Hello, and welcome to Checking The Vitals, a podcast powered by Specialty Care. I’m Todd Schlosser and today our guest is Dr. Alan M. Reznik, an amazing orthopedic surgeon out of the Orthopedic Group in Connecticut. In this conversation, we touch on his strange path to orthopedic surgery, how approaching situations with an inventor’s mind led to his many patents for medical devices and both the different types of Artificial Intelligence (AI) and how it is changing the industry. Enjoy the conversation.
Todd Schlosser: So I’d like to start off by asking what it was that drove you to want to have a career in healthcare? Was there a quintessential moment that you decided, “Hey, I want to be a doctor”?
Dr. Reznik: Yeah. This is one of my favorite questions to answer because I think sometimes in life things look like a straight line but really they’re really kind of jagged.
Todd Schlosser: Absolutely.
Dr. Reznik: And occasionally a serendipity occurs and then in a moment of serendipity, something changes and it changes the whole direction of your life.
Todd Schlosser: Absolutely.
Dr. Reznik: And for me, there were several moments of serendipity. One was I was probably about six or seven years old and I was sick. My doctor was seeing me as a house call and he was on black and white TV on Face The Nation at the same time discussing this new vaccine that he had worked on. I was watching him on TV as he was giving me an injection.
Todd Schlosser: Oh, wow.
Dr. Reznik: I was enamored at that moment of the whole thing.
Todd Schlosser: I’m sure.
Dr. Reznik: But not growing up in a family of any physicians or even anyone who ever went to an Ivy League school and almost everyone went to public college who went to college, and most of my father’s relatives didn’t even get to go, that was never something I thought was remotely possible. I finished high school with very lopsided math and English scores. Didn’t know why. Ended up going to engineering school because that was really the only good school I could get into because of such a discrepancy between my math and English abilities. And ultimately there, I needed to have a job because I didn’t have any money. So I had a bunch of different work study jobs. But eventually landed one as the engineer for a project studying sudden infant death syndrome at Columbia Presbyterian Hospital, actually a babies hospital. In that group, I did the computer stuff for them. I was an engineer and I did the computer stuff and I worked-
Todd Schlosser: When you say computer stuff, do you mean IT related stuff?
Dr. Reznik: Well, it wasn’t really IT at all. I mean, you think about computers now, you think about IT more than anything, right?
Todd Schlosser: Yeah, absolutely.
Dr. Reznik: But really what it was is we had a plastic container that was quite good sized and round and we actually put babies in the container and the container had controls for oxygen and carbon dioxide and sensors for those levels and also a way of measuring the baby’s breathing and their heart rate. In the process of doing that, you had to collect the data. And you can imagine back then, the computers weren’t that sophisticated.
Todd Schlosser: No.
Dr. Reznik: So we had a fixed drive and a floppy dive. And the floppy drive was this big around and it was actually a hard disk. And it could carry three megabytes of information, which was-
Todd Schlosser: Wow. That’s nothing.
Dr. Reznik: That was unbelievable back then. And in one three megabyte drive, we could get all the data from a one hour sleep study on babies. And they were studying sudden infant death syndrome, at the time crib death. Babies went to sleep, they didn’t wake up, we didn’t know why and they were trying to figure it out. And the whole thing about keeping your baby upside down or right side up is all related to those studies that were done back then. In the process of doing that, at some point along the way I’d written a paper about how you measure C02 in babies and mathematically errors in the computer model of what would go wrong with that based on the type of sensors and the chemical engineering I did, so the type of chemical engineering problem was. And then after finishing that, one of the doctors came to me and says, “Have you considered going to medical school?” I said, “No. I really didn’t think that was a possibility.” He said, “We think you’d be a very good doctor. We’d like you to apply and we’ll support your application.” So I was like, “Oh, okay.”
Todd Schlosser: But it’s something you had never considered up to that point?
Dr. Reznik: I mean, I had that magical moment when I was a kid and I never thought that it was possible so maybe in the back of my mind says wouldn’t it be wonderful to be that guy who gives the injection to the kid and he’s on TV at the same time. When you’re six or seven years old and you see that, that’s …
Todd Schlosser: Yeah, absolutely.
Dr. Reznik: Miraculous. Like, “Oh, he’s on TV and he’s in my room.” At that age, I couldn’t really figure that out. It always stuck in the back of my head. I said, “Well, what do I need to do?” He said, “You need to take the MCAT.” And being a chemical engineer, I’d already taken calculus and physics and chemistry and organic chemistry. All those classes were already under my belt. All I had to do was take two bio classes and one was genetics and one was biology, zoology, biology. And I did those classes and took the MCAT and got into Yale of all places.
Todd Schlosser: Oh, wow. Yeah. I’ve heard of that school, yeah.
Dr. Reznik: Yeah. So I got into Yale and that was fortuitous as well because it turns out the reason there was a big discrepancy in my scores and the reason why I ended up in engineering school as opposed to regular college, one of the main reasons was is that it turns out I’m dyslexic and I didn’t know it. So when I got to Yale and I was a student, the good thing about Yale is they have a lot of small seminars. And the way the classes are run, you have a large class and then you have small seminars. But all of your grading activity and everything was based on your performance in a small seminar.
Todd Schlosser: So not a big survey class?
Dr. Reznik: No. And it wasn’t big exams. There were no required exams in a sense but you were in a group, two professors and six kids, and they grilled you in the room. They knew what you knew and what you didn’t know. You didn’t have to do a written exam.
Todd Schlosser: Yeah. The student to professor ratio was so low, they didn’t want to track that.
Dr. Reznik: It was pretty intense. And some of the people were actually Nobel laureates or going to get a Nobel prize. I was enamored with that idea.
Todd Schlosser: Absolutely.
Dr. Reznik: So I purposely went out of my way to have a Nobel laureate as one of my preceptors, which was not always to my advantage, as you can imagine.
Todd Schlosser: Yeah. I’m sure.
Dr. Reznik: But there was a lot of fun things about that. I met my wife and we were dating and her sister was a special ed teacher. One day she’s giving me directions and I’m writing them down and she’s watching me write and I’m doing my usual scribble that no one can see, because I’m a doctor, right? And misspelling some words and whatever. She says, “You know, there’s something wrong with you.” I said, “What do you mean?” She says, “I think you’re dyslexic.” I said, “I don’t know what you’re talking about.”
Todd Schlosser: I mean, when was that? What-
Dr. Reznik: This was in medical school.
Todd Schlosser: Yeah. But what year was that? Sorry.
Dr. Reznik: What year was that? 1981 maybe. A long time ago.
Todd Schlosser: So dyslexia wasn’t really a commonly diagnosed thing until mid to late ’80s.
Dr. Reznik: No. I was not diagnosed. And not only that, I was always above grade reading level in English. No really knew this but what I was doing is I was memorizing words. I never was really reading them. And my first grade teacher thought I was brilliant. My mother tells this story. She thought I was brilliant. And what it was, as I look back now knowing what I was doing, I just memorized what people said when they turned the page in the book and I just repeated it when I saw the picture.
Todd Schlosser: Oh, wow. Yeah.
Dr. Reznik: So that’s what I was doing in first grade. You can get away with that. But in third grade you have to actually read and she was my same teacher again in third grade. She told my mother I was stupid. She said, “When he started, I thought your son was very smart, and now I don’t think he’s smart at all.”
Todd Schlosser: Yeah. That happens to a lot of dyslexic people. I’ve talked to some people who grew up with that and they had a friend who said he looked at the words and tried to memorize them as pictures because it helped him if he focused on the word the just being the picture of the word the and it has helped him focus on what it was, which I thought was very interesting. It’s a very difficult disability to struggle with.
Dr. Reznik: And I never really knew it. But math is easy and verbal communication was not hard. And even being creative verbally is not difficult for me. So I breeze through all that and then Yale was a seminar format, which was great for me because you don’t have to write things down. So Yale was the perfect medical school for me. I didn’t know that at the time. It’s again serendipity. I was just there and it turned out it was perfect for me. And then she diagnosed me and did all these tests and said, “You’re pretty badly dyslexic” but I’d already gotten so far and she said she doesn’t know how I did it. Truthfully, it’s because, like I said, things like a straight line, I went to the places that were easier for me by default but ended up looking like by design.
Dr. Reznik: So here I am an engineer with some computer expertise in medical school. Now what do you do? Let’s look at the different specialties that require some technology. At the time-
Todd Schlosser: Sure. Is that how you whittled your specialization to orthopedics?
Dr. Reznik: A little bit. I didn’t like dealing with mucus that much. I think I crossed that off the list. The other bodily fluids were a little unappealing. So I said, “Okay. What’s left?” There was radiology, there was ophthalmology, and there was orthopedics because all of those have a lot science. Ophthalmology has a lot of optics. Radiology uses different parts of the digital stuff. CT scan was just invented. It was very exciting. All that stuff. And then orthopedics was very mechanica. My dad was an engineer and also a master carpenter and he used to build furniture on the side. I learned how to build furniture. I took wood shop when I was a kid. All of a sudden I’m back in the tools shed fixing people’s bones with screws, nails, and plates.
Todd Schlosser: Right. The tool shed’s just a little bit nicer and cleaner.
Dr. Reznik: Much nicer, much fancier. A lot more expensive. And then I’m an engineer so the physics of it, the mechanics of it is all very appealing. All of a sudden I found myself, “Oh, this is perfect for me.” And again, more physical, more three dimensional reasoning, less words. I’m not memorizing 500,000 pharmacological numbers and names. I’m looking at structures and trying to figure out how they work. So for me, that became magical and that’s where I ended up.
Todd Schlosser: Yeah. It seems like with your specific skill set, it’s perfectly in your wheelhouse and you took a very strange pathway to get there but it seems like the way your brain works, it’s the perfect place for you to be, which I think is super interesting.
Dr. Reznik: Yeah. It just worked out that way. Everybody’s like, “Do what you love or love what you do” or whatever it is. Sometimes you get the marriage of things, what you can do and what you love to do ends up being the same thing so that worked out really well.
Todd Schlosser: Yeah, absolutely. And it’s interesting with that diagnosis of dyslexia you went on to write two books.
Dr. Reznik: Yeah. So that an overcompensation, right? One of my frustrations was I liked the idea of being able to put ideas down in words and be able to have some impact. At some point my son decided to become an organic chemist. I really wanted to go into medicine because whatever. I said, “Even if you become an organic chemist, why don’t you do an MD/PhD?” He said, “No way, dad. I’m not going into a hospital. I have no interest in hospitals.” And he said to me one day, he says, “Listen dad, you only help one patient at a time. If I make one molecule and cure a disease, I’m gogn to help millions of people.”
Todd Schlosser: It’s not bad reasoning.
Dr. Reznik: So when your son says something like that, you just say, “Okay. There’s the challenge.” How do you leverage what you know to help more people than you help now?
Todd Schlosser: Right.
Dr. Reznik: And there are several ways. You can make inventions, you can change the field that way. You can write something that’s helpful for patients. You can write articles on a national level. So the last 10 or 15 years I’ve really turned some of the things around in that sense. I’m looking at I know how to do certain things. Maybe I do them a little differently because of my background in engineering. But maybe some of that’s worth sharing on a bigger scale.
Todd Schlosser: Yeah. Absolutely. And you’ve done that. You’ve written the two books. And you’ve done a lot of work with AI, which I’d like to talk about at some point. But you also had a hand in creating new instruments, right?
Dr. Reznik: That’s right, yeah.
Todd Schlosser: So how did you get into that? Clearly it helps to have an engineering background, right?
Dr. Reznik: Yeah.
Todd Schlosser: But did someone come to you with problem and you were just excited to solve it or did you see a problem that you wanted to solve? How did the whole thing come about?
Dr. Reznik: So the thing for in invention and the way it happens is I’m doing something and I see it like, “I wish we could do this better.” And then so I look for people who are doing it better and maybe there are tools out there you don’t have and you kind of ding around. And then the problem just ruminates with me, like, “Ugh, this is so stupid. Why can’t we do this another way?”
Todd Schlosser: Yeah. There’s got to be a better way.
Dr. Reznik: There’s got to be a better way. And then usually I wake up with something.
Todd Schlosser: Wow. Really?
Dr. Reznik: Usually I’ll wake up in the morning, “Oh, this is interesting.” I’ll be in the shower and I’ll take another 10 minutes in the shower thinking about it. And then sometimes an idea will bite me and then I can’t let go of the idea. One of the earlier ones that actually J&J picked up was my grasper/grabber for surgery. There are several different small instruments for shoulder surgery to manipulate sutures. So initially when we did shoulder arthroscopy, you really couldn’t do much. We looked and then you open the patient. Then you looked, you do a little clean up and then you open the patient. Then you look. Maybe you can put some stitches in if you’re careful and you could do something. Then you open the patient anyway. At some point, we said, “Look, we should be able to do this and figure out how to do it.” But then you have to invent the whole new series of instruments to do it.
Todd Schlosser: Right. Because you can’t fix it with technique. You have to do it with an instrument.
Dr. Reznik: Well, you have to have techniques that match the instruments and the instruments match the techniques. I mean, there are certainly ways to MacGyver anything. There’s a little joke about me in the OR that they’re willing to give me almost anything, even if it’s broken because they know I’ll make it work, which is not what I expect, but occasionally in the operating room things happen out of your control and you have to MacGyver something. You have no choice. You face a problem that you just don’t have the instrument for because there is none.
Todd Schlosser: Absolutely. And you have to roll with it.
Dr. Reznik: You have to roll with it and you have to be creative. That’s a little bit of the creative side of me. I like that. But then every once in a while I finish that and I go, “Oh, I wish there was an instrument for that.” So one of the things that happened early on was we used to use suspension to hold the arm in the air. Originally, and no one believed this, but originally first doing shoulder arthroscopies, and this is in the late ’80s, we actually had a hook and a pulley on the ceiling and a rope. You tied a little gauze and muslin a hook and then you put that on the rope and the rope went to the ceiling and a weight on the wall and then you’d do traction and that held the arm in zero gravity, if you will.
Todd Schlosser: Sure.
Dr. Reznik: And with the arm in zero gravity, it’s easy to look inside the shoulder and move it around while you’re looking and check everything. So that was great except how do you do that all the time. Every operating room has to have a hook in the ceiling with a pulley and a rope. It doesn’t make sense. How do you make that portable? There were a couple of devices out there to do a little bit different version of that. But most of them were fixed angle devices. They were set up. You hooked it up. Just like the ceiling, it wasn’t moving. I’m thinking to myself, “Wouldn’t it be nice if that was totally variable, if you had full six degrees of freedom and you could put the pulley anywhere you want in space, and therefore position the arm?” So if you decide while you’re operating, “Gee, I’d like the arm abducted more, rotated this way, or pulled that way,” someone can adjust the device and the pulley would move and then the direction of traction would move based on where the pulley moved.
Dr. Reznik: And so I actually invented a shoulder holder that had full six degrees of freedom. You attach it to the operating room table and now it’s made out of aluminum so it’s very light weight. So then the nurses didn’t have to lug this equipment around.
Todd Schlosser: Also probably more sterile than rope.
Dr. Reznik: Well, actually, we don’t sterilize that. It’s interesting. You put it on the table, it’s a fixture of the OR table.
Todd Schlosser: Oh, okay.
Dr. Reznik: And there is a rope still but then there’s a hand piece that goes on that holds the hand that’s got … Another invention is force spreading gel so that when you apply the traction, it instantly spreads the force evenly over all the skin. One of the problems with a lot of the traction devices, it tends to bind up at the wrist and all the force is concentrated there and you take it off at the end of the case, the wrist is bright red. But if you have force spreading gel, it’s spread over the whole arm. It sets itself to the tension and you can apply traction for two hours and you’re not going to have a skin problem or much reduced problems with the skin. That was another invention later on.
Dr. Reznik: So anyway, so having done that, I got a patent on that. It was my first patent. Back in the day when I did that, the patent office wasn’t as sticky as it is now. All the drawings are hand drawings that are actually in the final patent, which they don’t accept anymore. It has to be a true mechanical drawing. I did it pretty much by myself.
Todd Schlosser: Do you have a shed somewhere that you go to tinker with stuff?
Dr. Reznik: So this one actually I had … The first one, the grasper thing, I posed the idea to J&J. Back then, they were very desperate for new instruments. So they would make you a prototype if you sent anything. Draw it on a napkin, they made it for you. And they made 27 versions of it before we finally made the real one. This other one, I befriended someone who had a marine shop nearby in Connecticut and I said, “You know, I have this idea. Do you think you could make it?” The guy goes, “Oh, yeah. We can make that. We make marine stuff.” So steel tubes and and flanges. And you cut the parts and put it all together.
Dr. Reznik: And then the second time I had it made there, they were not able to make it so I went to another guy and he misunderstood the weight requirement. So he thought that the arms were very heavy for some reason and the traction was supposed to very high. So he made it, you could put an elephant’s arm in the thing. It was so strong.
Todd Schlosser: It’s a bit much.
Dr. Reznik: But the nurses couldn’t pick it up.
Todd Schlosser: Oh, because it was so heavy.
Dr. Reznik: It was double walled steel. It weighed a lot. So then the third time, I actually found a local manufacturer in Connecticut called Innovative Medical Products, which is a small manufacturer that does a lot of medical supplies for positioning on tables and things like that. And they made a version that’s all aluminum and that one is very light weight and very flexible and usable. It’s used in a lot of places. That’s neat.
Todd Schlosser: And your inventions or innovations haven’t always been in the physical space, right? Didn’t you invent some stuff, some search engine optimization type things? You hold two patents in that realm as well.
Dr. Reznik: Yeah. Again, when you’re asleep at night and you wake up with a crazy idea, you can’t help yourself. But I have five patents in orthopedic stuff with fluid management, shoulder holders, things like that, cannulas for arthroscopy and those patents. And then more recently I decided that the internet is controlled by the wrong person. It’s controlled by the average advocate of everyone’s thoughts. You go and you look up something and Google tells you what the last million people wanted. So if you want to find something unusual, it’s actually quite hard.
Dr. Reznik: And the funny story is if you go to a classroom of kids and you say, “Write a paper about Hamlet.” And they all Google Hamlet and they all get the synopsis and they all write the same paper because they get the top 10 results and everyone gets the same top 10 results.
Todd Schlosser: Yeah. No one’s going past the first two pages of search results.
Dr. Reznik: Exactly. So I said to myself, “How do you get past the first two page and find things serendipity that you don’t know about that you’d want to know about?” So I came up with this idea that we uniquely know, and this I the first patent of the two patents, you uniquely know better what we don’t want than what we do want. So for example, if you’re going to a restaurants and you have six guys in the room. You say, “Where do you want to go?” The first guy rings in, “I do not want Chinese.” I mean, that’s what they say, right?
Todd Schlosser: Yeah.
Dr. Reznik: The other guy says, “Well, I’m not going to Indian. I had it last week. I’m sick to my stomach.” Whatever it is. I mean, no offense to any typical food, but someone might have … Or, “I need gluten free.” Whatever it is. But most people will tell you the things they don’t want. They won’t come to a consensus and say, “Oh, I absolutely want this tonight.” Because they don’t want to offend the rest of the crowd. But they won’t do it. So I said, “Well, what if we could do that idea in search? What if we could tell the engine what I don’t want from the answers you gave me and have AI figure out what that really means and scan a whole block of results and reorder that scan of results in a way that makes sense for what I’m thinking but I haven’t told you what I’m thinking yet.”
Todd Schlosser: Right.
Dr. Reznik: Okay. So I actually created an app for this and it’s available for the iPhone.
Todd Schlosser: What is the name of the app?
Dr. Reznik: It’s called Twittle It.
Todd Schlosser: Yeah. Twittle It.
Dr. Reznik: Twittle It Search.
Todd Schlosser: Did you come up with that name?
Dr. Reznik: Yeah. Because it’s to whittle it down.
Todd Schlosser: To whittle it down, yeah.
Dr. Reznik: Twittle it. So the app, what it does is you put a search term in and you see the answers and what you do is just swipe two or three of them. You can swipe right or left. If you like it or you don’t like it. Once you’ve done that, you hit the Twittle it button and the AI engine takes the information from the swipes, negative and positive, and it goes through the remaining results in the set of results and it reorders them based on your thoughts based on what you swiped by determining what it thinks is in common on the swipes and the remaining results.
Todd Schlosser: Right.
Dr. Reznik: And also what’s less common. And then it gives it a score, ranking. So the most negative is closest to the negative swipes and the most positive is closest to the positive swipes and it just ranks them down and it represents the same results in that order. So you can imagine what bubbles up to the top are things closer to what you want.
Todd Schlosser: Sure.
Dr. Reznik: It actually works. It’s crazy when you see it because it took a year to actually do it with a whole bunch of computer geeks and gurus.
Todd Schlosser: So do you actually go in there and program that code yourself or do you come up with the idea and then outsource it? I imagine you’re super busy.
Dr. Reznik: Yeah. I’m too busy.
Todd Schlosser: But building out that infrastructure from a programming standpoint would be difficult.
Dr. Reznik: When I was in high school and college I did a lot of programming.
Todd Schlosser: Yeah. You have a background in that.
Dr. Reznik: I have a lot of background. So I could have done it myself. But the amount of man hours to do it myself and practice orthopedics and do all the other things I’m doing was impossible. I actually hired a team of engineers. And then we would meet every week. We’d have a webinar meeting or live Zoom In meeting actually on Zoom. It actually was quite interesting doing that. We’d sit, the engineers and me, and I would review what they did and talk about what we thought and we’d look at the graphic presentation and how it worked and problems with certain areas that became glitches. For example, to do movies, if you do movies and you do it like a common search for movies, it isn’t really as satisfying as if you go to a movie database and just do the search in that.
Dr. Reznik: So if you put in movies right now, what it does is it gives you the standard Google type result. But if you ask for featured movies, it will go into a movie database and only give you about the movies themselves and take out all the extraneous movie results like the interviews with people. If you want to look at a certain genre, you could do it that way. Or you could swipe and it’ll sort based on your swipes, again the movies based on the preference. Let’s say I don’t want to see a horror movie, I swipe away that one, but I really want to see a movie about Donald Duck, I swipe that one. All of a sudden the Disney movies pop to the top and the horror movies pop to the bottom and then you can get a decision about that.
Todd Schlosser: Excellent.
Dr. Reznik: One of my partners said that it’s great for date night. You can’t decide with your girlfriend what kind of movie to watch but you could each get a couple of swipes and do the Twittle It button and then you could see what pops to the top which is a movie neither of you thought about but answers the question because it’s a movie that you haven’t seen and yet it’s something you could see that you might both enjoy.
Todd Schlosser: And technically counts as compromise.
Dr. Reznik: That’s right. It counts as a digital version of compromise.
Todd Schlosser: Yeah. Absolutely. So I grew up a science fiction nerd so when you say things like AI, I think of Isaac Asimov and his three rules and stuff like that. I don’t know that everyone will have that background, I guess. Can you high level when we talk about artificial intelligence, which is AI, what we’re referring to?
Dr. Reznik: So this is my favorite thing to talk about in a lot of ways because people will be shocked with what AI really is. AI truthfully is any activity that a device, a machine, or an object or something that does that would have been done by human thought. All right? So you could ask the question what is the first time you see AI in history? Some people say, “Well, AI was this big meeting in Dartmouth in 1950s and these guys got together and they decided to create this whole idea of a perceptron, which is an electronic version of a human neuron. And they decided if they wired them all together, we can create a brain that was electronic and we could teach the brain how to do stuff and we’re going to get artificial intelligence to be able to solve every problem in the world.”
Todd Schlosser: Right.
Dr. Reznik: And then it crashed because no one could make enough perceptrons, which was very expensive. I don’t know if people remember back, but if you had a megabyte of memory in the 1960s, it cost two million dollars. And a megabyte of memory today costs about a hundredth or a thousandth of a penny. I think it’s less than a thousandth of a penny now for one megabyte. It’s so little. So the difference between that and that cost is all the difference in the world. In the 1960s they couldn’t solve any problems really. They just didn’t have enough memory to do it.
Todd Schlosser: No. Absolutely. It blows my mind that we were able to get to the moon in the ’60s. I have much more data capability on my phone than they did then.
Dr. Reznik: Than the whole world had.
Todd Schlosser: Yeah. It’s amazing.
Dr. Reznik: One phone has more data capability than the entire world did in the 1960s. It’s uncanny the difference. Your phone would be billions of dollars today. Imagine if it was two million dollars per megabyte and you have a one gig phone.
Todd Schlosser: Yeah. You could buy countries.
Dr. Reznik: Yeah. You could buy a country for what your phone is worth. Yeah. That order of magnitude. It’s hard for people to wrap their head around that. But I’m going to say something that’s even crazier is that AI goes back even further. da Vinci had these things called the cart, which was like a little cart you could program and he made a mechanical lion for Venice’s 200th or 300th anniversary and he envisioned a mechanical knight and he had drawings for all these things. They would be artificially run. He would have a little program in it and it would do whatever.
Todd Schlosser: Like automatons.
Dr. Reznik: Automatons, right. Exactly. So da Vinci did that but then after da Vinci there was Babbage. I don’t know if you know that. Babbage was the guy who did the different calculator where he actually solved multi variable equations with a calculator that was mechanical gears. And then there was Ada Lovelace who is actually …
Todd Schlosser: She had a computer program language named after her.
Dr. Reznik: Right. She was the first person to see Babbage’s drawings of his calculating machine and she decided she would create an algorithm for that machine to solve bigger problems than it was designed for. So she’s credited as being the mother of programming. He’s the father of computers. So you say AI goes back to then. I say, “No, no. That’s not even true.” Go to Stonehenge, all right?
Todd Schlosser: Yeah.
Dr. Reznik: Think about this for a second. Stonehenge was erected and when the sun was in certain positions, everything would like up and the beam of light would come between the rocks and that would give you seasons, right?
Todd Schlosser: Yeah.
Dr. Reznik: And it would tell you when to plant and when to … Right? But if you had to look at the sun and calculate and check the stars and try to figure it out on your own, it would be a lot of brain work. So what if we created a device that did the calculation for us and we put these rocks up so when the photons of the sun hit the rocks in the right way, the answer would appear on the ground.
Todd Schlosser: Literally.
Dr. Reznik: Or a sun dial. That would be AI. A machine that does some thought for us that replaces human activity. So anything that does that, your watch does that.
Todd Schlosser: Absolutely.
Dr. Reznik: What time of the day it is, right?
Todd Schlosser: Yeah.
Dr. Reznik: Any computer does that regularly. Most of that we would consider shallow AI. So this is the distinction, right?
Todd Schlosser: Right.
Dr. Reznik: So shallow AI is the device. I programmed it. I know what it’s going to do. Easy one is you go to a BMI calculator. You want to know your body mass index, see if you’re dieting properly, whatever. And you put your height and your weight in and it gives you your BMI. It’s a formula, it’s a calculation. It’s AI. In a way, it does the calculation for you. That’s a simple one, right?
Todd Schlosser: Yeah.
Dr. Reznik: But others are much more complicated. But all of those are a human sat down and said, “Okay. We got this problem. Let’s get a piece of paper out. Let’s get a pencil. Let’s get a computer. Let’s figure out a formula. Let’s create an approximation to the answer or maybe even perfectly accurate answer and then we’re going to put it into a program and create a device and we don’t care what buttons are on it.” Your iPhone could have any button on it. A calculator has less buttons on it. But we don’t care what buttons you use to activate it, but it’s going to get an input and give me the output. Hey Siri.
Todd Schlosser: Right, absolutely.
Dr. Reznik: Now, Siri’s a little more sophisticated than that. But initially it wasn’t. Initially if you go back to the original cars that had voice commands, they understood six or seven words. There was an algorithm to make sure they understood those six or seven words it was to understand. Call Tom. It would call and the phone would activate. That was all you could do. But that was all done by algorithm. It was written and some guy figured it out.
Dr. Reznik: Now you go ahead and say, “Okay, that’s shallow.” There’s a ton of shallow AI. You can’t go two inches anywhere. You can’t miss shallow AI, it’s everywhere. Now yo go to deep AI. What’s deep AI? Deep AI is let’s emulate learning and let’s take these perceptrons and we’ll make a network in layers that look really like a retina. Because if you look at the retina, the way the retina works, you get light in the front, you have a layer of neurons that get activated. There’s another layer of neurons that process it. There’s another layer of neurons. There’s the optic nerve brings it to the optic ring. There’s more layers. And then at the end, you say, “Oh, it’s a cat.”
Todd Schlosser: Right.
Dr. Reznik: Or it’s a dog or whatever. So when AI is learning, it’s saying, “Okay, I’ll show you a picture of this article on AI, and the computer gets all this information in its visual cortex and then it says, “Oh, these are letters.” One layer. Another layer it says, “Let’s break them down, what letters they are.” And go on and on. And I could program that all in. But if I do it the other way and say I’ve got millions of letters and characters and words and I keep showing it to the computer over and over again and I keep telling the computer what the outcome should be, eventually it creates all these factors in the middle of all these layers of neurons to make it work.
Todd Schlosser: It does the learning for you.
Dr. Reznik: It does the learning just like your brain does. And then you give it a mystery one and it tells you the answer.
Todd Schlosser: And the true test is if it gets the answer right on that next one.
Dr. Reznik: Right and the one after that.
Todd Schlosser: Yeah.
Dr. Reznik: Now, there’s a lot of assumptions there, right? And I think as I’m writing my series of articles on AI, one of the things I’m writing about now is … The next article that’s going to come out is talking about some of the assumptions we’re making. The first assumption is the data we show it is representative of the real world. Sometimes we show it data that’s not really representative of the real world. It’s a fantasy set of information. And that data represents not only the real world now but some of the real world in the future because if I take a future problem, my assumption is words and letters aren’t going to change and if I show words and letters to the computer, the future words and letters are gogn to be same words as now.
Todd Schlosser: Right.
Dr. Reznik: If I change languages on the computer, it’s going to be lost. So the assumption number one is that I’m representing the real world I’m dealing in and assumption number two is that the real world I’m representing now is the same world I’m going to be asking questions about in the future. Those are the big, big assumptions. And then the data collection is similar. So in medicine one of the problems we might have is, for example, if I have an x-ray and I take it on an x-ray machine and I do AI and I say, “Find the tumor.” And I teach it how to find tumors perfect.
Dr. Reznik: Now I go to another x-ray machine and I do the same problem, but the voltage is different, the tube distance is different, the x-ray film that I’m using is a different digital CR recorder that records slightly differently. There’s different noise because one building is well shielded from RF and the other one is next to a radio station. Now all of a sudden the data I get in isn’t exactly the same, even though to the human eye it looks pretty similar, and AI would have trouble with that because I didn’t give AI to learn 25 different units to learn from representing all the x-ray units out there in the future. I just did my one machine’s worth of data and now I try to train it to do that problem and it does it well and now all of a sudden I present it with something that has a different set of noise and background.
Todd Schlosser: Yeah.
Dr. Reznik: And there’s more things we could talk about. But there’s a lot of assumptions underneath AI that people don’t wrap their head around so well. AI is great but then it has some pitfalls.
Todd Schlosser: So what are some of those pitfalls of AI?
Dr. Reznik: There’s so many ways to approach it. We talk about what happened when they made an AI bot for social media. It was a very popular concept, right?
Todd Schlosser: Yeah.
Dr. Reznik: So you make a bot and you say, “Okay. I have an AI bot and I’m going to pretend that I’m a person and I’m just going to read social media stuff and then I’m going to regurgitate appropriate answers when people write me.” So I’m pretending I’m on Facebook and I’m a person but I’m really an AI bot. What do I learn? I learn what I see. So quickly the AI bot becomes a bigot and prejudiced and angry and cursing and only tells horrible things because it amplifies what it sees.
Todd Schlosser: Garbage in, garbage out.
Dr. Reznik: Right. And it’s learned what you taught it but then sometimes you might be scared what we taught it is what we do, it reflects upon us, right?
Todd Schlosser: Yeah.
Dr. Reznik: So they’ve done some experiments like this and some of the AI bots come back with these results that are appalling because what’s out there isn’t so good. And the assumption is, and this is another assumption here, is that AI itself has no prejudice, it has no natural language, it has no presumptions, and it has no assumptions. It just learns from the data set. The problem is, a lot of the data sets have prejudice, assumptions-
Todd Schlosser: Built in. Yeah. It has all those things built in.
Dr. Reznik: Built in. Racial discrimination sometimes, sex discrimination. There was just an article the other day about women being discriminated by credit card companies but the AI bot is done on spending habits and payment histories. But it turns out that in their data set that might have been prejudice against women and so the credit card company’s not giving the same credit limit to women as men and it comes out as a discrepancy. But if they looked at the spending histories, they can’t justify the credit limit based on the spending histories traditionally. That wouldn’t be true in my house. My wife uses the credit cards. I hardly use them at all. I take $20, I’m good for the whole week. So if they did the credit history, actually her credit history is better than mine because she does most of the spending and paying and I don’t.
Dr. Reznik: But in some households, everything’s under the man’s name. So even if the woman does all the work and has a job and does the spending and pays the bills, it’s credit to him and not to her and if you do analysis of her credit risk, there’s no proof that she’s a good payer.
Todd Schlosser: Right, because it’s looking at just the data set, not the social norms that play out.
Dr. Reznik: Right. In my household we have two separate credit cards because I wanted my wife to have her own credit rating. It turns out in time hers is better than mine because she does most of the … I’m busy operating and she does the stuff that has to do with paying all the bills and everything else. So she’s a much better credit risk than I am even though I’m the one out there operating.
Todd Schlosser: So how does that play out, not with credit card companies but in the world of health? What are some problems that can factor in? I could imagine that if you’re looking at surgery outcomes but not taking into account social determinants like zip code. Is that stuff that you’re bumping into when you’re trying to I guess implement AI?
Dr. Reznik: Absolutely. And this is a two edged sword. Let’s say we take away zip code and race and all those things and strip the data and then we just look at outcomes. Let’s say because their zip code, someone doesn’t have access to certain services and doesn’t get a certain kind of care because they don’t have access-
Todd Schlosser: Yeah. That happens all the time.
Dr. Reznik: It happens all the time. So lack of access, they might end up with a discrepancy and then the AI bot may say, “Look, those people don’t do this, I’m not going to offer it to them.” Or worse, they’re not compliant with treatment because they don’t have the ability to do follow up. So then they’re at higher risk for complications. So if I ask the AI bot who should I operate on, who’s going to have the best outcomes? Tell me the patients I should avoid operating one because I don’t want to have bad outcomes. It’s going to kick out a whole bunch of people who are disadvantaged in many different way.
Dr. Reznik: So here’s the dilemma. If we use AI to improve outcomes in medicine, which is our big goal, and the insurance company takes this data and says, “Oh, you can’t operate on people with BMI over 40 because their outcomes are terrible.” Well, if we don’t go back and say, “Why is their BMI over 40 and what can we do about it?” Then we’re going to displace a whole bunch of people out of the healthcare system that are not going to get care.
Dr. Reznik: Moreover, if we use the AI bots to rate the physicians, and this is the big fear, because this is sort of happening already, and we rate the physicians based on their outcomes and someone is in an area where people’s general health is poor-
Todd Schlosser: Yeah. That could-
Dr. Reznik: Their outcomes are going to be poor, right?
Todd Schlosser: Yeah. And they could be the best surgeon in the world, they’re just-
Dr. Reznik: It won’t matter.
Todd Schlosser: Yeah, absolutely.
Dr. Reznik: Right. And in fact, what used to happen is the tertiary centers, a lot of them would get the worst cases. So I send all my bad cases to this one center where the outcomes … Let’s say there’s a 30% chance of a bad outcome on these patients because they’ve been referred after failed procedures and this and that, and in another area they’re just doing the primary procedures and none of the failed procedures because they send them all to the big university, the outcome profiles are going to look different.
Todd Schlosser: [crosstalk 00:36:28].
Dr. Reznik: They’re going to say don’t get operated at the university because you’re going to have a bad outcome. So how do you risk stratify? How do you look at data that way? I think, again, when we put the data in, we presume the data is neutral and the AI machine has no prejudice. But the data isn’t always as neutral as we think it is.
Todd Schlosser: Right. And we’re sort of building prejudice into the system even though it doesn’t have it inherently.
Dr. Reznik: Exactly.
Todd Schlosser: If the data has it, it’s only going to crunch whatever data it has.
Dr. Reznik: Sure. And then I wrote an article about the ethics of AI and one of the things, that’s one of the elements, right?
Todd Schlosser: Yeah, absolutely.
Dr. Reznik: One of the other elements of the ethical issues in AI is autonomy. Patients are theoretically in the United States at least allowed to have choice and allowed to actively participate in decision making and make informed consent. But what if the insurance company uses AI to deny procedures based outcomes and they tell your doctor what procedure he has to do even though that technique in his hands is not what he’s comfortable with. There might be three operations that are good for that problem and the guy’s been doing one for 30 years and he happens to be phenomenally good at it and in his hands, he has great results. But if someone is not skilled in that technique, this other technique might be easier for them and they may get good results with that technique.
Dr. Reznik: In shoulder surgery when we do arthroscopy, there are people who learned how to do shoulder arthroscopy in a lateral position and people who learned how to do it in the beach chair position. Purely a training artifact because where you trained and who you worked with pushed you in one direction or another. It turns out there are different considerations with each. There are advantages and disadvantages of each. It may play out that one might have a higher complication rate than another but half the surgeons are trained in one and half the surgeons are trained in the other. The ones who are trained and comfortable in beach chair are not necessarily trained and comfortable with lateral. And if you say now you have to do all your cases lateral as opposed to beach chair because you’re going to avoid this complication, there’s going to be a lot of resistance to doing that.
Todd Schlosser: Well and possible bad outcomes.
Dr. Reznik: And probably a lot of bad outcomes in the transition. There might be a heavy price paid for some patients having it done in a way that the surgeon’s not comfortable. It’s complicated, as you can see. It gets very complicated.
Todd Schlosser: So we’ve talked about some of the pitfalls of AI. Where do you see, and maybe it’s not there yet, but where do you see AI going that you’re excited about it?
Dr. Reznik: Well, there’s a couple of things. Right now I’m beta testing something from a company called Image Biopsy. That company makes a device AI driven where it takes an x-ray and makes measurements, standard measurements, but they’re actually not that easy to do accurately because it really requires a visual analysis. So I’m looking at something and I want to know is there signs of arthritis in the knee, meaning are there osteophytes or bone spurs? Is there sclerosis in the bone? Is there loss of joint space? So if I look at an x-ray, I say, “Oh, that looks narrower than the other side.” And I can physically take a caliper and try to measure that. It’s a little arduous, takes a bit of work. But if I can have an AI device that looks at the x-ray and gives me the measurements in every part of the joint that’s important and then gives me a sense of a scoring of how much sclerosis there is and how many bony spurs there are, that could happen very quickly, maybe in a minute.
Dr. Reznik: And as I’m testing it now, and I’m really literally this last two weeks it just got FDA approved. So literally in the last two weeks, this is a new thing. It’s not out there yet but I’m starting to try to use it a little bit and what I’m seeing is that I discovered today on a patient, they have more arthritis in one part of the knee than I realized when looking visually at it. I was a little fooled. It was almost like an optical illusion. And when you look at the actual numbers, it tipped me off to a problem that I wasn’t thinking about and it probably explains why the person has recurrent swelling. Just visually, I probably wouldn’t have come to that. It was a little bit of a Eureka moment. It actually happened today so I’m kind of excited about it.
Todd Schlosser: Yeah. That’s awesome.
Dr. Reznik: I can’t wait to report back to them that this happened because I think that in process when I first looked at it, I said, “Oh, how valuable could that be? I’m interested in AI so great, I’ll do this for you. I’ll look at these things. I’m not sure how valuable it would be.” But then I’m sitting in clinic and I’m looking at this thing and I said, “Wow. That made me think about it differently and now I have a different answer for the patient based on what I just saw.”
Todd Schlosser: Yeah. And a better answer.
Dr. Reznik: And a better answer. And also when I asked for the MRI study, I’m asking a different question of the radiologist.
Todd Schlosser: So it’s compounding.
Dr. Reznik: Exactly. So that was a little example of that. The other one that’s really interesting is another company I’m a little involved with is a company called In Hatch. And they’re out of New Jersey and they take two dimensional images and create a three dimensional AI version of it and they use it to predict ahead what prosthetic you’re going to use to replace the knee or the hip or even spine, put in what size screws you need.
Todd Schlosser: So it’ll take a 2D x-ray and 3D-
Dr. Reznik: Yeah. Plain x-rays from the office.
Todd Schlosser: Just make a model of it in 3D?
Dr. Reznik: The AI system makes a three dimensional model from the 2D images. You need a marker to make sure you get sizing correct.
Todd Schlosser: Right.
Dr. Reznik: But then it can go ahead and predict with reasonable accuracy what total hip or total knee you might get. And what that does is then it turns around and gives you a more accurate start in the OR, first of all. You know what you’re going to do. And you don’t need a tray of 27 total knees to figure out which one’s going to in. You maybe could make a tray of three or four different ones, bracket on either side of the size you predicted. The requirement of sterilization, let’s think about greenhouse gasses and all that stuff. Let’s say I’m sterilizing a tray of 27 things every time I do a total knee and I’m doing 100,000 of them a year. In the United States it’s probably a million of them a year. And then I can go to a tray this big and sterilize it and how much less energy, how much less cleaning, how much water I’m going to use less.
Todd Schlosser: How much less chemicals?
Dr. Reznik: Chemicals we’re going to use less. The cost of the inventory is going to go down. We’re going to do something indirectly through AI by figuring out ahead of time what size knee replacement to put in that’s going to help us with our environmental problems. So that to me is neat. It just blows me away. The implications of some of the things we do, we don’t really fully understand either. So you get this one idea, it’s really great. It’s going to make surgery easier, blah, blah, blah. But what’s the downstream effect of that? Well, we might save some energy. We might use less water. Might be better for the planet.
Todd Schlosser: Right, yeah.
Dr. Reznik: Right? So that’s neat. So I’m excited about those two right now. There’s other things on the horizon. I have some ideas about how I might use my Twittle It idea, of how to use that for medical diagnosis.
Todd Schlosser: Build a better Web MD.
Dr. Reznik: Right. Well, I mean, I think the problem for a lot of people is I go on Web MD and I don’t know what I’m looking at. I can’t figure out my way around.
Todd Schlosser: Right. And you’re terrified.
Dr. Reznik: And you’re terrified. My next patent has something to … Actually, I can’t really talk about it too much.
Todd Schlosser: Yeah. Don’t give it away.
Dr. Reznik: But my next one has something to do with that.
Todd Schlosser: Awesome.
Dr. Reznik: Approaching that problem.
Todd Schlosser: So I’d like to close with one final question. Because you’ve been in the industry so long and you have a unique path, I’d be surprised if you don’t have younger doctors coming to you for advice. So what advice do you give those who might be freshly out of their residency or fellowship about embarking on a career in this space?
Dr. Reznik: Yeah. This is the zillion dollar question, right?
Todd Schlosser: Right.
Dr. Reznik: Because people do ask me that.
Todd Schlosser: I’m sure.
Dr. Reznik: I think the frustration for some people is they’re trying to guess ahead what’s going to be 20 years from now. The truth is, no one knows. We have no window on it. If you asked me … In the beginning we talked about things I did when I started my practice and how I operate now, there isn’t a single procedure I do the same way. There isn’t anything I do the same way. All the techniques are different. Even if I invented some of the instruments, they didn’t exist so everything is very different. But the cool thing is, is everything is very different. So I think the advice I give people is you got to do what you like and love and you got to do things you’re good at so you can marry what you’re good at and what you love. It’s going to serve you well because it’s going to show on a daily basis. You’re there, you’re doing stuff. You’re exited about it. People know you’re exited about it. It’s infectious. People gravitate to people who are excited about what they’re doing.
Todd Schlosser: That’s absolutely true.
Dr. Reznik: And you’re better at what you do if you’re excited about it.
Todd Schlosser: Yeah.
Dr. Reznik: I always worry that people say, “Oh, this’ll be a great career because I’ll make a lot of money.” The people that make the most money are the people who went into something because they purely are crazy about it, loved it so much, worked so hard at it. They came up with some idea in that area because they were so excited about whatever they were doing that they became a gazillionaire because of that. Now, that hasn’t happened to me yet but that isn’t even important anymore for me at this point.
Todd Schlosser: Right.
Dr. Reznik: It’s funny. To me, it’s the process of inventing. It’s the knowledge of creating something new. It’s the hopes that other people might get some benefit for it. It’s a cool thing if I made a little instrument and J&J sells it and there’s a guy in Japan using it every day. I don’t even know the guy. The guy doesn’t know I invented it. But I know there are Japanese people getting shoulder surgeries where they’re moving sutures around inside their shoulder. And right now when we’re talking, there might be someone holding my instrument in their hand and doing something with it. It’s something. It’s a thing. How do you wrap your head around that? I can’t. It’s just a crazy thing. So there’s a huge pleasure in that. I think at the end of the day, if you can take those pleasures out of what you do, everything else doesn’t matter really because it makes things better for you.
Todd Schlosser: Absolutely. Well, Dr. Reznik, thank you so much for taking the time today.
Dr. Reznik: It’s my pleasure.
Todd Schlosser: We really appreciate you coming on the podcast.
Dr. Reznik: I’m so glad to talk to you.
Todd Schlosser: This was excellent. Thank you.
Dr. Reznik: My pleasure.