Welcome to the Tuesday session.

Okay, we are in a very exciting chapter on the modeling of projection geometries.

And this is something that will be heavily required once we start to discuss how computer

tomography systems work.

But before we dig into the problem that we started out to discuss yesterday, the projection

matrices, the 3x4 matrices and the computation of the 3x4 matrices, I'm going to summarize

where we are currently and I want to support your procedure filling out the evaluation

form.

There is one question asking for the storyline of the lecture and you have to say yes, there

is one and it's excellent because it's excellent.

We have a nice track of interesting topics.

We talked about modalities.

If I ask you which modalities do you know, James will answer MR because he sent me an

email yesterday.

What else do you know?

X-ray.

CT.

CT.

Spect.

Endoscopy.

Filtrosound.

Pet.

Pet.

Pet.

Pet.

Pet.

Pet.

Visual inspection.

Optical coherence tomography.

Say something.

No gestures.

Microscopy.

What else?

Homework.

So we know a lot of different modalities and they all have their advantages.

The images look completely different.

The images have different excitement levels.

For instance, MR has excitement level 30, I've learned.

Which one do you like the most?

MR.

Okay, but there you have one serious problem.

If you acquire MR images, you have inhomogeneous magnetic fields and this implies artifacts

and which types of artifacts do we observe in MR images.

That's a very serious question and we considered this in the second chapter of our nice lecture

where we discussed preprocessing methods and the term preprocessing means everything that

we do in between the sensor and the monitor that's summarized as preprocessing.

And how do we eliminate the inhomogeneities that are implied by the heterogeneous magnetic

field?

For instance, what is the KL divergence telling us?

It measures similarity of what?