Code & Cure
Decoding health in the age of AI
Hosted by an AI researcher and a medical doctor, this podcast unpacks how artificial intelligence and emerging technologies are transforming how we understand, measure, and care for our bodies and minds.
Each episode unpacks a real-world topic to ask not just what’s new, but what’s true—and what’s at stake as healthcare becomes increasingly data-driven.
If you're curious about how health tech really works—and what it means for your body, your choices, and your future—this podcast is for you.
We’re here to explore ideas—not to diagnose or treat. This podcast doesn’t provide medical advice.
Code & Cure
#59 - When a Humanoid Robot Removes A Gallbladder
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
What happens when the “robot surgeon” stops being a fixed set of arms and starts walking into the operating room like a person? We explore a fresh research result where a teleoperated humanoid robot performs laparoscopic gallbladder surgery in pigs, and we get specific about what worked, what didn’t, and why the details matter more than the headlines.
We start by defining what “humanoid” actually buys you in surgical robotics. The promise is less about a face and more about human-like dexterity and mobility: a robot that can reposition itself, approach from different angles, and potentially use standard human tools in a normal OR without the heavy, purpose-built infrastructure that platforms like the Da Vinci system require. That leads to a crucial reality check: this is not autonomous AI doing surgery. A human surgeon is still on the controls, and a bedside assistant is still in the room handling the constant small needs of a real case.
From there, we dive into the most revealing technical constraint in laparoscopic surgery: the remote center of motion, the fixed pivot point at the skin that instruments must rotate around to avoid tearing tissue. Classic systems enforce it with hardware; a humanoid has to enforce it in software. We talk through what their evaluation shows, including straight-line versus circular motion accuracy, speed tradeoffs, and why a measured 156 ms latency can be a big deal for operator feel and safety.
Finally, we unpack the pig surgeries and the unglamorous blockers that decide whether this scales: range-of-motion limits that force repositioning pauses, recalibration, overheating, and the sterilization problem when autoclaving can destroy sensitive electronics.
References:
In vivo feasibility study of humanoid robots in surgery
Liang et al.
Nature (2026)
Credits:
Theme music: Nowhere Land, Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 4.0
https://creativecommons.org/licenses/by/4.0/
The Big Question About Robot Surgeons
SPEAKER_01What if the next surgeon you meet isn't human? Researchers just had humanoid robots perform gallbladder surgery on a pig.
SPEAKER_02Hello and welcome back to Coding Cure, where we discuss decoding health in the age of AI. And my name is Vasant Sarathi. I'm an AI researcher and cognitive scientist, and I'm here as always with Laura Hagopian.
SPEAKER_01I'm an emergency medicine physician. And today we're talking about humanoid, humanoid robots doing surgery. But that like brings me to my first question because it's not just like any robot, it's a humanoid. So like what is what is that?
SPEAKER_02Well, first of all, uh to me, I think this whole business of robots doing surgery is really cool because you know you see these science fiction movies and someone's on a bed and you have all of these arms coming down and they're all doing different things simultaneously. They're like all perfectly precise, and you know, they help, you know, and if the human is behaving or misbehaving, they're able to like control the human back down and fix the human or whatever. And I'm always amazed at at that because it's like, whoa, that's that's the future, right? Um is it? That's the question of today, I suppose. Yeah, yeah. But you see that this is like one stream that you see in every science fiction movie, is that you have machines that are basically performing full-on surgeries, right?
SPEAKER_01Yeah, and not it's not like to that extent in reality, right? But um, there are robotic systems that are being used by surgeons right now. Like there's the thing called the Da Vinci system, which is probably the most well-known one. Yes, which is not humanoid, it's like robotic arms that perform the surgery with like a human operator kind of videoing in behind it.
SPEAKER_02Yeah. And so that brings us to what a humanoid is, right? Yeah, that imagine the vision of a robot that has 12 arms coming in. That's not a humanoid. A humanoid is a robot that looks like a human that walks around, has has a couple of arms, has two legs, uh has hands, has hands, often has a head and a face of some sort. Uh, maybe not perfectly um, you know, designed to look exactly like a human, like no skin and stuff. But it you can imagine when we think of a an a robot, a robot, modern-day robot, like in the movie iRobot or whatever, right? Those are those are humanoids. They're walking around. Um, and it gets a little bit, it's not like a fixed definition by any means. There's obviously variance to this. There'd be robots that have le don't have legs, but have wheels instead, right? But still have arms and and still have, and they might have, you know, three fingers instead of five, depending on what tasks they perform. But they're human-like.
SPEAKER_01That's the whole idea. It's not just like robotic arms that are isolated on their own. It's like a human-like entity.
SPEAKER_02Yes.
SPEAKER_01That is then going into the operating room, walking around and doing stuff.
SPEAKER_02Yes. And so I think the way to think about this also is what it affords us, right? Why do they even care? Why does it have to be human looking at all? Like what is that and why is that an important thing? And and I and I think it's important to realize that we're not saying it needs to look like a human, like in its face or whatever, right?
SPEAKER_01Uh, we're saying it's like able to act, like have the dexterity of a human or whatever.
What Makes A Robot Humanoid
SPEAKER_02Have the dexterity of a human, but you know, have mobility, right? Being able to move around like a human does in an in a in a in a in an operating room is important. And I think that's kind of the basis for this whole paper was that look, we have these surgical instruments out there, like the Da Vinci system. These are excellent, they do things really well, a limited number of things, um, albeit, but they do them really well, but they require a lot of infrastructure and a lot of setting up to do. And we'll talk about specifically in this paper what that means. But more generally, they just need to have a setup. The OR already needs to be set up for them, right? And so that costs a lot of money and that can be very expensive, and they're very purpose-built. They're often um, they uh they're they require some kind of docking, they require, like I said, space. They might have um, and and the other thing is they might only be doing one specific thing, right?
SPEAKER_01Right, or they might have specialized instruments for certain things.
SPEAKER_02For its own, right, for their own uh the idea of the humanoid is to say, look, we need more people doing these surgeries. We don't have as many humans doing this. Why don't we have instead of having a robot that requires this like bespoke setup, why not just have um a humanoid robot in an existing surgical OR using human tools? Yeah, exactly.
SPEAKER_01The same tools that a human would like manually be using. Yes, that that type of medical equipment. Exactly. This doesn't like you you don't have no humans involved, by the way. Like you have the humanoid robot, but they have a human assistant that's present in the operating room. Oh, it's more than that. And they have someone, yeah, you know, like a physician who's actually doing the surgery maybe remotely, like with the video screen. No, no, it's not maybe that's that's it. Maybe yeah, yeah. They are not not maybe doing surgery, but maybe they're remote. Like they don't have to be physically present. Exactly. But in all of these systems, it's not like the humanoid robot's just like doing it on its own.
SPEAKER_02No, no. And in fact, that's it's not a it's not a it's not an AI system. It's a humanoid robot, it's a machine that somebody's teleoperating. They literally have a joystick, something more than a joystick, but something more advanced.
SPEAKER_00It's like a video game.
SPEAKER_02It's like a just like a video game.
SPEAKER_00Just just removing gobletters. Then do you get like 10,000 points at the end when you've successfully removed it?
SPEAKER_02Yeah. Well, that's it. So they have uh it's all teleoperated. So let me just take a step back for a second because that is important to note. The as much as we've seen a lot of AI development over the last you know a few years and really cool stuff that AI can do, there are some very basic um robotic things that AI can't do even today. Like the AI automated robots, it's even simple problems are really, really hard. And just think about the simple example of grabbing a tool, right? Like a surgical tool or and really anything, there are so many different ways of grabbing it and grasping it. And uh the different ways might have different meanings, it depending on if the tool is dirty or clean or if you're passing it to someone versus whether you're using it. You can imagine holding it in different places, holding it at different angles. Um, and also you want to make sure that in addition to holding it so that it uh can serve its purpose, that you're grasping it so it doesn't fall out. Um, you know, and when you hand it over to somebody or they pass it to you, you grab it in the right way. There's like a whole bunch of different challenges just in the physics and the mechanism of just grabbing a thing, right? That's but beyond that, there's even more. There's like navigation, like moving around the surgical room by itself. Like if it were completely autonomous, it would need to know where it can go and can't go. It would need to be able to move around a surgeon and not bump them, right, while they're busy working.
SPEAKER_01Yeah, what happens if it falls over?
SPEAKER_02Right. Well, it can't it can't trip on something and fall over, which can happen, right? These robots are not like perfectly designed. I mean, I know we've seen a lot of really cool videos, like Boston Dynamic, for example, puts out these really cool um humanoid robot demo videos in which they like push down the robot and it like gets up and stands up, or it's able to like you know, do like parkour, right? Jump through things and do flips and things like that. All of that is great. But in a surgical setting and in many of these settings, it's way more precision and you know, like careful management of space and and movement.
SPEAKER_01The dexterity is like very important for very important, yeah.
SPEAKER_02So, like, so having so
Teleoperation Versus True Autonomy
SPEAKER_02let's just the whole issue of autonomy is out the table here. There's no way this is not doing it on its own.
SPEAKER_01Like we're literally just talking about a remote-controlled robot remote-controlled robot that a human is on the other end of that their tele operates.
SPEAKER_02Yeah, yes. And even that is challenging. And so the I think the really cool part of the study was that, hey, it's a good idea maybe to put a humanoid there because they can move around and up and not not require all of this specialized um setup. Uh, but then the next question is how do we use it for performing surgery? And what are the challenges there? And they focused on laparoscopic surgery.
SPEAKER_01Yeah, they did. And um, laparoscopic surgery is basically the surgery where you get like little tiny holes put in you, right? It's not like a big open surgery. Um, it is much tinier incisions where they put these little trochars through, they introduce these small instruments and they kind of inflate you with air and they do the surgery and then they close you up and you just have these like a couple of tiny little holes. Um, and so that is a very common form of surgery done today. Like if you get your appendix out, most people have it done laparoscopically. If you get your gallbladder out, most people have it done laparoscopically. So it's a great um, it's great because it's very common. It's very commonly used.
SPEAKER_02Right, right, right.
SPEAKER_01And so then the question was like, hey, when we're using a humanoid in this situation, what is it good at? Like it needs to have good dexterity, right? And good range of motion, whatever it is. And like, where, where might there be drawbacks? What might it not be good at?
SPEAKER_02Yeah, and this is where I think it's very interesting. I know we're gonna get a little technical here, but this is where it gets very interesting because of the nature of the problems. And I think it's a nice opportunity to talk about, you know, we've talked in abstract about how um uh the old surgical systems take up space and and the modern human eye robots can move around. But the issues are even more specific than that. So with laparoscopic surgery in particular, one of the challenges is because you're inserting it through such a small point, very small cuts, right?
SPEAKER_01It's not a large opening, and you use these very long, thin instruments and cameras through those openings. Exactly.
SPEAKER_02Yeah, no, exactly. And so, because of that, and because once it's inside, you need to be able to move around. Like there are things you're doing inside within uh within the confines of that small hole, which means um then there's this point called the remote center of oper uh motion or the RCM. Um, that's almost like a point that is the entry point into the skin. That is, yeah, exactly. And and that point, you you want to be able to kind of pivot and move around it without uh moving across it. Because the moment you start moving across it, you're tearing the skin more. So you want to keep that hole and keep that skin, but be able to move the tool kind of um kind of around it. And and I think we talked about this before, and you gave me a great analogy of of the ore in the in the in a in a in a canoe where you know uh or or a boat where the ore is attached to the side of the boat, and you can kind of like move the oar, but there's a fixed pivot point about which it it cannot move.
SPEAKER_01Exactly. And so the robotic instrument can like angle and rotate inside the body, but the point at the skin is going to stay the same. Yes. You can't change that and you don't want to rip that.
SPEAKER_02Exactly. And and in the old Da Vinci systems, the non-humanoid robotic systems, those things were actual metal things, bars and like a parallelogram or trapezium-shaped thing that held that fixed point in place. So the robot machine couldn't accidentally tear your skin. So it would be able to operate within that space. So that was like a constraint that was enforced physically before. But again, that requires a lot of setup and requires the OR to be equipped with something like that. But if you don't have that, now you have a machine, a humanoid robot that's walking around with a tool. How do you enforce that it doesn't move around along that pivot point, that it stays at that pivot point?
SPEAKER_01And that it can still reach everything it needs to on the inside. Like a human, for example, might be like, Oh, I can't reach that. Let me change my own angle, let me move to the other side of the patient, or let me move closer to the head, or whatever it is to get the angle that they want. And so this is one of the interesting things I think they found with the humanoid, which was like, hey, it was able to reach some stuff in the regions that we wanted it to, but then it in other in other areas it wasn't able to. So it couldn't reach everything that they that they wanted it to reach or essentially.
SPEAKER_02Yes, yes, exactly. And you know, uh they had to also constrain the the pivot point that you know, because they in the in the old Da Vinci systems, they had like a physical pivot attachment. For these systems, they don't. So they had to do that in software. So they had to figure out how to make it so that the robot
Laparoscopy And The Fixed Pivot Point
SPEAKER_02was able to manipulate and move the tool without uh ever changing that fixed point in space. Um and there's a lot of geometry stuff that comes into play there, and it's really cool. And they were able to figure that out first. And then once they were able to figure that out, the next thing was to evaluate the quality of how the system performed. And they had and the way they evaluated is to see can it draw straight lines inside, right? Can it make straight lines uh while sticking to the pivot? Can it make circles inside? Because these are the types of motions that are needed inside to be able to manipulate. And it was able to do straight lines like pretty well pretty well, yeah, yeah. Really well. Uh, but circles it had more trouble with, and it was a little bit uh, you know, kind of had a few errors in it, and it wasn't it wasn't great. And that just like is sort of setting up for the next stage of research, I'm guessing. And they also measured um not just the accuracy, but the latency that is the uh that the time it would take to do these things. And humanoids tend to be a little slow. That that's just the case. They are they are a little bit on the slower side, right?
SPEAKER_01But if you are, if I am, for example, I don't I don't do this kind of surgery, but if I am like, you know, the one on the other end of this, kind of directing the robot what to do, and there's I think they measured 150 sec 156 second millisecond latency. Sorry.
SPEAKER_02That's right.
SPEAKER_01Yeah, that means that the time that it takes from for me to the robot or the robot to me, that like that's there's a lag there. I I think I would get I think I would get really annoyed by that lag because it's not it's not real time. You're not getting real-time feedback, you're not seeing what the robot is doing in real time. And so I think that latency is something that would definitely need to be solved for from a user standpoint.
SPEAKER_02Yes, exactly. Exactly. So I think if you were to do what they said was like with manual laparoscopic laparoscopic surgery, uh humans were generally the worst in terms of accuracy, and the Da Vinci robots were the best, these are the non-humanoids with the best in accuracy, and the and and our humanoid robot was kind of in between, right? Uh it had some errors with the circle drawing piece of it, at least. Um, and then we also had speed, and and again, the humans, sorry, sorry, the Da Vinci robots were the fastest, uh, and the humans were kind of in the middle, and this one was the humanoids were a little bit on the slower side. Um, and again, the point is if it was only about accuracy, then the Da Vinci robot wins. If it was only about speed, the Da Vinci robot wins. But it's about the fact that practic practically speaking, it's really hard to get this everywhere.
SPEAKER_01Well, and I would say like the Da Vinci has been around for a while.
SPEAKER_02Yeah, it's true.
SPEAKER_01And they've probably made a bunch of different iterations on it. And so no wonder that it's that's not I'm not surprised that it does better at some of these things because they've probably been working on this for, I don't know, almost yeah, 25, 30 years.
SPEAKER_02Yeah, yeah, yeah. That's a good point. I mean, that's still certainly a good point. And humanoids has been in and out, it's been really challenging to work on humanoid robots. I've myself worked on them. Um, there's a lot of challenging aspects of just like movement and motion and things like that, in addition to the teleoperation and the dexterity and all of that. Um, and you know, hopefully doesn't trip on things, you know, like there's all kinds of other things that have to be have to come into play. Um, but if they can, if they're getting so close on its performance, um, we're already, you know, a long, you know, we're already a big step there. And these robots, uh humanoid robots are getting cheaper. I mean, some of these robots they're using is in the range of $15,000 to $16,000, which is which seems like a large number, but when you compare it to like, you know, the Da Vinci system, which is multi-million dollars, it it it becomes a lot more affordable to have um in your system, in your uh in your in your OR. Um, so I I think there's there this is just kind of setting up setting us setting us up for some very interesting future work, I'm sure. And
Accuracy Speed And Latency Tradeoffs
SPEAKER_02um, and there are some challenges, right? With these robots.
SPEAKER_01Yeah. And so they took it to the next level, right? In this article, they said, okay, we're not just doing like bench top research. We're gonna try this in a dry lab and then we're gonna try this in pigs and see what happens. And so the dry lab was like kind of normal surgical stuff that you that you might have um people do just as like practice, where you say, Okay, like there's uh these O-rings on a peg. I want you to grasp the O-ring and like transfer it to the other peg. Got it. Um, or I want you to grasp one peg and transfer it to the other side of this board. And so it was like regular transfer tasks where you'd have to like, you know, have the dexterity to grab it, pick it up, move it over, and get it on the next location. Yeah. Um, and and so they compared, like, hey, how is the human doing? How is the humanoid doing? And again, how is this, you know, DaVinci system doing? And not surprisingly, they found the results that we would expect based on what we've just been talking about, where, you know, it was slower for sure. Um, people said, hey, it has a less intuitive kind of control interface, and there was also some motion error in it, especially, you know, around this circular type of motion if it was needing to do it.
SPEAKER_02And the less intuitive interface is is also another another product of Da Vinci being around for a while. Yeah. Because they try to set it up with the Da Vinci setup, but it's obviously a little different. And so people are going to experience that kind of user interface and user experience type problems, right?
SPEAKER_01Totally. And so then the next step was like, hey, could we actually do surgery with this thing? Yeah. Um, and they did two gallbladder surgeries on pigs. They were 11 weeks and 16 weeks old. Um, and they had one surgeon who was on the console. So like they were not present in the room, but they were, you know, operating the robot. Okay. To do the operation. The robot was doing the operation, but they were operating the robot. Yeah. And there was a bedside assistant in the room who was a human who was doing things like, oh, you know, I something needs to be pulled back, something needs to be adjusted, something needs to be exposed, a camera needs to be cleaned. That's normal to happen in an operating room. Those are normal things. So that that person was human and was still in the room. And they found that there were, you know, pauses to do those kinds of things, like clean cameras. And then there were actually some longer pauses when the robot needed like recalibration or um physical repositioning because it wasn't able to do it on its own, um, like the the base or the arms relative to how it was, you know, going into the pig's body. Oh, interesting. So laparoscopic trocarcytes. And so those were much longer uh pauses because they would be like three minutes or or longer in the middle of the surgery.
SPEAKER_02That seems long for a pause. For I mean, maybe maybe it's not, because if you're used to the kind of um pauses throughout the surgery, then maybe it's fine.
SPEAKER_01That's that's long. I would say the short pauses are normal in a surgery. That long of a pause, um, I I don't want to say it never happens, but it would be like a bit more unusual. Right. But the surgeon, the surgeon who did it said, hey, like it was able to perform the tasks that we wanted it to. We could dissect, we could control the instruments. Um, you know, I think they did find, yeah, there were some range of motion issues, which is why they had to like stop and move the robot. Um, and that there were actually some some limitations in terms of its strength, which is interesting. Yeah. Um, and then it needed, you know, recalibrating. They actually said it it overheated sometimes. So they had to like deposit.
SPEAKER_00Right.
SPEAKER_01Um, which is again like a limitation, but they were able to actually carry it out. They were able to do these two gallbladder surgeries on these pigs.
SPEAKER_02That's incredible. And you know, some of these limitations
Dry Lab Tasks And What Breaks
SPEAKER_02are uh just like generalized robotic issues, right? That can be resolved across the industry, not just for for this use case. Um, but there are some use case-specific stuff, right? About, I know they talked about sterilization, and that was a big challenge as well was how do you sterilize these things?
SPEAKER_01Well, I think that was like, yeah, maybe future work that could be done because here they made it sterile by putting gloves over the arms of the robot. But if you think about an operating room, this is a place where everything has to be very sterile. Your your field, your field that you're operating in, it gets autoclaved. All the instruments um are very, you know, they're autoclaved. The the person you're doing surgery on is like draped and cleaned, everything is sterile. So I I mean, I'm not a robot expert, but I'm assuming that like when you have like sensors, um, calibration for motion, that kind of stuff. If you I don't I don't know that you can autoclave those things, you might wreck them.
SPEAKER_02I'm not even sure what what what the what is involved in autoclaving. That's interesting in and of itself. But I, you know, as far as these these machines are concerned, in my head, I always think of machines being more sterile than humans, right? We're we're I mean, in some level, it's just a piece of metal or pieces of metal, and in theory, they're they can be they can be sterilized. I mean, I always again going back to these like um science fiction movies, uh, have this notion of having um a a bunch of robots in a in a in a in a in a sterile chamber or in a sterile sterile room because they they don't they don't need to be, they won't pick up bugs and they don't, they don't they don't like you know how to do something like that. Well but they could, right?
SPEAKER_01If you're exposed to the universe, you could pick up anything. And like on top of the human could be anything too. Right. So the way that you clean the human and then you you drape them prior to surgery, like as a surgeon themselves, like scrubs, puts on sterile gloves that have never touched anything, puts on a sterile gown and a uh a hat and all of this stuff so that you don't introduce contamination. Um, autoclaving itself is done for like the instruments, and that's when you have like pressurized saturated steam to sterilize things. So I don't know. I'm like, in my mind, I'm like, oh, this combination of like water, water is not good for water and like no heat, right, um, and and robots. I in my mind, I'm like, oh, I don't know if the I don't know how how well that would work. But the whole point is like you want to destroy all the microorganisms so that you can reuse these critical instruments. This is done in your dental office, this is done in medical office, etc. And it's
Pig Gallbladder Surgery And Sterility
SPEAKER_01moist heat under pressure. That's how they do it.
SPEAKER_02Yeah, that's a huge problem, right? For robots. So, but again, it's a that's a materials and manufacturing problem, not necessarily an AI or health um specific to the healthcare sector, right?
SPEAKER_01Oh, for sure. It's not, I mean, you might need it in other locations, but in the healthcare sector is where you need things really clean. Yeah, you can't have contamination or else you're gonna get infections internally. Like, can you imagine taking out someone's gallbladder and like leaving behind a bunch of E. coli or something that then like they get an internal infection from? It's like not good, not good, not good. So I think this is it's important to call out because when if you're trying to scale something like this, it could be a very significant barrier.
SPEAKER_02That's a good, that's a good point. That's a good point, but very exciting work nonetheless.
SPEAKER_01Yeah, I think it's interesting. We haven't done a lot of episodes about robotics and uh specifically about humanoids. And I think it's interesting to dive into the specifics of like, hey, they were able to do this, but these were the limitations. No, I think specifically around like how it can manipulate an instrument, yes, um, or not, right? I think that was really interesting because I hadn't I hadn't necessarily thought of that before as an issue, but when you but like now that I now that I read this paper, I'm like, oh yeah, that seems very obvious that like it it's not gonna work quite the same as a human would. Um, and then you layer on top some of the things like, oh, it's taking too long to send the images back and forth, or um, we don't know how to make this sterile at scale. Uh I think there's a lot of opportunity and potential here, and there's a lot of problems that still need to be solved. Yeah.
SPEAKER_02All right. With that, we will end today. Thank you for joining us.
SPEAKER_01We'll see you next time on Code and Cure.
