Okay, so welcome everybody to the Tuesday session.

As usual, we will briefly sketch the storyline of the winter semester lecture on diagnostic

medical image processing.

And we started out with a very brief introduction.

In former days I made this very more detailed, but this year I made it very short.

We talked about different modalities for medical imaging, and you should keep in mind X-ray

is or the exploration of X-ray was basically the starting point of medical imaging.

And then people continued with ultrasound, we know about nuclear medicine methods, PET

and SPECT, we know about computer tomography, we have learned something about MR.

So we basically know there are hundreds, thousands of different modalities in the hospital.

I'm exaggerating.

And these modalities generate images, and they are required to be used for diagnosis

and therapy.

And in the first section, or first chapter, we looked at different types of pre-processing

methods.

Basically we looked into the physics, how are the images acquired.

Usually principle from physics are borrowed to...

Good morning.

You can sit here.

You look at certain principles from physics and you use this for imaging.

For example, if you...

Okay.

So let's think about principles from physics that are used for imaging.

For instance, you have your X-rays.

Your X-rays have a certain energy, they are propagated through the object.

The X-rays are attenuated according to Bayer's law, which says that we have this exponential

decay dependent on the tissue classes we are propagating through.

And at the end, we measure basically the leftovers and encode these in terms of intensity values.

Or in computer tomography, we take these different line integrals to do a reconstruction and

get the CT images.

Or in MR imaging, we use the magnetic properties of certain materials and the spin, you know,

and then the flips of the spin to visualize the human body.

Modern technologies, for instance, in phase contrast CT, we also make use of the phase

shift of the X-ray beams or of the X-ray waves that are propagated through material, then

the phase is shifted, and this can also be used for reconstruction.

Something we did not talk about here, but this is something that we consider in our

lab in terms of research and investigation.

So we use these principles, and we know that due to the sensors we are using, due to manufacturing

inaccuracies, we get images that are not perfect.

And in general, we can remember the rule of thumb saying as long as we have symbols, everything

is clean and nice.

And as soon as we use sensors and use measurements for processing, then we have to deal with

noise and corrupted signals.

And you know that pattern recognition is basically about the processing of signals, so we are

just working with dirty data in our lab.

So preprocessing is an issue, and we have considered a few preprocessing procedures.

For instance, we looked into X-ray imaging, and we looked into a quite old and well-established

technology to acquire digital images using an image intensifier, and we have seen that

this is like an old TV set with this vacuum tube, and there are electrons accelerated.

So we have an electron amplification device, and these electrons are moving in the earth's