All right, so let's get started with our treatment of parallel imaging.

And I just want to show you this as a motivation in the beginning.

This is the fundamental constraints that we are facing when we are doing MR imaging.

And in a way, you have heard this already in various bits and pieces throughout the

lecture.

This is just one slide that really summarizes it all.

And by the way, this is not only true for MR imaging.

This is pretty much the case every time you do a measurement of a physical process under

some conditions, but in many cases.

So we have this fundamental greater that we have between resolution, signal to noise ratio

and the scan time.

We can always maximize two, but never three.

Meaning we can make the scan time short and we can make the SNR really, really high if

we use a very low resolution.

If we are fine with having very low SNR, we can really push the resolution very, very

high and make the scan time very low.

If we want a very high SNR and a very high resolution, what we can do is can with multiple

averages.

Remember that the SNR is proportional simply to the amount of samples we take.

If you take multiple averages of the same image over and over and over and over again

and average them all together because we have Gaussian noise, the noise will average out

with square root of the number of averages.

But the problem that you then face is if your object moves a little bit during the scans,

you run into issues.

So in practice, you cannot average indefinitely.

You can't do it if you measure dead objects.

Like what people are doing is measure brains of dead people that have been extracted, put

them in some preservation liquid, measure for two weeks nonstop and they get super high

SNR.

That's not really something that is very practical.

But this is the situation we are dealing with.

We always would like to have all of these grades, but if, and that's the thing here,

if we find a way to improve any of these without compromising the other, then we have really,

really made a net gain.

If I am able to improve the SNR without effectively being scanned by the resolution, then I really

have a net benefit.

This is the example, for example, when you go to high-definition.

You can see either the scan time or SNR or the resolution as in a way as a currency that

you can spend to improve these three things.

So I'm for the rest of this lecture, mostly going to talk about reducing the scan time,

but always keep in mind, reducing the scan time is great, but it doesn't just mean that

you can measure faster.

You can also use it to measure in the same scan time if you're fine with it, but with

higher resolution or probably with higher SNR.

And that brings me to a concept of case-based understanding and why we would want to do

it.

So if you want to reduce the scan time, there are a couple of ways how you can do it.

We have seen one already when we talked about partial Fourier imaging, but just first of

all, think about it from a very, very basic point of view.

It is simply just the time that it takes us to traverse case-based scanning.