So now we are listening to Denisa Martonova, which is a postdoctoral researcher at the Department

of Mechanics of the University.

The stage is yours.

Thank you.

Okay, so hello.

I will bring you again a little bit back from maths and bring you closer to mechanics.

So we have seen a nice overview or presentation of how we can use AI in mechanics.

I will be even more specific and I will show you how we can use AI in biomechanics and

specifically for cardiac tissue.

So this is also a collaborative work of many people from many institutes and universities.

And yeah, I will start with a small motivation of AI in biomechanics.

Then I will just give you overview of cardiac model itself.

I will show you how we can model or how we can obtain passive material models.

Then I will introduce orthotropic neural networks, aerophysics in form neural networks.

Then I will discuss how we can use regression in order to get a little bit more generalizable

models.

I will compare with other models in the literature and lastly I will discuss fiber dispersion,

which you will see what it means.

Okay so let's have a look to this overview.

You can see different biological tissues you can find in your body and they are ordered

by the stiffness.

So in general the stiffest tissues are the easiest to model, whereas the softest tissue

are quite hard to model and it's very hard to find the proper material model.

As you have seen before, we engineers are always looking for some material models which

we can then feed into our equations.

So my colleagues have discovered or have found with AI different models for almost all of

the tissues, so for example for brain, then for arteries or skin or even as you also have

seen before for artificial meats like tofu or tofu key or whatever.

In my talk I will focus on cardiac tissue which is quite complex, so it's orthotropic

tissue and I will show you how we could discover the best model for this.

Why we are doing this?

It's quite important because cardiovascular diseases are the leading cause of the dead

worldwide.

So it's very good to have computational models which can help us to predict progression

diseases rate, to help by the therapy planning, to help by investigation how to design some

drugs or how to design some devices, for example in this specific case for heart pumps or other

devices helping disease hearts.

The latest computational model used AI and as you can see here it started about 2020

whereas the first models goes back to 50s and they are very simple model and it developed

a lot such that now we have quite advanced full heart models.

So you can see here a simulation of my model I have used in my PhD and this is rather simple

model it has just two chambers like it's called biventricular models and even this model has

a lot of ingredients.

So as you can see here first of all we need some clinical inputs in order of data like

these are measurement data and then with these we can build up our computational model.

Heart is a very specific tissue and we need a lot of ingredients again so we need a model

for passive mechanics but we also need some interaction with electrical parts so we need

a model for active mechanics, electrical excitation because in heart you have like electrical

signals which trigger the contraction then we need a model for hemodynamics and of course

a geometry this is obtained usually from MRI data and they include also fiber orientation.