It's easy to simulate the robotic manipulators - the dynamics are well known and they are rigid objects.

On the other hand, simulating say skin/muscle/blood or anything that closely resembles human body is near impossible. Without that, a simulator is pretty much useless and it'll probably easier to train the surgeons on real robot + some animal like pig

Not impossible, there are some good new algorithms for soft tissue deformation that aren't CPU killers like even a few years ago. Most of the surgical sim software out now is either so basic that it isn't good for much beyond cursory anatomical review, or it requires some major hardware installed in a permanent space with limited use-cases that nobody ever bothers to use because at the end of a resident shift they just want to go home and sleep.

I'm working for a small company that is trying to bridge the gap and make something good that can run on the consumer VR/AR hardware that's coming out in the next year or three. Lots of interesting problems to solve.

It's a solved problem

https://www.intuitive.com/en-us/products-and-services/da-vin...

Do you believe the marketing of this company?

In my experience in a related field, simulators are about 2% as useful as advertised. The cousin thread explains how difficult building a flesh simulator is, so I wouldn't expect surgical experience on a simulator to be very useful.

I've used it, it's pretty good.

I’m not even thinking of anything as grandiose as a full simulation of the human, it could be as simple as mimicking the movements from a previous surgery using an overlay of the simulated robot arms.

Is a physical object even required? Why can't it be like a flight sim and just be controls and some screens?

What do you mean? In a flight simulator you have your controls controlling a plane (right body, dynamics are well known) in the sky (simple environment, dynamics are reasonably well known, even with different atmospheric conditions like wind). The output is visualized on the screens as a rendering of the scene.

In the case of robotic surgery simulator, you are using the controls to control the arm, but to interact with what? If you just want to move the arm around and maybe interact with some rigid objects sure that's easy. Would that add any training value to the surgeon? Probably not. You can get value only when the simulator includes a simulation of something the surgeon would have to face eventually - organic mass of the human body. Simulating that is hard, and I doubt anyone would invest much into it when you can train surgeons on alternative physical objects like pigs.

it would control a software arm operating on a software body, just like a flight sim controls a software plane.

It is nearly impossible to accurately simulate a “software body“.

Why? Couldn’t one model the various tissues and bones, fluids, surface tensions, resistance to cutting etc?

Yes, you can produce a model incorporating nonlinear elasticity, viscosity, plasticity, frictional contact, subfailure damage, and fracture / cutting. Repeat for each tissue involved. Some of those methods are well-developed, others are not. Damage and failure in particular is poorly developed, and simulating deformable body contacts can be tricky. Then you have to estimate parameters for all of that, validate that the model produces realistic outcomes, and get it to run in real time. You can ignore most of the complexity and make approximations, like treating everything as an isotropic linear elastic material, setting cutting resistance to a fixed value per tissue, treating the scalpel like a lightsaber, and accepting that contact will be imperfectly enforced (surfaces may interpenetrate). A crude simulation can still be good enough to be useful as a learning tool though, but it tends to work more like a video game than a true physics-based simulation. We'll still eventually get the kind of accurate simulation you're asking for, probably.

I was specifically constraining the discussion to robotic surgery. Do all these variables apply in that case?

Not sure how to interpret the question. Using a robot vs. a person doesn't change anything about the physics of how tissue responds to perturbation. You can of course simplify the simulator and correspondingly limit its use to specific learning objectives, probably novice-level. To return to the thread's starting point—we understand flight well enough to make general-purpose flight simulators, we don't understand tissue well enough to make general-purpose surgical simulators. You can still make simple interactive training tools to help someone rehearse the motions of a procedure, but I wouldn't call such a tool a simulator.

Very very many variables... For example, burnt tissue, crushed tissue, injection trauma, infected tissue plus variable anatomy.

But don't we now have thousands of hours of real-use? Surely in the age of Dall-e, a system to use that as input to ape if not fully simulate the surgical environment is within reach.

While DALL-E might seem like something out of the world, what it outputs at the end of the day is pretty basic: a 2D image.

To run a high-fidelity simulator of human body that is useful for surgeons, you need a LOT more. And I doubt it can be data-driven. Data-driven simulators for things like autonomous cars are just coming up, and these are way simple as the agent (the autonomous car) doesn't directly change the environment when it's driving around (as in, you don't have to simulate the car colliding into a traffic pole and then breaking it, etc.)

To simulate a human body, you need to be able to capture the material properties of different layers, some of which are fluids, and then also the interaction between them and how different organs react to a surgical operation. It's a very hard problem.

Why do all that when you have animal corpses readily available? There's no need to create a problem that doesn't exist.

Content note: surgery gore

a) thousands of hours is far far far too little data. This is a total ballpark estimate, with only knowledge of the ML and basically zero knowledge of the surgery side, but I would expect three to six orders of magnitude greater, i.e. millions to billions of hours necessary to train a machine learning model to do something like that, to a standard.

b) the big problem is not the procedures themselves, but edge cases and things going a little bit wrong.

c) you are fine with tiny flukes in your generated images, but you won't be very happy with an ML model whose tiny flukes lead to internal bleeding.

d) unfortunately (with contemporary models) we can't easily explore latent space, which might possibly contain regions corresponding to catastrophic robot arm movements. the chance isn't that high, but it's not a chance most'd be willing to take at this point.

I think this overestimates what Dall-E is actually doing, and underestimates the complexity of the task.

It's because the interface console itself is a big part of the cost and complexity of the thing. The simulator requires the console and input manipulators to use, so they can't train on it while it's being used for surgury. Hospitals could buy a second console to use exclusively with the simulator but that's a multi hundred thousand dollar piece of equipment.

Surgeons can take a few practice simulations in the Sim, but it's just that doing so ties up the robot from doing real cases at that time also.

"multi hundred thousand dollar piece of equipment"

So, pretty cheap, it sounds like? How much is the hourly rate of surgeons?

Not really.

A surgeon makes on average $115 an hour in California, $280k a year. So the console plus simulator backpack costs a surgeon or two's yearly salary.

https://www.ziprecruiter.com/Salaries/Surgeon-Salary--in-Cal....

Only if they're spending 2500hours a year in the OR.

It'll be interesting to see the growth of simulation in other fields similar to that used in radiation oncology [0]. The trick about human bodies is that they aren't all the same and they don't say the same. Once the robot can perform sufficiently similarly in simulation as in real life, then a high risk and cost intervention like surgery will be simulated before performance.

An additional aspect of simulation is calibration and use of phantoms [1]. These are materials of known characteristics approximating human anatomical densities. I suppose for robotic surgery this would be used for both the imaging and the surgical tech.

0. https://radiationoncology.weillcornell.org/clinical-services...

1. https://www.nist.gov/physics/what-are-imaging-phantoms

The thing about simulation in radiation oncology is its more focused on simulation of the physics of the beam delivery, and not so much on simulation of the tissue or human body portion of the equation. This is more the latter.

When I saw the keywords in these posts it reminded me of a job I interviewed for. This was in 1991 and the state of the art was to plot iso-dose contours on top of X-ray images, presumably using facts about the beam and the tissues at hand. The plan was to evolve from there to 3D images, possibly rendering these planing data in arbitrary orientations in real time. The link provided above discusses tattooing the patient for future reference and other things, using a machine that was 'similar' to the beam machine but without the beam itself. So it is good to see that planning has moved on in 30 years.

Yeah, what you describe is common now on systems like cyberknife. etc.

I don't know for sure, but I suspect cadavers are cheap enough and much "higher fidelity" than any sim could ever be.

Yes and no, for example, they have no blood circulation anymore, thus one of the potentially fatal risks of any surgery - bleeding - can't be "trained" there.

It can, they can plumb things up with pumps to simulate whatever.

This same limitation applies to simulators though, so this application isn't an answer to the parent comments question.

I would think pumping blood through a cadaver is a quick way to get a visit from a handful of agencies.

There are lots of hoops to jump through to get to use cadavers for one's work, but the act of pumping blood itself isn't particularly of concern to any agencies.

Here is one example paper: https://pubmed.ncbi.nlm.nih.gov/29023350/

What makes you think they're not readily accessible? They are. Part of the issue is time. If you're a surgeon, one individual surgery is but a small part of your day. There really isn't any spare time to "prep" for a surgery other than reading the specific of that individual case.

The submitted article also discusses how a limiting factor is time rather than availability of simulators.

The author wrote: "The paper I published in 2019 summarized my findings, which were dismaying. The small subset of trainees who succeeded in learning the skills of robotic surgery did so for one of three reasons: They specialized in robotics at the expense of everything else, they spent any spare minutes doing simulator programs and watching YouTube videos, or they ended up in situations where they performed surgeries with little supervision, struggling with procedures that were at the edge of their capabilities. I call all these practices “shadow learning,” as they all bucked the norms of medical education to some extent. I’ll explain each tactic in more detail.

"Residents who engaged in “premature specialization” would begin, often in medical school and sometimes earlier, to give short shrift to other subjects or their personal lives so they could get robotics experience. Often, they sought out research projects or found mentors who would give them access. Losing out on generalist education about medicine or surgery may have repercussions for trainees. Most obviously, there are situations where surgeons must turn off the robots and open up the patient for a hands-on approach. [...] My data strongly suggest that residents who prematurely specialize in robotics will not be adequately prepared to handle such situations."

The author also listed examples of accessible simulators, notably one that uses virtual reality: "In the past five years, there has been an explosion of apps and programs that enable digital rehearsal for surgical training (including both robotic techniques and others). Some, like Level EX and Orthobullets, offer quick games to learn anatomy or basic surgical moves. Others take an immersive approach, leveraging recent developments in virtual reality like the Oculus headset. One such VR system is Osso VR, which offers a curriculum of clinically accurate procedures that a trainee can practice in any location with a headset and Wi-Fi."