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

OK, so welcome to the Tuesday lecture in the Interventional Medical Imaging Processing.

This is the second guest lecture in two days.

My name is Sebastian sindsger hypoc Geist, PhD student at the Pattern Recognition Lab.

And today we're going to talk about feature detectors, feature descriptors and its applications

in medical image processing.

So first of all, an overview about the next 45 minutes.

So I'm going to motivate you.

So why do we need features?

What is important for features?

And then we have a look at those detectors and descriptors.

First of all, some basic aspects.

So what do we require for a feature to be good?

What kind of invariances do we want to address?

And then I give you the big picture of the feature matching

pipeline.

So we have a look what is to be done in order to match points.

Then, and this will be the main part of this lecture,

we have a look at feature detectors and feature

descriptors.

I will give you some examples here.

And finally, we learn how we can match those descriptors.

At the end of the talk, I show some applications

in medical image processing, a short insight

in research at our lab.

So let us start with a motivation

from the field of stereo vision.

And stereo vision, here an application in radiation therapy.

So if a patient has cancer, he first gets into the hospital

and you do a CT scan.

So you do a CT scan of the patient.

You find the tumor.

A physician is labeling the tumor contours.

And based on the contours, you are generating a treatment plan.

So you have an initial session where you scan the patient.

And then you have several treatment sessions

where the patient is coming to the hospital

and has to be aligned very accurately with respect

to this planning city.

So you have to do a registration of those surfaces

in order to get an accurate treatment.

And this is where stereo vision comes into play.

So here is an example of a commercial system

from a British company.

They use stereo vision, so they acquire two images

from different viewpoints.

They generate a 3D surface from that.

And then they can use the surface

to match it to the city surface that is here showing green.

So you can align the surfaces.