So, good morning, Tuesday session. We had two lectures on math theory basically where

we have discussed the singular value decomposition, a mathematical tool that is quite powerful

and that we will heavily use throughout the whole lecture. So, do not be frustrated by

the theory, this is something that we are going to use over and over now. And today

we will now consider more concrete medical image processing problems and

especially we will talk about acquisition specific pre-processing.

And what does that mean? Basically you want to consider different imaging

modalities like x-rays, CTs, MR scanners and we will think about what are the

physical principles that are underlying these technologies and what are the

implications on image quality. And many of the artifacts we have can be reduced by

understanding the acquisition procedure and the drawbacks of the

acquisition device. And we will see soon what I exactly mean by that. First of all

let's do a few definitions. While I was browsing through the slides

yesterday I had the feeling it's a little bit overloaded these slides. I

don't like them anymore but anyways a few definitions will give you good

intuition what's actually done in pre-processing or what pre-processing

is for. Please do not start to learn these things word by word, that's

not what I expect. So what's meant by image pre-processing? Well image

pre-processing is in simple terms just an image to image transform. We

transform one image to another image and usually the new image should look much

better than the image before, that's image pre-processing. And image

pre-processing is the term that we use if this image to image transform is

done from the acquisition device to the monitor. So if the doctor in simple terms

if the doctor presses the foot switch and sees an image on the monitor,

in between the acquisition sensor and the acquisition device and the display

on the monitor pre-processing routines were applied to the images. Now we will

see a few examples later on. And image post-processing that's what's done

afterwards. Once you have acquired the image and you want to do some

segmentation for instance you want to find out important parts, you want to

find out a tumor or you want to emphasize the bones or you want to find

the bones and eliminate the bones or you just want to extract the vessels or

or or. These are all post-processing operations. That means you have the image

on the monitor and then you sit down and you do some post editing of these

images. This is post-processing. I am not sure whether these two terms are

consistently used in the literature. I mean this is a very young field medical

image processing and throughout the lecture we use these two terms

pre-processing from the sensor to the monitor post-processing once the data is

on the disk and you do some post editing you're doing post-processing.

What are obvious needs for image pre-processing? Of course and I

mentioned that last week already the image quality basically decides whether

people like a system or not. If you buy a camera you basically look at the images

and if the images look fine you like the camera right. So image quality decides

basically on the whole overall system quality and the image quality has to

meet the requirements of the physician. That means the radiologist he has to

like your images. If he says I don't like these images they will not buy the

system so you have to work hardly on this. You want to reduce noise. You might

know that in especially in medical image processing we have always to deal with

the trade of dose between dose and noise. What does it mean dose if you apply

x-ray to a patient? You know what x-rays are. These are the x-ray particles

and you know they interact with the DNA and they can destroy the DNA and can