Magnets, Spins,

and Resonances
An Introduction to the Basics
of Magnetic Resonance
siemens-healthineers.com
Magnets, Spins, and Resonances
MRI diffusion tensor tracto-
graphy of the human brain1
(2015) vs. bell pepper image
(1980)
1 MPI, Leiptzig, Germany
 MR Basics:
A Long
Road
Made Easy
For a Nobel laureate, it is
a long road to Stockholm.
Magnetic Resonance has
traveled a long road from
probing water droplets
“Before every big “It’s a different and paraffin blocks to
breakthrough, it is first type of image that imaging bell peppers and
a crazy idea.” carries with it much the human body, creating
more information about seven Nobel laureates
Paul C. Lauterbur,
Nobel laureate 2003 the disease process.”
along the way. This
Sir Peter Mansfield, brochure may shorten
Nobel laureate 2003
your road to understand-
ing the basic principles
of MRI.
 Table of 1 Tracing the MR Signal

Contents 2 MRI in a nutshell

8 Magnetizing the body

24 The beauty of spin

38 Creating the resonance

54 The technical side

iv Magnets, Spins, and Resonances

2 Echoes, Decay, and Relaxation 3 From the Signal to the Image

66 A first acquaintance with relaxation 96 Slicing by gradients

74 Longitudinal recovery 108 The matrix: resolution

80 Transverse decay 120 Introducing the pulse sequence

86 The spin echo 126 Meet the gradient echo

v
 4 A Study of Contrasts 5 The Strategy of Image Quality

132 Spin echoes and contrast weighting 158 Contrast, signal, and noise

142 Contrast with inversion recovery 170 About image size and resolution

148 Contrast with gradient echoes 176 Speeding up the measurement

vi Magnets, Spins, and Resonances

 6 Increasing Diagnostic Value

186 Packing echoes and slices 217 References

194 Take the echo train: Turbo spin echo 221 Index

200 Echo-planar imaging (EPI)

206 Ultrahigh fields

212 Hybrid imaging

vii
1
Tracing the
MR Signal
How does magnetic resonance imaging work ?
Let us follow the basic principles step by step.
The MR image is created by magnetic reactions
in the patient’s body that generate a measurable
signal.
1 TRACING THE MR SIGNAL

MRI in a nutshell
Let us start with a clinical question

How can we detect a lesion in the image?

Medical imaging depends on the ability to visualize anatom-
ical structures, exploiting some physical properties within the
human body. For diagnostic purposes, it is essential to
discriminate between normal and pathologic tissue.
MR is a non-invasive imaging technique that produces series
of slice images with arbitrary angulation, displaying the
structure and/or function of the head, body, or extremities.
An MRI system may also be used for imaging during inter-
ventional procedures.
How and why does MRI work?

2 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

It is all about contrast

Of course, good contrast resolution of the “Magnetic resonance measurements
MR image is of paramount importance for may be used as a method for discrimi-
a precise diagnosis. nating between malignant tumors
and normal tissue.”
MR imaging offers excellent soft-tissue (Raymond Damadian, 1971)
contrast that results from the combination
of several parameters. What is the source
in the human body that provides for this
contrast?
Let us start with an axial slice through a
humans head and trace back to where
image contrast originates.

3
1 TRACING THE MR SIGNAL

Tracing image contrast back to the source

A simplified picture of the MRI “value chain”

The slice image is an array of picture ele-

ments (pixels). Each pixel has a certain gray
value. More pixels in an image mean better
resolution.
To simplify things, we will start with very
low resolution and focus on four contrasting
Pixels pixels in the brain image (graphic left).
This is crucial: The pixels in the image corre-
spond to volume elements (voxels) in the
slice measured (graphic right).

4 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The patient is positioned in a strong

magnetic field. A distribution of
“magnetic forces” will build up in the
body (arrows left), visualized for
only four voxels of interest.
This distribution of magnetism,
known as magnetization, is not
uniform but depends on tissue
properties. Stronger magnetization
within a voxel is depicted by a
longer arrow (left).

Voxels
Slice

5
1 TRACING THE MR SIGNAL

In order to make the magnetic distribution

Pulse within the body visible, we have to tilt the
magnetizations with the help of an electro-
magnetic pulse (left). This behavior is the
hallmark of the magnetic resonance process.
The magnetizations begin to rotate and
generate an alternating electric current in
a receiver coil. This current is the MR signal
(right), a mixture of the magnetizations of
the contributing voxels.

Rotating
magnetizations MR signal

6 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

In a magnetic field, the human body

builds up a spatially distributed
magnetization. Once tilted, the
magnetization rotates and generates
an MR signal. From a series of signals,
an MR image can be computed.

Peaks Actually, the local magnetizations are made

to rotate with different speeds (frequencies)
and timing (phases). Hence, we can filter out
their individual contributions from the MR
signal and thus obtain separated peaks (left).
The information of original voxel locations is
conserved by the MR imaging method. Thus,
we can match individual peaks to corre-
sponding pixels in the image. Peak intensity
is assigned to a gray value (right).

Image

7
1 TRACING THE MR SIGNAL

Magnetizing the body

Preparing the patient’s body for imaging

Susceptible to magnetic fields

A prerequisite for MR imaging is that the patient is positioned Let us trace the origin of magnetism in the human body. The
in the strong magnetic field of a scanner. Why? human tissue not only consists of atoms and molecules, but
of elementary magnets as well. What kind of particles do
The human body is susceptible to magnetic fields: the strong
we mean by that?
external field enhances the magnetism already present in the
tissue.

External magnetic field

Elementary magnets

8 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Discussed in more detail:

As early as in the 19th century,

physicists theorized that ’elementary
magnets’ might be responsible for
magnetism. Today we know: in
ferromagnetism, it is electrons, in
magnetic resonance, it is atomic
nuclei.

Not only protons (hydrogen nuclei)

are suitable for use with magnetic
Hydrogen protons abound in the body
resonance. MR imaging and spectro-
Atoms consist of a nucleus and electron shells. Hydrogen scopy are possible with heavier nuclei
possesses the least complex nucleus: a single proton. as well.

MR tomography uses the magnetic characteristic of the The ability of matter to be magnetized
hydrogen protons to generate images. is known as susceptibility.

1. Hydrogen is an elementary part of water and fat, which

makes it the most prevalent element in the human body.
2. Of all the elements, the nuclei of hydrogen produce the
strongest magnetic resonance signal.

The human body contains elementary magnets and

is thus susceptible to magnetic fields.

9
1 TRACING THE MR SIGNAL

On magnets

Magnetism is a fundamental property of The strength of a magnetic field at each

nature and is seen both in very large and location in space is known as the ’magnetic
very small structures. induction’ (symbol: B). In MR technology,
the concept of magnetic field strength
A magnet creates a surrounding magnetic
measured in 1 tesla = 10,000 gauss is
field. A magnetic field can also be created
commonly used. 1 tesla is approximately
by electric currents and electromagnets.
20,000 times stronger than the magnetic
Each magnetic field exerts a force on mag- field of the earth.
netic and magnetizable particles. Strength
A magnetic field of uniform field strength is
and direction of a magnetic field can be
called a homogeneous field. The field lines
visualized by magnetic field lines.
of a homogeneous field are drawn as equi-
distant, straight lines running in parallel.
A magnetic field that does not change over
time is known as a static field.

10 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Magnets were known in ancient times

in the East and West : the Greek
philosopher and mathematician
Thales of Miletus (634–546 BC), and
the Chinese statesman Guan Zhong
(685–643 BC) both made reference to
the “loadstone” (magnetite) and its
mysterious behavior. With his work
De magnete (On magnets), British
astronomer William Gilbert (1544–
1603), brought a scientific under-
standing to magnetism. Gilbert was
the first scientist to realize : “the
terrestrial globe is magnetic … the
Earth itself is a giant magnet.”

The English physicist Michael

Faraday (1791–1867) conceived
of magnetic field lines after
observing the behavior of iron
filings placed on a table
between magnets.
The tesla unit is named after
the Serbian American physicist
and inventor Nikola Tesla
(1856–1943).

11
1 TRACING THE MR SIGNAL

The proton as a tiny magnet

The natural “ingredient” of magnetic resonance

A magnetic moment
We can conceive of the proton as a tiny sphere. As an ele-
mentary magnet, the proton possesses a magnetic field sim-
ilar to the Earth’s magnetic field, with a “north” and a “south”
pole.
We attribute to the proton a quantity, known as the
magnetic moment, which by definition points to “north”
(red arrow). The magnetic moment determines the magni-
tude and direction of the force this elementary magnet can
exert.
The large number of magnetic protons in the body are
responsible for the magnetic reaction of the human tissue
to the strong magnetic field of the MR scanner.

12 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The concept of magnetic moment was

introduced by the French physicist
Paul Langevin (1872–1946). He
theorized that, at room temperature,
molecular magnets display a weak
magnetism, also known, in contrast
to the strong ferromagnetism, as
’paramagnetism.’

The German physicist Otto Stern

(1888–1969) determined the proton
magnetic moment in 1933 using the
’molecular beam method’ (Nobel prize
As we are only interested in the proton’s magnetic moment 1943), further developed by Isidor
(m), we will let its field lines and sphere disappear for good… Rabi for magnetic resonance.

Hydrogen protons possess magnetic moments,

which are responsible for the weak magnetism
created in the human body with MRI.

13
1 TRACING THE MR SIGNAL

Vectors revisited: arrow arithmetic

As magnetism is a directed physical phe- Arrows are suitable for depicting vectors.
nomenon, it might be useful to review the The direction of the arrow corresponds to
meaning of vectors. the direction of the vector quantity, the
length of the arrow corresponds to the
Many physical quantities, for example,
magnitude of the vector.
temperature or mass, are known to be non- a
directional. They are sufficiently identified Vector quantities allow for spatial addition.
by their magnitude and unit (for example, The direction has to be taken into account
70 degrees Fahrenheit, 5 kilograms), so- and visualized by linking the arrows.
called scalars.
If the arrows point in the same direction,
On the other hand, many physical quanti- i.e. have the same sense, the magnitude
ties depend on spatial orientation (for of the vector sum is simply the sum of the
example, force or speed). Vectors are magnitudes (in this case a+a). a
excellent for defining these quantities,
which exhibit magnitude and direction.

a + a = 2a

14 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Please do not confuse physical quan-

titites with vectors or vectors with
arrows. A vector is a mathematical
model for a physical quantity. An
arrow is merely a tool for the visual
representation of a vector.

Vectors of the same magnitude but oppos-

ing direction cancel each other out:
a–a = 0
Just as you can add vectors, you can also
decompose them. Each vector, for exam- +a –a
ple, can be divided into separate compo-
nents. These are the projections of the
arrow along predefined spatial axes, typi-
cally a coordinate system. a–a = 0
In our example, vector c is the sum of
vectors a+b. It consists of a vertical compo-
nent a and a horizontal component b. b

a
a+b

15
1 TRACING THE MR SIGNAL

A different spin on protons

The essence of MRI

About compass needles and spin magnets

When we use a compass, the needle aligns with the Earth’s
magnetic field and, in principle, points north. The compass
needle aligns because:
1. The Earth’s field exerts a force on the poles of the needle,
which makes it deflect;
2. The needle can rotate freely on a pivot with low friction,
although not entirely without friction.
If there were no friction at all, the compass needle would not
align but perpetually swing like a pendulum about north.
Exposed to a magnetic field, a hydrogen proton does not
behave like a compass needle, but more like a spinning top.

16 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

A toy model
As children, we all liked tops. We know that when you tip
a rapidly spinning top, it does not fall over. Instead, it begins
to wobble because its rotation keeps it from falling over on
its side.
This is how the top behaves: its spin axis moves in the shape
of a cone about the direction of gravity.
This movement is called precession. ’Precession’ is the word
used by physicists to describe the wobbling movement of a
spinning top.
Precession is fundamental to magnetic resonance.

17
1 TRACING THE MR SIGNAL

B
Magnetic tops in the human body

We have learned that a proton does not behave like a com-
pass needle. Let us look at the movement of its magnetic
m
moment (m) subject to a magnetic field (B).
The magnetic moment cannot line up with the external field.
Just like a child’s spinning top, it is forced to precess: to follow
the shape of a cone around the field. Because the precession
of the proton’s magnetic moment is frictionless, this
movement does not stop as long as the magnetic field is
present.
Note that the proton itself is not precessing, unlike a child’s
top. It is rather the proton’s magnetic moment, visualized as
a rotating vector.

When exposed to a magnetic field, the magnetic

moment of a proton precesses like a spinning top.

18 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Perpetual spin precession…

As we are surrounded by magnetic fields, elementary The U.S. physicist Edward Purcell
magnetic moments always precess, for example, in the pioneered magnetic resonance in solid
Earth’s magnetic field. Even though the magnetic field “(The magnetic dipole) matter. “I remember, in the winter of
our first experiments, looking on
of the Earth is approximately 30,000 times weaker than responds like any snow with new eyes. There the snow
that of a 1.5 tesla MR magnet, it can, in fact, be used for
magnetic resonance.
rapidly spinning lay around my doorstep—great heaps
of protons quietly precessing in the
gyroscope: instead earth’s magnetic field.” (Nobel prize
For clinical imaging, magnetic fields tens of thousands
of lining up with the lecture, 1952)
of times the strength of the Earth’s magnetic field
achieve stronger MR signals and thus, better contrast field, the spin axis
resolution in a shorter measurement time. precesses around the
field direction.”
Edward Mills Purcell
(1912–1997)

19
1 TRACING THE MR SIGNAL

The build-up of magnetization

A wonder of nature exploited by MRI

The body under the looking glass

Typically, the magnetic field of an MR
scanner runs parallel to the longitudinal
axis of a patient positioned in the magnet
bore. Let us take a closer look inside a
volume element (voxel) of the body.
Resulting
magnetization

External magnetic field

20 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

It is sometimes believed that nuclear

magnetic moments can only align
parallel or anti-parallel to a magnetic
field. In general, however, the mag-
netic moments have an arbitrary
orientation. A static, homogeneous
magnetic field cannot realign a
proton’s magnetic moment, only
make it precess. It is the magnetic and
thermal interaction between protons
and their molecular environment (the
’lattice’) that builds up a weak align-
ment of precessing moments with the
The total of all proton spins within a voxel constitutes a large external field.
statistical number, known as an ensemble. As soon as the
The build-up of longitudinal magneti-
patient is positioned in the MR scanner, all magnetic
zation takes some time. In a 3 tesla
moments of the ensemble will precess around the external magnet, for example, it will take up to
field lines. They will precess with a nearly uniform 15 seconds for body fluids, in soft
distribution of directions, but seemingly, a small portion of tissue around 4 seconds, for the
the ensemble will tend to align with the external field. magnetization to build up. This pro-
cess is known as “relaxation” and will
So, what difference does that make? be explained in Chapter 2, Echoes,
Decay, and Relaxation.

21
1 TRACING THE MR SIGNAL

Partial alignment of spin magnets

A dynamic interaction takes place within the body, partially Perpendicular to the external field, the magnetic moments
realigning the precessing magnetic moments. Three “actors” cancel out because the distribution of precessing moments
enter the voxel: remains perfectly uniform. Only parallel to the external field
does a small excess of magnetic moments remain. This is why
1. The ensemble of magnetic moments;
the magnetization is purely longitudinal.
2. The external magnetic field;
The larger the number of protons within a voxel (proton
3. The molecular environment of the protons. density), the stronger the longitudinal magnetization. This
local magnitude is specific to the type of tissue.
The interaction of protons with their molecular environment
builds up a macroscopic magnetization parallel to the
external field (big arrow in the figure of page 20). We can
think of this weak magnetization as the vector sum of the Exposed to a magnetic field, the human body builds
whole ensemble. up a weak longitudinal magnetization, whose
magnitude is proportional to proton density.

22 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

REDUCED TO THE ESSENTIALS

When the human body is positioned within

the magnet bore of an MR scanner, it will
acquire a spatial distribution of magnetiza-
tion parallel to the external field. This longi-
tudinal magnetization is much weaker than
ferromagnetim but can be used for mag-
netic resonance imaging.
The magnetization is caused by the com-
bined effect of hydrogen protons inter-
acting with their molecular environment
(relaxation). Hydrogen is the most preva-
lent element in the human body.
The magnitude of the magnetization is pro-
portional to the number of protons per unit
volume (proton density within a voxel). This
local magnitude is specific to the type of
tissue.
When protons are exposed to a magnetic
field, their magnetic moments will precess
like spinning tops. Magnetic moments are
the vector quantities that determine the
magnetic force of the particles.

23
1 TRACING THE MR SIGNAL

The beauty of spin

A deeper look into the MR microcosm

The source of magnetic resonance

As we have seen so far, the physical property of a proton rel-
evant for MRI is the vector quantity known as the magnetic
moment. The reason why the behavior of a proton is differ-
ent from a compass needle or a bar magnet is explained by
the property of spin. The spin makes nuclear magnets pre-
cess like spinning tops.
How can we picture the property of spin? In fact, the spin of
a proton cannot be attributed to intrinsic rotation. As an anal-
ogy, we can visualize the spin as something like a spinning
wheel. A magnetic spinning wheel.
Let the spinning wheel tilt and wobble, and we have preces-
sion.
The speed, or frequency, of precession is of great importance
in magnetic resonance: We excite many proton spins to tilt
and precess “in phase,” and thus generate an MR signal.

24 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

In 1921, the American physicist

Arthur H. Compton (Nobel prize 1926)
supposed “that the electron is spin-
ning like a gyroscope (and) possesses
a magnetic moment.” It was not until
1926 that Dutch physicists Uhlenbeck
and Goudsmit reestablished the idea
of “self-rotation” of particles in order
to explain the hyperfine splitting of
spectral lines. In terms of physics, the
notion of self-rotation of subatomic
particles cannot be upheld, but the
word “spin” has been coined and is
here to stay.

25
1 TRACING THE MR SIGNAL

On frequencies and phases

What do we mean by frequency? It is the number of revo-

lutions or oscillations of a periodic movement per unit time. 50 Hz
You know this from your car and the tachometer. For exam-
ple, the tachometer shows 3,000 revolutions per minute.
This is nothing more than the frequency.
3,000 rpm are the same as 50 revolutions per second. The
unit for revolutions per second is hertz (Hz). In our case, this 100 Hz
means a frequency of 50 Hz.
When we apply oscillations to a time axis we obtain the peaks
and crests of a sinusoidal wave. Oscillation with double the
frequency (100 Hz) is shown as a compressed sinusoidal
curve.

26 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

You can compare a phase to the move- western mountain central eastern
ments of the hands on a watch or clock. time time time time
The hands show you the offset in time
between one state of rotation and another.
If your watch is an hour fast, you could
think of this as a “phase shift” of 1 hour
relative to local time. You could correct
this by setting your watch accordingly,
that is, by setting the hour hand back by
30 degrees. This is not possible when we
talk about the phase shift between San
Francisco and New York. The three-hour
time difference between the two cities is of
a lasting nature. If you travel in a plane
across large distances, you can experience
New York
this kind of time shift as jet lag. Most oscil- San Francisco
lations, for example, radio waves, contain
this type of “jet lag.”
Later, we will describe how both shifts in
frequency and phase are used to compute
images from the MR signals acquired.

27
1 TRACING THE MR SIGNAL

Precession revisited
Proportional to field strength

B
The Larmor relationship

The precession frequency of spins is also known as the
Larmor frequency. It depends on the type of nucleus and the
strength of the magnetic field applied.
m
The Larmor frequency  increases proportionally with the
magnetic field B. The following expression applies:

 = B

(The constant factor  is known as the gyromagnetic ratio

of the nuclei.)
In the earth’s magnetic field, spins precess relatively slowly at
approximately 2,000 Hz (2 kHz).

28 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The quintessential movement in MRI

In the magnetic field of an MRI scanner, spins will pre- Spin is a quantity that adds to the total
cess at frequencies comparable to radio frequencies. angular momentum of a particle. In
This means that the spins precess at several million “The influence of this book, we don’t look at spin as
such, which is a complex object
revolutions per second. The Larmor frequency in a magnetic field on residing in a cone-like probability
a 1.5-tesla magnetic field is 64 MHz, in a 3-tesla field
it is twice that, i.e. 128 MHz.
the motions ... is cloud, describable by quantum
physics. In MRI, discussion of spin is
precisely the same as reduced to observable components of
Larmor precession is at the heart of magnetic reso-
that of a rotation with spin along a quantizing axis (“spin up”,
nance. It is the physically most adequate description “spin down”). The relevant aspect of
of spin dynamics in magnetic fields. angular velocity  ...
spin for MRI is the averaged value
around the axis of (expectation value) of the spin mag-
the field.” netic moment, visualized as the spin
polarization vector, also known as the
Larmor precession of spins is proportional to Joseph Larmor “Bloch vector.” This vector is the pre-
magnetic field strength. (1857–1942) cessing quantity shown. And this is
what, in MRI, is meant by ’spin’ for
short.

29
1 TRACING THE MR SIGNAL

Nuclear spin beyond protons

Sources for magnetic resonance imaging

Nuclei with net spin

MRI predominantly uses hydrogen nuclei (protons). Atomic Atomic nucleus with
nuclei in general consist of protons and neutrons. Both have proton and neutron
the property of spin. spins

Nuclei with an uneven number of protons or neutrons have

a net spin known as the nuclear spin. Common examples
are: carbon 13C, fluorine 19F, sodium 23Na, or phospho-
rus 31P. Two thirds of the isotopes found in nature have a net Net nuclear spin
nuclear spin, making them suitable for use with magnetic
resonance.
As explained before, the Larmor frequency is not only pro-
portional to the external field strength but depends on the
type of nucleus as well (identified by the gyromagnetic ratio).
Exposed to a 1.5-tesla field, hydrogen protons precess with a
different Larmor frequency than, for example, phos-phorus
nuclei.

30 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

In the atomic nucleus, two identical

particles cannot be in the same state.
They have to align their spin anti-
parallel to each other, and the net spin
of this “couple ” of particles cancels
out. This rule of nature is known as the
Pauli Verbot (or ’exclusion principle’).
It is the residual particle that creates
the nuclear spin. The presence of
nuclear spin does not mean that the
nucleus is spinning but has a net
magnetic moment.
Nuclei without spin
Nuclei with an even number of protons and neutrons do not
have a net nuclear spin. They are magnetically neutral.
Examples of these are oxygen 16O (with 8 protons and
8 neutrons each) or carbon 12C (with 6 protons and
6 neutrons each). These isotopes are not suitable for use in
mag-netic resonance.

Atomic nuclei with an uneven number of protons or

neutrons have a net spin and are suitable for
magnetic resonance imaging.

31
1 TRACING THE MR SIGNAL

The small difference that makes MRI possible

Focusing on the “suspects” that are responsible for the magnetization

H
Introducing the concept of excess spins O
H

Magnetization is built up in a spin ensemble because the

ratio of spins precessing in and against the direction of
the external magnetic field is not 50:50 but slightly biased. 100 000 000 000 000 000 000
There are slightly more spins pointing in the direction of the Protons
exter-nal field than against the field. We ignore the majority
of spins that cancel each other out and focus solely on these
excess spins : the relatively small number of spins consid- 600 000 000 000 000
ered to add up to longitudinal magnetization.
Excess spins
The number of excess spins is a function of different factors.
Their number grows:
• in proportion to the proton density
• with the strength of the external magnetic field
• as the temperature decreases.
At body temperature and a field strength of 1 tesla, there are
approximately 6 excess spins, or 0.0006 % of 1 million pro-
tons in a voxel of water.

32 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The excess, or population difference,

of precessing spins in a magnetized
volume is expressed by the well-
known Boltzmann distribution of
statistical physics.

Calculating the excess

Since it takes many zeros after the period to express it in
percentages, there is another way of expressing this small
numerical relationship. The unit of measurement for very
small quantities is ppm or parts per million. Coming back to
our example, at 1 tesla, the number of excess spins is approx-
imately 6 ppm.
As you can see, the number of excess spins is relatively small.
That we are still able to obtain a measurable effect is due to
the large number of protons in the human body.
For example: a tiny voxel of 1.5 microliters of water is popu-
lated by approximately 1020 hydrogen protons (a 1 followed
by 20 zeros). At 1 tesla with 6 ppm excess spins, this trans-
lates into 600 trillion small spin magnets that add up to
a macroscopic magnetization.

Excess spins are the small number of spins within

a voxel that add up to net magnetization.

33
1 TRACING THE MR SIGNAL

The xy-plane
Let us agree on the following conventions:
Z
In the usual xyz coordinate system, we place the z-axis by
definition in the direction of the external magnetic field.
We will call the plane running transverse to the field lines
the xy-plane.

34 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The spin packet z

We are going to concentrate on the excess
spins of an ensemble. Let us reduce them
for clarity’s sake to a spinning ”six pack”,
with identical precession cones but differ-
ent phases.
While all spins of the spin packet have the
y
same precession frequency, they point in
different directions. The components trans- x
verse to the magnetic field (blue arrows),
that is, within the xy-plane, cancel out to
zero.
Consequently, magnetization will be built
up along the z-axis only, as a statistical sum
of the z components of the spin vectors.
This is the longitudinal magnetization.

35
1 TRACING THE MR SIGNAL

REDUCED TO THE ESSENTIALS

The precession of magnetic moments of

nuclei is due to its spin property. We can
picture spin as a kind of magnetic spinning
wheel.
Along with hydrogen, two thirds of the
isotopes found in nature have a net spin,
making them suitable for use in magnetic
resonance.
The velocity of precession, known as the
Larmor frequency, is a function of the
magnetic field strength applied. In MRI,
we concern ourselves with field strengths
that correspond to the high-frequency
range of radio waves (in Megahertz).
A spin packet is a model for the excess of
spins precessing at the same frequency in
a voxel, which are considered to add up to
macroscopic magnetization.

36 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

37
1 TRACING THE MR SIGNAL

Creating the resonance

Making spins generate the MR signal

The excess spins within a voxel create a static longitudinal The magnetization can be tilted by applying a short electro-
magnetization. In order to generate the MR signal, we tilt magnetic pulse, the RF pulse.
the magnetization so that a component will precess in the
xy-plane.

38 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The RF pulses transmitted during

a pa-tient examination use a circu-
larly polarized wave that contains
a rotating magnetic field.

Magnetic frisbees
What is an RF pulse? Imagine a magnetic frisbee that is sud-
denly flying through the field. The frisbee acts like a rotating
magnet. How does a rotating magnetic field tilt the magneti-
zation?

39
1 TRACING THE MR SIGNAL

Satisfying the resonance condition

For maximum impact, the RF pulse acts perpendicular to the
external field and is in resonance with the precessing spin
packet of the voxel of interest.
Physically, the resonance condition means:

The frequency of the RF pulse applied has to match

the Larmor frequency of the spins.

Returning to our analogy: the rotating magnet has to rotate

at the same speed as the spins precess.

40 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

On resonances

Resonance stimulation in MR can be com-

pared to the oscillations created by a tuning
fork. When the tuning fork is struck, it
begins to oscillate and generates a specific
sound. The pitch corresponds to the oscil-
lation frequency of the acoustic wave.
When you introduce a second tuning fork
tuned to the same frequency, it will oscil-
late in response to the acoustic waves
emitted by the first tuning fork. The two
tuning forks are now in resonance.

41
1 TRACING THE MR SIGNAL

Tilting the magnetization

A rotating frame

The story of the merry-go-round

What exactly is happening during magnetic resonance? Let M
us look at yet another analogy:
Imagine yourself as the rotating magnet (that is the RF pulse)
running in resonance with the precessing spins.
For this purpose, you are running around the spin carousel
throwing stones into a rotating “spin scale.“ Your time is
limited. If you run either too fast or too slow around the
merry-go-round, you are no longer in step. Now, you can
only catch up with the scale after one complete revolution
and you can only throw one stone. But if you are in step with
the spin scale, you can throw as many stones as you want
into the scale.

42 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

As a result of your efforts, the spin scale loses its The American physicist Isidor Rabi
equilibrium and its axis simply flips over. (Nobel prize 1944) discovered mag-

This is exactly what the RF pulse does: it tilts the

“To the system ... we netic resonance in 1938, using the
technique of molecular beams. The
magnetization (M). apply an additional tilting of spins around the field axis

In magnetic resonance, the ’merry-go-round’ is known as

magnetic field which of the RF pulse, perpendicular to
Larmor precession, is known as ’Rabi
the rotating frame of reference, introduced by Isidor is much smaller than precession.’
Rabi and Felix Bloch. H0 and perpendicular
“After the first experiment, every-
to it in direction”
thing was easy. Since you’re doing
experiments of that sort, there ought
Isidor I. Rabi
to be something aesthetic about
(1898–1988) them.”

43
1 TRACING THE MR SIGNAL

Pulses and flip angles

The stronger the energy of the RF pulse, the farther the mag-
Z
netization will flip or tilt. The final tilt angle is known as the
flip angle ().

A 180-degree pulse flips

Z
the magnetization into
the opposite direction of 
the z-axis.
180°

A 90-degree pulse flips

the magnetization exactly Z
into the xy-plane.
90°

44 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Decomposing the magnetization

Just like a vector, the tilted magnetization can be decomposed
Z
into two components located perpendicular to each other:
Longitudinal magnetization Mz is the portion of the vector
in the z-direction, that is, along the external magnetic field. M
MZ
Transverse magnetization Mxy is the component of the
vector that rotates about the external field in the xy-plane.
How fast does it rotate? The rotating transverse magnetization
is the sum of the spins that rotate in phase in the xy-plane,
matching the Larmor frequency. Therefore, transverse mag-
netization also rotates with the Larmor frequency.
MXY

After magnetic resonance, the resulting transverse

magnetization will rotate with Larmor frequency.

45
1 TRACING THE MR SIGNAL

A closer look at the resonance situation

The viewpoint of the spin packet

z
By applying a 90-degree pulse, magnetization flips in the
transverse direction, that is, in the xy-plane. As long as the
RF pulse is present, two magnetic fields are active: the static
field, and for a short time, the rotating RF field.
y
Using a special trick, we can make the static field disappear: x
all we need to do is climb onto the spin carousel (the rotating
frame) together with the spins. Once on it, the spins effec-
tively “feel” only the rotating RF field (the frisbee magnet).
Since this field rotates in resonance with the spin packet, its
axis appears static to the spins.
How do the spins react to this magnetic excitation?
Before the 90-degree pulse

46 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The spin packet starts to precess “downward” to the xy-plane. z

After a 90-degree pulse, they are concentrated in the hori-
zontal direction (to the right in our example).
As you can see from our picture, the xy components of the y
spin packet do not point randomly in all directions any more, x
instead they all share more or less one direction. Now they
precess coherently and thus generate a transverse magneti-
zation.
As you can imagine, when a 180-degree pulse is applied,
the spin packet will rotate until it is flipped downward by
180 degrees. Accordingly, longitudinal magnetization will
At the end of the 90-degree pulse
be oriented antiparallel to the external field.

47
1 TRACING THE MR SIGNAL

Here comes the MR signal

Oscillating with Larmor frequency

Rotating transverse magnetization 90°

The transverse magnetization Mxy, subject

to the static magnetic field, behaves like
a rotating magnet. You can move a coil into MXY
the rotating magnet and induce an electric
voltage in it.
The course of the voltage over time is the
MR signal. The stronger the transverse
FID
magnetization within a voxel, the stronger
the MR signal.
By acquiring different signals from different
voxels, image contrast is created.

48 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Free induction decay

Note that the MR signal decays relatively quickly. Since trans- As the transverse magnetization does not rotate forever but
verse magnetization starts to decay, the induced signal will decay, too.
• precesses freely The reasons for the decay of the MR signal are explained in
Chapter 2, Echoes, Decay, and Relaxation.
• induces a signal, and
• decays immediately
after the end of the RF pulse, this MR signal is called free
induction decay or FID for short.

49
1 TRACING THE MR SIGNAL

On induction

From electrical engineering we know that a magnetic field We use this induction every single day. For example, a bicycle
changing in strength or direction generates an electric dynamo contains a rotating magnet driven by the wheel of
voltage in a coil. This is what we call electromagnetic the bicycle which continuously changes the direction of its
induction. magnetic field.

50 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

These changes in magnetic field induce a flow of Electromagnetic induction is a near-

current in the coil, which can be used to light up the field effect and can be described by
bicycle lamp. The faster the dynamo magnet rotates, “If a single wire be classical electrodynamics. Sometimes
the MR signal is believed to be a radio-
the higher the induced electric voltage and the brighter moved like the spoke frequency wave, which is received by
the light generated by the bicycle
of a wheel near a mag- the RF coil like a radio signal. But in
MRI, the radio wave of the spins is
netic pole, a current negligible. The Faraday field at the coil
of electricity is deter- is much larger and is responsible for
the signal detected.
mined through it ...”
Michael Faraday
(1791–1867)

Michael Faraday not only explained

magnetic field lines but also discovered
the phenomenon of electromagnetic
induction in 1831, also known as
“Faraday induction” in honor of his
name.

51
1 TRACING THE MR SIGNAL

SUMMARY

In a strong magnetic field, a locally varying

distribution of magnetization is created in
the human tissue.
The external field forces the proton spins to
precess at the Larmor frequency, which is
proportional to the strength of the mag-
netic field.
The molecular environment forces the
spins to realign, breaking the uniform
distribution along the external field.
A weak magnetization is built up parallel to
the external field.
When excited by a 90-degree RF pulse,
longitudinal magnetization is tilted into
the xy-plane and rotates at the Larmor
frequency.
The rotating transverse magnetization
generates the MR signal in a receiver coil
and decays quickly (free induction decay,
FID).

52 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

53
1 TRACING THE MR SIGNAL

The technical side

The components that make MRI possible

The main system components

A typical MR system consists of three components or sub-
systems:
• a magnet with a main magnetic field,
• a gradient system, and
• the radio-frequency (RF) system.
The gradient system is needed predominantly for the locali-
zation of slices and voxels and will be explained in Chapter 3,
From the Signal to the Image.

54 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The computer system

To generate and evaluate high-quality MR images, the three
subsystems have to be controlled and the measured results
displayed. For this purpose, the high-performance computer
system includes the following:
• the image processor for computing the MR images from
raw data acquired
• the host computer with console for the technician, and
• the control as well as evaluation software for performing
measurements and creating image results of diagnostic
value

55
1 TRACING THE MR SIGNAL

The main magnet

A strong magnet Superconducting magnets

The homogeneous magnetic field required for MR imaging is This is how a superconducting magnet works: At normal
generated by a strong magnet. This magnet constitutes the temperature levels, an electric conductor is resistive. Without
most important as well as the most expensive component of a constant power supply, an electric current injected into
the MR system. a circuit would begin to decay because of its loss in energy.
In general, superconducting electromagnets are used. Superconductors are materials that have no electric resis-
A strong magnetic field is generated by the electric current tivity at very low temperatures close to absolute zero
flowing in large coils. (0 Kelvin = – 273 °C). A constant, high current (above
400 amperes) will flow for years without an electrical poten-
The homogeneous magnetic field is located in the center of
tial or voltage. For this reason, the superconductor has to
the magnet bore, which runs parallel to the longitudinal body
be kept at very cold temperatures. Liquid helium is used as
axis of the patient.
a coolant.

56 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

The RF system

The radio-frequency (RF) system RF coils

The proton spins of the body tissue are excited by briefly The body transmit coil is integrated in the magnet bore. Local
applied alternating magnetic fields in the radio-frequency RF receiver coils come in all shapes and sizes. The shape of
range. These RF pulses are transmitted, the MR signal the coil is determined by its area of application. Depending
generated by the rotating transverse magnetization has to be on the body region to be examined, they are positioned
received. locally on the patient's body.
The RF system of an MR installation consists of the following: The better a local coil is adapted to the geometry of the human
body, the less unwanted signal from surrounding tissue can
• RF coils
influence the imaging process.
• RF transmit amplifier for sending RF pulses
Array coils are used for examining larger measurement areas.
• RF receive amplifier for amplifying the received MR signal They include several independent smaller coil elements that
can be combined according to the area under examination.

57
1 TRACING THE MR SIGNAL

The magnetic field: safety aspects

Biological effects
The strong magnetic field of the MR scanner affects the tis- The distribution of surface currents present during an ECG
sue as well as all other magnetizable material in the vicinity alters in the magnetic field (magnetohydrodynamic effects).
of the magnet. Cardiac functions are not affected by it, only the observed
ECG signal is.
Since the introduction of MR tomography, a number of studies
have been performed to determine the biological effects of
the static magnetic field. Known effects include, for exam-
ple, dizziness, stomach upsets as well as a metallic taste.
Most of these effects only occur at field strengths above
3 tesla. These are short-term effects, that is, they occur
exclusively in the magnetic field or shortly after leaving
it. To date, no biological long-term effects have been
observed.

58 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Magnetic effects on devices and material

Magnetizable materials, for example iron, are attracted by The functionality of hearing aids may be compromized in
the MR magnet. This constitutes a potential source of hazard strong magnetic fields.
to the patient or the operating personnel. Considerable
In each case, the patient must be interviewed prior to the
forces may be generated, attracting even large iron masses
actual MR examination. If there are reasonable doubts, other
and causing them to accelerate as they move toward the
examination methods should be used.
magnet. The force exercised is proportional to the mass
involved. The functionality of mechanical devices and electric compo-
nents is not ensured in the vicinity of the magnet. The func-
Metal parts in the patient are also a source for hazard. Metal
tions of clocks, respiratory devices, as well as monitors,
splinters, clips, screws, or injection needles may be moved in
infusion pumps, and other devices may be affected by the
the body by the magnetic forces.
magnetic fringe field. The same applies to computers and
Especially critical are electrical implants such as pacemakers magnetic data carriers. Also, the encoding on credit cards
or hearing aids. As determined by national and international may be deleted in the vicinity of the magnet.
recommendations and guidelines, the safety/exclusion zone
for pacemakers has been established at a field strength
of 0.5 mT outside the magnet.

59
1 TRACING THE MR SIGNAL

7
0.05 mT

6
0.1 mT

0.2 mT 5

0.5 mT
4
1 mT
3 mT 3

1
Effect of the fringe field
The typical field strengths of magnets used in today's whole-
body MR tomographs are up to 3 tesla and may exceed 1 2 3 4
7 tesla in special cases. The MR magnets not only generate
the desired nominal field in the area under examination, they
also generate a fringe field outside the magnet.
The strength of the fringe field as well as its spatial distribu-
tion depends on the configuration of the magnet, its size, as
well as its basic field strength. 3 mT
1 mT

0.5 mT

0.2 mT

0.1 mT

0.05 mT

60 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Shielding the fringe field

For super-conducting magnets, the fringe field is shielded by
using measures that limit the external safety zone.
Today, active shielding is used. The fringe field is largely
compensated by additional super-conducting coils wound
in the opposite direction on the field generating coils.

61
1 TRACING THE MR SIGNAL

The RF field: safety aspects

Tissue warming
RF electromagnetic waves generate currents in electrically If the receiving RF coil is in resonance with the transmitter, it
conducting tissue and stimulate molecules in the tissue. The may act to increase the RF field close to the coil. This increase
resulting oscillations lead to tissue warming. Usually the in field strength is of particular concern when it occurs near
increase in temperature is less than 1 degree Celsius. the eyes. To eliminate this effect, the system decouples the
receiver coil during transmission.
The specific absorption rate (SAR) is the RF output absorbed
per time unit and kilogram. The RF field may induce AC currents in metal implants or
cables routed close to the patient (for example, ECG cables),
For safety reasons, the RF power emitted by the system into
resulting in local warming.
the body is monitored and the respective SAR values are
limited accordingly. The IEC limit values are 4 W/kg (whole System-specific warnings, labels, or notices have to be
body), 4–10 W/kg (partial body), 20 W/kg (local SAR head, observed at all times.
trunk), and 40 W/kg (local SAR extremities).

62 Magnets, Spins, and Resonances

 MRI in Magnetizing The beauty Creating The technical
a nutshell the body of spin the resonance side

Interference caused by other systems

The RF field emitted by the transmit coils may be super-
imposed on the voltage in external devices and lead to
interference.
Conversely, external interference (for example from radios,
cell phones, electronic controls, electro motors) may emit
interfering signals into the MR system and degrade the image
quality.
To provide the best possible protection in both directions,
MR systems are installed in RF-sealed rooms made from
conducting materials (Faraday cages).

63
2
Echoes, Decay,
and Relaxation
After RF excitation, the rotating transverse
magnetization decreases rather quickly and the
MR signal decays. We can create echoes of the lost
MR signal for a while. It takes some time before the
longitudinal magnetization fully recovers. These
processes are known as relaxation.
2 ECHOES, DECAY, AND RELAXATION

A first acquaintance with relaxation

Two seemingly independent processes

Decay and recovery

MXY MZ
After the longitudinal magnetization has
been tilted in the xy-plane by an alternating
field (RF pulse), it immediately rotates at
Larmor frequency and induces the MR
signal (FID).
Two things happen:
1. Transverse magnetization Mxy (red)
decays. “T2” “T1”

2. Longitudinal magnetization Mz (green)

recovers.
You might think that the following would
occur: the magnetization vector simply tilts
back to its original direction parallel to the
external field.
However, this is not true.

66 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

Two additional contrast mechanisms

Transverse magnetization Mxy decays more rapidly The Swiss-American physicist Felix
than the time required by longitudinal magnetization Bloch (Nobel prize 1952) was the first
Mz to recover. Both processes run exponentially, but “Relaxation can be researcher to establish the ’Bloch
equations,’ which describe the motion
with different timing. seen to act like of the macroscopic magnetization
A certain time is required for the longitudinal magneti- a friction which after resonance. The solutions to these
zation to recover (represented by “T1“). Transverse equations describe both relaxation
counteracts the tilt processes independently, resulting in
magnetization, however, disappears in a shorter time
produced by the expo-nential recovery for longitudinal
(represented by “T2“). magnetization and exponential decay
alternating field.”
That is why there is no simple tilt back. for transverse magnetization, with
different time constants.
Felix Bloch
But what is more important: Beyond proton density,
(1905–1983)
MRI has two additional contrast mechanisms that
enhance diagnostic value.
These mechanisms are known as relaxation.

67
2 ECHOES, DECAY, AND RELAXATION

On exponential processes

Many natural as well as social processes can Value

be expressed mathematically in a rather
easy way: they are exponential. The increase
in bacteria population, the reduction in
radioactivity, compounded interest—all
these are exponential processes. The same
applies to spin recovery. This is reason
enough for us to get involved.
Compounded interest is a good example of
unchecked growth. Let us assume you have
stocks or fixed rate funds in the value of
$10,000 invested at a 10 % interest rate.
After 10 years, your nest egg has grown to
approximately $26,000, after 20 years to as
much as $67,000, and after 50 years, you Time
are a millionaire. By then, your savings have
grown to $1.2 million.

68 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

Inflation is a good example of exponential Value

decay. Let us say you start with a cash bal-
ance of $100,000 and a rate of inflation
as high as 10 %, the value of your money
would be down to approximately $34,000
in 10 years, hence drop to $12,000 in
20 years and be practically worthless after
50 years.

Time

69
2 ECHOES, DECAY, AND RELAXATION

Relaxation made easy

Timing is everything

Equilibrium
The physical meaning of a time constant
100%
An exponential process is expressed by its time constant T.
After T, recovery of the physical quantity has reached 63 % of
its remaining differential value, after 2T, it has reached 86 %,
63%
after 3T, it has reached 95 %, and after 5T, the process is
nearly complete.
The strength of relaxation is a function of how far a physical
quantity is from its point of equilibrium. The closer the sys-
tem is to equilibrium, the weaker the relaxation. Thus, the T 5T t
process of returning to equilibrium slows down until a satu- Non-equilibrium
ration value has been reached.
You can compare this to a rubber band: the higher the
tension, the faster the rebound.

70 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

A falling box Kinetic energy

We can compare the two different relaxation processes in MRI “T2“
to a falling box. Imagine throwing the box from a high tower.
It will travel with increasing speed toward the earth. The
reason for this is the earth’s gravity. So far so good.
If the box is thrown from an airplane, two simultaneous
“forces” are at work: 1. gravity, 2. the kinetic energy in the
direction of flight. Gravity

The actual movement of the box is the superposition of two

“T1“
motions performed independently of each other. While the
box is falling toward the earth, coming closer and closer,
it continues in the direction of flight but slows down in
a parabola.

71
2 ECHOES, DECAY, AND RELAXATION

It is faster getting down the mountain than getting up

Let us summarize: while longitudinal magnetization is recov- The time constants are known as T1 and T2.
ering, transverse magnetization is decaying. And as men-
Normally, the T2 constant is considerably shorter than the
tioned before, the decay of the transverse magnetization is
T1 constant.
usually much faster than recovery of the longitudinal mag-
netization.

T1

T2

72 Magnets, Spins, and Resonances

A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

73
2 ECHOES, DECAY, AND RELAXATION

Longitudinal recovery
Moving back toward the initial maximum value with T1

90°
Return to equilibrium
The build-up and recovery of the longitudinal
magnetization is an exponential process
known as longitudinal relaxation. The MZ
time constant is known as T1. 100%
Please note: T1 is not the time span of total
magnetization recovery, but only the time
constant for its exponential growth: After 63%
time interval T1, the longitudinal magneti- MZ=100%
zation Mz has recovered to approximately
63 % of its final value. After 5 T1 times,
recovery is practically complete.
What is important is that the T1 constant T1 5T1 t
depends on the tissue type.

74 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

Fat White matter

Mz
Gray matter

100%

63%

240 680 809 2500 t[ms]

CSF

The T1 constant under the magnifying glass

Different types of tissue show different relaxation times. This
is key to the sharp image contrast obtained with MR. As the
table shows, the T1 constant also depends on the field as
well.

T1 constants (in ms)

0.2 tesla 1.0 tesla 1.5 tesla

Fat 240
Muscle 370 730 863
White matter 388 680 783
Gray matter 492 809 917 Fat has a short T1, water has a long T1.
CSF 1,400 2,500 3,000

75
2 ECHOES, DECAY, AND RELAXATION

Fat
White matter
Causes of longitudinal recovery Mz
Gray matter
Hydrogen protons in the body continuously 100%
“feel” local magnetic fields and their fluctu-
ations caused by molecular motion. These 63%
minute field fluctuations add to the exter-
nal field. The strongest effect is produced
by field oscillations transverse to the main t[ms]
field that match the Larmor frequency of
the protons. These “microresonances” 240 680 809 2500
cause the spins to randomly change their
orientation with respect to the main field. CSF

76 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

The T1 process is frequently referred to

as “spin-lattice relaxation” because in
solid state magnetic resonance the
molecular surroundings of nuclei are
called a “lattice.”

This process occurs after interference

from an RF pulse and as early as
during the build-up of the longitudinal
magnetization, after the patient has
been moved into the magnetic field.

The environment of the protons frequently consists of larger

molecules (lipids) and macromolecules (protein). Hydrogen
protons inside a relatively slow-moving fat molecule as well
as protons bound to proteins feel strong local field fluctu-
ations: they quickly change their spin direction. This explains
the relatively short T1 relaxation of fatty tissue, for example.
In fluids, the molecular mobility of water is considerably
faster than most field fluctuations. Resonances with oscil-
lating magnetic fields are less frequent as well as weaker: the
protons do not change their spin direction as quickly. This is
why pure water and CSF show a relatively long T1 relaxation.

77
2 ECHOES, DECAY, AND RELAXATION

Getting a taste for T1 contrast

Since different types of tissue show different T1 relaxation,
this difference can be shown as MR image contrast. How this
happens is explained in detail in a following chapter.
Diagnostic use in a nutshell: pathological tissue shows a dif-
ferent concentration of water than the surrounding tissue—
and this means a different relaxation constant. The difference
in relaxation is visualized as contrast in the MR image.

With T1 contrast, CSF appears dark in the MR image

78 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

REDUCED TO THE ESSENTIALS

After resonance, the longitudinal mag-

netization fully recovers over a time that
depends on the type of tissue and the field
strength. This process is known as longitu-
dinal relaxation.
The longitudinal relaxation follows an ex-
ponential course of growth characterized
by the time constant T1.
As the T1 constant depends on the type of
tissue, this characteristic can be employed
for the MR image contrast.
The source of T1 relaxation are local mag-
netic field fluctuations generated by molec-
ular motion. Transverse field fluctuations in
the range of the Larmor frequency show
the strongest effect and cause the proton
spins to change their direction randomly.

79
2 ECHOES, DECAY, AND RELAXATION

Transverse decay
The lifetime of the resonance phenomenon (T2)

The signal must decay 90°

Directly after the RF pulse, the spins are

what is called phase-coherent. They act like
one large magnet that rotates in the xy-
plane. That’s why they can induce a signal
in the receiver coil.
However, the rotating spins get out of
1 2 3
phase again due to unavoidable molecular
interactions, and the transverse magneti-
zation begins to decay. 37%

1+2+3

MXY = 0
2

3 1

T2 5T2

80 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

As the transverse magnetization decays, the induced MR

signal deteriorates. This is what we call transverse
relaxation. Its time constant is known as T2. As we will see
later on, this is only the maximum time for which magnetic
resonance will persist. In reality, the MR signal (FID) decays at
a more rapid rate.
After time interval T2, the phase coherence of the spins
has dropped to approximately 37 %. After 2 T2, it drops to
approximately 14 % and after 5 T2, phase coherence has
just about disappeared.

81
2 ECHOES, DECAY, AND RELAXATION

Mz Fat
White matter
100%
CSF

63%

Gray matter

84
1400 2500 t [ms]
92
101

The T2 constant under the magnifying glass

T2 is also tissue-specific, but largely field-independent.

T2 constants (in ms)

Fat 84
Muscle 47
White matter 92
Gray matter 101 Fat has a short T2, water has a long T2.
CSF 1,400

82 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

Reasons for the loss of coherence

There are two major causes of transverse relaxation. These slightly different precession frequencies are an addi-
tional reason for the loss of phase coherence: As a result,
1. As we have shown, fluctuating local fields are responsible
transverse magnetization decays before the longitudinal
for longitudinal relaxation. And it has also an effect to the
magnetization recovers.
transversal relaxation: when the proton spins change their
direction randomly, their phase coherence is lost as well. This Although the interaction between the spins is not the
is the “T1 contribution” to T2. only source of transverse relaxation, the term “spin-spin
relaxation” has made its entrance and is here to stay.
2. In addition, fluctuating magnetic fields in the direction of
the external field make the local field vary by a small amount
of approximately 1 millitesla. Accordingly, neighboring
protons precess with slightly different frequencies of approx-
imately 40 kHz around the normal Larmor frequency.

83
2 ECHOES, DECAY, AND RELAXATION

Getting a taste for T2 contrast

Since different tissue types show different
T2 relaxation, these differences are shown
as MR image contrast. A detailed explana-
tion is provided in a following chapter.

T2 contrast shows CSF

bright in the MR image—
opposite to T1 contrast

84 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

REDUCED TO THE ESSENTIALS

Immediately after the RF pulse, the rotating

transverse magnetization starts to decay.
1. Spin precession gets out of phase until all
spins are distributed uniformly within the
transverse plane.
2. At the same time, longitudinal magneti-
zation starts to recover.
Transverse relaxation follows an exponential
decay curve characterized by the time con-
stant T2, which is a measure of the speed of
spin dephasing.
The T2 constant is also tissue-specific and
contributes to the image contrast.

85
2 ECHOES, DECAY, AND RELAXATION

The spin echo

The magical moment: getting back the lost signal

The true decay of the FID 90°

We might expect the MR signal (FID) to

decay with the time constant T2. However,
the FID decays much more quickly, with a
shorter effective time constant T2* (see
figure). FID
The static magnetic field as felt by the spins T2
is not the same everywhere, in fact it is T2*
somewhat inhomogeneous. Unlike the
relaxation processes that lead to the T2
decay, we are dealing here with purely
static differences in the magnetic field that
remain constant over time within a specific
location. These are mainly spatial field var-
iations caused by the patient’s body as well
as technical inhomogeneities of the main
magnet.

86 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

T2* decay
Static magnetic field inhomogeneities add to the The U.S. physicist Erwin Hahn, born in
fanning of the spins: they dephase more quickly than 1926, discovered the spin echo in
with T2 relaxation. This is T2* decay. “The echo effect is 1950: the recovery of the MR signal
because T2* decay is reversible. Echoes
It seems as if the phase coherence of the spins is irre- brought about by can also be created by using inversion
vocably destroyed faster than T2 relaxation “allows.” subjecting the sample pulses different from 180 degrees,
also called “Hahn echoes.” Hahn
But this is not in fact the case, as demonstrated by the to two r-f pulses in originally used a pair of 90-degree
spin echo effect. succession. At time pulses.
(TE) after … the
second pulse the
echo signal appears.”
Erwin L. Hahn

87
2 ECHOES, DECAY, AND RELAXATION

180-degree flip

1
Spins flipped like an omelet 2

Since static magnetic field differences remain constant with

respect to space and time, we can cancel out their effect with
a “magic trick.” 3

We give an “about-face” command via a 180-degree pulse.

The 180-degree pulse apparently flips the spins just like an
omelet: the order of the spins is reversed, the direction of
precession remains, of course. 1
2
The faster spins (1), now placed behind, catch up with the 3
slower (3) precessing spins.

88 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

Erwin Hahn originally explained the

echo effect from the analogy of a
team of runners with different run-
ning speeds. During the race, the
group of runners fan out more and
more, seemingly in a random manner.
They are asked to turn around by 180
degrees and then continue running in
the opposite direction. In this way,
they will return together precisely at
the starting line. We are experiencing
an “echo” of the start.

But there is a catch: the spins don’t

turn around by a 180-degree pulse.
This would only happen by reversing
the main magnetic field. In fact, the
spins always precess in the same
direction, for example clockwise. That
is why we prefer the “flipped omelet”
analogy.

89
2 ECHOES, DECAY, AND RELAXATION

90° 180°
Here comes the echo
This is the effect of the 180-degree pulse: the out-of-phase
spins go back into phase, and a new MR signal is generated—  
the spin echo. FID
The 180-degree pulse is switched after the 90-degree pulse
after the run time . The spin-echo signal initially increases Spin echo
and reaches its maximum after double the run time (2). This
time span is known as the echo time (TE). The spin echo will
decrease after this time.
Since the FID begins to decay immediately after the 90-
degree pulse, it is difficult to measure its strength. For this TE = 2
reason, echoes are the preferred signals for imaging.

90 Magnets, Spins, and Resonances

 A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

90° 180°

Please note: The spin-echo signal decays with T2*, its

strength (amplitude, maximum), however, depends on
the T2 relaxation curve. TE

FID T2

T2*

91
2 ECHOES, DECAY, AND RELAXATION

REDUCED TO THE ESSENTIALS

The MR signal (FID) decays according to the

very short time constant T2*, smaller than
T2. The cause of this fast decay are static
magnetic field inhomogeneities. They force
spins to go out of phase more quickly.
We can recover the MR signal with a second
pulse, usually a 180-degree pulse. This MR
signal is known as the spin echo.
Rule of thumb:
T2* < T2 < T1

92 Magnets, Spins, and Resonances

A first acquaintance Longitudinal Transverse The spin
with relaxation recovery decay echo

93
3
From the Signal
to the Image
How do we generate an image from an MR signal that
shows spatial structures as different gray values?
3 FROM THE SIGNAL TO THE IMAGE

Slicing by gradients
Creating “zip codes” for locations in the body

The imaging question

In tomographic imaging, we create slice
images at specific positions of the human
body.
Remember that, in the magnetic field of the
MR scanner, a spatial distribution of longi-
tudinal magnetization is built up in the
body. If, for example, we simply excite all
the spins within the head with one RF
pulse, the averaged transverse magneti-
zation in the head would generate the MR
signal. We would not obtain any spatial
resolution.
The question remains: how can we distin-
guish between MR signals from different
locations in the body?

96 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

The trick with the gradients

As the magnetic field of an MR scanner is not perfectly Professor Lauterbur’s insight:
homogeneous, spins will precess at different frequen- why not make the proton spins from
cies. But only those spins that precess at the correct “The distribution of different locations precess at dif-
ferent frequencies on purpose?
Larmor frequency will resonate with the RF pulse. In magnetic nuclei such Instead of placing the body in a
order to make the field more homogeneous, a techni-
cal process known as ’shimming’ is applied.
as protons ... may be uniform magnetic field, we let the
field vary from one location to
obtained by imposing another. We can use the different
The revolutionary idea, which started MRI, was to
magnetic field gradi- frequencies that make up the mixed
’unshim’ the magnetic field in a controlled way: by MR signal as a “zip code” for cor-
switching gradients. ents ... and measuring
responding anatomical locations.
the intensities as func- Paul Lauterbur applied this idea for the
As a result, spins at different locations in the body will
precess at different frequen-cies, which is intended. tions of the applied spatial encoding of voxels within a
slice. The idea of slice selection by
magnetic field.” gradient fields was introduced by Sir
Peter Mansfield.
Paul C. Lauterbur
(1929—2007)

97
3 FROM THE SIGNAL TO THE IMAGE

On gradients

A gradient is an incline comparable with the

incline of a road. From a mathematical
point of view, a gradient defines the
strength and direction of a magnitude
changing in space.
Translated into MR technology this means:
a magnetic field gradient is a controlled
change in the magnetic field strength in a
certain direction, that is, a linear increase or
decrease.

98 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

Locally changing the field strength

In a homogeneous magnetic field, the field strength is the  B
same everywhere by defi-nition (B0). For this reason, all
proton spins show the same precession frequency 0 pro- nt
Gradie
portional to the field strength. The magnetic resonance is the  B0
same everywhere.
When a magnetic field gradient is applied, the field shows a
linear increase. The precession of the proton spins in this
direction varies accordingly. They precess more slowly in a weaker field as before stronger field
weaker field and more quickly in a stronger field. In sum they
show different resonance frequencies.

99
3 FROM THE SIGNAL TO THE IMAGE

Defining the slice position

Slice positioning and slice thickness

The slice-selection gradient

For slice selection, a gradient is switched in the z-direction,  B
for example, simultaneous-ly with the RF pulse. This gradient
is called the slice-selection gradient (GS).
Now the field has its original strength B0 at one location only  B0
z0. If the RF pulse had only one single frequency 0, it would
only excite the spins at the resonance location z0. This is the
selected slice position.
Z0
However, this does not really suffice. We would get a “slice,”
but without thickness. The slice would be paper thin and the Z
signal too weak, because only a few protons would be excited
in this thin area. What we need is a certain resolution in the The strength of the homogeneous static magnetic field is B0.
z-direction. And this is what we call slice thickness. The associated Larmor frequency of the proton spins is 0.

100 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

RF

Gs

How do we select the slice thickness? 

0
The RF pulse has a certain bandwidth of neighboring fre-
quencies about its center frequency 0. So it can excite
the desired spatial area of the slice thickness (z0). z0

As an alternative: the slice thickness can be modified by Z

keeping the RF pulse bandwidth constant while changing
the gradient strength. A steeper gradient ramp (a) excites
a thinner slice (za), a shallower gradient (b) excites a thicker
slice (zb).
Whatever technique is used: a slice is a defined resonance
area of the proton spins. Outside the slice, the spins are
not excited by the RF pulse. Transverse magnetization  a
(and therefore an MR signal) is generated only within the b
selected slice.


A slice is defined by switching a magnetic field

gradient.

Z
Z a

Z b

101
3 FROM THE SIGNAL TO THE IMAGE

This is how gradients are generated

Gradient coils
The MR system has three gradient coil arrangements for all
three spatial directions (x, y, and z), situated around the
magnet bore. The gradient coils do not generate a perma-
nent magnetic field, but are switched on briefly and multiple
times during the examination.
The gradient coils are operated via special power supplies,
known as gradient amplifiers.

102 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

Gradient performance
Gradient coil design The performance of a gradient system is characterized by the
for an MRI scanner minimum rise time required to obtain the maximum ampli-
tude (=gradient strength). The rise rate is calculated from
these two parameters. These characteristic data are also
known as the slew rate (SR) and allow for a quick comparison
of the performance of gradient systems.

103
3 FROM THE SIGNAL TO THE IMAGE

How do gradient coils work?

B
As a principle of physics, a magnetic field
B0
is created as soon as an electric current
flows through a conductor or a coil. When Z
the direction of current is reversed, the
orientation of the magnetic field changes
as well.
In MR, gradient coils in pairs are operated
in a specific direction, for example, the
z-direction, with the same current strength
but opposite polarity.
One coil increases the static magnetic field,
the opposing coil reduces it. This means that
the magnetic field with its original strength
B0 changes like the incline of a road.

104 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

Y oblique

double-oblique

Free slice positioning

coronal
The huge advantage of gradients in MRI:
Y
they allow us to position arbitrary slice
planes. axial

For sagittal slice positioning, we need to sagittal

switch the x-gradient, for coronal slice X
positioning, the y-gradient.
To obtain oblique slices, we have to switch
Z
two or three gradients simultaneously.
Their effect is superimposed.
A single oblique slice is generated by two
gradients, for example, in the z and y-direc-
tion. For a double-oblique slice, the system
switches all three gradients simulta-
neously.

105
3 FROM THE SIGNAL TO THE IMAGE

The gradient system: safety aspects

Noise Pacemakers
As with a loudspeaker, strong mechanical forces are gener- Pacemakers are critical with respect to gradient fields. The
ated by the gradient coils, resulting in knocking noises during control as well as the programming of pacemakers may be
the examination. These noises are attenuated by using suit- adversely affected by high-speed switching gradient pulses.
able measures, for example, earplugs or special sequences,
for example, whisper sequences or “quiet sequences.”

106 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

REDUCED TO THE ESSENTIALS

By switching gradients, we can

make spins precess at different
frequencies in a controlled
manner.
The slice-selection gradient
Physiological stimulation allows us to generate a slice,
a spatial region where proton
At certain thresholds for the rise time and amplitude of the spins will resonate. Outside the
gradient fields, the induced voltages may be large enough to slice, the spins are not affected
cause peripheral nerve stimulation. The muscle fibers con- by the RF pulse.
tract involuntarily. This is not hazardous to the patient's
health but may, however, be uncomfortable for the patient. By switching two or three gra-
dients simultaneously, we can
In the safety standards for MR systems, the maximum field position arbitrary slice planes.
changes are defined as a function of the switching time.
Normally, these threshold values are not exceeded by the
imaging methods used today. However, with some sequen-
ces, which use extremely fast gradient switching, for example,
EPI sequences, these thresholds can be exceeded. But for
safety purposes, gradient pulses are limited in routine
applications.

107
3 FROM THE SIGNAL TO THE IMAGE

The matrix: resolution

Encoding and decoding the “zip codes” that make up the image

Defining the image matrix

We have defined the slice position for an image. Now we
want to define the image matrix and its resolution. What
interests us: how do we get an image from a slice?
Let us assume we want to generate a tomogram in a 256256
matrix size. In this case, each row and column has to be
divided into 256 locations. We use the MR signal to create an
image with 256 different values, that is 65,536 voxels. This is
spatial resolution. 256

256

108 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

Following Monsieur Fourier

What is important here is: the image is not directly The French mathematician and physi-
generated from the measurement procedure. Instead, cist Jean-Baptiste Joseph Fourier deve-
by switching additional gradients and using some “A general problem loped his series of sine and cosine curves
while investigating the propagation
signal processing, raw data are generated first, in our consists in developing of heat. This mathematical tool helps
example 65,536 data points. The final image is com-
puted from the raw data.
any function whatso- to analyze and construct a multitude
of phenomena, including the MR
ever in an infinite signal.
These processes include some signal encodings and
series of sines and
mathematical transforms.
cosines...”
Let us follow these processes step by step.
Joseph Fourier
(1768–1830)

109
3 FROM THE SIGNAL TO THE IMAGE

On prisms and spectra

Almost all natural and technically-generated signals comprise The same applies to hearing: Most sounds consist of a multi-
a mixture of oscillations of different frequencies. tude of acoustical waves and pitches, which our sense of
hearing can differentiate, for instance, in a piece of music.
White light is a mixture of light consisting of different wave
lengths or frequencies. A prism helps us to differentiate the This spread across frequencies is called a spectrum. The
various frequencies contained in white light—in this case mathematical method for investigating such a sprectrum
the colors of a rainbow. is known as Fourier analysis.

110 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

We can build a given structure from the

“building blocks” of sine curves. The more
curves we use, the finer the results. The
curve showing the profile is the result of
32 superimposed sine curves : Fourier
synthesis.

111
3 FROM THE SIGNAL TO THE IMAGE

A stripe image
“Zip code” by frequencies
Fourier transform

The piano analogy

Imagine a piano keyboard, the white keys representing eight
different notes, i.e. frequencies, of an octave. Each key is
simultaneously hit with a different strength, resulting in a
mixture of notes. Can we, just by listening, know which notes
make up this audible experience, and what the respective
amplitudes are?
The MR scanner can. The MR signal is actually made to be
a mixture of the signals of all excited spins along, say, the
x-axis. At a resolution of 256 voxels, an echo includes not
only eight but 256 “notes” of different frequencies.
Frequency encoding

112 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

RF

GS

GF

This is frequency encoding

During the measurement of the echo, we switch a gradient in The Fourier transform allows us to determine the signal
the x-direction. What is going to happen? contribution of each frequency component (as shown in the
graphic by the height of the peaks). The individual frequen-
As you already know from gradients, the spin ensembles of
cies are reallocated to their location of origin along the x-axis.
the individual voxels precess along the x-axis at an increasing
The individual signal strength obtained determines the gray
frequency. This is known as frequency encoding. The
value of the allocated pixel.
associated gradient is known as the frequency-encoding
gradient (GF).
The echo is a mixture of signals. It contains many sine and
For the definition of a line of voxels, the frequency-
cosine waves. By knowing the gradient strength, we know
encoding gradient is switched.
the frequencies left and right and at all positions along the
gradient.

113
3 FROM THE SIGNAL TO THE IMAGE

Piling up stripes for a complete picture

“Zip code” by phases

This is phase encoding Phase

We cannot use the same frequency-encoding trick in the
y-direction to encode a two-dimensional image. Why not?
In this case, two different voxels could have the same fre-
quency and thus could not be differentiated. Obviously
we have to choose a different method.
During the time between the RF pulse and the echo, a gradient
is switched briefly in the y-direction. As a result, the spins
precess at different speeds for a short time. After the gra-
dient is switched off, the spins along the y-axis show differ-
ent phase shifts directly proportional to their locations.
This process is called phase encoding. The associated gra- Frequency
dient is the phase-encoding gradient (GP).

114 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

RF

GS ... ... ... 256 times

GF

GP

We can use the Fourier transform to filter out these phase 8

shifts as well. There is one slight drawback, however, it takes
more time. We have to generate 256 MR signals with differ-
7
ent phase encodings for 256 different locations. In other
words, this means 256 phase-encoding steps. This is why
the pulse sequence has to be repeated 256 times for a 6
256  256 matrix.
Thus, a signal matrix is filled line by line with the echoes
5
(shortened to eight in the graphic). The lines represent fre-
quencies, the columns represent phases.
4

For the definition of multiple lines of voxels, the 3

phase-encoding gradient is switched.

115
3 FROM THE SIGNAL TO THE IMAGE

Transforming to and from k-space

ky
A matrix of stripe patterns
The signal matrix is preprocessed, resulting in the raw-data
matrix, also known as k-space (a notion adopted from wave
physics). The axes (kx and ky) of the raw-data matrix desig-
nate so-called “spatial frequencies”. What do you think they
are?
kx
Just as temporal oscillations combine waves of different
frequencies (sines and cosines), so an image can be com-
posed from different spatial waves or stripe patterns.
The raw data simply determine whether and how intensively
a certain stripe pattern contributes to the image. A rough
stripe pattern (close to the center) shows low spatial fre-
quency. A fine stripe pattern (far from the center) shows high
spatial frequency.

116 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

ky

k-space Image space

An example will help to illustrate our case: by simply super-
imposing the horizontal and vertical stripe pattern shown,
we generate a complex gray value pattern. You can see that
a weighted superimposition of stripe patterns will result in
a more complex image. kx

This is exactly what, as the last step in image formation,

a two-dimensional Fourier transform does. It uses the
raw-data values in the k-space to calculate the gray value
distribution in the image, that is, the weighting of the many 256 Fourier transform 256
stripes. Subsequently, the associated gray value is assigned to
each pixel. And there you have your image.

The MR image is created by a 2D Fourier transform

from the raw-data matrix.

117
3 FROM THE SIGNAL TO THE IMAGE

Raw data versus image data

This is interesting: a data point in the raw-data matrix
(k-space) does not directly correspond to a pixel in the image.
Each part of the raw-data matrix contains information about
the whole image—comparable to a hologram.
The center raw data determine the rough structure as well as
the image contrast.
The outer portions of the raw-data matrix provide informa-
tion about margins, edge transitions, contours in the image—
in short they show finer structures and in the final analysis
determine the resolution.

Center raw data : structure and contrast

Outer raw data: resolution

118 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

REDUCED TO THE ESSENTIALS

The MR imaging technique does not create

the image directly. To localize the individual
voxels, phase encoding and frequency
encoding is used. For this, gradients are
switched.
A signal matrix is filled with the resulting
echoes and converted to a raw-data matrix,
known as k-space.
The MR image is computed from the raw
data by means of a final two-dimensional
Fourier transform.

119
3 FROM THE SIGNAL TO THE IMAGE

Introducing the pulse sequence

The complete picture

90° 180° 256 times

Understanding pulse diagrams
RF
As you will remember, a spin echo is created by a 90-degree
pulse, which generates the FID, followed by a 180-degree
pulse, which generates the spin echo in echo time TE. This is
a simple pulse sequence. GS

The pulse sequence is repeated with repetition time TR until

the raw-data matrix is filled with echoes. The number of GF
phase-encoding steps (that is, raw data lines) corresponds to
the number of repetitions of the sequence. The scan time is GP
determined to a large degree by the resolution of the image
in the phase-encoding direction.
Scan time = NP  TR
(NP : number of phase-encoding steps) FID
Spin echo

TE
TR

120 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

90° 180° 256 times

Slice encoding
RF
The slice-selection gradient GS is switched simultaneously
with the 90-degree pulse.
What does the additional bar pointing downward GS mean? GS
The gradient dephases the spin phases along the slice thick-
ness. We need to compensate for this with a gradient of
opposite polarity and half the duration (rephasing gradient). GF

During the 180-degree pulse, the slice-selection gradient

is switched again so that the 180-degree pulse only affects GP
the spins of the previously excited slice.

FID
Spin echo

TE
TR

121
3 FROM THE SIGNAL TO THE IMAGE

90° 180° 256 times

Phase encoding
RF
As you will remember, a phase-encoding gradient super-
imposes a different phase on spins from different locations,
for example along the y-direction. The phase-encoding
gradient GP is switched on briefly between slice selection GS

and spin echo.

GF
For a matrix consisting of 256 columns and 256 rows,
gradient switching of the spin-echo sequence is repeated
256 times with repetition time TR—with the phase-encoding GP
gradient increasing step by step.
The phase-encoding steps in the pulse diagrams are frequently
represented by a multitude of horizontal lines in the bar.
These represent the different gradient step amplitudes—
FID
positive or negative. Spin echo

TE
TR

122 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

90° 180° 256 times

Frequency encoding
RF
During the spin echo, the frequency-encoding gradient GF is
active. Since the spin echo is read out during this time, this
gradient is also known as the readout gradient.
GS
If we only applied the readout gradient, the spins precessing
in the direction of frequency encoding would begin to
dephase. During echo time TE, the spins would be fully GF
dephased, leaving us without a spin echo. However, we can
circumvent this problem by applying a dephasing gradient GP
before the 180-degree pulse. This makes the readout gra-
dient rephase the spins in a way that causes the spins in the
center of the readout interval to be in phase again at the time
of the maximum spin echo.
FID
When we switch on the dephasing gradient before the 180-
Spin echo
degree pulse, the gradient has the same polarity as the read-
out gradient. The 180-degree pulse reverses the phase of the
spins. TE
TR

123
3 FROM THE SIGNAL TO THE IMAGE

The principle of MR imaging

By switching gradients, we obtain the signal mixture for
a slice image in two steps:
• We only excite spins within a certain slice (slice selection).
• Subsequently, we acquire a 2D scan matrix via frequency
and phase encoding in the slice.
With the help of two-dimensional Fourier transform, the
MR system reconstructs the MR image from the raw data
measured.

124 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

125
3 FROM THE SIGNAL TO THE IMAGE

Meet the gradient echo

Spin echo’s counterpart

Gradient pulses again

Let us recap the effect of a magnetic field gradient. When  B
a gradient pulse is switched in a specific direction, spins will
precess in this direction with linearly decreasing and increas- nt
Gradie
ing frequencies.  B0
As a result, by using a gradient pulse we can destroy the free
induction decay faster than it would if undisturbed.

weaker field as before stronger field

126 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

A simply different echo

RF
A gradient pulse (–) directly after the RF pulse artificially
dephases the spin frequencies. Since they are now preces-
sing at different speeds (faster on one side, slower on the
opposite side), they lose their phase more quickly, that is, GF +
they are being dephased. The FID is eliminated considerably –
faster than it would under normal conditions.
When the polarity of the gradient is reversed (+), the preces- FID
sing spins will be in phase again, that is, they are rephased. T2*
We measure an echo during the rephasing of the FID. Since
this echo is generated by a gradient, it is called the gradient
echo. Gradient echo

TE

127
3 FROM THE SIGNAL TO THE IMAGE

Only a little time for the echo time

The echo time TE is typically shorter for a gradient-echo
sequence than for a spin-echo sequence. Why is that?
The 180-degree pulse is omitted in gradient-echo technology.
This means that we do not cancel the static T2* dephasing
mechanism as we do in spin-echo technology. Instead, we
use gradient pulses to quickly destroy the FID and build it
back up again, all within the T2* decay.
The echo time for a gradient echo has to fit into the T2* time.
This is why the gradient-echo technique is faster than the
spin-echo technique.

128 Magnets, Spins, and Resonances

 Slicing The matrix: Introducing the Meet the
by gradients resolution pulse sequence gradient echo

REDUCED TO THE ESSENTIALS

A gradient echo is generated by switching

gradient pulses of reversed polarity.
The echo time has to be short because the
gradient echo can be generated only within
the T2* decay.
The gradient-echo technique is typically
faster than the spin-echo technique.

129
4
A Study of
Contrasts
MR imaging is unique in the way it can control image
contrast and therefore expand the diagnostic range.
The art of MR application lies in the choice of pulse
sequences and the combination of acquisition
parameters.
4 ECHOES, DECAY, AND RELAXATION

Spin echoes and contrast weighting

Demonstrating the three most important types of contrast in
MR imaging: T1 contrast, T2 contrast, and proton density contrast

What determines image contrast?

How do we obtain an image with the largest possible
contrast between different tissue types? Different tissue
types have different transverse magnetizations. Where the
signal is strong, the image shows bright pixels; weaker
signals result in darker pixels.
What determines the signal strength? Clearly to a large
degree, the proton density in the respective voxel: the
greater the number of protons contributing to the mag-
netization, the stronger the signal.
But even more important for medical diagnostics is the effect
of the two relaxation constants T1 and T2 on the image
contrast.

132 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

90° 180° 90°

Important parameters: TE and TR 256 times
Remember the spin-echo sequence? A 180-degree pulse is
applied at time  after a 90-degree RF pulse, a spin echo is  
generated after echo time TE = 2. FID
This pulse sequence, 90 degrees—180 degrees has to be
repeated until all phase-encoding steps of the scan matrix
have been acquired (for example, 256 times). The time inter-
Spin echo
val between the repetitions is called repetition time TR.
TE and TR are important parameters for controlling the con-
trast of a spin-echo sequence. Let us follow these two tempo- TE = 2
ral parameters and see how they affect image contrast. TR

133
4 ECHOES, DECAY, AND RELAXATION

Proton-density contrast
Based on differences in the number of protons per voxel

The maximum value of magnetization

The diagram shows three different tissue Mz Mxy
types (1, 2, 3) with different relaxation 1
2
times.
3
Longitudinal relaxation begins immediately
after the 90-degree pulse. The longitudinal
magnetization Mz of the three tissue types
recovers at different speeds. Their maxi-
mum values correspond to the proton
densities, that is, the number of hydrogen 500 3000 TR [ms] 15 60 90 120 TE [ms]
protons per volume unit.
By means of a repeated 90-degree pulse
after repetition time TR, the actual longitu-
dinal magnetizations are converted into
TR long TE short
transverse magnetizations Mxy and gener-
ate signals of different strengths. PD contrast

134 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

If we select a sufficiently long repetition time TR, the differ- Proton-density contrast:
ence in signal in the tissue after a repeated 90 degree pulse TR long (2,500 ms),
depends mainly on the proton density of the tissue because TE short (15 ms).
of the nearly complete longitudinal relaxation. The greater the proton
density of a tissue type,
Should we decide to generate echoes very shortly after the the brighter it appears in
repeated 90-degree pulses, that is, with a shorter echo time the PD image.
TE, we will obtain a proton density-weighted image (PD for
short).
In actual application, the TR of a spin-echo sequence is rarely
longer than two to three seconds. However, this also means
that tissue types with longer T1 constants, for example, CSF,
will not have recovered completely after this time period.

135
4 ECHOES, DECAY, AND RELAXATION

T2 contrast
Based on the speed of transverse relaxation

Signal strength depends on T2 decay

Let us stay with the long repetition time TR. Mz Mxy
What happens when we also select a long 1
echo time TE? 2

3
The signal curves decrease due to T2 re-
laxation and even might begin to intersect.
The proton density contrast is lost. With
longer echo times, the curves begin to
diverge and the contrast is controlled by the
T2 relaxation. We obtain a T2-weighted 500 3000 TR [ms] 15 60 90 120 TE [ms]
image.
The signal strength of the spin echo typi-
cally depends on the T2 decay.

TR long TE long

T2 contrast

136 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

The comparison of images shows T2 contrast with increasing

echo time TE.
As the echo time increases, at some point the proton density
no longer influences contrast. The T2 contrast depends
strongly on the TE selected. The optimal TE of a T2-weighted
image is the mean value of the T2 constants of the tissue
to be displayed (in our case between 80 ms and 100 ms).
If the echo time is too long (last image), transverse magneti-
zation decays to a level where the signal of some tissue types 60 ms 90 ms
disappears in the unavoidable signal noise.
Image comparison with
respect to T2 contrast:
TR long (2,500 ms), TE is
increasing. CSF with a
long T2 appears bright in a
T2-weighted image.

120 ms

137
4 ECHOES, DECAY, AND RELAXATION

T1 contrast
Based on the speed of longitudinal relaxation

Signal strength depends on T1 build-up

What happens when we select a short Mz 1 Mxy
repetition time TR so that the T1 relaxation 2
is not complete. The signals will be much 3
weaker and the contrast will decrease rap-
idly with increasing echo time. For this
reason, we have to select the shortest pos-
sible echo time TE.
A short TR reduces the effect of the proton
densities, a short TE cancels the effect of
T2 relaxation. The difference in signal 500 TR [ms] 15 60 90 120 TE [ms]
strengths depends largely on the previous
longitudinal magnetizations, that is, from
the T1 relaxation of the tissue. We obtain
a T1-weighted image.
TR short TE short
T1 contrast

138 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Image comparison with respect to T1 contrast: TR short (500 ms) ;

TE is increasing. CSF with a long T1 appears dark in a T1-weighted
image. The optimal TR corresponds approximately to the average
T1 constant of the tissue type to be displayed. This means between
400 ms and 600 ms for 1.0 to 1.5 tesla.

The image comparison shows good T1 contrast when TR and

TE are both short.
With longer echo times, both the T1 contrast and the measur-
able signal are reduced. The combination of short repetition
time and long echo time is obviously completely unsuitable.
Normal soft tissue types differ only slightly in their proton
density. However, they do show different T1 relaxations. For
this reason, T1-weighted imaging is highly suitable for anato-
mical displays. 15 ms 60 ms

90 ms 120 ms

139
4 ECHOES, DECAY, AND RELAXATION

REDUCED TO THE ESSENTIALS

TR
The comparison of images shows the three
important combinations of TR and TE as
well as their resulting contrast weighting:
• T1 contrast (TR short, TE short)
long
• T2 contrast (TR long, TE long)
• Proton density contrast (TR long, TE short)
In spin-echo imaging, the effects of T1 and
T2 are reversed: tissue with longer T1 PD T2
appears darker in the T1-weighted image,
tissue with longer T2 appears brighter in
T2-weighted images.

short

T1

short long TE

140 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

141
4 ECHOES, DECAY, AND RELAXATION

Contrast with inversion recovery

A spin-echo sequence with a 180-degree preparation pulse

180° 90° 180° 180°

Inversion time determines contrast
The inversion recovery sequence (IR) is a method for creating
a signal depending primarily on T1. The IR sequence applies
pulses of 180 degree—90 degree—180 degree. The longi- FID
tudinal magnetization is first flipped by the 180 degree
preparation pulse in the opposite direction—that is, it is Spin echo
inverted. Transverse magnetization is therefore zero and we
TI TE
do not receive an MR signal.
TR
The interval between the 180-degree pulse and the 90-degree
stimulation pulse is known as inversion time TI. During this
time period, the longitudinal magnetization recovers.
The 90-degree excitation pulse converts the momentaneous
longitudinal magnetization into transverse magnetization.

142 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Strong T1 contrast 90° 180° 180° 90°

While the standard spin-echo sequence

provides for T1 contrast and in particular an
excellent T2 contrast, the inversion recov-
ery sequence is known for its higher
T1 contrast. Mz Mz

As the longitudinal magnetization relaxes

its negative value following inversion, the
magnetization of different tissue types
reaches zero at different times. The inver-
sion of the magnetization improves the
dispersion of these TI curves, leading to
better T1 contrast. By selecting a suitable
inversion time TI, the contrast is optimized.
TE TI
We can use IR sequences to display the
most minute T1 contrasts, for example, in
the brain of a newborn. The disadvantage is
the longer measurement time. Depending
on TI, fewer slices are measured than with
a T1-weighted spin-echo technique.

143
4 ECHOES, DECAY, AND RELAXATION

The zero point

Making tissue disappear

Gray on gray and the zero signal 180° 90°

Let us look at the curves of the longitudinal

relaxation for a special case. Because of the
TI selected, the faster relaxing tissue (a) has Mz

already passed the zero point, while the a

slower relaxing tissue (b) has not.
This can be very confusing if only the mag- b
nitude of the signals is used for image con-
trast. No differentiation is made between Zero
the positive and negative longitudinal
magnetization. Tissue types with widely
different T1 constants would be displayed
with the same gray value.
TR

144 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Image comparison with respect to contrast with inversion recovery:

TI is increasing. The signal of white matter decreases as the inversion
time TI increases and goes through the zero point at TI = 300 ms.
At TI = 400 ms, the signal of gray matter (with longer T1) has reached
zero point, while the signal of white matter is increasing again.

The comparison of images shows the effect of inversion time

TI on the contrast in the brain. The signals from white or gray
matter may disappear.

100 ms 200 ms

300 ms 400 ms

145
4 ECHOES, DECAY, AND RELAXATION

True inversion recovery

T1 contrast all the way

Taking into account the sign of

the magnetization 180° 90°

How can we ensure contrast between dif-

ferent tissue types? By considering the
Mz
orientation of the longitudinal magneti-
zation. a

The positive and negative longitudinal

magnetizations are converted by the
b
90-degree excitation pulse into transverse
magnetization with a 180-degree phase
shift. If we consider both the magnitude as
well as the phase difference of the signals,
it is possible to allocate the signals to the
original positive or negative longitudinal
magnetization. This will ensure maximum
T1 contrast.

146 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

This technique of phase-sensitive reconstruction gives Image background,

the true longitudinal magnetization and is also known as usually black, is shown
in mid-range gray
true inversion recovery. Its preferred application is in the
when phase-sensitive
area of pediatrics.
reconstruction is used.

147
4 ECHOES, DECAY, AND RELAXATION

Contrast with gradient echoes

Increased signal and shorter measurement times with
a reduced flip angle

Shortening the repetition time without signal loss

There are different types of gradient-echo sequences applied
Z
in clinical practice. A major benefit of a gradient-echo
sequence is the possibility to decrease repetition time TR.
The magic “trick” is the ’low-angle’ tech-nique: a reduced flip M

angle  of less than 90 degrees. In this case, we do not have Mz

the effect of the entire available magneti-zation M in the 
xy-plane, but rather only part of it is converted into a
transverse magnetization Mxy.
On the other hand, the longitudinal magnetization is not zero
after such an -pulse, but continues to have a reduced
magnitude Mz. Mxy

148 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Boost in signal
For example, an RF pulse with a flip angle of 20 degrees already
generates a sufficiently high transverse magneti-zation of
34 % of its maximum value. In this case, the remaining
longitudinal magnetization reaches 94 % of its maximum
value.
At the next pulse, a high longitudinal magnetization is avail-
able again. At very short repetition times (less than T1), after
several 20-degree pulses, the magnetization will reach a
’steady state’. The generated MR signals are even larger than
with 90-degree pulses at the same short TR. This boost in
signal allows for very short repe-tition times and a greatly
reduced measurement time.
The contrasts generated with a gradient-echo sequence
rather than with a spin-echo technique are very complex.

149
4 ECHOES, DECAY, AND RELAXATION

The optimal flip angle

Balancing the flip angle between T1 and TR

Ernst angle and the steady state

For a tissue type with a specific T1, a maximum signal is Mz
generated at a defined flip angle, the so-called Ernst angle.
 = 90°
This optimal flip angle is a function of the selected repetition
time TR.
 < 90°
We know that the longitudinal magnetization recovers in
proportion to its size (exponential growth). By flipping it by
an angle  the remaining longitudinal magnetization is
smaller than before (at 20 degrees, it is 94 % of 94 %, etc.).
But it also recovers accordingly faster. After repeated -pulses,
an equilibrium is obtained between the opposing tendencies; TR
the longitudinal magnetization remains the same after each
pulse. This equilibrium is also known as steady state.

150 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Calculating the Ernst angle

Let us calculate the optimal flip angle, for example, in Swiss MR researcher Richard R. Ernst
the brain: With T1 = 800 ms and a TR of 2500, an angle (Nobel prize 1992): “I am not aware of
of 87 degree results which is not very different from “The optimum angle any other field of science outside of
magnetic resonance that offers so
90 degrees. But with a TR of 100, the optimal flip angle for producing the much freedom and opportunities for
is only 28 degress.
maximum magnetiza- a creative mind to invent and explore
For proton-density weighted images, an angle smaller new experimental schemes that can
tion in the xy-plane be fruitfully applied in a variety of
than the Ernst angle is desirable.
is easily found. The disciplines.”
average optimum flip
angle is given by
cos  = exp(TR /T1)”
Richard R. Ernst

151
4 ECHOES, DECAY, AND RELAXATION

FLASH
Spoiling the transverse magnetization on purpose

Exploiting the steady state of longitudinal

magnetization  

The FLASH gradient-echo sequence (Fast Low Angle Shot)

uses the equilibrium of longitudinal magnetization.
With a very short repetition time TR, we have a residual trans- GS
verse magnetization prior to each emitted -pulse. This
remaining magnetization would contribute to the next ac-
GF
quired signal, in a FLASH sequence it is typically eliminated
by a strong “spoiler” gradient pulse.
GP
RF spoling can also be done by applying randomized phases
of the subsequent -pulses.
FID

Gradient echo

TE
TR

152 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Contrasts with FLASH

FLASH allows for very strong T1 contrasts using an extremely
short TR.
• T1 contrast:
TR short (40–150 ms)
TE short (5–10 ms)
 medium to large (40–80)
• T2* contrast:
TR long (500 ms)
TE relatively long (18–40 ms) T1 T2*
 small (5–20)
Image comparison showing
• Proton density contrast:
FLASH contrasts
TR long (500 ms)
TE short
 small (5–20)

PD

153
4 ECHOES, DECAY, AND RELAXATION

FISP
Utilizing a uniform transverse magnetization

Refocused gradient echoes  

With the FISP sequence (Fast Imaging with Steady-state Pre-
cession), the remaining transverse magnetization is not elim-
inated before the next -pulse. Instead, it contributes to the
GS
signal along with the longitudinal magnetization.
To obtain a uniform transverse magnetization, the dephas-
ing gradients in the phase-encoding direction (GP) are GF
compensated after the echo by applying gradients of reversed
polarity. The magnetization is flipped in the opposite direction GP
after each repetition time TR with a negative -pulse. Subse-
quent -pulses have alternating polarity.
FID

Gradient echo

TE

TR

154 Magnets, Spins, and Resonances

 Spin echoes and Contrast with Contrast with
contrast weighting inversion recovery gradient echoes

Contrast with FISP

The strength of the longitudinal magnetization depends on T1, Contrast with 3D FISP
the amplitude of transverse magnetization depends on T2*.
Contrast is a function of the ratio of T1 to T2* and is generally
not dependent on TR.
• T1 /T2* contrast:
TR short
TE short
 medium
Repetition time TR should be as short as possible. A long TR T1 /T2*
makes FISP behave like FLASH.

155
5
The Strategy of
Image Quality
Image quality is most important in MR imaging. The
secret lies in optimizing image quality in relation to
the measurement time required.
5 THE STRATEGY OF IMAGE QUALITY

Contrast, signal, and noise

The most important criteria for image quality: a strong signal,
low noise, good contrast, as well as sufficient resolution.

Let us start with a strong signal

The MR image consists of a multitude of image pixels. Each
pixel has a specific gray value. The pixels in the image repre-
sent the individual volume elements (voxels) in the slice Pixel
measured.
During measurement, each voxel in a slice contributes to the
overall MR signal.
Voxel
First and foremost, the signal strength per voxel depends on
the quantity of signal-generating protons, i.e. the local
proton densities.
256

The more protons contribute to magnetization, the

stronger the signal.

158 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

SA Difference in signal

SB

Contrast
This is how contrast is generated
Hyperintense pixels in the image represent stronger signals,
weaker signals result in hypointense pixels. For simplicity‘s
sake, our example shows a single row of eight pixels.
The contrast in the image is the relative difference in the
signal strength between two adjacent tissue types, A and B.
In other words, contrast is equal to the difference in signal:
Contrast  Difference in signal  SA  SB
Contrast resolution is the ability to distinguish differences in
grayscale colors of adjacent structures. Each type of tissue
emits individual signal strengths. This allows for anatomical Image comparison: Low T1-contrast (left), high T1-contrast (right)
differentiation in the image and, in the final analysis, differ-
entiation between pathalogical and healthy tissue.

159
5 THE STRATEGY OF IMAGE QUALITY

Signal versus noise

Noise is unavoidable but has a strong adverse affect
on image quality

Noise

If noise takes over

The NOISE in the image appears as a grainy,
random pattern similar to snow on a TV Signal
screen. It represents statistical fluctuations
in signal intensity that do not contribute to
image information. There are basically two
sources for this effect.
Noise is generated throughout the human
body through Brownian motion of mole-
cules. To that we add the electronic noise of
the receiver technology.
We are faced with a problem when the
signal from a slice is too weak. In this case,
the signal may be “washed over“ perma-
nently by noise.

160 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

The signal-to-noise ratio

An important criterion for MR image quality is the relation-
ship between the intensity of the wanted signal and the
statistical noise, the signal-to-noise ratio (SNR) :

Signal
SNR = ----------------
-
Noise

A higher SNR means a less grainy image.

Image comparison:
The two images on the left were
acquired in the same way and
subsequently subtracted from
one another (= pixel-by-pixel
difference in gray values).
What remains is background
noise (on the right).

161
5 THE STRATEGY OF IMAGE QUALITY

Amplifying the signal

A strong signal as the first step toward good image quality

Increasing slice thickness SNR

SNR ~ d
Let us assume we enlarge the voxel by measuring a thicker 3
slice. As a result, signal intensity increases, since more
protons are contributing to signal strength. The show- 2
stopper is: The portion of noise remains the same since it is
not just coming from the slice, but rather overall from the
patient’s entire body (or more precisely, from the sensitive 1
volume of the receive coil). Also: the thicker the slice, the
stronger the signal. And the higher the SNR.
1 2 3 d [mm]

The SNR is directly proportional to the voxel size.

162 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Disadvantages: Increasing slice thickness reduces spatial

resolution in slice direction. This may result in partial volume
effects that distort image results (for example, bones pro-
truding into soft tissue).
Please note: Doubling the matrix size to im-prove the in-
plane resolution will decrease the voxel size by a factor of
four. Consequently, SNR will be decreased by a factor of four
as well.

Image comparison: The figure on the right shows a slice that is three
times as thick as the slice in the left image. As a result, the SNR is
three times as high.

163
5 THE STRATEGY OF IMAGE QUALITY

SNR
Increasing the number of acquisitions SNR ~ ACQ
We do not have to select an undue large slice thickness. The 2,8
SNR can be improved by other methods as well: by measuring
2
one slice several times (several acquisitions) and by adding
the results in a single image. 1,4

However, the SNR is not increasing linearly, it is rather getting 1

to be less.

1 2 4 8 ACQ

SNR is proportional to the root of the number

of acquisitions.

For example: If four acquisitions are measured and averaged

on one slice, the total SNR is twice what is was before.

164 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Disadvantage: The measurement time increases as the

number of acquisitions increases.

Image comparison: left 1 measurement, right 4 measurements.

Results: The SNR on the right is double the SNR on the left.

165
5 THE STRATEGY OF IMAGE QUALITY

The effective contrast

A quality criterion that relates to what we see in the image

Combining contrast and noise

The contrast we actually see and evaluate in the image is
more than differences in signal intensity: This effective
contrast is also related to the noise level.
A high signal-to-noise ratio (SNR) alone does not guarantee
easy differentiation between two structures in an image.
The contrast-to-noise ratio (CNR) in an MR image is the
difference between the signal-to-noise ratios between two
tissue types, A and B:
CNR  SNRA  SNRB
Since CNR equals effective contrast, it is a better quality Image comparison: good CNR, poor CNR
criterion than SNR.

166 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

When contrast is too noisy

SA Difference in signal
Let us assume we have a noticeable difference in signal
between two tissue types A and B. In this case, we could
obtain good contrast. However, if we set this difference in SB
signal in relation to high noise, the contrast drowns in noise.
A B
Our example: Although the difference in signal is higher in
the first than in the second case, the CNR and consequently
the effective contrast is lower.
To obtain good image quality, the difference in signal be- SA Difference in signal
tween two types of tissue has to be significant despite the
noise. SB
A B

167
5 THE STRATEGY OF IMAGE QUALITY

REDUCED TO THE ESSENTIALS

SNR and CNR are important criteria for

MR image quality. They relate signal and
contrast to noise.
The signal strength is in part determined by
the proton density in the respective voxel.
The more protons contribute to magneti-
zation, the stronger the signal.
Contrast is the difference in signal strengths
between two relevant tissue types.
The signal-to-noise ratio (SNR) describes
the relationship between the intensity of
signal and noise. SNR can be improved by
increasing both the slice thickness as well as
the number of acquisitions.
The difference of the signal-to-noise ratios
between two types of tissue is the contrast-
to-noise ratio (CNR). It is a better quality
criterion compared to SNR.

168 Magnets, Spins, and Resonances

Contrast, signal, About image size Speeding up
and noise and resolution the measurement

169
5 THE STRATEGY OF IMAGE QUALITY

About image size and resolution

Controlling the spatial resolution of the structures acquired
in the image

The matrix determines resolution Y

The matrix size determines resolution and measurement

time. If you halve the number of rows (the phase resolution),
you also halve the measurement time, since the number of
time-consuming phase-encoding steps is also reduced by
50%. If you use twice the phase resolution, the measurement
time accordingly increases by 50%.

Measurement time  phase resolution 

 TR (repetition time)  number of acquisitions
256 X

Example: For a phase resolution of 256 sampling points,

500 ms TR and one acquisition, the measurement will be
completed in 128 seconds.

170 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Matrix size and signal-to-noise ratio

The matrix size also affects the signal-to-noise ratio. Because
with larger voxels, more protons contribute to the signal.
If the measurement matrix is enlarged without changing the
other parameters, resolution will increase accordingly. The
voxels become smaller and thus SNR is reduced as well. SNR
is proportional to the size of the voxel. This means that at a
constant slice thickness, SNR is proportional to the pixel size.

Matrix Relative SNR Image comparison: Matrix 128 (left) low resolution and better SNR,
128 4.0 Matrix 256 (right) higher resolution and decreased SNR
256 1.0
512 0.25

171
5 THE STRATEGY OF IMAGE QUALITY

The field of view

Optimizing in-plane resolution and measurement time

What is the field of view?

The field of view, abbreviated to FOV, is the square or rect-
angular image area to be measured (in mm). In short, the
FOV determines what you can see in the MR image.
To save time and to obtain maximum resolution, the FOV is
optimally adjusted to the area under examination.

Image comparison: FOV  230 mm (left),

FOV  330 mm (right) is unnecessarily large

172 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Pixel size and in-plane resolution

The smaller the field of view at a fixed matrix size, the higher
the in-plane resolution. The number of pixels per in-plane
unit increases while the pixels themselves decrease in size.
Conversely, for a given matrix size and larger FOV, the pixels
are greatly enlarged. The resolution is reduced accordingly.
FOV
Pixel size = -----------------------------
-
Matrix size

Smaller pixels mean an improved in-plane

resolution. FOV 256 mm FOV 350 mm
Matrix 256x256 Matrix 256x256

Right image is magnified to show the same size of the head. Pixels
FOV (mm) Matrix size are larger, resolution is lower.
256 256256
256 128128
128 128128

173
5 THE STRATEGY OF IMAGE QUALITY

REDUCED TO THE ESSENTIALS

Matrix size, field of view (FOV) and slice thickness affect

resolution as well as measurement time and the signal-to-
noise ratio.
Changing these parameters has a number of effects. The
most optimal solution is a compromise—primarily between
image quality and measurement time.
Summary of the effects of the parameters:

Measurement time Resolution SNR

Matrix
FOV 
Slice thickness 

174 Magnets, Spins, and Resonances

Contrast, signal, About image size Speeding up
and noise and resolution the measurement

175
5 THE STRATEGY OF IMAGE QUALITY

Speeding up the measurement

Reducing measurement matrix and field of view for a faster
measurement

Y KY
Decreasing phase resolution
To accelerate the measurement, you can
select a reduced measurement matrix with
a lower phase resolution, for example, in-
stead of 256256 you select 128256. The
pixels are now rectangular.
128 128
Why is the measurement faster? The phase KX
resolution of the measurement matrix cor-
responds to the number of phase-encoding
steps (NP). This means it is directly propor-
tional to the measurement time (NPTR). X
256 256
A halved phase resolution (for example,
128) corresponds to half the number of
phase-encoding steps. The measurement
time is halved.

176 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Filling the raw-data matrix

As the reduced measurement matrix is rectangular, signals
are acquired only for the central raw data lines. The missing
outer rows of the square k-space are filled with zeroes.
Why does this work? As you might remember, fine structures
are shown in the outer regions of k-space. The central rows
provide the important contrast. When an image is recon-
structed from the raw data, the image pixels are interpolated
in the phase-encoding direction.
Image resolution is reduced with phase resolution (for exam-
ple, phase resolution reduced by 50%, image resolution by
50% in this direction). Since the voxels are larger, SNR is Image comparison:
improved. Phase resolution 100% (left) and 50% (right)

177
5 THE STRATEGY OF IMAGE QUALITY

Cutting the field of view

Cutting the FOV in phase-encoding direction accelerates
image acquisition

KY
Rectangular field of view
If the object to be measured does not fill a
256
square image, we can select a rectangular
field of view. If we cut halve the FOV in the
phase-encoding direction, we only need
half as many phase-encoding steps. Only ...
every other row is filled with raw data, the 256
KX
others contain zeroes only.
3
The measurement time is directly pro-
portional to the number of phase-encoding
1
steps. For this reason, the measurement
time is reduced by 50% with half the FOV. 128 Phase

178 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Measuring faster at the same resolution

At half the FOV and half the number of phase-encoding
steps, the voxel size remains unchanged and so does the
resolution.
SNR is decreased.
So a rectangular FOV is an acceptable choice for accelerating
image acquisition.

Image comparison:
FOV Phase 100% (left) and 50% (right)

179
5 THE STRATEGY OF IMAGE QUALITY

Measuring only part of the raw data

Utilizing the symmetry of k-space

Y KY
Half-Fourier
With the half-Fourier technique, only half of 132
the raw data matrix (k-space) is filled with
.
data in the phase-encoding direction. The
missing data are reconstructed sym- .
metrically. .
256 4 KX
Unavoidable small magnetic field inhomo- 1
geneities lead to phase distortions. For this
reason, slightly more than half of the
phase-encoding steps are acquired for
phase correction.
As a result, the measurement time is almost 256 X 256
cut to half.

180 Magnets, Spins, and Resonances

 Contrast, signal, About image size Speeding up
and noise and resolution the measurement

Partial raw data

The partial Fourier technique works in the same way as the
half-Fourier technique: Only part of the k-space is filled in the
phase-encoding direction (5/8, 6/8, or 7 /8).
What is the image quality like? Because the voxel size has not
changed, the resolution has not deteriorated. SNR decreases.
In most cases, there will be a barely discernable difference
between images with and without half- or partial Fourier.

Image comparison: Normal (left) and half-Fourier (right)

181
5 THE STRATEGY OF IMAGE QUALITY

Summary

The measurement time is shortened by reducing the phase

resolution or both the phase resolution and the field of view.
The examples provided demonstrate the effects on SNR and
resolution:

SNR Resolution
Field of view 100% Better Less
Phase resolution 50%
Field of View 50% Less Unchanged
Phase resolution 50%
Half Fourier Less Unchanged

182 Magnets, Spins, and Resonances

Contrast, signal, About image size Speeding up
and noise and resolution the measurement

183
6
Increasing
Diagnostic Value
Since the invention of MRI, progress has never stopped.
Let us introduce some small, some big, some “old” and
some new techniques and innovations that help to
enhance the diagnostic value of MRI.
6 INCREASING DIAGNOSTIC VALUE

Packing echoes and slices

A multiecho experience
Multiple echoes with different degrees of T2 weighting

90° 180° 180° 180°

Echoes following one another
We have discussed the generic pulse
sequence for image acquisition in MRI.
There is no reason to generate only one TE
echo within the repetition time TR. In a
multiecho sequence, several 180-degree
pulses follow each other in sequence, thus FID T2 1 2 3
several spin echoes with different degrees T2*
of T2 weighting are generated.
Signal height of a multiecho sequence
reduces with transverse relaxation: The
longer the echo time, the smaller the echo.
The echo time-dependent signal decay
follows the T2 relaxation time of the tissue.
We can repeat this until transverse mag-
netization is irrevocably decayed.

186 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

Creating a pure T2 image

The echo time-dependent drop in signal of a multiecho
sequence can be used to calculate a pure T2 image, a so-
called T2 map.

187
6 INCREASING DIAGNOSTIC VALUE

Multislice imaging
The standard in MRI


Measuring several slices at once
1
In order to cover a body region, we can measure a series of 3 2
4
slices in 2D, or alternatively a 3D cube (see next section). Z1
Z2
The linear slice-selection gradient of a pulse sequence allows Z3
Z4
us to distinguish different slices which can be excited by RF
pulses in quick succession. Z

As echo time TE is always considerably shorter than repe-

tition time TR, we can excite additional slices (for example, z1
through z4). by using a multislice sequence.

188 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

90° 180° 90°

The multiecho sequence uses a series of 90-degree and 180- 1

90° 180°
degree pulses with varying resonance frequencies that fit
with the defined slice positions. This method provides us with 2
90° 180°
all the slices necessary for examining a particular area or
region during a measurement. 3
90° 180°
These slices do not need to be in straight axial orientation. By 4
switching the gradient fields properly, we can scan slices with
arbitrary orientations, even varying across slices.
TR

189
6 INCREASING DIAGNOSTIC VALUE

Volume imaging: 3D
Creating spatial views from a region of interest

Generating 3D data
Fast 3D pulse sequences with short repe-
tition time TR allow acquisition of three-
dimensional data. The entire measurement
volume, the 3D slab, is excited and not just
single slices.
Different phase positions can be accurately
located in space. This is the basic principle
of phase encoding. In order to generate 3D
data, we superimpose an additional phase- z
encoding gradient in the direction of slice
selection (“z” in our example).
Through additional phase encoding per-
pendicular to the image plane and conti-
guous images, we obtain information
regarding the defined slab. The planes of
this volume are known as partitions.

190 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

Creating spatial views

With the help of standard post-processing software available,
the 3D data sets generated can be used to create spatial
views (for example, from the vessels).

Spatial image reconstructed from 3D data, a Maximum

Intensity Projection (MIP)

191
6 INCREASING DIAGNOSTIC VALUE

REDUCED TO THE ESSENTIALS

A mulitecho sequence generates mul-

tiple echoes by applying several excita-
tion pulses in sequence, This is possible
as long as T2 relaxation continues.
Multislice imaging (2D): A multislice
sequence creates a series of slices for
a region of interest.
Volume imaging (3D): Fast 3D pulse
sequences allow acquisition of 3D data
sets from which spatial views can be
created.

192 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

193
6 INCREASING DIAGNOSTIC VALUE

Take the echo train: Turbo spin echo

Shorter measurement times and improved tissue contrast

The emergence of turbo spin

echo
Conventional spin-echo techniques can
take several minutes per slice. Turbo spin-
echo sequences (TurboSE, TSE) consider-
ably shorten acquisition time and have
replaced spin-echo techniques to a large
extent. The use of TurboSE imaging has
become routine in MRI today.

Image comparison: T2 spin echo T2 turbo spin echo

194 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

Turbo spin echo is based on RARE

imaging, invented by the German
clinical MRI researcher Jürgen Hennig
in 1986: “The world (and the world of
MR in particular) is full of ’impossible’
things which turned out to become
reality.”

Strong in T2 contrast
In the majority of cases, TurboSE sequences
are used for T2-weighted imaging. The
most noticeable difference between
TurboSE and spin-echo techniques is the
bright fat signal in strongly T2-weighted
images. A T1-weighted TurboSE technique
is possible also, frequently used for imaging
the spine.
TurboSE sequences offer a far better con-
trast between white and gray matter. It is
difficult to imagine neuroradiological imag-
ing without the high-resolution possibilities
offered by TurboSE sequences.

195
6 INCREASING DIAGNOSTIC VALUE

The turbo factor defines speed 90° 180° 180° 180° 180° 180° 180° 180°
TurboSE is a fast multiecho sequence:
During the time it takes for a spin-echo
sequence to acquire just one single echo, T2
1 2 3 4 5 6 7
the TurboSE sequence acquires an entire
series of echoes, known as the echo train.
Each echo of the echo train is given a
different phase encoding (GP) and fills one FID
row of the raw-data matrix. The length of
the echo train determines the maximum
time savings or turbo factor. Example:
At a turbo factor of 7, the TurboSE sequence
measures 7 times faster than a spin-echo
sequence with comparable parameters. GP

The central echo, when the phase-encod-

ing gradient is zero, determines the image TEeff
contrast. The time interval between the
90-degree pulse and the central echo is the
effective echo time TEeff.

196 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

36 1 Segment

36 Rows = Number of excitations

1 2 3

Segmented raw-data filling ky

TurboSE works because of the ’segmented k-space’ tech-

nique: Within the repetition time TR, an entire series of raw-
data rows is acquired, whereas with the conventional tech-
nique, only one raw-data row is acquired.
The raw-data matrix (k-space) consists of, for example,
7 segments (= turbo factor) of 36 rows each. The total num-
ber of rows is a whole-number multiple of the echo train
length (7  36 = 252). kx

This means that within each echo train, one raw-data row is
filled for each segment, creating a “comb” of filled-in rows.
This “combing” has to be repeated 36 times, as shown by our
example.

7 Echoes

197
6 INCREASING DIAGNOSTIC VALUE

A further development of TurboSE

techniques involves the combination
with an inversion pulse (Turbo Inver-
sion Recovery, TIR), with half-Fourier
imaging (Half Fourier Acquired Single
Shot Turbo Spin Echo, HASTE), or the
addition of gradient echoes (Turbo
Optimizing parameters for TurboSE Gradient Spin Echo, TurboGSE). Fastest
The longer the echo train at a fixed TR, the shorter the meas- turbo spin-echo techniques use a
complete echo train of 256 or more
urement time. As a result, fewer slices can be acquired. At the
echoes (single-shot TurboSE or RARE,
same time, T2 decay is stronger, which reduces the reso-
Rapid Acquisition with Relaxation
lution in the phase-encoding direction, especially when Enhancement).
tissue with a short T2 is examined.
To ensure detection of small hemorrhages, for example, in
the brain, a longer TR and a higher resolution are used for
contrast improvement. The turbo factor may be reduced, for
example, from 15 to five, but all in all this is still a significant
increase in speed.

198 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

REDUCED TO THE ESSENTIALS

A TurboSE sequence accelerates image

acquisition by gener-ating a series of spin
echoes for each excitation known as the
echo train.
The raw-data matrix (k-space) is seg-
mented. If an echo train of, for example,
15 echoes (= turbo factor) is used, only
17 excitation pulses are required. The result
is a significantly reduced acquisition time.
The contrast of TurboSE sequences is
dominated by T2-weighting.

199
6 INCREASING DIAGNOSTIC VALUE

Echo-planar imaging (EPI)

Acquiring an entire image in one “shot”

From line scanning to plane scanning

In most of the conventional acquisition techniques discussed
so far, like spin-echo and gradient-echo techniques, the raw-
data matrix (k-space) is filled line by line with the generated
echoes.
Echo-planar imaging (EPI) is a “single-shot” method. By that
we mean that an EPI sequence uses a single excitation pulse
to acquire an entire image, i.e. the whole 2D plane by gener-
ating an echo train of a sufficient number of gradient echoes
to fill the raw data matrix.

EPI image

200 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

An order of magnitude faster

EPI is one of the fastest techniques for acquiring MR Echo-planar imaging was introduced
images. The acquisition time for one slice is the repe- in 1977 by Peter Mansfield (Nobel
prize in 2003).
tition time, i.e. one TR period, a fraction of a second. With EPI, “the speed
The EPI sequence is repeated not for the phase- of image formation
encoding steps, but for the number of slices chosen.
can be increased by
EPI is the method of choice for diffusion and cranial an order of magni-
perfusion as well as for functional neuroimaging (BOLD
tude or so over that
imaging).
of the single line-
scanning method of
imaging.”
Sir Peter Mansfield

201
6 INCREASING DIAGNOSTIC VALUE

This is how EPI works

RF
A readout gradient is switched in bipolar
fashion. Within the FID, it generates an
entire echo train of ascending and descend- GF
ing gradient echoes with alternating alge-
braic signs. The number of gradient echoes
is the EPI factor. GP

The fast T2*decay of the FID leaves only

approximately 100 ms to generate the T2*
echoes. For this reason, the readout is
FID
limited to between 64 and 256 echoes.
The EPI matrix is therefore between 64  64
and 256  256, accordingly the EPI factor is
between 64 and 256. TEeff
The effective echo time TEeff coincides with
the maximum signal.

202 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

kY kY
GRE EPI
Conventional Meandering
To move to the next raw-data row, the 1 1
phase-encoding gradient is switched 2
2
briefly between the individual gradient
echoes (blips). The raw-data matrix is . 3
sampled in a “zig-zag” pattern, i.e. filled . .
in a meandering manner. kX
. kX .
This is how EPI sequences acquire .
diagnostic images in as little as 50 to .
100 ms. These images are completely 128
void of motion artifacts, making EPI
especially suitable for examining dyna-
mic processes or generating diffusion-
weighted images, which are sensitive to
motion on a molecular level.

203
6 INCREASING DIAGNOSTIC VALUE

Combining EPI for different contrasts

EPI is essentially a readout module, a kind of “add-on” to
a pulse sequence. The EPI method can be combined with
freely-selectable preparation pulses (spin echo, inversion
recovery, etc.). This allows us to obtain different contrasts
with EPI sequences. Since the echoes decay with T2*, the
images contain a T2*-weighting component that varies with
the basic contrast. As a single-shot procedure, EPI does not
show any T1 contrast.
EPI-FID sequences generate good T2* contrast that increases
with the echo time.
EPI spin-echo sequences can be compared to conventional Image comparison: Strong diffusion contrast (left), weak diffusion
spin-echo sequences with an infinitely long TR. A long T2 contrast (right)
generates sharp images. For tissue with a short T2, the image
may not be as clear.
EPI diffusion sequences add additional diffusion gradients.
They are sensitive to molecular motion and show the diffu-
sion of water in tissue. EPI acquisitions have the advantage of
freezing motion that would create artifacts in conventional
sequences, obscuring the diffusion contrast.

204 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

REDUCED TO THE ESSENTIALS

Echo-planar imaging is one of the fastest A further development of Echo-planar

imaging is segmented EPI, a multi-
acquisition techniques. It uses a single
shot technique. The raw-data matrix is
exitation pulse to acquire an entire image
sampled segment by segment.
within a fraction of a second: The whole
plane in one “shot”. The complete echo RESOLVE is a special sequence with
train is created by a “blipped” gradient segmentation in readout direction.
pulse. The EPI factor (for example, 128)
defines the increase in speed.
The EPI method can be combined with
preparation pulses, allowing for different
contrasts (EPI SE, EPI diffusion, etc.).

205
6 INCREASING DIAGNOSTIC VALUE

Ultrahigh fields
Welcome to clinical 7T

Magnetic field strength translates to clinical outcome

In MRI, higher field strengths of 7T or more provide a higher
signal-to-noise ratio, enabling increased spatial and spectral
resolution. Acquisition times can be reduced, and certain
contrast mechanisms (e.g., susceptibility-weighted imaging,
SWI) are enhanced as field strength increases.
The primary reason to increase the field strength is to im-
prove the MR signal. The magnetization M increases pro-
portionally with the strength of the magnetic field.
The Larmor frequency is proportional to the field strength.
At 7 tesla, proton spins will precess at a frequency of
approximately 300 MHz. The corresponding wavelength is
comparable to the dimensions of the human body. As we
will see later, this has consequen-ces for the design of the
pulse sequences and the coil setup.

High-resolution SWI of the brain

206 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

More than double SNR, better contrast

With higher field strength, the signal increases and the Contrast-to-noise ratio
signal-to-noise ratio (SNR) is improved. Compared to 3 tesla, maps in resting state
with a 7 tesla scanner we receive more than the double SNR MRI: low values (cold
which means a significantly improved image quality. colors) for 3 tesla (top),
higher values (warm
Some relaxation constants at 7T are dif-ferent to 3T. For some colors) for 7 tesla
applications, this results in an increased contrast. BOLD (bottom)
imaging, for example, benefits from an increased SNR and an Consortium
increased BOLD contrast simultaneously. For some appli- The Human Connectome
Project. CMRR, Minnesota,
cations, averaging is no longer necessary, and acquisition USA; Washington University
times are signifantly reduced. St.Louis, USA; Oxford
University, UK

207
6 INCREASING DIAGNOSTIC VALUE

Submillimeter resolution
The stronger MR signals of a more powerful scanner can be
used to acquire images in the submillimeter range, making it
possible to visualize previously unseen structures.
Let us assume we reduce the voxel size. As a result, the signal
intensity decreases, since less protons are contributing to
signal strength.But this is compensated by the higher field
strength. At 7T, it is possible to acquire high-resolution mor-
phological MR images up to 0.2 mm in-plane resolution.

T2 TSE of the head with 0.2 mm in-plane resolution

208 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

Multichannel transmit and receive

Challenges associated with higher fields include increased
inhomogeneity of the RF field (B1 inhomogeneity) and the
static magnetic field (B0 inhomogeneity). Additionally, at 7T,
the RF energy absorbed by the human body, quantified by the
specific absorption rate (SAR), is increased. Furthermore,
other unwanted effects, such as susceptibility artifacts, scale
with the field strength.
Increased B1 inhomoheneity can be addressed by using
multichannel parallel transmit (pTx). Due to the flexible
exploitation of the degrees of freedom in the pulse design,
the RF field can be homogenized over a predefined region
and the SAR values can be significantly decreased.
Susceptibility effects can be addressed by using parallel
imaging and a higher number of B0 shim terms (e.g.,
3rd order shim terms). At higher field strength, a higher
acceleration factor can be achieved, allowing the imaging of
challenging body regions.

Parallel transmit (pTx) enables homogeneous excitation and

extended coverage

209
6 INCREASING DIAGNOSTIC VALUE

Clinical benefits of higher fields

High-resolution imaging allows to visualize anatomical
details and functional information, for example, for ana-
lyzing tissue metabolism. This enables better lesion con-
spicuity. Some examples from many applications:
• Depiction of small structures not visible at lower field
strengths
• More power for diffusion MRI
• Visualizing subcortical activations with BOLD fMRI
• Innovative metabolic brain mapping
• Exploring physiology with multinuclear MRI
• Ultra-high resolution venography and cartilage imaging
• High-resolution neuroimaging and MSK morphology
imaging
• Parallel transmit technology enabling robust homogeneity
and extended coverage

TOF angiography with 0.3 mm isotropic resolution

210 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

Coronal PD TSE with 0.1 mm in-plane resolution

211
6 INCREASING DIAGNOSTIC VALUE

Hybrid imaging
Combining different imaging modalities (MR-PET)

What is PET?
PET (Positron Emission Tomography) is an imaging technique
for localizing tissue with high metabolic activity which is
often an indication of tumors. The tissue can be indirectly
localized through metabolite accumulation of a radioactive
tracer.
As the tracer decays, it emits a positron (the antiparticle of an
electron). The positron loses energy through interaction with
surrounding tissue. Depending on the endpoint energy of the
tracer or isotope utilized, the positron will travel a certain
distance until it annihilates with an electron. Subsequently,
two gamma rays, consisting of high-energy photons, are
emitted in nearly opposite directions. If a detector pair
registers both back-to-back photons within a certain time
window, a ’line of response’ (LOR) is formed between both
events and registered as a ’coincidence’.
From a large number of coincidence events, three-dimen-
sional images of the tracer distribution in the body can be
Detection of photon emission
calculated.

212 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

PET (on the left), fused, and MR image (on the right)
MR-PET advantages ZEMODI, Bremen, Germany

Let us compare PET with MRI. As the generation of PET sig-

nals is based on events of positron annihilation, the resulting
images are comparatively noisy and of lower spatial reso-
lution.
However, PET provides metabolic details which cannot be
generated with less sensitive MR imaging. Thanks to better
localization and higher resolution, MR brings structural and
additional functional information to the image.
In this respect, the combination of PET and MR is highly
complementary. Despite the exposure to ionizing radiation,
no lack of diagnostic information occurs.

213
6 INCREASING DIAGNOSTIC VALUE

MR-PET team play

How do these modalities work together?
MR can affect PET performance, due to the high magnetic
and radiofrequency fields. MR image quality, however, can be
impaired by either radiofrequency noise introduced by the
PET electronics or magnetic field inhomogeneities.
One approach is sequential acquisition: PET and MR images
are processed separately and aligned afterwards. In this case,
there is no mutual interference. However, sequential imag-
ing holds the risk of involuntary patient motion in between
the two examinations and, therefore, may increase the
chance of local misalignment. In addition, the examination is
time-consuming.
This disadvantages are compensated for by acquiring PET and
MR images simultaneously during identical (functional)
states.
Whole-body images: MR (on the left), attenuation corrected PET,
fused MR/PET (on the right)
ZEMODI, Bremen, Germany

214 Magnets, Spins, and Resonances

Packing echoes Take the echo train: Echo-planar Ultrahigh Hybrid
and slices Turbo spin echo imaging (EPI) fields imaging

PET signal attenuation

The signal of emitted photons is weakened by material or
tissue. For example, RF coils, the patient table, and the
patient (soft tissues, air, or bones) weaken the PET signal.
Therefore, acquired PET data must be corrected for artifact-
free PET imaging. This is done with specifically post-pro-
cessed MR images known as attenuation correction (AC)
maps. These AC maps are used to create corrected PET
images.
While attenuation-corrected images are generally more
reliable, the correction process itself leads to artifacts. As a
result, both corrected and uncorrected images are always
reconstructed together. For appropriate diagnosis, the
uncorrected images have to be checked as well.

Whole-body PET images: uncorrected (on the left), attenuation

corrected (on the right)
ZEMODI, Bremen, Germany

215
References
 REFERENCES

A. Abragam. The Principles of Nuclear Magnetism. Clarendon W. Gilbert. De magnete. Short, London, 1600. engl: On The
Press, Oxford, 1961 Loadstone and Magnetic Bodies. Quaritch, London, 1893

F. Bloch. Nuclear Induction. Phys. Rev. 70:460, 1946 E. L. Hahn. Spin Echoes. Phys. Rev. 80(4):580, 1950

S. Brandt, H . D. Dahmen. The Picture Book of Quantum E. L. Hahn. Free Nuclear Induction. Physics Today 6(11):4,
Mechanics. 4th Edition. Springer, New York, 2012 1953

R. W. Brown, Y. N. Cheng and E. M. Haacke. Magnetic J. Hennig, A. Nauerth and F. Friedburg. RARE Imaging: A Fast
Resonance Imaging: Physical Principles and Sequence Imaging Method for Clinical MR. Magn. Reson. Med. 3:823,
Design. 2nd Edition. Wiley & Sons, Hoboken, NJ, 2014 1986

C. Cohen-Tannoudji, B. Diu and F. Laloë. Quantum J. Hennig. Ultra high-field MR: useful instruments or toys for
Mechanics. 2 Volume Set. Wiley-VCH, Weinheim, 1992 the boys? Magn. Reson. Mater. Phy. 21:1, 2008

A. H. Compton. The magnetic electron. J. Franklin Institude D. I. Hoult, B. Bhakar. NMR Signal Reception: Virtual Photons
192(2):145, 1921 and Coherent Spontaneous Emission. Concepts Magn.
Reson. 9:277, 1997
R. Damadian. Tumor Detection by Nuclear Magnetic
Resonance. Science 171:1151, 1971 J. Larmor. A Dynamical Theory of the Electric and
Luminiferous Medium. Part III. Philos. Trans. R. Soc. Lond.
M. J. Dawson. Paul Lauterbur and the Invention of MRI. MIT 190:205, 1897
Press, Cambridge, MA, 2013
P. C. Lauterbur. Image Formation by Induced Local
R. R. Ernst. Nuclear Magnetic Resonance Fourier Transform Interactions: Examples Employing Nuclear Magnetic
Spectroscopy. Nobel Lecture, 1992 Resonance. Nature 242 :190, 1973

R. R. Ernst, W. A. Anderson. Application of Fourier Transform P. C. Lauterbur. All Science is Interdisciplinary—From

Spectroscopy to Magnetic Resonance. Rev. Sci. Instrum. Magnetic Moments to Molecules to Men. Nobel Lecture,
37:93, 1966 2003

M. Faraday. Experimental Researches in Electricity. Phil. M. H. Levitt. Spin Dynamics: Basics of Nuclear Magnetic
Trans. R. Soc. Lond. 122:125, 1832 Resonance. 2nd Edition. Wiley & Sons, Chichester, 2008

J. B. J. Fourier. Théorie de la chaleur. Didot, Paris, 1822. engl.: Z. P. Liang and P. C. Lauterbur. Principles of Magnetic
The Analytical Theorie of Heat. Cambridge University Press, Resonance Imaging: A Signal Processing Perspective. IEEE
Cambridge UK, 1878 Press, New York, 1999

218 Magnets, Spins, and Resonances

P. Mansfield, P. K. Grannell. NMR Diffraction in Solids? J. Phys. Picture credits:
C: Solid State Phys. 6:L422, 1973
Page 10–11: Bar magnet with iron filings, Awe Inspiring
Images, fotolia.com
P. Mansfield. Snap-Shot MRI. Nobel Lecture, 2003
Page 25–25: Roulette wheel in motion, fergregory,
P. Mansfield. The Long Road to Stockholm: The Story of fotolia.com
Magnetic Resonance Imaging. Oxford University Press, 2013
Page 184–185: Landeanflug, Stefan Lenz, fotolia.com
Interview with Sir Peter Mansfield. Nobel Media AB 2014.
http://www.nobelprize.org/mediaplayer/index.php?id=554

M. Munowitz. Coherence and NMR. Wiley & Sons, New York,

1988

E. M. Purcell, H. C. Torrey and R. V. Pound. Resonance

Absorption by Nuclear Magnetic Moments in a Solid. Phys.
Rev. 69:37,1945

E. M. Purcell. Research in Nuclear Magnetism. Nobel Lecture,

1952

I. I. Rabi, S. Millman, P. Kusch and J. R. Zacharias. The

Molecular Beam Resonance Method for Measuring Nuclear
Magnetic Moments. Phys. Rev. 55:526, 1939

I. I. Rabi, N. F. Ramsey and J. Schwinger. Use of Rotating

Coordinates in Magnetic Resonance Problems. Rev. Modern
Phys., 26(2):167, 1954

V. M. Runge, W. R. Nitz, M. Trelles and F. L. Goerner. The

Physics of Clinical MR Taught Through Images. 3rd Edition.
Thieme, Stuttgart, 2014

C. P. Slichter. Principles of Magnetic Resonance. 3rd Edition.

Springer, Heidelberg, 1990

S. I. Tomonaga. The Story of Spin. University of Chicago Press,

1997

219
Index
 INDEX

D
180 degree pulse, 44 Dephasing, 127 FISP, 154
3D imaging, 190 Diffusion sequences (EPI), 204 FLASH, 152
90 degree pulse, 44 Flip angle, 44

E FOV (Field Of View), 172

A Echo time TE, 90 Matrix size, 173

Active shielding, 61 Echo train, 196 Rectangular FOV, 178

Addition Effective echo time TEeff, 196, 202 Free induction decay, 49

Vectors, 14 Elementary magnets, 8 Frequency, 26

Arrow Ensemble, 21 Bandwidth, 101

Depiction of vectors, 14 EPI (Echo-planar imaging) Frequency encoding, 113

Atomic particles, 30 Diffusion sequences, 204 Frequency-encoding gradient GF, 113

FID sequences, 204 Fringe field, 60

B Spin echo sequences, 204

Bandwidth, 101 EPI factor, 202 G

Ernst angle, 150 Gradient amplifier, 102

C Exponential processes, 68 Gradient coils, 104

Center raw data, 118 Gradient echo, 127

CNR, 166 F GS (Slice-selection gradient), 100, 121

Components Faraday cages, 63

Vectors, 15 FID, 49 H
Contrast, 132, 159 Field gradient Half-Fourier

Contrast-to-noise ratio, 166 Magnetic, 98 k-space, 180

Coordinate system, 15 Field lines, 10 SNR, 181

XY-plane, 34 Field of View (refer to FOV as well), 172 Hertz, 26

Z-axis, 34 Field strength, 10 Homogeneous magnetic field, 10

222 Magnets, Spins, and Resonances

I
Inhomogeneous magnetic field, 86 Matrix, 170 Phase-encoding gradient GP, 114, 122
Inversion recovery sequence, 142 Reduced matrix, 176 Pixels
Inversion time TI, 142 Matrix size, 108 Pixel size, 173
MR signal, 48 Precession, 17

K Multislice sequence, 188 Preparation pulse, 142

k-space, 116, 177 Mxy (Transverse magnetization), 45 Proton density, 32, 134

Half-Fourier, 180 Mz (Longitudinal magnetization), 45

Partial-Fourier, 181 R
N Raw data, 109

L Noise, 160 Along the margin, 118

Larmor frequency, 28 Nuclear spin, 30 Center, 118

Longitudinal magnetization Mz, 45 Raw data matrix, 115

Longitudinal relaxation, 74 O Readout gradient, 123

Oblique slices, 105 see also Frequency-encoding

gradient, 113
M Relaxation, 66
Magnetic field P Repetition time TR, 120, 133
Homogeneous, 10 Partial volume effect, 163
Rephasing, 127
Inhomogeneous, 86 Partial-Fourier, 181
Resonance, 41
Static, 10 Partitions, 190
RF pulse
Magnetic field gradient, 98 parts per million, 33
180 degree, 44
Magnetic field lines, 10 Phase, 27
90 degree, 44
Magnetic field strength, 10 Phase encoding, 114
Rise rate, 103
Magnetization, 22 Phase encoding steps, 115

223
 INDEX

S U
SAR (specific absorption rate), 62 Spin ensemble, 21 Ultrahigh field, 206
Shielding Spins
Active, 61 Dephasing, 127 V
Signal, 158 Rephasing, 127 Vectors
Contrast-to-noise ratio, 166 SR (slew rate), 103 Addition, 14
Noise, 160 Static magnetic field, 10 Components, 15
Signal difference, 167 Steady state, 150 Depiction, 14
Signal intensity, 162 Voxel, 21
Signal-to-noise ratio, 161 T
Single-shot method, 200 T1 (Longitudinal relaxation), 74 X
Sinusoidal curve, 26 T1-weighted image, 138 XY-plane
Slab, 190 T2 (Transverse relaxation), 81 Coordinate system, 34
Slew rate, 103 T2-weighted image, 136
Slice, 101 TE (Echo time), 90 Z
Slice position, 100 TEeff (Effective echo time), 196, 202 Z-axis
Slice thickness, 100 TI (Inversion time), 142 Coordinate system, 34
Slices Time constant, 70
Oblique, 105 TR (Repetition time), 120
Slice-selection gradient GS, 100, 121 Transverse magnetization Mxy, 45
SNR, 161 Transverse relaxation, 81
Spatial frequency, 116 True inversion recovery, 147
Specific absorption rate (SAR), 62 Turbo factor, 196
Spin echo, 90

224 Magnets, Spins, and Resonances

225
Manufacturer Siemens Healthineers Headquarters
Siemens Healthcare GmbH Siemens Healthcare GmbH
Henkestr. 127 Henkestr. 127
91052 Erlangen 91052 Erlangen
Germany Germany
Phone: +49 9131 84-0
siemens-healthineers.com

Print No. MR-07000G.643.03.01.02 / © Siemens Healthcare GmbH, 2023