💡 Download the complete guide to AI-generated synthetic data!
Go to the ebook
Episode 42

On the future of digital health - and how to get there with Dr. Meshari Alwashmi

Hosted by
Alexandra Ebert
In this episode, we are joined by our esteemed guest, Dr. Meshari Alwashmi, a prominent Digital Health Scientist whose expertise spans not only extensive research but also a successful track record as a serial entrepreneur and trusted advisor to digital health initiatives. Prepare to be enlightened as we uncover the latest trends and advancements propelling the digital health industry forward. Discover the remarkable potential for progress and the direction in which this dynamic industry is heading. Moreover, we will delve into the invaluable contributions that synthetic data in healthcare can make to this ongoing revolution. This episode offers much more than just a glimpse into the future of digital health. Tune in for actionable advice that will guide your organization toward embracing innovation and achieving success in the realm of digital health.

Transcript

[00:00:09] Alexandra Ebert: Welcome to episode 42 of the Data Democratization podcast. I'm Alexandra Ebert, your host and MOSTLY AI's Chief Trust Officer. Last episode, we took a deep dive into how synthetic data can unleash innovation in health insurance. In today's episode, we will fully focus on innovation in healthcare, more specifically, innovation in digital health, how to achieve it, in which direction the industry is moving, and how synthetic data can contribute to all this.

My guest for today's show is digital health scientist, Dr. Meshari Alwashmi, who not only has an extensive research background but also successful track record as a serial entrepreneur in the space of digital health and an advisor to many digital health initiatives. So tune into this episode not only for an outlook on what's on the horizon for digital health but also actionable advice on how to get your organization there.

[music]

[00:01:10] Alexandra: Welcome to the Data Democratization podcast. Meshari, it's so nice to have you here, and I'm very much looking forward to the topics we're going to talk about today. Before we dive into digital health, AI and emerging technologies in healthcare, can you briefly introduce yourself to our listeners?

[00:01:25] Meshari Alwashmi: Thank you, Alex, for having me. I'm delighted to be here. I am a Digital Health Scientist with over a decade experience in the field. My work in digital health includes using technology to improve wellness, prevent diseases, as well as manage chronic diseases including diabetes, asthma, and heart disease. I'm also an Adjunct Professor at Digital Health, and as a Digital Health consultant, I help stakeholders in the digital health strategy, research, and development space.

My work involves advising on how to invest in digital health projects or identify key areas of innovation, and helping design and execute studies to help with enhancing health outcomes, but most of my time goes into founding and building digital health startups. I'm actually now currently working on a new venture that leverages deep technology and the detection and monitoring of several conditions.

[00:02:19] Alexandra: That sounds exciting and like you have a lot on your plate. Could you also share, because our listeners always love to know it, what makes you passionate about being in this space? What drives you to do this work that you do?

[00:02:30] Meshari: Of course, digital health has a potential to transform the way we prevent and manage diseases, and I'm committed to helping advance the field through my work and saving potentially millions of lives.

[00:02:44] Alexandra: That's obviously a noble and important goal. Very great to have you here and curious to learn more about digital health and this entire space. Before we kick off, could you also define what is digital healthcare for those who haven't looked into it so far?

[00:02:57] Meshari: Of course. Digital health refers to using technology to improve the delivery of healthcare services. It can range of so many technologies, including telemedicine, mobile health, or using your smartphone to deliver healthcare, wearable devices like the smartwatches and the smart rings. A good way to understand digital health is to think of the concept of the digital twin. Each human is going to have a digital twin part that's made up of data.

Right now, it's mostly made up of your electronic healthcare record data, but in the future, we'll be able to integrate more data sources like the data collected from your wearables, your genetic data, data from the environment, and then will help us understand how these different factors impact your life and help us predict if you're going to get sick or not, to be able to intervene sooner, to help you in managing your conditions.

[00:03:49] Alexandra: This means the promise of digital healthcare is basically that we can go from a population-level understanding or maybe gender-based understanding of healthcare much more granular towards precision medicine, if that's the correct term to use, and really figure out what does my individual composition of, I don't know, my gut microbiome, my genes, whatever, require me to have in terms of treatment or in terms of prevention? Is this correct?

[00:04:16] Meshari: This is definitely correct. It's helping us to shift from not only sick care, to also healthcare, focusing on the individual and how does my certain body interact with the different environment and how would it interact with certain drugs. Another easy way to help understand this is to think of the check engine light in your car. When there is a problem with your car, you get that check engine light that tells you that there is something wrong, so you'd go and try to fix it before actually the car breaks down. This is a similar way that what we're trying to imitate in digital health.

[00:04:49] Alexandra: Makes sense, and potentially it could even go one step further than this kind of check engine light in the car and tell you, "Well, now you're doing good, now you're in the right direction in terms of nutrition, movement, sleep quality, something like that, to actually prevent you from even having this light show up and indicating that it's more likely to have a problem.

[00:05:07] Meshari: Absolutely. One of the challenges of digital health was that we have this massive amount of data, but we need to digest it and analyze it to put it in an actionable format to bring it to patients. It's been a challenge to do it, but now, with the advancements in generative AI, I believe the future is bright.

[00:05:27] Alexandra: I can also imagine. Just coming back to wearables, since you mentioned, there are so many wearables out there from, I don't know, the Oura Ring, the WHOOP Strap, and so many different things that can tell you something about your sleep quality, how you're recovering. Are there any wearables that you personally use or that you think everybody should use to already take care of their health? Or are there many things where you actually need to work with a skilled practitioner to interpret the type of data that you're getting from it?

[00:05:53] Meshari: No, absolutely. For me personally, the way I use it, I use, as you can see, the Oura Ring. I'm wearing the watch. I may take it, some people say, maybe to an extreme, but I like to see what-

[00:06:05] Alexandra: [chuckles] It's your job.

[00:06:05] Meshari: -works, what makes it easier, so I can help suggest it to patients or suggest it to clinicians or researchers in that space. I recently published a paper in Forbes discussing the differences between the smartwatches and the smart rings, but going back to how I use it, I have a scale, a smart scale. The weighing scales have been terrific in helping me monitor my weight over the past, about seven years now. I have that historic data to see how does my body fluctuate, how does it increase or decrease based on different things. I noticed that it decreases throughout the weekday, and the weekend happens, it brings me back to where I was.

[00:06:44] Alexandra: Interesting.

[00:06:44] Meshari: That helps me do certain changes to my diet.

[00:06:47] Alexandra: Quick interruption. Smart scale, for those who are not aware, I assume are those which give you more details in your body composition in terms of how much fat percentage, fluids, muscle mass, something like that. That's the smart part about it, isn't it?

[00:07:00] Meshari: Definitely, that's the smart part about it, but also it helps capture every single measurement, so you don't rely on your subjective or on your memory to remember every time you weigh yourself. You can see how that changes. I think what makes it beneficial is how certain changes in your exercise and your diet could impact the weight. I wouldn't say go in and focus on the numbers. It's more about focusing on the trend. How did changes to my diet this month or my exercise this month impact my body composition?

[00:07:34] Alexandra: Understood. I've also read this once, that with the different wearables, you not necessarily get the most accurate measures compared to, I don't know, if it would be one of these sleep devices where you have thousands of cables attached to you to measure how your sleep actually was, so there are some deviations, but that obviously is not the absolute numbers, but the trends are something that could be insightful in terms of how are you developing, how are certain changes that you implement in your lifestyle, in your exercise regime actually then impacting you on this level?

[00:08:04] Meshari: Absolutely. I'd say wearables helped me so much with-- another aspect would be my sleep, and the trend of my sleep and matching how that trend of sleep impacts my productivity, and there is this notion of the Readiness Score that's calculated based on your sleep and your physical activity, and it's fascinating how that would correlate to my productivity. I can sit down and write with a low Readiness Score, it will take me hours, but if I have a better mental state, if I slept well and moved well, then the words flow much smoother.

[00:08:38] Alexandra: Sounds interesting. Basically, since we're the Data Democratization podcast, this is actually a movement towards democratizing healthcare data and putting it in the hands of the individuals to make it actionable for them. Are there also some examples where you say there's collaboration between healthcare practitioners and patients wearing certain types of wearables using certain measurement devices where digital health already shows impact and positive results?

[00:09:04] Meshari: Absolutely. Actually, I published a paper back in 2019 where we used smartwatches and smart scales to help patients that are prediabetes. Their weights, they had an increased weight, and they were more prone to having diabetes, so we wanted to go in and intervene to help them in managing their diet and managing their weight before they get a diagnosis of diabetes, and it was successful.

We were able to show that it reduced their weight by 5% within the first four months. It was a startup company called Blue Mesa Health. Later on, Virgin Pulse came in and acquired them, and we've worked with Virgin Pulse to help them also build on that exciting work.

[00:09:49] Alexandra: Very interesting. You already mentioned a few of the benefits of digital health care from going to reactive to actually more proactive and health preservation and also democratizing and making it more actionable for individuals, for people. What else would you say are benefits of digital health?

[00:10:06] Meshari: Usually when we think of digital health, we think of the Quadruple Aim. The first one is health outcomes, like the ones I mentioned. The second one would be reducing healthcare costs. That can be complex to measure because you'd spend a lot in digital health in the beginning to develop it and integrate it, but you start to reap these rewards in the future.

The other two aspects of the Quadruple Aim is the experience. We want to enhance the experience for the patient and enhance the experience for the clinician as well. These are the four aims. Improve outcomes, reduce healthcare costs, enhance experience for clinicians, and enhance healthcare experience for patients.

[00:10:45] Alexandra: Makes sense. Could you give us a few tangible examples for enhanced experiences? I can imagine from the patient perspective that if I could get a wearable as opposed to having to sit in a hospital to obtain certain measurements, this obviously would be a kind of timesaver and then much more convenient. What would be the enhanced experiences for the practitioners?

[00:11:03] Meshari: As you see, practitioners are burned out, and with COVID, there is a lot of burnout that happens, and when we bring in so many technologies to them in addition to the workload that they're dealing with, it's not easy to do. We're trying to find and make technologies much easier to help them, for example, go in and find the medical information quickly that they would need to, to treat their patients. That would be a good example.

Examples of different technologies that were included for nurses to help them in doing certain tasks that they do in the hospital. For example, a small one would be, they used, let's say Alexa or smart speakers, putting them in rooms. Instead of every time a patient wants to do something, they will ring a buzzer, the nurse will go in and then see what they want and go back, but having that speaker would help digitize, for example, turning off the lights, turning on the lights, or different aspects as well that would save a lot of the nurse's time that can be put towards something that's more impactful.

[00:12:13] Alexandra: Understood. Basically, also when we talk about digital health, we started off with wearables and more consumer-facing devices, but does the space also encompass technology like AI tools that are used in radiology to help practitioners understand which areas of an image might be worth their attention? Is it also solely focused on diagnosis and practitioner work tools that come into digital health?

[00:12:38] Meshari: Absolutely. There is a lot of diagnostic imaging. We've seen a lot of algorithms that came out and digitized different senses. One of them is Eko. They built what you would call the Shazam for heart sounds. That will help them understand what's going on with the heart. You can use that technology with someone with very minimal healthcare experience, can leverage it. We've seen a lot of similar technologies in AI implemented in MRIs and X-ray machines to help us understand what's going on. The other side of the spectrum is in drug discovery and finding the right drug therapeutic targets to be targeted to create drugs.

[00:13:21] Alexandra: Very exciting. What would you say, how far along are we on this journey towards more digitized health? Is there a lot already implemented? What is on the horizon? Where are we going with this?

[00:13:33] Meshari: It's really hard to say or put a number to it because the field is advancing rapidly and we continue to get new technologies that just push our sphere of knowledge. Once we reach the destination where we want to go, a new technology, new exciting advancements happen that push that further, that make us work harder to reach it. It's not easy to put a number to it, but there is definitely so much work that needs to be done.

[00:14:03] Alexandra: Understood. One thing I would also be curious to learn more about from your experience, we have many listeners, some coming from the healthcare industry, others being more located in financial services, insurance, telecommunications, but many of these organizations struggle to actually engage with digital transformation initiatives as of now. You already mentioned it's difficult, particularly with overworked practitioners to get them to be open to new technologies. Can you share a few success stories in this space or even how you and your colleagues were capable of actually introducing new technologies and transforming healthcare organizations in that sense?

[00:14:42] Meshari: I would say similar to, whether it's healthcare or it's outside of healthcare, it goes back to the power of selling and building the right business case. If you build a strong business case of why you're bringing the technology, bring in the right stakeholders to come in, sit on the table to understand what is happening and what we're doing and why are we doing it, I think that will help a lot to adopting these new technologies. There is also the notion of pilot projects.

If you've seen a new technology that you really like and you want to implement, reach out to this technology, reach out to this company, and see if you can reach an agreement where you can do a small pilot to test what they're offering, that will help. Starting small, seeing the benefits, it will help you understand if this actually works, if they're saying what they're claiming is true and right, would it work if you bring it to the real world? I would say building a strong business case involving the right stakeholders and building pilot projects to test, and if everything goes well, I think the right technology will sell itself.

[00:15:50] Alexandra: Understood. From your experience, are there different arguments as details of the business case that work better with senior management of let's say a hospital versus with the actual practitioners, nurses, doctors using this technology? Is it important to always have benefits for both of them to actually make a project a success?

[00:16:11] Meshari: Absolutely. I think each one has a certain thing that they care about more. Let's say executives, they care about how can we make the hospital more efficient and how can we make more money essentially. Where the clinician cares more about the patient and the patient experience, and how can we make sure that we're taking good care of the patients but also think about the experience itself. I always say if you want someone to do something, make it really easy for them to do so. If you're bringing a technology to a clinician or a bank or whatever industry it may be, I think usability plays a critical role that we generally oversee.

[00:16:51] Alexandra: Understood, but still coming back to this question on where we're going with this and what's the status quo, at least in Austria where I'm based, there are not many health insurances working with wearables. There are not many hospitals utilizing this type of digital health technology, I don't know, as a patient, what's happening in the backend, how much AI and other emerging technologies are used there. What still needs to change in the industry as a whole to get them further down the road of digital healthcare adoption?

[00:17:22] Meshari: There is definitely a lot that needs to be done in several parts of the world. I did a project where I brought nurses, pharmacists, physicians, and the key player, which is patients, around the same table to help us understand what would be the barriers and facilitators to bring in and integrate more digital health technologies into the workflow. That work was done before COVID, and the results that came out are still relevant today. Some of them were ease of use, I mentioned usability.

If you develop the best technology in the world and it's really hard to use, nobody is going to use it. Lack of awareness, you'd be surprised, we're talking about these different technologies, but if you go to actually speak with different clinicians or even different industries, there's a lot of people that just hear about it but don't actually understand what it does and how to use it.

I think our educational system, it takes a very long time for the educational system and universities develop a course to bring to the students. We're developing really quickly and to the point where the pace of innovation is moving faster beyond education, it's also moving faster beyond regulation. The regulatory bodies are struggling to keep up with this rapid phase of development. Two more, I would say, barriers that we'd need to overcome would be the adoption, in terms of engagement.

I think if you bring a technology to health care, it's tricky because you need to have double adoption. You need to have a clinician to prescribe it. You need to have the patient to use it. You mentioned the insurance companies earlier, we can dig deeper into that, but that's one aspect, is how can we make sure that we bring the right thing for different people? I would say interoperability is also a really big issue that we're currently struggling with.

[00:19:22] Alexandra: I can only imagine, since you mentioned adoption also from the patient side, how do you evaluate the component of trust? Is there usually skepticism when new technologies, AI is introduced, or do you feel that patients are much more open to this nowadays?

[00:19:37] Meshari: No, absolutely. They definitely know about it, they hear about it, but they won't use it or they won't feel more comfortable until they use it and start seeing benefits from it. I've had patients where I spoke with them about, let's say, the digital scales I mentioned earlier, a lot of people have skepticism or even friends have skepticism, but when I gave it to them, coming back two years after, they're still using it on a regular basis, so sometimes you just have to go in and use it for yourself to see what happens there, but they like it because it makes them feel like they are playing an active role in their health care and they're contributing to their health care, not only relying on the clinicians to do it.

[00:20:23] Alexandra: Makes sense. Are there in general any areas of human's health that you feel have currently the most attention in terms of digital health innovation? Is it to prevent heart attacks, getting new insights on cancer? Any fields that you feel have the most innovation, or is it all over the place?

[00:20:42] Meshari: There is a lot of innovation in different places, but the field that would get the most innovation is the one that has the cleanest data. That's why we see a lot in terms of, let's say, imaging. We've seen a lot of X-ray and MRI images, radiology has advanced significantly because all the images are captured in a standard way in the back system, and that have gotten a lot of innovation. Other parts are keeping up and trying to keep up with it because-- and realizing the importance of having a clean record and clean data to innovate more.

[00:21:20] Alexandra: Makes sense. Similar to other industries also, data management and how you collect data is an important part to even be in the position to make use of new technologies.

[00:21:29] Meshari: Absolutely.

[00:21:31] Alexandra: In the call we had in preparation to our conversation today, you also mentioned that the advancements that we have in drug discovery with new technologies, with AI are something that are simply stunning. Can you elaborate a little bit on what has changed and what new technologies now facilitate?

[00:21:46] Meshari: We're seeing a lot of advancements in the sequencing machines and now we have a lot of-- I'll just take a step back. Sequencing machines are the machines that help us read the human DNA. If you go back to, let's say, the year 2000 to read the whole human DNA of one person, would take a very long time, and it would cost I would say around probably millions of dollars.

[00:22:14] Alexandra: That's a lot.

[00:22:15] Meshari: It is. Where now I think it's coming to less than $100, so that significant change is democratizing it more. More patients are using it. It's giving us more data about the genetic makeup that we can understand, how can we tailor and personalize different procedures or different medications based on that genetic makeup. Even if you look at the FDA actually, and the new drugs that are being accepted now, a lot of the newer drugs that are coming to market are coming out because of that personalized medicine and knowing these new novel targets.

[00:22:53] Alexandra: Understood. Since you mentioned also DNA of people and we oftentimes also cover privacy topics on the Data Democratization podcast. How do you see this that currently there are so many startups offering also DNA sequencing and giving you access to your DNA data with sometimes the recommendations on, "Okay, this is your gene setup, this is how you should behave" are not necessarily grounded in that much science as necessary to make claims like that but still more and more customers are getting their DNA sequenced, which introduces potential privacy risks or security risk to this data, given that our DNA obviously is not something like a password that we could change if it's exposed at some point in time, but just potentially in the future, much more harm that could happen.

How do you see this whole space currently?

[00:23:47] Meshari: I'll speak to my experience with a company Newfoundland, Canada. We created the Newfoundland Genome Project, which is similar to different population health projects, all of us in the US, the Genomics England project, these are population-based projects to help us find essentially new novel targets, improve drug discovery. There is a lot when it comes to security and privacy, and we spent a lot of time to get this right, and we had to, yes, make sure that the data is extremely protected, that we are only collecting what we need.

We don't collect more than what we need. We spent a very long time not only making sure all of that is in a good consent form where we sit down with everyone who is coming in to explain what we're doing and also be more transparent. Everything that we were doing was in front of them, and they were able to agree or disagree to it. I think it's important for anyone who is considering to participate and get their genome sequence or DNA sequence to go in and read that fine print to understand what's being done and what's not being done.

I think there are two also in the US, it's called the Genetic Nondiscrimination Act. In Canada, there is Bill S-201, that will help, for example, in making sure that, let's say, insurance companies or even banks do not limit you from accessing certain things because of your genetic records. I would say read the fine print, make sure that you understand it, take your time before doing it, and know what that data is being used for essentially.

[00:25:26] Alexandra: Makes sense. For those who haven't yet looked into this topic and ever thought about having their genome sequenced, there's always, when you look into new technologies, into new mechanisms that pop up over-exaggerated type versus what's actually realistic, what's happening. If you look into the future, what would you say are valid concerns in terms of having genome data floating around somewhere and potentially getting into the wrong hands. How could this impact individuals if you would exclude banks and insurances not giving you access to services because you're literally using this type of data, what else could be inferred from this?

[00:26:06] Meshari: Well, one thing would be, let's say this startup is not doing the right job on not bringing good, right, accurate data and saying that you have a certain condition or a certain disease, and it's all out there in the public. If someone comes in and looks at that, that could impact essentially your relationship-- or how you would-- even your relationship with your potential partner in the future, so I think making sure that they're doing it right and also making sure that, yes, if that could be another significant thing, is if they're talking about information that's not accurate, getting in the wrong hand could have the wrong consequences.

[00:26:49] Alexandra: Makes sense. Since we are at the topic on privacy and data security, coming back to the digital transformation processes within healthcare organizations, is privacy usually a big stumbling stone or something where you feel that organizations don't want to move forward in fear of not complying with some of the many privacy laws that are out there, or is it not that big of a topic?

[00:27:10] Meshari: No, it is. For the ones that want to do it right, it is a big topic, and that's where the notion of privacy by design came in. When you're small, you have certain privacy things you need to make sure that you have and be educated about them. As the organization grows, you go in and employ and add in a lot more privacy regulations that you would need to go into the next stage. Let's say you got patients' data now. Before you even get this data, what do you need to have in place?

Once you have that data, what are the additional privacy considerations or security considerations that you need to have? It is critical that you need to think about it from the get-go, but it should not be a barrier to innovation. That's why working with the right privacy expert will help you become more agile in terms of what are the requirements that you would need to move forward.

[00:28:03] Alexandra: Completely supporting that, so privacy by design definitely can help here. I also really liked the analogy you shared when we had our preparation call. I think it was about rocket science, wasn't it? Or not being afraid to launch. Can you maybe elaborate again on this?

[00:28:17] Meshari: Absolutely. Usually in digital health, we're afraid to move forward because we're afraid to hurt a human. In medicine, we're taught not to make mistakes and make sure that everything is safe before we move forward. It reminds me of rocket science similar to what NASA and SpaceX have done. NASA would stay behind and think and take years and decades to find ways to go into space, and they couldn't do it very well, but SpaceX came in, and we were able to launch rockets quickly, to go and learn and see what's happening in a safe environment without having or endangering lives, but learn so much that help us take that industry on orders of magnitude to the future.

I think we can employ a lot of that in digital health, of imagining what's possible, going out there and doing it, and make sure that it's in a safe environment. The notion of sandboxes came out now to help us exercise and try these craziest ideas that will help us revolutionize healthcare and make sure that patients are safe as well. It simply can be just having a clinician in the end that says yes or no based on whatever the AI or the technology recommends.

[00:29:34] Alexandra: Understood. Not being afraid to launch, as you said back then. You also talked about iteration, and this obviously is an important element of each innovation, each digital transformation process. Do you feel that there is a different weight attached to iteration just given the fact that with healthcare it's oftentimes about patients' lives and therefore obviously you have to act much more carefully as opposed to, I don't know, building the next retail company or retail service or retail app. How do you feel this iteration handled and should it be handled in digital health?

[00:30:08] Meshari: Iteration essentially is important. Capturing the right baseline in the beginning is key. Capturing, "Okay, if you employ the technology, what are the outcomes?" We reduce someone's weight by 5% let's say. In the future, we will go in and see, "Okay, how can we increase that number from 5% to 10% or whatever it may be?" You do changes in your intervention to see the impact of that.

This is what we meant by iteration. A good framework to show how can we do that in healthcare, a paper I published is called Iterative Convergent Design, which combines quantitative and qualitative data together to help us understand complex phenomenon. It focuses on usability, but a lot of that aspect can be done towards let's say the launch of different healthcare initiatives.

[00:30:57] Alexandra: Understood. Then, we were talking a little while ago about the new wave of technology and where digital health is going. What are your thoughts on generative AI? Is ChatGPT in the future going to be used in hospitals or in digital health settings?

[00:31:13] Meshari: Absolutely. It's funny you asked, yesterday Microsoft and Epic, one of the largest electronic medical record companies, announced a collaboration that they can use generative AI to essentially fill in missing information from electronic healthcare records. It will suggest diagnoses and even predict future health outcomes based on the historic data that's captured in the electronic medical record.

I think this is going to be very exciting in the field of health care, and we will be able to see a lot come out of that. Just last week, Google launched their Med-PaLM 2 which can safely answer medical questions in the US Medical Licensing Exam. This is just the beginning. There is much more that's going to come out of this, and it's exciting. Going back to that digital, to an idea of having all of these different sources of data, I believe generative AI can now go back and turn that data source into an actionable data to help us do something about as opposed to just collecting it and leaving it sitting there.

[00:32:19] Alexandra: I can imagine that this is going to happen, but it also, of course, brings up question in terms of data and AI ethics, knowing from other spaces that oftentimes services just don't perform as well for minority groups or not enough data was collected initially. We know from the healthcare space that still to date there is this disbalance between everything we know about the male body to date versus the female body, and obviously also in certain AI projects, it was also highlighted that skin changes couldn't be detected as well in dark-skinned individuals.

What are your thoughts on the AI ethics, on the data ethics on this side, and any best practices on how to deal with that to make sure that services are more inclusive and perform well for everyone?

[00:33:05] Meshari: Absolutely. I think with that, it's good to go back to the fundamentals, the fundamentals of biostatistics. These fundamentals are the ones that helped us develop drugs and develop different interventions. This is what we're currently using, so it's no different when we bring in AI. We need to make sure that if we're creating data, if we're using synthetic data, if you're using AI to come up with a certain thing, go back to what data sources did we use to feed that AI? Was it a generalizable sample? Did it have males and females, that have people from different ethnic backgrounds, from different ages?

Having that generalizable sample and using the fundamentals that we learned before will help us in avoiding some of that. We won't be able to completely avoid it, but it will definitely become more accurate and at least it shows that we did what we control. We're controlling the controllable.

[00:34:02] Alexandra: Understood. I'm also wondering if this moving towards wearables is actually then something that could help us to close data gaps much quicker because with studies, I believe it was regulation in the '90s or so that required that also females would be sampled in medical studies? I'm not 100% sure, but still, of course, the data we have for the female body is not as extensive as for the male body, if I'm not mistaken. Please correct me if I'm wrong.

I'm just wondering if we go towards wearables and basically insurance companies and others giving them out, then we should be much quicker in collecting representative data on a large sample of the population. I'm curious if digital health and new technologies can bring us to a more diverse and fair world in medicine.

[00:34:48] Meshari: I believe in medicine, what was shocking that I came to learn, it goes way before that when it comes to not having data from women, unfortunately. When they use rats to come up with let's say drug therapeutic targets or try drugs on them in labs, most of the sample was male rats as opposed to female rats because there are different hormonal imbalances that occur in women that they did not want to take care of or it was much harder to deal with. A lot of the experiments were done on male rats, that can come and becomes a therapeutic target that's used to develop drugs, but it wasn't generalizable.

It didn't have the right sample from the get-go. That was a big issue. I'm not sure if it's even resolved completely, but now with technology, we're trying to have more of a balance, and they think it's much easier to. Once it leaves that aspect of the lab, it becomes much easier in the future to have a generalizable sample. From my studies, I actually have more women than males. They tend to let's say care about their health more and are more proactive than males.

[00:36:00] Alexandra: I can imagine. Really fascinating and super curious to see where we are going with this. One other topic I want to cover with you today is synthetic data in healthcare. You wrote a piece about this in Forbes a while ago, and I was just wondering how you see the role of synthetic data in healthcare, but before we talk more about synthetic data, knowing that it's not a clearly defined term, what's your understanding of this technology? How do you define synthetic data?

[00:36:24] Meshari: Absolutely. Synthetic data is data that's artificially created or generated by a computer or an algorithm. That will help us a lot in variety of purposes. It can help us in testing and validating our machine learning models or training certain algorithms or even conducting simulations of clinical trials or whatever it may be.

[00:36:48] Alexandra: For you, the space of synthetic data is not only a privacy-preserving technology where data already exists, but you just want to get a synthetic representative version thereof. It also goes beyond that towards data augmentation, upsampling, simulating elements.

[00:37:05] Meshari: Absolutely. We use synthetic data to synthesize more data. There's a lot of problems with data scarcity and whether it's a certain image that I'll talk about now, or also certain patient populations. We can use synthetic data to help us with that. Practically, one of the things that we're currently using is we're using data augmentation on the thermal images.

There is not a lot of thermal images that are captured clearly. We found a dataset of let's say 100 patients. We were able to turn that into thousands of patients, and by doing that augmentation, we're resembling what it could look like in the real world. Tilting the image, going closer, zooming out, and that helped us improve the accuracy of our models. There are many examples in the literature that have used synthetic data to help improve the accuracy.

[00:37:58] Alexandra: Very interesting things. You mentioned thermal imaging. What is this used for as opposed to, I don't know, me being finally able to prove that I have colder feet than my boyfriend. But what is it used for in medicine? To have thermal images?

[00:38:10] Meshari: You hit the nail on the head, actually. I think that was one. Key one is thermal imaging essentially is related to blood flow and how blood flow goes to certain parts of the body. You usually just take the temperature as 37 Celsius, okay, normal. Above that is a problem. Below, it's okay. There is usually different temperatures based on if there is a problem or not.

Let's say if you have an injury on your left leg, your left leg would be much hotter than your right leg. With regards to the feet, we're working on finding ways to actually detect foot ulcers, and there are different ways to do it as well, and a lot of research is being done to have many applications. The field existed for decades, and I think with computer vision, we can unlock and democratize that field further.

[00:39:06] Alexandra: That's very fascinating. One question that oftentimes comes up when one is discussing synthetic data for data augmentations are the limitations, because obviously you can't use synthetic data in a manner that you say, "Here's a bunch of medical data from male patients. Now please use your synthetic data generator, create me some medical histories for women," because it just wouldn't know what to do correctly.

In that scenario, you described the thermal images, is it that it's already very well understood, how it would look like in the body, and then it's more about having medical images to train an algorithm on it, so is it basically the knowledge that already exists with the experts that allows to perform this task? Or is there a danger of creating a synthetic thermal image that's not realistic because, I don't know, you would have super hot ears, but I don't know, something that doesn't exist in real life in certain scenarios?

[00:39:58] Meshari: I will give you an example that many people could understand better. Would be researchers from Stanford University use synthetic data for skin cancer, and one of the things that they have done is they generated synthetic data to simulate a variety of skin tones and different skin types, and that algorithm allowed them to detect and classify skin lesions among different populations, but as you mentioned, there is some limitations with the synthetic data as well.

[00:40:30] Alexandra: Understood. I also wanted to ask you about the benefits of synthetic data in healthcare. You already mentioned a few ones, particularly sample size and the right population and data scarcity. Is there anything else that you would see as a benefit from synthetic data?

[00:40:47] Meshari: I would see these would be the key ones. If we're able to use these ones and help us train our AI models to become more accurate, then that's significant.

[00:41:01] Alexandra: Of course, that would definitely be a huge impact. I believe in the Forbes piece you wrote, there was also a study that you mentioned. I'm not sure which specific study it was. Could you elaborate on what was done here with synthetic data?

[00:41:13] Meshari: There were multiple studies that have done work with synthetic data. I mentioned the Stanford University study, there is also the University of Michigan. They use synthetic data to develop and test predictive models for cardiovascular disease. It helped them refine their model to achieve better accuracy in predicting heart disease. That's an interesting study. Mayo Clinic has used synthetic data as well for their new cancer treatments, trying to see how that treatment would work on different patients, allow them to see how it works in a cost-effective manner.

We think of it as a new technology, but taking us out from digital health and putting us back into fintech, think of the stockbrokers and how they used to use existing data to predict what's going to happen, what stocks are important this time of the year, why will that increase, why will that decrease? There is a lot of deep analytics that happen, and there is a lot of advancements there. If we can take some of these technologies and employ it in healthcare, it will help us understand so much in predicting what would work for the human body.

[00:42:22] Alexandra: Truly fascinating. Here again, I know you don't have the crystal ball, but how do you predict the future of synthetic data in healthcare? Which fields and uses are you particularly excited about?

[00:42:33] Meshari: I believe AI, machine learning because we need a really large amount of data to help in training these models, and if we use synthetic data a lot to help us in doing that, that's great. Clinical trials will help us understand as well, especially in the design phase, having a synthetic control arm or what do you think would be the factors that would affect that trial? You're doing and designing your trial now without that. Having that additional layer of information, whether it's accurate or not, but that would help you understand, that as we move in the future, we'll be able to see and understand better if it's working or not, and continue to fine-tune it.

[00:43:14] Alexandra: Understand. Synthetic control arm sounds interesting. What would that be?

[00:43:19] Meshari: Pretty much just like any other clinical trial, based on the clinical trial, you'll have the--

[00:43:26] Alexandra: The intervention group and the group that's not getting any intervention, but with synthetic data--

[00:43:33] Meshari: You can have millions of people in a control arm that resemble the healthy group that you have and see how that would affect it, but with that said, I would like to mention that by the end of the day, it is synthetic data, and it may not capture the nuances and the complexities of the real world patient population or medical scenarios. We can synthesize it, we can have it there. It's an additional piece of knowledge, but think of it as, "Okay, this is lab data.” I would use it that way, similar to developing technology in a beta trial or alpha trial. The results are promising, but we won't actually know how good it is until we ship it to the real world and see what happens there.

[00:44:20] Alexandra: Makes sense. It's a tool in the toolbox but definitely not a replacement for thorough analysis and thorough studies, but very cool to hear what's already happening here in terms of studies. Meshari, before we close off, we learned a lot today. I want to ask you two things. First, can our listeners reach out to you? Where can they find you? Where can they read more of the content that you produce? Is it LinkedIn, is it Twitter? Where are you reachable?

[00:44:43] Meshari: Definitely LinkedIn and Twitter. You can find me in both. I post mostly on LinkedIn. I try to do a bit better in Twitter, but that's where you can find me.

[00:44:54] Alexandra: Perfect. That's good to know. My final question for you would be, are there any final remarks, anything that you absolutely want to share with our audience about digital health or even for the digital healthcare practitioners and organizations listening, any last lessons on how to succeed in this space?

[00:45:11] Meshari: I believe I mentioned that earlier, and I want to reiterate it, is that not being afraid to make mistakes, because the moment we stop making mistakes, we will no longer be at the frontier of science and innovation. We need to make mistakes in a safe environment to help us push the sphere of knowledge to better understand how can we make humans live longer, healthier, and better.

[00:45:35] Alexandra: Awesome. Thank you so much for being with us today. I really enjoyed learning about digital health from you. Thank you so much for your time.

[00:45:41] Meshari: Thank you, Alexandra, for having me, and have a wonderful day.

Ready to try synthetic data generation?

The best way to learn about synthetic data is to experiment with synthetic data generation. Try it for free or get in touch with our sales team for a demo.
magnifiercross