Physics for understanding fMRI artifacts: Part Three

Coffee break! Time for a few tangents

In this post we're going to do a whistle-stop tour of some background concepts that you should have seen before. None of the information in today's series of videos is essential to understanding what's coming up later, when we get to k-space, the EPI pulse sequence and artifacts, but it's interesting and useful to review. Besides, these videos are well made, entertaining and are available free so we might as well use them! So, if you have the time, go grab a coffee and spend the next hour being reminded of things you probably knew at some point in a dim and distant past. You might even learn something about scanner hardware you didn't know before.


The anatomy of a miniature scanner

Don't worry too much about following every detail in today's first video, which dissects a miniature MRI scanner. It contains the same basic components as your fMRI scanner. Below, I've given a few explanatory notes on the coils and components that are most relevant to us.

The outer coil of this mini-scanner is a prepolarizing coil which, although it is turned on and off over a few seconds, achieves the same result as the permanent superconducting magnet that we use for human MRI; it generates bulk polarization of the hydrogen nuclei, as shown in previous videos.

In the middle of the mini-scanner is an RF coil, a coil that generates and detects an oscillating magnetic field at the Larmor frequency. Recall that on most human MRI scanners the RF transmission is done with a large, body-sized coil whereas the signal is received with a head-sized coil.

In between the RF coil and the prepolarizing coil is a set of gradient coils which impose a linear spatial dependence (a gradient!) on top of the otherwise homogeneous polarizing magnetic field. There is one gradient coil for each of the three orthogonal lab axes, X, Y and Z, thereby allowing images in any plane through the object.

Here's a homework question for you. When this mini-scanner acquires an image it does so almost silently. How can that be? Our fMRI scanners are very loud indeed! (Tip: Think about the relative size of the magnetic fields and the way the spatial magnetic gradients are created by driving a current into the copper gradient coils.) A short answer is here.


A simple NMR experiment 

As before, don't worry about understanding every point Prof Callaghan makes in this next video. Just enjoy...

We will look in more detail at time and frequency domains and the Fourier transform in a later post. What's a free induction decay (FID), you ask? It's just a physicist's term for a signal (an induced voltage) that dies away exponentially with time.


Shimming: It's not just for wobbly restaurant tables

Now we're getting back into issues that will be critical to recognizing and understanding some of the artifacts in fMRI. In this video, Prof Callaghan looks at magnetic field homogeneity. It is the perturbation of magnetic field homogeneity on a local level - around blood vessels - that gives rise to the BOLD effect. As you're watching, think about how the rate of signal decay and its corresponding width in the frequency domain might affect imaging resolution. And, if I tell you that the frequency domain is also the spatial domain of an MRI - we label the image axes with cm but the actual units are strictly Hz - then perhaps you can see that a fast-decaying signal will be broad in the frequency (image) domain, and this will lead to blurred, or lower spatial resolution, images.

How about that wrench experiment? Did you see the way it ruined the NMR signal? Well, that's what you could expect to get if your fMRI subject has metal in or around his head (lots of metal amalgam fillings in his upper teeth, say). But it's also akin to the presence of the skull and venous sinuses adjacent to the frontal and temporal lobes, and which cause drastically reduced signal from these brain regions. We'll investigate these effects more in later posts. Just remember that it's ridiculously easy to destroy NMR signals, either regionally or globally, by introducing certain types of material adjacent to our signal-yielding nuclear spins.


Spin echoes to provide a 'virtual shim'

We don't usually use spin echoes for fMRI at the typical polarizing field strengths (1.5 and 3 T) used these days. The whole point of a spin echo is to remove the effect of magnetic field inhomogeneity at the time of signal detection. And if the contrast mechanism we're using - BOLD - relies on local magnetic field inhomogeneities then a spin echo will eliminate global signal imperfections we don't want - as might arise from metal fillings (or a wrench!) - as well as those that we do, around small blood vessels. Doh! Spin echoes can be used for fMRI at field strengths above 3 T, but that's another topic entirely. They have also been used at 3 T but the sensitivity of the fMRI experiment is much reduced. (Remind me and I'll do a post on spin versus gradient echoes for fMRI at a later date.)

So why bother to include this video now, if we're not going to be looking at spin echoes in fMRI experiments? Simple. The spin echo is a useful step in understanding the concept of signal phase. When we come to look at the EPI pulse sequence we will be using trains of gradient echoes, which can be understood in terms very similar to the spin echo.

If you have a moment, go back and watch the last three minutes of the video again. Prof Callaghan repeats the (Hahn) spin echo trick in a train, creating signal after signal using 180 degree pulses equally spaced in time. In the video this train of spin echoes - named after Carr & Purcell, who came up with the idea - produces signals that decay away with time constant T2. In later posts we will look at the EPI pulse sequence, which involves generating a train of gradient echoes, signals which die away with time constant T2*. What does all this mean? If you don't already know about T2, T2* and gradient echoes don't worry about it now, just remember that you've already seen the concept of a signal train. The concept is directly applicable to EPI.

(PS Erwin Hahn, inventor of the spin echo, turns 90 years young this year and is still doing science at Berkeley. What a guy!)


Finally, it's time to relaaaaaax

When we relax it generally involves sitting in a comfy chair, or sipping a glass of something. That's not quite how it works for nuclear spins. For starters, we quantify the rate at which the spins relax; bet you don't measure your exponential relaxation rate on a Friday night in the pub... Or perhaps you do.

In any event, in today's final video Prof Callaghan describes how spins relax, i.e. lose energy. For our conceptual convenience we separate the entire relaxation process into two categories: one where energy is exchanged between spins but where there's no net loss of thermal energy, and another where the polarized spins give up their energy to "the lattice" in order to return to their ground state, as existed before the excitation (90 degree) RF pulse.

Okay, coffee break over. Back to work. Next time we'll look at how we encode spatial information into the spins and make images. We are rapidly closing in on the time we must encounter (dah, dah, daaaah!).... k-space. (The left ear, the right ear and the final frontier.)