Event Recap


Aug 19
Alley Team
Event Recap


Aug 19
Alley Team
Event Recap


Aug 19
Alley Team
Community Over Everything



Smart technology touches every facet of our lives, especially in the age of COVID-19. Having more technology means having more data, which isn’t necessarily a bad thing.

Big data is transforming the health sector, making it more equitable and accessible to all, and helping doctors treat, and diagnose, patients in real time. Gone are the days where doctors had to wait on patient charts or test results—with 5G and mobile edge computing (MEC) technology, patient data can be processed, and accessed, instantaneously, catalyzing the treatment process from start to finish.

Increased access to health data can shorten wait times, offer more accurate diagnoses, and even help eliminate language barriers between patients and providers. Even better, big data, coupled with MEC technology, can strengthen machine learning to help physicians make diagnoses with near-real time analysis, fundamentally changing the diagnostic process.

Hear how the increased availability of data, coupled with MEC and 5G connectivity, is unlocking solutions in healthcare, eliminating barriers and making quality care more accessible to all.

Joshua Ness
Verizon 5G Labs
Randi Foraker, PhD, MA
Institute for Informatics
Bruce Lampert
Avesha, Inc
Rohit Shirish Saraf
5G Technology, Verizon
Jon D. Morrow, M.D.


Joshua Ness  0:00  
Hey everyone. Thank you for joining us here on our Wednesday afternoon. My name is Joshua Ness and I am a Senior Manager with Verizon 5G Labs in New York City. We're very excited to host this event as part of Verizon 5G Labs virtual event series. For those of you unfamiliar with 5G Labs, we work with startups, academia, enterprise teams, and many others to build a 5G powered world using the practical application of emerging technologies. And part of that mission includes having conversations like this one that address barriers to digital inclusion and create opportunities for communities to thrive and grow. If you're interested in learning more about our work at Verizon 5G Labs then you can visit verizonglabs.com I promise, I'm not going to be saying Verizon 5G Labs the entire time. We also want to thank Alley for helping us to create the space to host these events. Alley is a community agency that unites rich and diverse communities around the country with corporate partners to provide the resources and catalysts to drive positive change in technology and in the broader world. So without further ado, I want to turn it over to our incredible panelists who are with us today. We have a rock star lineup of folks in the healthcare and technology spaces. And I'm excited to invite them to introduce themselves and their organizations. Jon, let's start with you. Oh, you're on mute. I think

Jon D. Morrow, M.D.  1:28  
I was saying I'm Jon Morrow, but I would I should have said is I'm muted Hi, Josh. Thanks. My name is Jon Morrow I am with MDClone. I am an obstetrician gynecologist in New York and a medical informaticist and an MDClone. I am the Senior Vice President and Physician Executive. MDClone is an Israeli based company. We have a product that brings data to the masses. We aggregate all different sorts of data into a data lake and then surface it to ordinary users in the clinical setting either administrative users or clinicians in order to do hypothesis testing, quality improvement, and then also generate synthetic data off of the data set to allow to remove the privacy implications from idea generation and looking at data.

Joshua Ness  2:28
Fantastic. Thanks, Jon. Randi, turn it over to you.

Randi Foraker, PhD, MA  2:32  
Hi, everyone. My name is Randi Foraker, and faculty in the School of Medicine at Washington University in St. Louis, and at Washington U. I direct the Center for Population Health Informatics. And the Institute for Informatics is designed as a horizontal institute that serves not only the School of Medicine campus, but the main Washington campus as well. And our role is to serve as the hub for Informatics Innovation. And we have faculty that work in the translational bioinformatics space, the applied clinical informatics space, as well as population health informatics. And a lot of my work focuses on optimizing clinical data for risk prediction, and also developing and testing clinical decision support tools at the point of care that are embedded in the electronic health record.

Joshua Ness  3:35  
Wonderful. Thank you for that. Bruce, let's go to you.

Bruce Lampert  3:39  
Yeah, thank you, Josh. So I'm Bruce Lampert, the SVP of business development and strategic partnerships for Avesha. I've been involved in eight startups over my career. My expertise is building new ecosystems for emerging technologies running from video streaming, to networking. To robotics, Avesha is an exciting new technology. It's a new startup. And we are transforming how applications traverse the network meaning how applications move from the cloud to this new cloud edge or MEC as Verizon calls it. And we use a very innovative technology called application slicing, which takes the complexity out of the network in puts tools in the app developers hands so they can simply port their applications from the cloud to an edge environment. Today I'll be discussing Avesha's AI inferencing platform. We use the 5G we use the MEC Verizon's edge for real time polypdetection in workflow automation.

Joshua Ness  4:50  
Very cool. I think we're going to talk some more about that here in a little bit. Rohit, let's round it out with you.

Rohit Saraf  4:56  
Thanks, Josh. Oh, have you been My name is Rohit Saraf And I am a Senior Strategic Partnerships Manager in the 5G Technology team at Verizon. I work on external partnerships with companies, from startups to large corporations, academic and research institutions to evaluate come up with sort of initial proof of concepts and subsequently do trials at our 5G Labs. And ultimately, the aim is to develop, you know, go to market solutions in various verticals, such as healthcare, robotics, AR VR, and some combination of these. So specifically, in regards to health care, I've been looking at various use cases from telehealth and remote patient monitoring to AR VR based medical training, things like workflow optimizations for medical staff, leveraging technologies, such as computer vision, AI, big data and analytics, right? So pleasure to be part of this panel today, and the experts in the healthcare field and our partners. So excited to be here.

Joshua Ness  6:03  
Yeah, likewise, Same here. Before we kick things off, everyone in the audience, please make sure to drop your questions in the Q&A feature here in zoom, we're gonna monitor that and insert those into the conversation where applicable. Also, FYI, this event is being recorded and will be made available in the next few days. So I want to frame things here really quickly, where there's there's several words that are being thrown around in terms that you're going to see used, things like mech, and edge, and cloud and things like that 5G included. 5G is the is the fifth generation of wireless connectivity, and you see it in the news, you see it on TV, it's kind of everywhere, right now. It's kind of one of the one of the technologies that a lot of people were talking about, including, as well as technologies like cloud, we saw that Microsoft's third quarter profits went up like 47%, because they have such an increase in their cloud business. Well, that's not going away. None. Certainly, it's certainly if anything, it's advancing. And so when we use terms like edge and MEC, what we're talking about is taking that cloud and being able to bring it closer to the edge of the network where the customers are using unless it customers that could be a customer like you or me, or it could be an enterprise customer, it could be a doctor who's working in a hospital or a clinic, that is using applications, wireless applications that are writing on 5G that are living on that connectivity medium, that are then using edge servers to process and analyze that data in almost real time. And we're talking single or only double digit millisecond latencies. And that's very, very fast. And it becomes really important to, to create the opportunity for those speeds when we're talking about situations like we're going to be talking about today. And so when you hear when you hear any of the panelists talking about MEC it's an acronym, it stands for multi access edge computing, it's fairly synonymous in conversation with the term edge. And that really just means taking those cloud servers that usually could be anywhere, and bringing them really, really close to where the applications actually being used. And, and then 5G is a wireless medium that transports that data very, very quickly, and with very low latency. And so that's how, as we're thinking, that Verizon and beyond, as we're thinking about how to provide value back to enterprise customers, or ultimately to consumers as well, we're really talking about the the marriage of these two technologies to create the the space for some of the folks on this panel here today just start creating new applications to start creating new value for people in both the business and the consumer space. So we're really excited to be talking about how these are impacting various industries, including healthcare. And that's what we're going to be talking about today. So I want to talk about how having data at the the right kinds of data at the right time, how that impacts the work that you're doing, and some of the partnerships that are being created. And so I'd like us to, to kick this off and talk about the how having that data at the right time and being able to create inferences from it and being able to create new value out of it, and how it relates to the work and the partnerships that you have with one another. And I'm hoping we can just talk really quickly about that. Randi, I'd like to start with you and to talk about how that's actually impacting the work that you're doing.

Randi Foraker, PhD, MA  9:27  
Sure. Thanks, Josh. As I mentioned in my introduction, I design and implement clinical decision support tools that are embedded in the electronic health record. And these tools capture and visualize relevant data at the point of care, so they populate automatically with electronic health record data. And in this way, health care providers have the right data at the right time, and quite frankly in the right place during a patient provider encounter in order to inform decision making. And the alternative strategy would be to hunt around the electronic health record in order to extract the lab values that you need and the demographics that you need in order to more comprehensively understand a patient's health or even their risk for a given disease. So we really seek to streamline that information and get the necessary data to the right person at the right time.

Joshua Ness  10:29  
Hmm. And, Jon, it sounds like some of this, some of this speaks directly to some of the work that you're doing with MDClone.

Jon D. Morrow, M.D.  10:38  
It does, and there we go. It does. And it also speaks to the work that MDClone is doing. with Randi and with with, with her organization, we, in the old days, if you wanted to ask a question about the clinical data, if you wanted to get a sense of how you were doing on a, you know, with a quality measure, or with identity with closing gaps in care, you know, let's actually talk about that if you want to, if you want to find out where your gaps in care, or where your inequities and care are, that would, in the old days take a lot of specialized knowledge in the database structure, it would be removed from the user, the user would have to open a request, perhaps do an IRB request or do a business intelligence request, in order for some programmers somewhere who are familiar with the data structures to generate a query in Sequal or some other language, which would then result in a report that may or may not resemble what you were looking for in the first place, and may or may not be clinically or administratively useful, by bringing the information to the users. And you know, by making that throughput faster, and by making the data more amenable to be used by the users, you can spur innovation, the end user, the clinician, in this case, or the administrative person in a hospital can ask a question of the data in a plain language or with a visual interface at the point of care or, more likely, or more specifically, at the point of inspiration of innovation, you have an idea, you immediately go test it, you identify where you need to look further or changes you need to make you make the changes, and then you act on them again. And so that you know bringing the information and the ability to query the information closer to the hands of the users, allows those users who are the influencers to influence the outcomes, and you get a really tight cycle and improvements in care and MD clone tries to does bring that information and in a way that that it can be accessed directly.

Joshua Ness  12:42  
Interesting. All right. It definitely is, is good too as we're creating these, these new types of data, and these do these new sets of data, and we're gathering it in new ways, and we're creating like you mentioned inferences off of that to be able to make these things available at the time that they're needed the most by the people who need them. And I think we're going to touch on that a little later. Bruce, let's let's kick it over to you. Where Where does does the the way that the the algorithms and the technologies that Avesha is running? How does that how does that impact? Or how is it impacted? And how does it help to create some of this value that Randi and Jon were just talking about?

Bruce Lampert  13:21  
Yeah, we talked about getting the right data at the right time as a catch phrase. And it's very relevant for us. Because what we're doing is we're we're grabbing the video signal in real time from the device, we're sending it through 5G to the edge to the MEC, and then we have an AI inferencing engine that's on the edge. So it's close to the hospital. We then through AI inferencing engine, our model will, will do over a millions and millions of permutations to find an object of interest, as we call it. And that's looking for a polyp, we then send it back through the edge through 5G, and we superimpose it on the monitor. And then the doctor can actually see there's an object of interest. And this is where the data starts. We have a voice activation where the doctor says take a picture, that means that overlay on top of the image is important. And they want to start either biopsying it or gathering information on it. We then extract information and we take all this raw data. And this is where Jon and Randi's technologies come into play as part of the ecosystem. We don't know what to do with this data. we're capturing it, but it needs to go somewhere it needs to go somewhere to create more value for the for the institution for the doctor for the patient. And we would then through API's integrate with their companies, and they would have all of this great information that we're capturing in real time that the doctor would have this extra pair of eyes. So that's how we would sort of play within the data capture. And then let those guys do all of their Advanced Data lakes and data research to enhance clinical insights. And you know, in really what's important is, is performance measurements. AI is going to raise the bar. And if they can create real performance reports and performance measurements, I think that's going to just help raise the bar for health care overall.

Joshua Ness  15:24  
Yeah, and we've we've done a little bit of work demonstrating that after we wrote it, can you talk a little bit about some of the work that Verizon has done with the Avesha and how we put some of this into practice?

Rohit Saraf  15:34  
Yeah, so what we did with Bruce and his team in his journey, you know, leverage the power of our edge computing, we refer to it earlier, Josh, MEC, right, multi access edge computing, we have announced that we are alive in some cities with our AWS, waveland edge compute resources. So we took AI inference engine, what, what Avesha has, and that really, you know, besides in the MEC, and on the think of it as we are the solution by that higher level, it's about democratizing healthcare. It's about enabling expert advice to wherever possible. So in Trauma Way Waltham um, 5G innovation labs, we had an end to end system where we could capture the video, in real time, it's streamed at a very high def, onto the cloud. And within split seconds are really, essentially, the old time updates less than 10 frames that you're getting an expert advice with a green box around at, let's say a polyp telling you that yeah, this is this is a detection of a polyp. So now, so this was done in Waltham, and I'm gonna imagine hospitals or doctors who are not that, you know, equipped, or they are not, at that level of expertise can still use or leverage the expertise of someone remote, and still do really near real time, precision detection of something like polyp and you can expand this to Avesha, is one of our partners, but, you know, you're working with several other companies for various use cases. And I always talk about, you know, value based patient journey, they start from home, come to, you know, doctor's office and go to the facility. So there are various pain points today, in, in patients that wide availability of this data will help unlock right there are various use cases that will help unlock remote patient monitoring, fault detection, chronic chronic diseases. And Avesha is one of the example where How can leverage the power of AI and further democratize the whole, you know, healthcare and make quality health care accessible to all.

Joshua Ness  18:01  
You mentioned that democratizing healthcare and I want to I want to touch on that just a quick follow up and is, is the idea there that with, with technologies like Aveshas running on a platform like 5G and edge computing that certain diagnoses can be made, even if a I don't use the word adequate. But even if a doctor doesn't specialize in that particular part of healthcare is, is administering that test. For instance, if somebody who isn't as quite adept at detecting colon polyps, they can still perform the procedure and then the technology can then deliver the same result to any type of doctor who's performing the procedure, regardless of how, how well versed they are getting the procedure, is that correct?

Rohit Saraf  18:48  
That's correct. And there's one way to look at democratizing the healthcare aspect, right. But when you talk about AI and the general availability of huge compute resources, a lot of verticals in a doctor's flow, like workflow optimization you can do that enables doctors to, you know, send patient data to EHR as rapidly. One part is what we did with Avesha second part is medical based training for medical staff and nurses. So, all these holistically if you take a look, Bill, they all need real time, you know, data processing capability, the ability to send data in near real time or ability to process data as close to the source of the data as possible and send results back to doctors or patients to to to satisfy your particular use case. So that is what I mean by that of helping Democrat in the healthcare and making quality health care accessible to all.

Joshua Ness  19:48  
Thanks for that. Just addressing a quick question there in the in the Q&A. Regarding edge computing, or MEC, it does mean Jim that a certain type of server perform processor does need to be at the edge. And the benefit of the of the cloud is that it started off in, you know, in the in the 70s and 80s started moving things away from on site mainframes into the cloud where things didn't mean you need to have that that storage and processing capability there on premise. Now we're talking about bringing some of those compute resources and capabilities closer get closer to the user, not necessarily on site. So it's not mandating that a company or an organization needs to operate its own mainframe technologies and have people in the CTOs office to do that. Now, some of these things that might be a commercial network, it is a it's available for several different entities to use. And there can be certain organizations that get certain slices of that capability. Or in some instances, it might actually be private, and then placed at the at the facility, but then co maintained by the Verizon and the customer there. So hopefully that clears up some of the some of the confusion. But it's a good question. I'm glad you asked. And thanks a lot.

Jon I want to get back to something that I think you mentioned, it was the global network that MDClone is building. Why is a collaboration to like this vital as technology continues to march ever onward in the healthcare industry? Like we know that it's showing no signs of slowing down. And especially if the pandemic is any indication, healthcare is going to continue to be top of mind for people moving forward in the technology space and how technology can be used to advance healthcare capabilities. Tell me me how this global network is going to be something that that is quite necessary as this occurs?

Jon D. Morrow, M.D.  21:40  
Thanks, Josh. Yeah, so we're building a the global network, which is a gathering a consortium group, if you will, of researchers around the globe, who have common interests, or who will develop common interests in research and share projects and data and information. In order to facilitate research, research can't be done in a crucible. we've all learned very well over the past eight months, unfortunately, how a local problem can become very quickly a global problem, and you can't solve a global health problem, whether it's Coronavirus, or whether it's diabetes, if you're acting in a you know, in a microscope locally, you need to disseminate the information. And so we're building a community and infrastructure to share information and knowledge and projects. And you start by by opening those doors and those avenues. And, you know, we have a tool that allows data to be liberated and data in all different sorts of forms and formats to become accessible. So make it accessible on a broader scale, not necessarily by sharing the whole data warehouse, but by making information exchange and the exchange of ideas available. You know, I heard, Rohit mentioned before, about the democratization of data and the democratization of information and healthcare. And you really have to look at that democratization from a couple of viewpoints. There is the democratization that you make it accessible to the users that I touched on before. But if we're addressing, if we're addressing health disparities, you really have to make health information and access to information and input democratized among the users of the health system among the patients. And if you think about the amount of information that is out there, and it's starting to come in, and it will come in even more through wearables and other technology, through apps, the data will flow much more rapidly. And then you have to ask yourself, well, that's nice that the data coming in the pace is faster, but you have to have the environment to to turn that new acquisition of knowledge into practice, and into actually influencing change and resolving disparities once you identify them. And that, you know, that's where a global community, a global network really comes into play. You're How do you affect change and behaviors once you identify them?

Joshua Ness  24:18  
Interesting. Rohit I'm wondering if you can follow a follow up with that. How do you see 5G and MEC playing a crucial role in this type of this type of use in these types of collaborations?

Rohit Saraf  24:32  
Yeah, so I think I'll extend to what Jon was saying earlier, like, right now healthcare is undergoing a tremendous digital transformation and it's only accelerated with COVID. So a lot of hospitals medical centers, urgent cares are looking for innovative ways to keep their patients connected, monitor their health, accurately diagnose and provide treatments in fastest possible way. So called in all of this collaboration is a central essential tool, whether it be through EHR through widespread availability of data. And to achieve all of this and to develop more such use cases, right, we need tons of connected devices, they may be sensors, cameras, lots of rapid back end processing systems, with the ability to scale and offer tremendous compute capability, and really the information flow amongst different sort of, you know, you take some example, something that Avesha has with someone like, you know, video conferencing tool, and then you marry those two, and, and all of this end to end you need, you know, infrastructure and connectivity at near real time capability to make the entire thing very seamless to doctors as well as patients to provide really value, you know, quality care, I would say. So 5G in that sense, it is truly transformational, because it's not, it's not just providing faster data pipe, right. But it's also supporting lots of connected devices, think about scale, right? It's not limited to one hospital thing or blanketing the entire in all hospitals having this facilities, and capabilities. And there's an entirely new paradigm which you mentioned, Joshua, your MEC stands for multi access edge computing, what it essentially is doing, it's providing the necessary storage and compute resources to run all these intensive compute intensive applications, right near the doctor, the patient, the source of the data and real time response, which is just not possible in in today's wireless networks, right? Maybe you can do one or two things, but talk about scale, today's networks is not able to handle that. And MEC is not even, you know, a concept. So as we know, as of today, but with 5G, it's all it's changing in pretty much real time. So to me, 5G MEC accelerate the transformation to value based patient journey. And not from doctors to clinics, to hospitals, and even post discharge. Right, that's also an important piece in this entire journey.

Joshua Ness  27:10
Yeah, and you mentioned on that last point, there, there are other solutions that can also live in work on on that 5G connectivity platform, something like collaboration tools like Bluejeans, or for example, which is a company that Verizon saw a lot of value in and we acquired this as a teleconferencing solution. That is current that is HIPAA approved, and it's currently being used by doctors all over the country to to have these telehealth sessions with their patients. So that's it's a it's interesting to start thinking about it from an end to end standpoint. And I did want to talk real quick about that Rohit mentioned, performance and in the ability for for many devices to be connected using 5G and MEC and, and he talked about how 5G is really allowing that to open up and start really seeing some really interesting uses that come out of it. And one of the reasons is because it properly configured and well structured. Next setup that uses 5G as its connectivity medium can oftentimes actually outperform the compute capabilities of the device of devices themselves. Which means that by sending information via 5G into a MEC environment, you could actually get responses back in your your your information could be able to be acted on faster than if it was actually processed on the device itself. And that's not in all cases, but it certainly isn't some, so it's really going to be interesting to see how this technology continues to evolve.

Randi, a lot of your work is focused on health inequities. We talked about this already a little bit. And I'm happy about that. Things like democratization as well as addressing disparities. I was wondering if you could share more about this and what it takes to create timely and sustainable solutions that that we're going to be able to implement moving forward. For sure.

Randi Foraker, PhD, MA  28:58  
And this has been a great discussion around democratization of data, and then also how we capture and integrate data in real time to inform clinical decision making, because we all know that there are health inequities in how care is delivered and also health outcomes among underserved populations. And the work that I do in clinical decision support is squarely in that space. The rationale for it is that we're designing these tools to level the playing field so that evidence based care can be equitably distributed across a patient population. And I think a challenge to us moving forward is how do we do a better job of capturing and acting on this information in the clinic. So when I say this information, it could be social and socio economic factors that impact Health, it could be as Rohit was suggesting earlier data from wearable devices and sensors, we need to know more about the patient's context. And that means what their health is like outside of a healthcare encounter, but also what their social environment is, and things that either promote health or get in the way of them achieving optimal health. And so I think that we really need to use this technology to more effectively not only integrate those data that can tell us more about the patient context. But also we could integrate that additional information with helping the healthcare provider know what to do with that information. So what is the prescription that a provider could give, if you will? If a patient lives in a food desert? Or has food insecurity? What how can we make those data actionable? And how can we use this technology and the additional information to improve the care of all patients?

Joshua Ness  31:12  
Do you see that varying across certain areas of the country or certain populations? Like I'm just trying to think about how, because it can't be a one size fits all solution? Maybe it is, but I see, especially if certain populations don't have the same access to health care as it is right now, then the solutions for helping what what medical providers that do have, helping to get the right kinds of information to them, or various types of information to them about the populations that they're serving, it might not be one size fits all. And it might be might be, might be a mix am I on the right track here?

Randi Foraker, PhD, MA  31:53  
That's right. And I think that there's great potential with telehealth and our aptitude for using that type of technology has really exploded during the pandemic, we're getting much better at caring for patients from a distance in that regard. So I think that that should certainly be leveraged. In terms of health inequities, what I worry about are the technology gaps that still exist in certain populations of patients. So not only access to care, but also access to the technology needed to collect these data, and broadcast these data, if you will, to the point of care so that the health care provider is in a better position to act on it. And so I think that's where my concern is in terms of where the gaps still are.

Joshua Ness  32:46  
And we had a question from Jim again, in the audience. And he was asking about, you mentioned individuals social health, how is it how something like that measure? And and how could it be measured in a, in a telehealth type of environment?

Randi Foraker, PhD, MA  33:01  
Right, so when I think about social considerations, or even socio economic considerations, I think about the most effective way to make those measurements. And oftentimes, at least with medical data, we have to rely on proxies for socio economic status. So we commonly use things like type of health insurance that a patient has, which can tell us about their employment status and their poverty level. However, we also want to take into account their environment. And so how do we get more information about where they live, work and play. And I think in that case, we rely a lot upon their residential address, and what we can find out about their neighborhood, from for example, census variables, and that can tell us a lot about access to care, as well. And as I mentioned before, whether the patient lives in a food desert or not, so having those very granular address level data can tell us a lot about the the patient context.

Joshua Ness  34:14  
Interesting. Yeah, it's I, I envision, like folks who don't have that ready access, being able to like my mind goes to them, being able to have certain types of sensors, whether it's a Fitbit or whether it's certain, whether it's a prescription, if you will, for certain types of wearables, that are specifically gathering the information and sending it directly to just their health care provider and thinking about how to make these things secure. And like mine is going really interesting direction about how a lot of systems especially the kind that Bruce is running can be used to keep that information secure. And draw inferences from that information on almost a real time basis so that the person can get care at the time that they need it even if even if they don't know that they don't know they need it. On that note, Bruce Avesha's is focused on making cloud applications edge aware and wondering, what are some some health related use cases? In addition to some of that we've already talked about how do you see this expanding in years to come as we as as long as technologies begin to really impact the healthcare space?

Bruce Lampert  35:19  
Great question. And very exciting, Randi, what you're doing in terms of bridging a lot of that health inequity. It's really exciting stuff.

I just want to further talk about that, sort of the technology gap. And then I'll talk about the opportunities and all the opportunities we see within medical. We did in worked and collaborated in building an AI investing engine for colonoscopy. So it was just one example of using an AI model to detect polyps. What's interesting is, we found that every single doctor that that performs, colonoscopies has a an EDR, and a Nomar rating detection rate. It's basically a percentage that they're given based on every case they do in how many polyps they detect, based on a national average. And you'll see that some of the great, doctors can get as high as 95, 96%. they detect everything. Well, other doctors, maybe their fellows, maybe their new internist, could be as low as 60%, you're seeing a huge gap in terms of the competency, the level of quality of care between doctors. And what we're trying to do with AI is bridge this technique using technology to bridge this gap. So if you can have an AI and use AI inferencing, during the procedure in real time, in help some of these doctors that don't have his higher ADR rating, and bring them from 60% to 90%, to 95%, just like the best doctors are, and you're in your best research institutions, you've done an amazing job of using technology to bridge that inequity. And to do this, you really need a network that's distributed, where everyone has access to be able to access this type of data in this type of technology. And we're starting to see that as you start to see the edge, allowing it hospitals and local ambulatory surgical centers and in other healthcare facilities all access sort of real time data, you're going to see a real tsunami of AI applications come out to do this exact thing is to bridge this gap and allow we call it an extra pair of eyes. Because we'll never replace a doctor, it's just not going to happen. But you can assist the doctor with this extra pair of eyes to be more efficient to catch more things. And we see this applying and building AI models for for neurology. We see a lot of great applications in cardiology, EMT, oncology, orthopedics, we've been just engaged with a large OEM to work with in the biliary path, detecting lesions, very similar polyps. So I think you're gonna see an endless amount of opportunities over the next five years or so using a lot of this technology in bridging this technology gap technology bridging this gap.

Joshua Ness  38:24  
Yeah, and it's, it's, it's interesting to think about, if you're thinking about the journey that the data takes it, it's pulling in information from the exam, not necessarily just the polyp exam, but in the case where you're talking about the extra set of eyes, it's pulling information from the exam, and it's making an inference that it can then give to the physician who is performing an exam or the person who needs that information in real time. And so a question that's been been that's been on my mind honestly, for the past week or so, is that what happens as an AI creates in AI takes a whole bunch of ones and zeros, which are ideally covered under HIPAA, right? There's a privacy associated with what what that test is, but then the result of that test is a is a completely new inference. Like it's taking a bunch of ones and zeros, and it's returning a result that says, you likely have cancer. Right now it that is a new piece of data to new piece of information that was inferred from existing data that was captured. And so how does that fit into the whole mix? Because that's something brand new that we haven't even really seen before. Up until this point, even though I've seen MRI machines or CAT scans, they're just giving information that a doctor then has to has to assess what this what the result of this actually are, but when an AI is doing it for you, what is that? What are the implications and have you run into any HIPAA considerations? Who owns that inference? And who has access to it? I actually like to open that up to everybody but but Bruce, because you were just talking I'd like to kick it off with you.

Bruce Lampert  39:58  
Yeah, well, it's actually probably more appropriate for Jon and Randi and Rohit to talk about this, we we have a VPN. So we we use a VPN tunnel directly to the to the edge to protect the processing of it. But then once it gets back to the hospital, we would hand it off to one of their platforms. And that's where you have to be really concerned about HIPAA and privacy. So with that said, I sort of turned her over to the those experts.

Joshua Ness  40:26  
Randi, what are your thoughts on that?

Randi Foraker, PhD, MA  40:29  
Well, I saw that Jon unmuted himself. So I'll pass the baton to him, I do have a thought about this. And Bruce, I just want to amplify something that you said earlier, around having having the data transmitted securely. And oftentimes, when, when I'm doing when I'm designing and building clinical decision support tools, it's in the context of the electronic health record. So the data aren't pushed anywhere else, or pulled from anywhere else. So the data in there are validated by the health care providers who entered them in there in the first place. And so I think about these tools and systems operating within that environment, and I'm interested in other people's thoughts about how this could work outside of that environment and still maintain the the HIPAA rules and regulations surrounding those data.

Jon D. Morrow, M.D.  41:34  
I'm happy to continue on that thought, Randi, thanks. I'm very observant of you to see that I had unmuted. I'm not an attorney. So it's hard to answer this specific question. But let me let me do it this way. We need to break down everything that we're talking about into a few buckets. There's, there's data in one bucket, and the data clearly are protected. There's no question that data, particularly it's specifically pH I, but data are protected, they're attached to a particular patient. The next step from the data is knowledge, you're generating knowledge from the data. And then the third step is what you do with the knowledge, it's it's the care that you provided the outcomes. So you have data, you have knowledge, and you have outcomes, the outcomes, you know, that's population that's there, there is no privacy implication there, the data clearly are protected. It's the the knowledge generation, that's sort of the, you know, the transition zone. And we're in that knowledge generation, do the data go from being attached to an individual and private and protected to being something that, you know, is a thought that has been generated. And it's interesting, if AI is generating that knowledge from the data, you know, that that's not a person doing that, but somebody had to train the AI. So there is, you know, if you're talking about intellectual property, there is some intellectual work that went into creating or training the AI to generate that knowledge. But when does the knowledge become not connected to the patient? In the early days, and I'm talking about, you know, way back maybe 10 years ago, we would do a lot of this work with de identified data where you take data, you remove very sensitive information, social security numbers, names, dates, phone numbers, exact locations of patients, you remove that. And then you have this sort of D personified, or de identified data that you could use to generate knowledge, and then have outcomes. we've now moved beyond that to the Europe synthetic data, where you're not just stripping information from data into your knowledge, you're actually creating a new population that's modeled on the original population. So a synthetic data set models a population, but you might have 10,000 people in an original data set, and maybe 10,000, or 10,005, just to keep it interesting, fake people or synthetic people in a synthetic data set, but there's no one to one correspondence. Grandma's not in there. But in a whole population, you do have the average age and the means and the standard deviations, all of that that are maintained. So you could then take that knowledge that's clearly not protected, that's not private information, and then use that synthetic knowledge to create your outcomes. And that's really where bringing together global partners can do you know, can really help to create this synthetic data set on a large population to drive knowledge. You know, Randi said something very interesting before about telehealth and you know, telehealth has boomed because of COVID. But there have always been these needs you know, right now we're all impaired we all can't get to the doctor it is it is health disparity that covers everyone equitably. We, you know, in in lockdowns, nobody can get to the health access to health system. Once COVID is finished and before COVID there were subpopulations who had trouble accessing the system, maybe because they have to take a bus and the buses aren't running or maybe you know, they live in a transit desert or you know, their car is not working, or they're working three jobs and have two kids. And how do you identify those populations? There's all these data that are now coming in. Yes, the old way to do it. And, you know, perhaps the current way to do it is by looking at, you know, at maps and looking at census data, and other sort of secondary indicators, once you start to have sensor data, and wearables. And once you can use AI to make inferences about where people go, where they live, and what are their barriers to care, then you create a synthetic data set around that so that you remove the individual privacy implications. And then you can generate from that new knowledge, and innovate innovative interventions in order to remove those disparities and bring the same level of care to everyone via telemedicine, be it care planning via outreach in the community, and so on beat economic inequities that need to be solved at a much larger level. So, you know, I think that's really important is to recognize the journey, that information takes from Data to Knowledge to outcomes, and figure out where along the way are the roadblocks and how do we democratize the data along the way, by the identifying, synthesizing and making it accessible globally, that's really the magic. And that's where we're going. And that's somewhere I think that COVID has actually pushed us a bit with the explosion of telehealth, it's pushed us to recognize that we have these tools to erase these disparities and to improve access, we just need to know where to apply them. And that's really where I think the these technologies do come in.

Joshua Ness  46:46  
Yeah, the idea of using synthetic data is really interesting. I'm curious, Bruce, like, does the does the idea of synthetic data does that? Does that do anything to increase the time or the efficiency? Or sorry, decrease the time or increase the efficiency efficiency that these AI models need to be effective? Like, is there is there a case to be made for being able to spin up new AI models very quickly using some of the synthetic data that Jon was just talking about? And Rohit, I'm curious what your thoughts are on 5G and next ability to to help to help with this process as well?

Bruce Lampert  47:23  
Yeah, that's a great point. I mean, it's absolutely helpful. The biggest challenge we have and we spinning up AI models, is for the health institution to want to give over data, getting them getting permission to get this autonomous data is the trouble because you want to load your model with as much data, reliable data that's curated as possible to teach the model how to how to detect what you're looking for. And it's just a challenge, we find it a challenge getting the hospitals to want to collaborate to give the data. So that to us, that's the biggest impediment.

Joshua Ness  48:01  
And Rohit does just does 5G and MEC play a role in that.

Rohit Saraf  48:05  
Yeah, it does. So I mean, to elaborate on Jon's point, right, we talk about AI, if you look at it, at an abstract level, it's prediction, right. But when you when Jon talks about synthetic data, that's really going into the whole deep learning part of it, when you keep it, you're becoming more smarter, the models are becoming more smarter, and you're never going to replace an occasional short term doctor is going to be this is all tools to help augment doctors, either augment the capability or reduce the time it takes to make certain decisions. It's all about confidence levels, right? If you if something is with a high confidence, and it's just about accepting it, if it's about medium confidence, and doctors really spend a little bit more time to give their own final decision. And, and from a MEC and 5G standpoint, all these synthetic data sets are the training data sets are really heavy, right? So you can do everything on the far cloud or the public load, as you know today. But then every, let's say you take example of synthetic data set for a particular use case, every hospital, it's possible, it's possible that they have certain other requirements or exceptions that you need to really generate a new set of data to train your models, deploy them faster. Now imagine you everything you do on the edge is much more real time, much more, less disruption in the service, much more confidence and it helps to even it helps for even doctors and medical staff to rely more and more on these systems once they have a certain trust. Right. So so it's a combination of all of the above right did the 5G MEC for here there's a synthetic data training model set and then ultimately, it's about to come mination of using AI is a tool to help doctors augment their decision save time. And really, you know, the part the physicians per capita in us is extremely low, right? It's low compared to other countries, comparable countries. So this, I mean, all these techniques will only help augment the healthcare and they're going to be different care in a timely manner. That's, that's crucial.

Joshua Ness  50:26  
Yeah, and sorry, Bruce mentioned the the roadblock of doctors not wanting to or hospitals not being reticent to share some of that data. And I'm curious, what, is there a moment? Or is there is there a, a tipping point where hospital systems are convinced about the ability of these of these models of AI to be used actually provide benefit back to their patients? Bruce, I'm curious if you've had experience with that, and then Randi and Jon, curious about what what any concerns that you have around AI being used in public health more generally. And if any of these concerns and some of these hospital systems have been well founded?

Bruce Lampert  51:09  
Yeah. So we found the best way of getting data from large teaching institutions is to collaborate with them. These usually specialists on board that have some research that they've been doing with some rare disease or something that they they spend an inordinate amount of time and their expertise on. And those are the doctors that really are passionate about wanting to solve and cure in create a, you know, this, bridge the gap. So they are very interested in wanting to build their AI models. And they from an institution, they could build an AI model, you have tools to do that. But they don't know how to do is commercialize it, how do you get it out there with tools that Jon and Randi have to share those models? So it's more than just a few experts in one hospital? How do you share this incredible amount of education and knowledge they have? Certainly, the Verizon edge is a great way of having people get access to it and using a Avesha for real time information. So that the pieces are here, but it's the collaboration with the actual hospital to get them to use the models, which I think is a critical. It's a roadblock that we have to overcome.

Joshua Ness  52:27  
Yeah. And Jon and, Randi, I'm curious, are there any? Are there any roadblocks that we should be looking out for in the industry as AI begins to play a more prominent role?

Randi Foraker, PhD, MA  52:38  
Well, being trained as an epidemiologist, I think about data a lot. And I worry about some of the blind spots that are inherent in AI and applying these tools to healthcare, and making decisions based on those models and solely on those models, not because I don't think the models are good, I think that the underlying data themselves can be biased. So when you think about, for example, a model that would predict hypertension control among hypertensives, you have to consider that of all the people with high blood pressure in the universe, how many are actually getting to a physician's office, how many are diagnosed with hypertension, how many are treated with hypertension, lowering medications, and then of those, how many are controlled. So what you have is a leaky pipeline, and the patients that you end up with in the end in your data set, who you're trying to predict for, don't look like the population of patients more broadly in our communities that have hypertension. And so I would just warn the the healthcare community in the public health community to understand your data and know who's underrepresented in your data before you put too much stock or emphasis on using those data for clinical care or public health recommendations.

Joshua Ness  54:12  
Yeah, that seems like it makes a lot of sense. There's there's a question that actually came into the q&a about how often just synthetic data training data need to be updated over time. So that it actually reflects the populations the growing and evolving populations that it's meant to, to be addressing. In Jon, what do you what do you think about data about the efficacy of the data that we're using to draw these inferences and how that data itself changes over time?

Jon D. Morrow, M.D.  54:39  
Yeah, well, well, clearly data do change over time. And the the synthetic data set to be clear, it's not it's not being synthesized by any AI that is trying to model a dynamic population. It's it is a statistical process that is modeling a static population. So if you want to regenerate a synthetic data set, you you, you regenerate it from The now new snapshot, I suppose we could think of a process as this evolves where you have a dynamically growing synthetic data set. But that's not quite where we are. But you know, there is. I want to echo what Randi said, as Randi spoke as an epidemiologist, let me speak as an obstetrician. You know, I could certainly I would love to see some AI that would, for example, help me interpret fetal heart tone waveforms, you know, there's a, there's an art to reading a fetal monitor tracing, just as there's an art to reading an echocardiogram, or an electrocardiogram and separately for any EG. But, sorry, electrocardiogram or EG, but I want to make sure that we don't lose the sight of sight of the entire patient. You know, as Randi alluded to, when I came of age in medicine, holistic medicine was sort of that was the talk point, you know, that we want to look at the patient not as a collection of organs, or collection of disease, but as an entire person. So I do want to make sure that, that you know, that we don't let the AI become the physician. Because we do need to keep the view of the entire patient, not just her her physical status or her waveform, but also where she lives and what her education level is, and what is her motivation to care for herself, what are her resources, who's around her, and that said, it, you know, it there, I actually see a question in the Q&A hat I want to jump to before even the moderator jumps there about the idea, the patient's own their own data, you know, and who owns the data about a patient, the 21st century cares act is about to kick in, what five days from now, the day after the election. So on November 4, patients around the country will start to have full access to their their charts, essentially, it's being implemented differently to places but essentially, you will be able to read all of the notes and see all of the labs, not after your doctor sees it, but when your doctor sees it, it probably before your doctor sees it. So if some AI engine is running, and if the AI engine decides to use Bruce's example, you have cancer, you know, that is a little bit dangerous, that the patient is going to see that on his or her phone, before the physician can put it into context. And not only the context of Let me tell you, Mrs. Jones, what that means, yes, it has that scary word cancer, but let me put it into context. But also the big picture of the whole patient and what the patient's problems are, you know, if somebody has end stage renal disease, and then there's a diagnosis somewhere of you know, cancer of skin cancer, that's not you know, that's really not going to be top of mind, because you need to put that in the big context. So yes, the patient does own the private information about the patient. But the patient needs a physician or another health care provider to help put the information into both the microscopic and microscopic context that's necessary to interpret and act on the data. So I think that answers to questions ago, but also wanted to address that that question about, about the ownership of the data. Now with synthetic data. That's where it gets interesting. Because once you make a synthetic data set on the population, it's no longer an individual's private data, just like census data belongs to everyone, it belongs to the people, if you will, and not to the individuals who who contributed their information to the data set.

Joshua Ness  58:22  
Right. All right, that's, that's really interesting. It's gonna be interesting to see how healthcare in and health providers and institutions begin implementing that here at the end of the year. That brings us to time, I do want to end real quick on a on an aspirational note for the future. If I'm going to go around the room, if we could just take 10 seconds and tell me what you're most excited about in the field and the advancing field of healthcare. It doesn't necessarily need to be what we're talking about today. But I do want to get just a glimmer of hope for the future. It's a tough week for folks right now. And so tell me tell me a little bit. Bruce, let's start with you. What are you most excited for? As the healthcare field field advances?

Bruce Lampert  59:03  
Yeah, well, I'm, again, I'm drinking the Kool Aid, but it's all about AI. I mean, the advancements in AI is just absolutely mind boggling. It can touch every single field of medicine.

Joshua Ness  59:16  
Yeah. And on the topic of technology, Rohit, what are you most excited for? As you as you work with all these healthcare providers who come into our labs and do testing with what are you seeing that really gives you hope?

Rohit Saraf  59:27  
I mean, I'm just seeing the sheer number of innovation that's coming out of all our partners and the entire ecosystem that we're building, right? I mean, I spent all my career in high tech, but this is a real chance to do something when you talk about technology for good, or as Bruce would put it, ai for good, right? This is really to take what we have and really enable and do something for the society and just seeing tons and tons of innovation in healthcare space and and we haven't even scratched the surface as of right now, watch out for this for use in next five years, it's going to be

Joshua Ness  1:00:06  
For sure. Randi how about you? What do you what are you most excited about?

Randi Foraker, PhD, MA  1:00:09  
I like seeing all of these people coming together and collaborating, because this is a big gnarly problem to solve. And it takes people from different disciplines to come together and really fix things and make healthcare run like it should. And so I'm excited to be a part of this group, and quite frankly, excited to see what what we will achieve in the coming years.

Joshua Ness  1:00:36  
And we're definitely excited to be working with you as well. Jon, take us home. What are you most excited about?

Jon D. Morrow, M.D.  1:00:40  
I'm excited about the speed to value here that we are able now you know, we're on the edge to use the word we're on the edge of being able to really turn these innovation cycles into innovation moments. Instead of innovation epics, we're able to take the data can come in faster, the knowledge can be generated faster, through you know, more eyes, more brains on the information and then being able to propagate the lessons learned back out to the community. The speed is just mind boggling how quickly things are going to change in the next several years because we now have the technological abilities to harness the brain power and to harness the communities that we've built.

Joshua Ness  1:01:20  
Yeah, for sure. That's that's going to be super exciting. that is a wrap. Ladies and gentlemen, thank you to our speakers and our audience for joining in. Like I said, this conversation has been recorded it will be available I think tomorrow anybody would like to share this content with their communities. For more information about some of the work that we're doing at Verizon 5G Labs, you can you can visit us at verizon5glabs.com And if you're interested in some of the great work that Alley is doing, you can check them out at alley.com I'm Joshua Ness with Verizon 5G Labs. If you haven't already, please remember to go vote this week, or by Tuesday on Election Day, and super important. Thank you again to all of our panelists, and thank you again to everybody who joined us here today. Thanks a lot. Have a good day.

Read Next