Well, welcome everybody.

It's a pleasure to see so many familiar faces in the audience.

I would like to start my talk by giving a rough agenda and also for people who are not

entirely familiar with my work, a little bit of context.

So first of all, the beginning of my talk will be about prerequisites to understanding

my actual contributions.

Those will be the epipolar geometry and the law of x-ray attenuation.

I'm presenting a medical C-ARM scanner on the right.

Just for those who are not familiar with those devices, there is an x-ray source on the bottom

emitting x-rays and a detector on the top.

Typically the patient would be placed between those with x-rays emitted at the source and

being detected at the detector.

So followed by these basics, I will give a little bit of theory and implementation of

the epipolar consistency conditions.

And specifically I'll take a look at a cost function to measure consistency and redundant

information in images.

In the second half of my talk, I will address three exemplary applications that all relate

to motion estimation in x-ray images.

Specifically I'll be tracking an object, an unknown object under fluoroscopy.

I'll estimate gating in rotational angiography of the heart.

And I will also present briefly a 3D-3D registration algorithm that does not use the reconstructed

images but rather works on the raw data of the acquisition.

So to start, I'll briefly define what data consistency conditions are.

So first of all, two projection images in a CT, for example, computer tomography scan,

are not independent but they fulfill certain consistency conditions.

That's because there is redundant information in the two of them and it's sort of like a

checksum.

So if there is an error in the acquisition or some assumptions are not met, then this

redundant information will not be exactly identical in the two images.

Redundancy is a real issue in analytic reconstruction algorithms where proper weighting of that redundant

information has to take place.

And this is also where some of the very important theoretical basics for the work that I'm presenting

today have been rooted.

We will be using consistency conditions for artifact reduction or tracking in motion estimation

in x-ray images.

Specifically redundancy can also be exploited to correct for motion, for example, but also

other measurement errors.

So that's the same idea as the checksum.

If you have redundant information, you can verify that it is the same.

If it is not, you can also sort of adjust some of the parameters of the acquisition

so to make the redundant information as similar as possible.

And this is also a work by Cassine Debler, which is probably the most important reference

in this talk because it's where the idea for the epipolar consistency conditions came about.

When I saw this work, they did not understand it is an epipolar geometry setup.

However, using their work and then adding epipolar geometry and all the mathematics

from computer vision in resulted in the work that I'm presenting today.

So just briefly what publications I'll be covering, there is mostly theoretical publication

concerning epipolar consistency and transmission imaging that was published in IEEE Transactions

Medical Imaging.

And then there is some implementation related work, so very practical considerations on