Category Archive For "Podcast"
To build on last week’s conversation and further illustrate the diversity of AI applications, we’ve interviewed three startup executives from the AngelMD network about how they are using AI to solve a problem in healthcare: Zeenat Ali of Mammoth Technologies, Marc Cohen of Kinometrix, and Nick Delmonico of Strados Labs.
If you prefer reading to listening, check out the full transcript below.
Last week we talked to Mohsen Hejrati about the potential AI has in healthcare. To further illustrate that and the diversity of AI applications I’ve interviewed three startup executives from the AngelMD network about how they are using AI to solve a problem in healthcare. Zennat Ali of Mammoth Technologies, Marc Cohen of Kinometrix, and Nick Delmonico of Strados Labs. I’ll let them introduce themselves.
Zee: Hello, my name is Zee Ali and I am founder and CEO of Mammoth Technologies. We are a block-chain AI company currently building applications in healthcare. I have a background in application development for ten years, where I worked with companies like Lockheed Martin, Children’s Hospital of Los Angeles, UCLA Health, and Sony in developing application. Specifically in their IT department and in Indie Health verticals.
Marc: My name is Marc Cohen. I have degrees in Biomedical Engineering and Electrical and Computer Engineering. I’m the CTO of a start-up company that is about one-year-old now. The name of the company is Kinometrix. And what we do is provide objective measures to physical rehabilitation Phoenicians so they can more effectively evaluate and define the protocols for rehabilitating a patient. And this is Neuromuscular rehabilitation and Muscular-Skeletal rehabilitation.
Nick: My name is Nick Delmonico and the CEO and found of Strados Labs. And Strados Labs is bringing better respiratory health to patients in the ICU and transitions of care to try and improve clinical optimization and reduce cost of complication.
Susana: I know with three guests it can get a bit confusing so if you hear them back-to-back they’ll be in the same order you just heard; Zee, Marc and then Nick. One thing that stuck out to me after these interviews is just how large the scope of AI is, even within a specific industry. Though all three of these companies use some sort of AI technology, they’re all doing it in different ways to solve different problems.
Zee: So we take a look at a physician’s credentials, specialties and expertise and we match them with the best possible opening, locum opening, or job that is provided through a healthcare organization such as a hospital, an urgent care or a clinic. And so how AI ties into all of this and deep learning and neural networks is that we have a technique that allows for a human-like deductive reasoning, sort of like an inference and decision making skill. So, we do that by applying algorithms. These algorithms allow us to use technical and complex based statistics to allow us to create smarter matches.
We use the element of probability theory, signal processing and even from elements of psychology as far as a physician’s background or what the healthcare organization maybe looking for in decision. And so what we do is we provide something called a “Machine Learning as a Service” and this concept comes from being able to have flexibility to do customize what you are looking for as a physician and what the healthcare organization may be looking for as well. And so those types of customization can work in public or private sectors and can be extremely flexible when it comes to customization.
Marc: So, in physical rehabilitation up til now there have not been any objective measures used; they’re all subjective. So typically the clinicians says to the patient, “so on a scale of zero to ten, how are you feeling today?” And what we’ve found is we’ve actually created a sensor and a cloud platform that enables the clinician to collect objective data based on the patient’s movements. Okay, so what this means is that we can collect a whole lot of data on patient’s with different types of injuries. We can use demographics to try and distinguish between the different patients and the different conditions and then we can use machine learning, which I guess is a sub-set of Artificial Intelligence, to help us figure out what the best course of treatment would be for a patient suffering from a particular condition and that patient is in a certain age group and is of a certain height and a certain weight. That is one way we use Artificial Intelligence.
We also use it to determine the rate that we think the patient is going to get better. So by sort-of looking at the progress and predicting ahead we can say, “okay, well we think that this patient will only need three more sessions with their clinician.” And this helps insurance companies determine reimbursement.
We’re also trying to do which makes the job easier for the clinician, is to use Artificial Intelligence machine learning to determine what kind of movement the patient just performed. So, instead of the clinician having to stand there and watch them do ten repetitions of a movement they could say, “I want you to do this, this, this and this,” and then our algorithms will automatically determine what kind of movement they performed based on the signals from our sensors.
Nick: So, we use Artificial Intelligence by integrating our acoustic lung sound classification system on to our big data platform which allows us to, number one identify changes among health in real-time for patients; and then to be able to provide that information in a structured way back to care teams. We also integrate that data with a variety of other information about patient or their environment so that we can get a much better holistic picture of somebody’s respiratory health at any given time. And that’s particularly beneficial for patients who suffer from either airway conditions or have come out of surgeries and have a higher risk of pulmonary related complications like pneumonia. So we are trying to get information that otherwise goes unnoticed back to care teams.
Susana: There was definitely a common thread with all three interviewees. They all saw the huge potential of AI, and they saw it from the perspectives of everyone involved in the care process. It wasn’t all, “this will make the doctor’s lives easier.” There was a real recognition of how it affects patients and insurance companies too. And this is why innovators decided to go the route of AI; it can impact the care process from all sides. Here is Zee, Marc, and Nick talking about why they decided to use AI.
Zee: Well, there are many different technology avenues available but what we were looking for was something that’s more robust and there are more capabilities of customization in the back-end to run something like predictive models. The decision primarily comes from the fact that, number one primarily my expertise lies in block-chain, and the integration of block-chain into AI. And so because of my growth with Mammoth Technologies in that area of expertise I was able to, it was a little bit more accommodating to figure those areas.
Also, I do firmly believe that the future in part lies in AI and the development of AI. We are certainly heading in that direction and it is truly a matter building a strong foundation right now, building that foundation in block-chain and creating it so that we are able to build on top of that to go in the direction of Artificial Intelligence.
Marc: Prior to this company, Kinometrix, one of my colleagues of mine and I started another company called Ergonametrix, and the idea was somewhat related. The idea was to put the sensor on a worker, a manual laborer, and measure their movements during the course of their work shift. And then what we had to do is from those measurements we had to give feedback to the worker and say, “okay, you are bending incorrectly. You’re not bending at the knees, you’re bending at the waist. You’re twisting your body instead of turning your body, when you want to move something from left-to-right or right-to-left. And these put you at risk for lower back injuries.”
After a lot of investigation, I mean this was over the course of three years, I did tell them that a couple of machine learning tools were very good at predicting the kind of movements that these workers had done. So when we started Kinometrix, I already had the idea that these kind of machine learning techniques would work well.
Nick: Yeah so, when we were thinking about how to develop the platform and develop the device that ultimately allows us to gather acoustic lung sound data. We understood that using Artificial Intelligence and machine learning methods would be really an optimal way for us, not only to identify and classify adventitious breathing sounds and other events that are important to respiratory health, but then be able to more effectively turn that data into information and ultimately into sort of wisdom, which is the ultimate goal.
And particularly for lung sounds and acoustics around pulmonary condition it was important for us to be able to use methods to learn about not just the specific sounds but also the environment that that sound is coming from. Or other variety of information that can be pertinent to a patient. So that was really why we wanted to use Artificial Intelligence, was to make it a little bit smarter and be more predictive and really get some interesting data back to care teams.
Susana: As Nick Delmonico of Strados touches on there, AI has to add to a physician’s ability, not take away from it. A key to getting to that point, which all three innovators discussed, was physician feedback. Though all had a different way of incorporating this into their product, it was something they all did and made a priority. And it is especially important when you’re working with a technology like AI that a lot of doctors aren’t familiar with.
Marc: Well we’ve worked, one of our co-founders is a MD and his specialty is Pain Management. He has a chiropractor in his office as well. And one of the other co-founders is an Athletic Trainer. So we started off in his office and we started looking at his patients and then taking data on his patients and giving feedback from him and the chiropractor. Then we moved on to some, I’d guess you’d call them elder care places, where a lot of in-patients experience falls. And what they asked us to do was try and assess the risk of someone falling, which we are working on right now.
We’ve also been able to use a particular billing code for making these measurements, which is reimbursed by insurance. We have been reimbursed 100% by all private and other government insurance companies. So we haven’t, we’ve worked with a lot of providers but we haven’t worked with a lot insurance carriers. Basically, the practitioner submits for the work that they’ve done. They use this particular CBT billing codes and this particular CBT billing code is accepted by insurance carriers as a functional test. So the clinician gets reimbursed for the use of our platform. Which is, helps enormously in getting clinicians to adopt it.
Nick: So it has always been really important to us at Strados, and particularly myself too on a personal level, I myself am not a clinician but I’m fortunate enough to have parents who are both in clinical medicine, and a brother and sister who are both in clinical medicine as well. So I’ve sort of known first hand that it is extremely important to get information back to care teams in a way that is simple and easy to use and makes sense. Can’t add more challenges to their day, can’t add more clicks to their workflow, it has to be something that adds value without adding a lot of time and cost.
So we’ve been very, very focused on integrating both patients as far as how they would integrate and use our device and which they’d be interested in wearing it while also working with care teams to figure out what data do you need or want and what’s most important for to be able to make better clinical decisions or to improve your clinical workflow, and in what way would you want to see that information. So that’s been something that we’ve done over the vivid lifecycle of the product and will continue to going forward.
Susana: In the last episode, Mohsen talked about how he identified AI as an emerging trend by the companies that were initially investing in it. Earlier this month, CBS Insides released a comprehensive report of how Google’s parent company, Alphabet, is planning to utilize AI to “reinvent the three trillion dollar US healthcare industry.” If I, or any of the guests I’ve had on the podcast, haven’t done enough to convince you AI will become a huge player in healthcare THAT should.
But for me, it is really these startups that get me excited about the potential and they’re pretty excited too.
Zee: I think the most exciting thing is being able to take a look at patient data and have that level of analytics with patient data. I think that there is a huge intelligence behind healthcare and AI and it is very well thought out. I think that there are some challenges that do exist with that, but I think that the future truly does lie in structured and unstructured data. There, you know, with healthcare specifically that there is an increase availability of healthcare data and its rapidly progressing to analytics techniques.
Marc: I think its a huge space for machine learning and AI. I think so much can be done with machine learning and AI. I think helping, the whole idea would be to help the patients. How can you help the patient?
Well, you can help the patient by directing them to the right kind of treatment. You can help the patient by directing them to the right kind of practitioner. You can help the patient by giving them feedback on how they’re doing during a course of treatment. As soon as you as you engage a patient with numbers and measurements they become much more involved in their own treatments. They’re much likely, more likely, to come back for second, third, fourth, fifth treatments then if you just say, “okay, we’re going to give you about twelve treatments cause that is all your insurance allows.”
What people are finding, we’ve talked to a lot of physical therapists. They say that people stop after three or four visits. Its not because they’ve recovered, they’re just bored. They’re not sure whether they are getting better or not getting better; some days they feel good, some days they feel bad. But it just becomes too much of a hassle for them to come and do the, and see the practitioner. But as soon as you give them numbers they say, “oh, I want to see if my numbers this week are better than my numbers two weeks ago.” I think it somewhat inherent human nature to be competitive, even if you’re competing with yourself.
So I think AI and machine learning, on so many levels in medicine has potential in so many different ways in medicine.
Susana: Thank you for listening to On Call with AngelMD. Visit us on AngelMD.co for more information. You can follow up on Twitter @AngelMD_inc or on Facebook at /angelmdinc and you can find us on LinkedIn as well. I’m also on Twitter @smacha1995. We’re a pretty new podcast so we would love any feedback you have. Tweet us with the hashtag #AngelMDOnCall and let us know what you thought of the episode. Thank you for listening, we hope you join us again.
So I interviewed three separate people for this episode of the podcast and unfortunately I wasn’t able to include all of their great responses I asked them. If that is something you’re interested in hearing, a full uncut interview with each of these guests or any of our previous guests, then tweet at me or the podcast and let us know.
Artificial Intelligence is one of the fastest growing fields out there and it’s no wonder, anywhere there’s data, there’s a potential for AI. We spoke to Mohsen Hejrati, CEO of ClusterOne, who has been working on Machine Learning and Artificial Intelligence his entire career. Mohsen shares his AI first philosophy and likens the tech to the arrival of electricity.
If you prefer reading to listening, check out the full transcript below.
About nine months ago, I wrote a post for the AngelMD blog about the artificial intelligence revolution in healthcare, and since then A.I. has further expanded its reach. Just a week ago, the FDA approved a software that can diagnose diabetic retinopathy, and therefore streamline the process by removing the need for consultation of an eye specialist. In China, the tech is being used to automate the most mundane tasks in the care process. According to Mohsen Hejrati, CEO of ClusterOne, A.I. will affect every industry. Considering he has a Master’s and Ph.D. in Computer Science, and has worked on A.I. projects like Alphabet’s Waymo, I’m inclined to believe him.
Mohsen Hejrati: I think the A.I.’s first philosophy is about understanding how A.I. is going to change the way businesses work, and life is going to change. If people believe in that… if people believe that it’s going to disrupt everything about our lives, then you’re going to start building companies, organizations, and products completely differently. Imagine you had electricity or you didn’t have electricity. So we are building everything completely differently, all of our processes, all of our tools have changed because of that. That’s the same for A.I. A.I. first is about changing… making A.I. the core of every organization, so that you can [inaudible 00:02:03] the applications of A.I. and [inaudible 00:02:05] the benefits of it within your organization. Otherwise, it’s going to be extremely hard for companies and organizations to compete in that race. We are talking about small startups taking over big companies because they’re just better at adopting A.I. I think that’s the A.I. first philosophy. People have to change and make A.I. the core of their organizations.
Susana: Because of the rapid progress of A.I. tech, many have been left confused or unsure about what A.I. is and how it will impact them. Mohsen can help us answer that first question, and defines machine learning in the process.
Mohsen: So A.I. is pretty much any program that does something intelligently. I remember when I was a kid I was writing programs that multiply things or ask questions, and that’s an A.I. Now you can see more advanced examples. When you play computer games, there is something intelligently happening inside the game, that’s A.I. in general. So pretty much any computer program is artificial intelligence because it does something intelligently. Machine learning is where you have data and you use data to do something intelligently, when we use data to do something better than without that data, whereas you can use data to build models that predict the future or classify things or cluster things. So those are machine learning applications. So machine learning is a subset of A.I. that refers to programs that use data to do intelligent stuff.
Susana: So A.I. and machine learning have a similar relationship to rectangles and squares. A square is a rectangle, but a rectangle is not a square. Just like machine learning is A.I., but A.I. is not necessarily machine learning. Something Mohsen touched on that is interesting and important to note is A.I. is not a new technology. It’s been around for decades. The term was actually coined way back in 1956. But, A.I. has been growing fast. According to the A.I. Index, a group at Stanford that researches trends in A.I., the number of active U.S. startups developing A.I. systems has increased 14 times since 2000, and VC investment in A.I. startups has increased six times during the same period. Similar growth patterns appear in conference attendance, course enrollment, and academic papers being published. The A.I. Index report, which I’ll link in the description, also details the improvements made to A.I. like drops in error rate for image labeling and increased accuracy in speech recognition.
A different report from Accenture about A.I. in healthcare sums up the distinction between the new age A.I. and existing systems well. “Unlike legacy technologies that are only algorithms or tools that complement a human, health A.I. today can truly augment human activity, taking over tasks that range from medical imaging to risk analysis to diagnosing health conditions.” This level of ability is coming to every industry.
Mohsen: It’s hard to imagine any specific industry that’s not going to be revolutionized, so pretty much anywhere that you have data or you’re producing data, you’re going to be able to use that data to do things better. Many times you can do things radically different and much better so that is going to be a big disruption. So I think it’s going to be every single industry out there that’s going to change. But as I mentioned, I am really really interested in life science because I think the impact that A.I. can have in life science is really big. It’s already one of the… it’s the single largest segment in every economy. There’s a ton of data produced in life science, health care, pharma, diagnostics, genomics, every aspect of it. There’s a ton of things that can be improved. I don’t think it’s a single industry, I think every industry will be affected
Susana: Like Mohsen said, any industry with a wealth of data is a prime opportunity for application of A.I. So let’s do a little exercise and vastly oversimplify to get the point across. If we think in terms of healthcare, we can think about data points from electronic medical records, which include a patient’s entire health history, or wearables, which provide day-to-day measurements on biometrics like heart rate. Or, we can think about genetic testing, which tells us if we have genetic markers for certain disease. The potential of A.I. in healthcare is being able to throw all those data points at a computer and for it to output a treatment plan that is both decipherable and useful. That’s obviously the ideal, and though technologically it might be feasible to get there, we’re far from the ability to do that with the current set of data.
Mohsen: We know that there is a lot of data out there, but the quantity of the data… while the quantity is high, the quality is not that high. So, we have been collecting a lot of data, but we didn’t know how we want to use it. So, we have not been collecting it the right way, and I think that’s one of the biggest challenges moving forward as people start to realize this data has to be used for A.I. and machine learning, then they will start collecting data in a more thoughtful way and collect better quality data. And that’s number one.
Susana: With no central database for patients, some companies in the A.I. diagnosis arena have turned their focus from obtaining info from the doctor to getting that info from the patient. Recently profiled in WIRED, AIDA Health does just this, with a conversational A.I. leading patients through a questionnaire and then connecting them with a doctor via video on completion. Compatibility and collection of data is one of the bigger challenges to adopting A.I. in healthcare. Mohsen identifies a lack of reproducible research and the overwhelming amount of players as to others.
Mohsen: So you will see a lot of work out there from academia or industry, and these are all exciting stuff, but they can’t be adopted in real world because [inaudible 00:08:40] not following the rigorous procedures that are required to build a product that go out in the field. It’s one of the other challenges, figuring out how to take research from ideas to production, and that I think is one of the big gaps that exists today. I haven’t seen any peer-reviewed publications using a machine learning technique in life science. It’s very hard right now and that’s something that everybody’s starting to talk about. Since this is a very new thing, you need to educate a large number of players, especially in the healthcare and life science. You have so many different players, it’s not like robotics or self-driving cars where you are going to be talking to one or two main players in that ecosystem. In healthcare you have to talk to providers, insurance, patience, doctors, pharmacists, and everyone, to bring them on board and educate them how to use this new technology. I think that’s one of the major challenges of the… are trying to drive more adoption of A.I.
Susana: The point Mohsen makes about players is important, and it is something that has come up on this podcast time and time again. You cannot design a solution to a problem without constant input and review by the stakeholder, or in this case, the provider.
Mohsen: There’s a very nice quote about how A.I. is the new electricity, and I think everybody should really try to understand the depth of how A.I. can impact different areas. The providers should really understand opportunities that A.I. can bring, like electricity. It will bring a lot of new types of tools, it will enable different procedures, methods, it will improve efficiency radically. It can change the value proposition deeply. So, it will really change the way people work, and providers need to focus on something different than A.I. is going to help them. Dumb example, you may not need someone to hold candles in a surgery room anymore, but you will need a lot more people to do more advanced stuff. So, A.I. is going to be the thing. It may eliminate some of the jobs, but those jobs… but it will enable a lot more jobs that increase the value significantly. So, I think that’s the biggest thing that people should realize and try to go embrace this new technology fully.
I really think that A.I. does create more jobs than it eliminates. That’s one of the areas you can think about how a more intelligent software can help my business, or my company, or staff to do more stuff, and if I have the technology, how can I reinvest the money that I gain from that efficiency into bringing talent that does more things. So that’s one of the areas, and then I think people who are A.I. practitioners should also focus on problems. I see a lot of researchers focus on technology and ideas but not try to understand the problems that are out there. I think that’s very important. It’s very important for them to focus on the problems and understand the problems. Sometimes these are so obvious and so simple that we don’t see them. I think that’s one. Then, also important to understand the limitations of the A.I. systems, and make sure that it’s clearly communicated to the end users. A lot of times I see that people want to publish papers or get into the media. It’s natural to overpromise the capabilities of A.I. systems. I think that’s very important to make sure the capabilities and limitations are clearly communicated with the end users so that it can be actually adopted, and not passing it as a research idea.
Susana: Despite these challenges, Mohsen remains optimistic about the future of A.I. in life sciences and beyond.
Mohsen: It’s very exciting to see more and more people and organizations starting to think about this and I think it’s very exciting time. I’m personally very eager to see the future in applications of machine learning and life sciences.
Susana: Make sure to check back for the next episode, as we’ll be continuing on the topic of A.I. in healthcare and speak to some people developing the technology. Thank you for listening to On Call with AngelMD. Visit us at angelmd.co for more information. You can follow us on Twitter @angelmd_inc. We’re on Facebook at /angelmdinc and you can find us on LinkedIn as well. I’m also on Twitter at smacha1995. As On Call is fairly new, we’d love to hear from you. Tweet us at the #AngelMDOnCall and let us know what you thought of the episode. Thanks for listening, we hope you join us again.
If you’ve read any story coming out of Silicon Valley in the past few years, you might be familiar with the term unicorn in reference to a super-promising, hyped-up startup. With valuations of $1 billion or more, unicorns are hailed as the holy grail, but research and some influential individuals have started to question the mindset.
In this episode of On Call we’ll explore why unicorns are such a big deal, how they can be misleading, and whether chasing unicorn status is really the right move for a healthcare company.
Despite owning 38 percent of businesses (and making up roughly half the population) female entrepreneurs only receive about 2 percent of all venture funding. Even in a sector like healthcare where 78 percent of the workforce are women, that number isn’t much higher.
But Gloria Kolb, founder of Elidah, and Mylene Yao, founder of Univfy, didn’t let those statistics deter them from starting their own companies after they each recognized a problem desperately needing to be solved.
Gloria and Mylene are just like any other entrepreneurs: driven, enthusiastic, and full with a sense of purpose. Listen to how they got to where they are now in this week’s episode of On Call.
Every startup’s journey is different, but they all start with an idea. Some begin as a side-hustle or passion project and then often reach a turning point, or an “I have to pursue this” moment. That’s when a someone becomes a founder. Every entrepreneur has one of those moments and a lot can be learned from how they got there and where they went after it.
On Call will feature these founders every couple of weeks starting with Raymond Cloutier of NovApproach Spine. Raymond was in a unique situation when he started his company — he was working for another one.