The following content has been provided by the University of Erlangen-Nürnberg.

So welcome to the Tuesday morning session. As usual, for the 90 minutes we will see the storyline of the lecture.

And I do this to make clear that there is one.

So we talk about medical image processing, and in particular we are looking for the diagnostic part.

That means we study procedures that can be done in systems that are applied for diagnostic purposes.

And we have seen several modalities, and I'm not telling you something new here.

Typical modalities are X-ray systems, CT systems, MR systems, ultrasound systems, and many more.

And now we have three major chapters that basically provide the columns of the lecture where the lecture builds on.

And this is pre-processing.

So what can we do to single images to enhance the quality?

How can we use the physics that is used during the acquisition and the implied artifacts?

How can we eliminate those?

And then we will talk about reconstruction. That means I have multiple projection images.

How can I generate out of different views of a patient?

How can I generate three-dimensional information out of that?

And the last part of the lecture deals with image fusion or registration.

And this is the problem. How can I map images of multiple modalities or images acquired by multiple modalities into a joint coordinate system?

And in the context of pre-processing, we currently talk about X-ray acquisition, and we spend way too much time for it because it's already mid-November.

We talked about image intensifiers and the implications.

We have seen that the usage of image intensifiers leads to images that are distorted due to several reasons.

One problem is, for instance, the earth magnetic field that implies a distortion of the image.

And this distortion is due to the fact that image intensifiers use vacuum tubes.

In the vacuum tube, there is an electron beam, and the electrons are accelerated and, of course, deviated by magnetic fields.

And currently we talk about FDs, flat panel detectors.

The short term is FD, flat panel detector.

And we have seen that one common problem in FD technology is that we have pixel defects.

So we have a failure of pixels, and the question is how can we interpolate the image information for those pixels that are providing no intensity information?

And in between, we do some math, right?

And you might think this is...

Okay, here I should...

Oh!

Let's store it first.

Let's stop it.

And let's restart.

We learn so much cool stuff here.

So let's try whether I can change the color.

Which color do you want, Alex?

Blue.

Okay, very interesting from a psychological point of view, by the way.

In between, we do some math.

Now you're thinking, right?

I just said it out of the blue.

I have no clue what that means.

I just said it's maybe important or interesting.

Good.

So we did some math.

What did we do at the beginning?

We had a very short refresher course on matrix calculus linear algebra, right?

And in this context, we talked about the SVD.

And whenever you see a matrix, you should remember, okay, the columns, they span the image space.