- [Steve] So hold on.
This is my heart, is it? - [Steve] Exactly. - I am on my way to get
an MRI scan of my heart. Not because there's
something wrong with it, but because I know someone
who can do something really interesting with the data. This video is paid for
by Dassault Systems. So after half an hour
in a very loud magnet, I've got one of these. I don't know what I was expecting
the results to look like, but really, it boils down to a series of black and white images, some white pixels, some black pixels, but mostly some shade of gray in between. Here, for example, is a slice roughly through the middle of my heart. You can see the left ventricle there, and the right ventricle, but actually a series of images were taken at different positions. So we can sweep through my
heart, look past the valves, and into the atriums, or
is it atria? Who knows? Scans were also taken at time intervals, so you can see my heart pumping. But what if you wanted to
take those grayscale 2D pixels and turn them into a 3D model? Well, you need to do something
called edge detection, and it's something we, as
humans, can do quite easily. Look, here's the edge of the heart. You can tell it's the edge of the heart, because, well, you're doing
some object recognition there, but also there's a big contrast between these pixels and these pixels. You can teach a computer
to do that, as well. This video isn't about that. Instead I'll just show you the result. Here, thanks to Dassault Systems,
is a 3D model of my heart. Wait, what is that? Some sort of lump? Oh gosh, I'm going to
worry about this lump now. Should I show it to a doctor? I don't know. If only I
knew a good heart doctor. Oh, of course. Rohin. I can ask him. (phone rings) - Hey Steve, how's it going? - [Steve] Hey, Rohin. Yeah, I've just got a question. Do you know any to get heart doctors? What if we wanted to animate the heart? Well, we could use some animation software to move bits of it around. But what if we wanted to simulate the way the heart behaves under
different conditions? Like, what if we want to
simulate how my heart behaves when I'm 80, with the addition
of a pacemaker, for example? Well, to do that, we
need to map out not just the position of all that muscle tissue, but other properties that go with it, like stiffness, electrical conductivity. With all that information
applied to this 3D model, perhaps we could send a
virtual signal to the heart and see if it will pump. - Sounds like a good idea, but how do you go about building a heart? There's tons of knowledge out there. Instead of all of us working in isolation, I pitched this idea to the
experts in each domain, that if they would just
share the piece of the puzzle that they understood, that I would take the
responsibility to put it together. And within one year, we actually had a fully working human heart. - [Steve] That was Steve
Levine from Dassault systems. How do you bring these disparate
professionals together? The people who understand the
electrical side of the heart, people who understand the
mechanical side of the heart, people who understand the fluid dynamics of the blood in the heart. Like, how'd you get them
together on a technical level? Well, there's one method that
can pull information together from these different
disparate disciplines. It's called the finite element method. You may have heard of the
finite element method before. It's that thing that
tells you whether a bridge is safe or not, and finite element analysis is that. But it has the potential
to be so much more. Let's take a look at that
bridge example, actually. Like if I wanted to figure
out what's the maximum weight I can apply to this
bridge before it breaks. Well, if the bridge was simple, like it was made of two
rods bolted at each bank and joined together in the middle, I could apply a weight in the middle. And if we assume that the
rods are somewhat elastic, they will lengthen until
the restoring force matches the weight of
the mass that we applied. So then, it's a simple case
of just balancing the forces, and doing a bit of
trigonometry to work out what's the stress in those rods. But this bridge isn't made of rods. It's made of stones and cement. It doesn't have these discrete parts that we can easily analyze. The opposite of discrete is continuous. It's a continuous medium. The forces at every
single point in the bridge need to be analyzed, and there are infinitely many points in a continuous medium,
an impossible calculation. What do we do? The first step is called discretization. You have a continuous medium, and you're going to create
a model that is discrete. It has a finite number of parts. So, great, now you can
just solve the problem at each one of those nodes. Except it's not that simple, because each node influences
the neighboring nodes, and those nodes influence
their neighbors as well. You have to solve these
problems simultaneously. you might know from your past experience with simultaneous equations, that if you have, for
example, two unknowns, you need at least two equations. Or if you have three unknowns, you need three simultaneous equations. But if you have a thousand
nodes in your mesh, and each node has three
degrees of freedom, then you have 3000 unknowns, and you need 3000 simultaneous equations, each one having 3000 terms in it. Actually you can simplify
things quite a bit, because a lot of the
terms ended up being zero. But the point is it's
still a huge undertaking. There are many different
approaches to solving that problem. But one option is simply to take a guess at how much and in what direction did each one of those
nodes in your mesh move. You then plug that guess into
the simultaneous equations and see if they were right. If they weren't, then you
tweak them a little bit, and see if it got you any closer. In other words, you iterate
towards the right solution. That's all very well for a bridge, but how do we make this work for a heart? Well, there's a few things you have to do. The first thing is you add data to each one of those nodes in your mesh. So, as well as the stiffness of your heart at each one of those points, you're also interested in
electrical conductivity. And how much does the muscle
distort, and in what directions when you apply a voltage. With all these additional properties, you can now not only apply
a force to the heart, like we did with the bridge, but you can also apply a
voltage to see what happens. And so look; here is Dassault Systems' simulation of my heart, when a repeated electrical
signal is applied. How cool is that? By the way, the calculations
become much harder, simply by adding a time element to it. Like when we were analyzing the bridge, the bridge was in equilibrium, but now we've got this dynamic model where you need to consider acceleration and damping forces and things like that. In fact, a complete heartbeat
might take eight hours of computation time, though they are looking at
artificial intelligence, trained on a massive
library of simulated hearts to see if they can get
important results more quickly. It's all good fun to see
a simulation of my heart, but these models are actually
being used with real patients. For example, patients
that have a blockage. - And so if somebody has a blockage, we can introduce that blockage, and see how that actually
affects their tissue. A doctor can say, "Okay, well that's where
our limitations are. Here's how I should fix it." Surgeons working with children usually are seeing unique cases. Once they open the child,
they see what's happening. They have maybe 10 minutes to decide, which will affect the child's entire life. It's staggering when you think of it, the pressure they feel, and you can see that when
they look in your eyes and say, "Anything you can do." I can't sleep the night
before that surgery, knowing what will happen. You know, these surgeons are brilliant at visualizing what's happening. If you enable them to
do the actual surgery on the virtual patient first, when they get in there,
"I've seen this before. I know what to expect. I know what to do," and they can execute. And so we're actually doing
this and saving lives. - [Steve] This model is also useful when you're investigating medical devices like stents or artificial valves. Like, imagine you're developing a new type of artificial
valve and you want to bring it to market. To do that, you have to
test it on human hearts, and that's a long and dangerous process. But imagine if you could first test it on hundreds or thousands
of virtual hearts. You recreate your device in a computer, you add it to the model and you apply finite element analysis
to the whole thing, to see how it behaves. Eventually, you have
to test on human hearts as validation of the virtual research, but by that point, it's
safer and it will be quicker. I also asked Steve about
the way doctors interact with these models. - We actually have been working
with a holographic tablet, and you can actually just
interact with the holograph floating right in front of you. The doctors can figure out
what the right surgery is. They can then show it to the entire staff. The surgeon normally has it in their head. And then they can show the family. They say, "See, this is where the hole is in your son's heart, and
we're going to put a patch, and here's how it's going to fix." A company take the output
of our simulations, and they use that as the input
into their printing process, and they change the material
properties of the print, so that its deflection
matches what we predict. So you can print ... It actually feels and moves
just like the real object. So if you have, for example,
a calcification in an artery, it'll be hard in that area
and then softer in other areas and you can actually feel that. - And that would be really
useful for a surgeon. You can even use the model
to explore drug interactions, for example, beta blockers. They block electrical signals. So all you have to do is
change those variables related to electrical signals
at the nodes in your model and see how the heartbeat changes. - [Steve] We can tell you what will happen to the blood flow in that heart based on what that drug is doing inside the cell. - [Steve] Something
interesting happens when loads of practitioners and researchers are working with the same model. - So we now have over 150
hospitals, research organizations, universities in 24 countries,
all using the same model. So someone's introducing heart disease. Someone is putting in a valve problem, and they're all sharing the results so that the heart model
just gets better and better and better. We've been so successful, now, all the other communities
are starting to nucleate. So we now have a living brain community. We have a living lung community. I've come to realize that,
with the proper motivation, people really do want to come together to solve society's problems. Give them the vehicle to work together, they'll rise to the occasion. - Thank you to Dassault Systems for all their help on this video. The YouTube algorithm thinks
you'll enjoy this video next. That's interesting, isn't it? Thanks also to my
patrons on Patreon there. That's it for this video. Don't forget to subscribe
if you're not already, and I'll see you next time.
