has a predictive power but needs to be engineered into a solution that can be used by clinicians.

00:11:05:21 - 00:11:21:16
Griff Partington
What Sameer is trying to do is create the A.I. system that can predict diseases before a clinician would even flag it. Obviously, we all know that early detection is key for curing a lot of diseases. He's not trying to develop the Alexa Doctor.

00:11:21:16 - 00:11:26:23
Yamila El-Khayat
Griff, wait, did you just say the Alexa Dr.? No, come on.

00:11:27:13 - 00:11:32:03
Griff Partington
He's also not trying to replace doctors like some of the sci-fi shows that I've seen.

00:11:32:03 - 00:11:34:16
Yamila El-Khayat
Thank goodness.

00:11:34:16 - 00:11:48:18
Griff Partington
He's trying to provide additional information to a doctor so they can help guide the patient. Now that we fully understand what he's doing, what his research is. Here's Sameer talking about some of the specific examples of what he's doing research on.

00:11:49:07 - 00:12:20:01
Dr. Sameer Antani
My research is in advancing machine learning and A.I. for significant diseases, diseases such as cervical cancer, sickle cell disease and tuberculosis with HIV. Detecting and treating them at an early stage automatically could improve clinical care and save lives. So take cervical cancer. A lot is known about cervical cancer. If a machine could detect cervical pre-cancer, which is the direct precursor to cancer, we could actually treat the woman and give her quality of life back.

00:12:21:00 - 00:12:51:16
Dr. Sameer Antani
Now, sickle cell disease, which overwhelmingly impacts African-Americans. One of the risk factors is Cardiomyopathy or cardiac muscle disease, which leads to stroke and perhaps even death. So analyzing cardiac echo videos, using A.I. could be a solid predictor, along with blood tests to improve the chances of survival. I'm also studying the expression of tuberculosis, particularly for children that are HIV positive in the part of the world where CT machines are difficult to come by.

00:12:51:16 - 00:13:11:20
Dr. Sameer Antani
And chest X-rays is all you got. My A.I. automatically separates TB from pneumonia with high accuracy. The two look very similar on chest X-rays and require radiology expertise to separate them. By separating the two at an early stage automatically, we could provide the right treatment at the right time and save lives.

00:13:12:23 - 00:13:19:02
Griff Partington
Now, with your research, do you take a certain pride by the fact that you may be helping and improving and saving lives in the future?

00:13:19:03 - 00:14:19:15
Dr. Sameer Antani
Oh, absolutely. I think I think we have an opportunity before us where all our findings, once translated, would reduce the barriers to access to medicine, improve treat-ability, improve early detection, which leads to better quality of life. And in many parts of the world, seeing a physician is is one of the greatest challenges for the US in particular, where you have rural areas where you have probably have one or two physicians that are serving a community of thousands and much of the disease that is there is of the routine nature. That can be monitored with devices and and it is not an impossible future to imagine telehealth services which are not being observed or viewed by necessarily by a human nurse, but are being monitored by automated processes that are looking for changes in patterns. In a remote, rural region.

00:14:21:14 - 00:14:27:01
Griff Partington
Sameer, will you ever be satisfied with or is there always more more research to be done?

00:14:27:07 - 00:14:54:12
Dr. Sameer Antani
There's always more research to be done. I think we are at the cusp, at the beginning of trying to understand human bodies and the information that is contained in it. Each day, different parts of the NIH are discovering new things. I think that presents NLM with an opportunity to keep toe to toe and advance further so that the two can work together. There is always a new tomorrow.

00:14:55:14 - 00:15:05:18
Griff Partington
As humans, I think we're all looking for a new tomorrow with improvements to our health and longevity. So thank you NLM and Dr. Sameer Antani for your continued research.

00:15:06:14 - 00:15:33:09
Yamila El-Khayat
Subscribe to our podcast, leave us a rating and comment, and share these episodes to help us grow our audience. For more information about NNLM the NLM and supporting information about the story, check out the links within this episode's description. The NNLM offers free training partnerships and many resources that help to improve health and wellness. Learn more at NNLM.gov.