chapter 7

Resonance — Nuclear
Magnetic Resonance
GUSTAVO MARRERO CALLICO

CHAPTER 7 1
 Overview
 Principles of nuclear magnetic resonance (NMR) as a
method to generate internal body signals.
 NMR is particularly focused on detecting water content
and properties within biological tissues.
 The technique relies on the resonance behaviour of
protons (hydrogen nuclei) when placed in a strong
magnetic field.
 Radiofrequency (RF) waves are used to interact with
these protons, causing them to resonate.
 This resonance produces signals that can be detected and
analysed to infer tissue characteristics.
 Non-ionizing technique, ideal for soft tissues.

CHAPTER 7 2
 Nuclear Magnetic Resonance
 Resonance-based imaging is likely the least familiar method
for generating internal body signals.
 The focus is on Nuclear Magnetic Resonance (NMR) imaging,
although the term “nuclear” is typically omitted in clinical
contexts.
 The name “NMR” itself reveals the core principles:
• Nuclear: Signals originate from atomic nuclei, not radioactive decay.
• Magnetic: Requires strong magnetic fields to manipulate nuclear
behaviour.
• Resonance: Uses radiofrequency waves to exploit the resonance of
nuclei.
 NMR is widely used in chemical analysis to study:
• Types of nuclei in a sample.
• Interactions between nuclei via chemical bonds.

CHAPTER 7 3
 Nuclear Magnetic Resonance

 However, NMR does not provide spatial information

about where nuclei are located.
 This chapter is divided into signal generation, and image
formation.
 A basic NMR experiment involves:
• Transmitting an RF wave into a sample within a magnetic field.
• Recording the re-emitted signal from the sample.
 Although NMR operates at the atomic level, its principles
can be explained using classical (Newtonian) physics.
 Quantum theory is more accurate but not essential for
the scope of this chapter.

CHAPTER 7 4
 Microscopic Magnetization

 Nuclei with odd atomic weight or atomic number possess a

quantum property called spin.
 Spin can be visualized as the charged nucleus rotating
around an internal axis.
 The number of energy states for a spinning nucleus is
determined by the formula: 2S + 1, where S is the spin
quantum number.
 For nuclei with S = 1/2, there are two
energy states:
• +1/2 (parallel) – aligned with the magnetic
field (low energy)
• −1/2 (anti-parallel) – opposed to the magnetic
field (high energy)

CHAPTER 7 5
 Microscopic Magnetization
 A spinning nucleus behaves like a tiny dipole magnet,
generating a microscopic magnetic field.
 The magnetic moment vector (μ) is defined by:
𝝁𝝁
𝝁𝝁 = 𝛾𝛾𝜱𝜱
• 𝛾𝛾 : gyromagnetic ratio (cte.). Depens on the particle.
• 𝜱𝜱: angular momentum (rotation of the particle)
 The strength of this magnetic moment depends on
the external magnetic field, measured in Tesla (T).
 Gyromagnetic ratios are often expressed in MHz/T
and vary by nucleus type.
 Hydrogen nuclei (protons) are most commonly
used in imaging due to their abundance in the
human body.

CHAPTER 7 6
 Microscopic Magnetization

Common gyromagnetic ratio for spin 1/2 systems

CHAPTER 7 7
 Microscopic Magnetization
 Other nuclei like ¹³C may be used in research but are less
common in physiological systems.
 In biological tissues, water is the main source of protons,
making it the primary target for signal generation in
NMR/MRI.
 The term “spins” is frequently used to refer to these hydrogen
nuclei in imaging contexts.

CHAPTER 7 8
 Precession
Proton Precession in a Magnetic Field
 When a proton is placed in an external magnetic field (B₀), it
begins to precess or “oscillate" around the axis of that field.
 This precession is a secondary motion that occurs in addition to
the proton’s intrinsic spin about its own axis.
 The spin axis of the proton is not aligned with the direction of
the magnetic field, leading to a tilted rotational behaviour.
 The resulting motion is a combination of
spin and angular displacement, forming
a circular path around the magnetic field
axis.
 This behaviour is analogous to a
spinning top, which wobbles as it
rotates due to gravitational torque.

CHAPTER 7 9
 Precession
Precession of protons in an applied magnetic field B0
In the absence of a magnetic field (left), orientation of the protons
magnetic polarity is random. When a field B0 is applied, the protons
precess around the direction of the applied field, either aligning with
(lower energy) or against (higher energy) the direction of the field.
Individual protons will not be in phase with each other

Absence of a magnetic field Presence of a magnetic field B0

CHAPTER 7 10
 Precession

 The rate of precession is described by the Larmor equation:

ω₀ = γ · B₀
• Where:
o ω₀ is the Larmor frequency (angular frequency of
precession, in radians/second),
o γ is the gyromagnetic ratio, a constant specific to
each type of nucleus,
o B₀ is the strength of the external magnetic field.
 The Larmor frequency determines how fast the proton
precesses and is critical for tuning radiofrequency pulses in MRI.
 Understanding this motion is essential for signal generation in
NMR/MRI, as it underpins how energy is absorbed and emitted
by nuclei.

CHAPTER 7 11
 Macroscopic Magnetization
 Without an external magnetic field, nuclear spins in a
material are randomly oriented, resulting in no net
magnetization.
 When an external magnetic field (B₀) is applied, spins begin
to align, creating a macroscopic magnetization.

 The magnetic field is

typically represented as:
B₀ = B₀ · ẑ
where B₀ is the field
strength (in Tesla) and ẑ is
the unit vector in the +Z
direction.

CHAPTER 7 12
 Macroscopic Magnetization
 In a spin-½ system, each nucleus can align in one of two
ways:
• Parallel to B₀ (+ẑ) – lower energy state (“up” direction). μ has a
component in +ẑ.
• Anti-parallel to B₀ (−ẑ) – higher energy state (“down” direction). μ has
a component in -ẑ.
 The x–y orientation (phase) of
spins around z-axis is random,
so there is no net transverse
magnetization.
 A slight preference for the low-
energy (parallel) state leads to
a small net magnetization in
the z-direction.

CHAPTER 7 13
 Macroscopic Magnetization

 Macroscopic magnetization (M) represents the collective

magnetic behaviour of many individual nuclear spins. It is
defined as:

𝑴𝑴 = � 𝝁𝝁𝒏𝒏 = � 𝛾𝛾𝜱𝜱 = 𝛾𝛾 · � 𝜱𝜱 = 𝛾𝛾 · 𝐉𝐉

where,
• γ is the gyromagnetic ratio,
• J is the bulk angular momentum,
• μ is the microscopic magnetic moment,
• n is the number of spins.
 Due to the random x–y orientation of individual magnetic
moments (μ), there is no net transverse magnetization (i.e., M
has zero x and y components).

CHAPTER 7 14
 Macroscopic Magnetization
 Over time, if the sample is left undisturbed, the system
reaches equilibrium magnetization (M₀) aligned with the external
magnetic field (B₀).
 The magnitude of equilibrium magnetization is given by:
𝐵𝐵0 𝛾𝛾2 ħ2
𝑀𝑀0 = · 𝑃𝑃𝐷𝐷
4𝑘𝑘𝑘𝑘
 where:
• B₀ is the magnetic field strength (Tesla),
• γ is the gyromagnetic ratio,
• ħ is the reduced Planck’s constant (or Dirac constant),
• k is the Boltzmann constant,
• T is the absolute temperature,
• PD is the proton density (number of mobile nuclei per unit volume).
 Higher B₀ or PD leads to greater M₀, which is useful for enhancing
signal strength and tissue contrast in MRI.
 Measuring M₀ can provide insights into proton density, helping
differentiate between tissue types.
CHAPTER 7 15
 Macroscopic Magnetization
 Boltzmann’s equation describes the ratio of protons in each
energy state:
𝑁𝑁𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎−𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 −
Δ𝐸𝐸
= 𝑒𝑒 𝑘𝑘𝑘𝑘
𝑁𝑁𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

• where ΔE = γ · ħ · B₀ is the energy difference between states.

 This equation shows that more protons align in the lower-
energy (parallel) state than in the higher-energy (anti-parallel)
state.
 Even in a strong 3 Tesla field, the difference is small—only
about 10 more protons per million in the lower-energy
(parallel) state.
 Despite the small imbalance, the large number of protons
ensures sufficient magnetization for signal detection.
 This magnetization effect also exists in the Earth’s magnetic
field but is much weaker and harder to detect.

CHAPTER 7 16
 Transverse Magnetization

 Initially, only a static magnetic field (B₀) aligned along

the z-axis has been considered.
 This causes nuclear spins to precess around the z-axis at
the Larmor frequency, resulting in:
• A net magnetization (M) aligned with B₀, with no transverse
component (i.e., no x or y components).
• A longitudinal component M0.
 But, when a time-varying magnetic field B(t) is applied,
the magnetization vector M(t) experiences a torque.

CHAPTER 7 17
 Transverse Magnetization
 This torque is described by the equations:
𝑑𝑑𝑱𝑱(𝑡𝑡)
= M(t) × 𝑩𝑩(𝑡𝑡) → change in angular momentum.
𝑑𝑑𝑑𝑑
𝑑𝑑𝑴𝑴(𝑡𝑡)
= 𝛾𝛾M(t) × 𝑩𝑩(𝑡𝑡) → change in magnetization over time.
𝑑𝑑𝑑𝑑

 Where, J(t) is the bulk angular momentum

 These equations assume a short time interval and rely on
the relationship between magnetization and angular
momentum.

CHAPTER 7 18
 Transverse Magnetization

 If B(t) = B₀ ẑ and M is initially oriented at an angle α wrt. the z-

axis, the magnetization evolves over time as:
• Mₓ(t) = M₀ · sin(α) · cos(−ω₀t + φ) → x-component
• My(t) = M₀ · sin(α) · sin(−ω₀t + φ) → y-component
• Mz(t) = M₀ · cos(α) → z-component
 These equations describe how the magnetization vector
precesses in 3D space, forming a cone-like motion.
 The angle α determines the tilt of the magnetization vector from
the z-axis.
 The phase φ represents the initial orientation in the transverse
(x–y) plane.
 This dynamic behaviour is essential for signal generation in MRI,
as transverse components are what induce measurable signals
in receiver coils.

CHAPTER 7 19
 Transverse Magnetization
 Magnetization system described using
polar coordinates

Mₓ(t) = M₀ · sin(α) · cos(−ω₀t + φ)

My(t) = M₀ · sin(α) · sin(−ω₀t + φ)

Mz(t) = M₀ · cos(α)

CHAPTER 7 20
 Transverse Magnetization
 The precession frequency of the net magnetization
vector M is the Larmor frequency (ω₀), the same as for
individual protons.
 When the tilt angle α = 0, the magnetization
is fully aligned with the z-axis:
• Mₓ = My = 0 → no transverse magnetization.
• Mz = M₀ → full longitudinal magnetization.
 When α ≠ 0, the net magnetization vector
M precesses around the z-axis (B₀ direction)
at the Larmor frequency.
 This means the entire magnetization vector,
not just individual protons, is rotating in
space.
CHAPTER 7 21
 Transverse Magnetization
 The magnetization vector M(t) can be decomposed into:
• Longitudinal component Mz(t) → aligned with B₀.
• Transverse component Mxy(t) → lies in the x–y plane, orthogonal to
B₀.
 The transverse magnetization can be expressed as a complex
number:
• Mxy(t) = Mₓ(t) + j·Mᵧ(t)
o Where j is the imaginary unit.
o This form simplifies mathematical treatment of rotating vectors.
 The phase angle φ of the transverse magnetization is given
𝑴𝑴𝒚𝒚
by: 𝝓𝝓 = 𝒕𝒕𝒕𝒕−𝟏𝟏
𝑴𝑴𝒙𝒙
 The full expression for transverse magnetization becomes:
𝑴𝑴𝒙𝒙𝒙𝒙 𝒕𝒕 = 𝑴𝑴𝟎𝟎 · 𝒔𝒔𝒔𝒔𝒔𝒔 𝜶𝜶 · 𝒆𝒆−𝒋𝒋(𝒘𝒘𝟎𝟎−𝝓𝝓)
• This shows that Mxy rotates in the x–y plane at the Larmor frequency.
• The exponential form captures both magnitude and phase of the
rotating vector.

CHAPTER 7 22
 Transverse Magnetization
 Transverse magnetization rotates rapidly in the x–y plane,
generating a radiofrequency (RF) signal.
 This rotating magnetic field induces a voltage in a nearby
coil, which is the measurable signal used in MRI.
 To maximize the transverse magnetization, the
magnetization vector should be tipped by α = 90°:
• This places the entire magnetization in the transverse plane,
maximizing signal strength.
 The Larmor frequency of precession is the same as the RF
frequency used in MRI systems.
 In clinical MRI scanners (e.g., 1.5 T or 3 T), the Larmor
frequency is similar to FM radio frequencies.
• [@ Bo= 1.5 T, γ= 42.58 MHz/T] ω₀ = γ · B₀ = 63.87 MHz
• [@ Bo= 2 T, γ= 42.58 MHz/T] ω₀ = γ · B₀ = 127.74 MHz

CHAPTER 7 23
 Transverse Magnetization

 The signals generated by NMR are very small, due to

the tiny imbalance between protons in the parallel and
anti-parallel states.
 Because of their low strength, these signals are
susceptible to interference from external RF sources.
 To prevent interference, MRI rooms are shielded using
a Faraday cage:
• This blocks stray RF signals from entering the scan environment.
• Evidence of this shielding can be seen in the fine mesh
embedded in the glass between the scan room and control
room.

CHAPTER 7 24
 RF Excitation
 At equilibrium, the magnetization vector M(t) is fully aligned
with the static magnetic field B₀, meaning the tilt angle α = 0°.
 To generate a measurable signal, we need to tip M(t) away from
the z-axis, i.e., α ≠ 0°.
 This is achieved by applying a secondary
magnetic field B₁ (RF), which is:
• Much smaller than B₀.
• Oriented orthogonally to B₀ (e.g., along the x-
axis).
 The application of B₁ disturbs the equilibrium,
causing M(t) to begin moving toward the x–y
plane.
 However, once disturbed, M(t) begins
to precess around the z-axis, just like
individual spins.
CHAPTER 7 25
 RF Excitation
 To effectively tip M(t) into the transverse plane, B₁ must
be applied in sync with this precession.
 This is analogous to pushing someone on a swing:
• If you time your pushes with the swing’s natural motion, you
can amplify the movement.
• Similarly, synchronized RF pulses (B₁) are more effective at
rotating M(t).

 This process is the basis of RF

excitation in MRI, where precisely
timed pulses are used to control the
orientation of the magnetization
vector.

CHAPTER 7 26
 RF Excitation
 To effectively tip the magnetization vector M(t) into the
transverse plane, the applied RF field B₁ must rotate at the
same frequency as M’s precession:
• This frequency is the Larmor frequency (ω₀).
 When B₁ and M(t) are synchronized, the rotating B₁ field
continuously pushes M(t) downward into the x–y plane.
 This technique is known as linearly polarized RF excitation:
• The B₁ field is fixed in one direction (e.g., x-axis).
• It interacts with M(t) only when their orientations align.
 A more efficient method is circularly polarized RF excitation:
• Uses two orthogonal RF components (e.g., sine and cosine
waves).
• This creates a rotating B₁ field that continuously interacts
with M(t).
• Most modern MRI coils use this approach for stronger
and more consistent excitation.

CHAPTER 7 27
 RF Excitation

 The RF field can be mathematically modelled as

a complex magnetic field in the transverse plane:
𝑩𝑩𝟏𝟏 𝒕𝒕 = 𝑩𝑩𝒆𝒆𝟏𝟏 (𝒕𝒕) · 𝒆𝒆−𝒋𝒋(𝒘𝒘·𝒕𝒕−𝝓𝝓)
o𝑩𝑩𝒆𝒆𝟏𝟏 (𝒕𝒕): amplitude envelope of the RF pulse.
oω: angular frequency.
oφ: initial phase of the RF field.
 Under this excitation, M(t) spirals from the z-axis toward
the x–y plane in a clockwise direction, achieving the
desired transverse magnetization.

CHAPTER 7 28
 RF Excitation
 The flip angle (α) determines how far the magnetization
vector M(t) is rotated away from the z-axis into the
transverse plane.
 This angle depends on both the amplitude and duration
of the applied RF pulse 𝑩𝑩𝒆𝒆𝟏𝟏 (𝒕𝒕).
 The general formula for calculating the flip angle is:
𝝉𝝉𝒑𝒑
𝛂𝛂 = 𝜸𝜸 · � 𝑩𝑩𝒆𝒆𝟏𝟏 𝒕𝒕 𝒅𝒅𝒅𝒅
𝟎𝟎
• γ: gyromagnetic ratio
• 𝑩𝑩𝒆𝒆𝟏𝟏 𝒕𝒕 : envelope of the RF field amplitude
• τₚ: duration of the RF pulse

CHAPTER 7 29
 RF Excitation

 A 90° flip angle rotates the magnetization fully into the

transverse plane, producing the maximum measurable
signal.
 A 180° flip angle is called an inversion pulse, which flips
the magnetization to align with the −z direction.
 For short RF pulses with constant amplitude, the flip
angle can be approximated by:
• α = γ · B₁ · τₚ
 Controlling the flip angle is essential for:
• Optimizing signal strength
• Manipulating contrast
• Designing pulse sequences in MRI

CHAPTER 7 30
 The rotating frame
 To simplify the analysis of magnetization dynamics, we can use
a rotating frame of reference:
• This frame rotates at the Larmor frequency (ω₀), matching the
precession of the magnetization vector M(t).
 In this rotating frame, the coordinates transform as follows:
• x′ = x · cos(ω₀t) − y · sin(ω₀t)
• y′ = x · sin(ω₀t) + y · cos(ω₀t)
• z′ = z

 These transformations
effectively remove the
time dependence of the
rotating magnetization
vector.

CHAPTER 7 31
 The rotating frame
 Another familiar example is watching versus riding on a
merry-go-round:

Merry-go-round viewed from the Merry-go-round viewed from the

laboratory frame rotating frame

CHAPTER 7 32
 The rotating frame
 The transverse magnetization in the rotating frame is
expressed as:
𝑀𝑀𝑥𝑥 ′ 𝑦𝑦′ 𝑡𝑡 = 𝑀𝑀0 · sin 𝛼𝛼 · 𝑒𝑒 𝑗𝑗𝜙𝜙 = 𝑀𝑀𝑥𝑥𝑥𝑥 𝑒𝑒 𝑗𝑗𝑗𝑗
• Where:
o Mxy = M0 sin(α) is the magnitude of the transverse magnetization.
o φ is the phase angle.
 In this frame, the magnetization vector appears stationary,
simplifying calculations and signal interpretation (now the new
axes rotates at the Larmor frequency).
 When measuring MRI signals, we obtain both:
• Magnitude of the transverse magnetization (used for image
contrast).
• Phase information (often discarded in basic imaging but useful in
advanced techniques like phase-contrast MRI).
 When viewed in the rotating frame of reference (which spins
at the Larmor frequency), the behaviour of the magnetization
vector M(t) appears simplified.

CHAPTER 7 33
 The rotating frame

 In this frame, the rotational motion is

effectively removed, allowing us to focus on
the net change in orientation.
 Upon applying an RF excitation pulse (B1),
the magnetization vector tips away from the
z-axis toward the x–y (transverse) plane.
 This tipping motion is easier to visualize in
the rotating frame, where M(t) appears to
move in a smooth arc rather than spiralling.
 The rotating frame helps us understand
how RF pulses manipulate magnetization,
making it a valuable conceptual tool in MRI
physics.

CHAPTER 7 34
 Relaxation
 When an RF pulse (B₁) is applied, the magnetization vector M begins
to spiral from the z-axis toward the x–y (transverse) plane.
 This spiral motion represents the tipping of M into a new orientation,
away from its equilibrium alignment with the static field B₀.
 In the stationary frame of reference, this process appears as M tipping
from the z’-axis down toward the x’–y’ plane, where it remains.
 After the RF pulse, M continues to precess in its new orientation within
the transverse plane.
 However, this motion is not sustained indefinitely—it is subject
to damping due to two key relaxation processes:
• Transverse relaxation (T₂): loss of phase coherence among spins in the
transverse plane.
• Longitudinal relaxation (T₁): recovery of magnetization along the z-axis toward
equilibrium.
 These relaxation mechanisms are fundamental to signal decay and
recovery in MRI and are crucial for image contrast and timing.

CHAPTER 7 35
 Transverse Relaxation: T2

 Transverse relaxation, also known as spin–spin

relaxation, describes the loss of phase coherence among
spinning protons in the transverse plane.
 It occurs due to microscopic interactions between
neighbouring magnetic moments:
• These interactions cause some spins to speed up or slow down,
leading to dephasing.
 Nearby protons alter the local magnetic field
experienced by each proton, slightly changing their
Larmor frequencies.
 As a result, protons begin to precess at different rates,
increasing the phase difference over time.

CHAPTER 7 36
 Transverse Relaxation: T2

 When the signals from all protons are summed together,

the misalignment reduces the net transverse magnetization,
weakening the overall signal.
 This decay in signal is observed as a damped sinusoidal
waveform, known as Free Induction Decay (FID).
 The decay can be mathematically modeled as an exponential
function:
𝑡𝑡
−
𝑀𝑀𝑥𝑥𝑥𝑥 𝑡𝑡 = 𝑀𝑀0 · sin 𝛼𝛼 · 𝑒𝑒 𝑇𝑇2
• Where:
o M₀ · sin(α) is the initial transverse magnetization,
o T₂ is the transverse relaxation time constant,
o t is time.
 T₂ relaxation is a key factor in determining signal strength and
image contrast in MRI.

CHAPTER 7 37
 Transverse Relaxation: T2

Process of T2 relaxation shown in the rotating frame

after a 90° flip angle. Over time, the individual
contributions from different protons get out of phase
with each other.

CHAPTER 7 38
 Longitudinal Relaxation: T1

 Longitudinal relaxation, also known as spin–lattice relaxation,

describes how the magnetization vector Mz(t) returns to its
equilibrium state M₀ along the z-axis.
 This process occurs after an RF excitation pulse, which tips the
magnetization away from the z-axis.
 The recovery of Mz(t) is modelled as a rising exponential
function, indicating gradual realignment with the static magnetic
field B₀.
 The general equation for longitudinal magnetization recovery is:
𝑡𝑡
−
𝑀𝑀𝑧𝑧 𝑡𝑡 = 𝑀𝑀0 · 𝑐𝑐𝑐𝑐𝑐𝑐 𝛼𝛼 + 𝑀𝑀0 · 1 − 𝑐𝑐𝑐𝑐𝑐𝑐 𝛼𝛼 · 1− 𝑒𝑒 𝑇𝑇1
• M₀ · cos(α) is the initial z-component after excitation with α-pulse.
• T₁ is the longitudinal relaxation time constant.

CHAPTER 7 39
 Longitudinal Relaxation: T1
 This can be rearranged into a more common form:
𝑡𝑡 𝑡𝑡
− −
𝑀𝑀𝑧𝑧 𝑡𝑡 = 𝑀𝑀0 · 1 − 𝑒𝑒 𝑇𝑇1 + 𝑀𝑀0 · 𝑐𝑐𝑐𝑐𝑐𝑐 𝛼𝛼 · 𝑒𝑒 𝑇𝑇1
 After excitation, both T₁ recovery and T₂ decay occur
simultaneously, but they are independent processes.
 The magnetization does not simply retrace its excitation
path; instead, it follows a unique recovery trajectory.

Reduction in the magnetization

due to T1 decay

CHAPTER 7 40
 Longitudinal Relaxation: T1
 Typically:
• T₂ decay is faster than T₁ recovery.
• This results in a curved recovery path.
 For biological tissues:
• T₁ values range from 250 ms to 2500 ms.
• T₂ values range from 25 ms to 250 ms.
 These relaxation times are crucial for image contrast,
pulse sequence design, and tissue differentiation in MRI.

CHAPTER 7 41
 Longitudinal Relaxation: T1
 With the introduction of typical relaxation times for T₁
(longitudinal) and T₂ (transverse) processes, we can now define
what is meant by a “long” time in NMR experiments.
 A “long” time refers to a duration sufficient for both T₁ recovery
and T₂ decay to occur, allowing the magnetization to return to
equilibrium.
 Understanding these timescales is essential for:
• Designing pulse sequences
• Timing signal acquisition
• Optimizing image contrast
 This future discussion will show how differences in relaxation
times can be used to distinguish between tissues, beyond
just proton density variations.
 These principles are foundational for contrast-enhanced MRI,
enabling detailed anatomical and functional imaging.
CHAPTER 7 42
 The Bloch Equations
 The Bloch equations integrate the three key properties of the
spin system:
• Proton density
• Longitudinal relaxation (T₁)
• Transverse relaxation (T₂)
 These equations describe how the magnetization vector M(t)
evolves over time under the influence of magnetic fields and
relaxation effects.
 The general form of the Bloch equation is:
𝑑𝑑𝑴𝑴
= 𝛾𝛾 · 𝑴𝑴 𝑡𝑡 × 𝑩𝑩 𝑡𝑡 − 𝑹𝑹 · 𝑴𝑴(𝑡𝑡 − 𝑴𝑴0 )
𝑑𝑑𝑑𝑑

• γ: gyromagnetic ratio
• B(t): total magnetic field (static + RF)
• R: relaxation matrix
• M₀: equilibrium magnetization

CHAPTER 7 43
 The Bloch Equations
 The total magnetic field B(t) is composed of:
• B₀: static magnetic field
• B₁(t): time-varying RF field
• So, B(t) = B₀ + B₁(t)
 The relaxation matrix R accounts for the rates of decay and recovery:
1/𝑇𝑇2 0 0
𝑹𝑹 = 0 1/𝑇𝑇2 0
0 0 1/𝑇𝑇1
• T₂ governs decay in the x and y directions (transverse plane).
• T₁ governs recovery in the z direction (longitudinal axis).
 The Bloch equations are fundamental for:
• Modelling spin dynamics
• Designing pulse sequences
• Predicting signal behaviour in MRI systems

CHAPTER 7 44
 Free Induction Decay (FID)

CHAPTER 7 45
 Spin Echoes and T2*

 T₂ relaxation refers to the dephasing of spins due to random

interactions between neighbouring magnetic moments.
 In practice, the observed signal decay is even faster than T₂ due
to additional dephasing effects, known as T₂*.
 T₂* arises from:
• Microscopic magnetic field inhomogeneities.
• Chemical environment variations.
• These cause protons to precess at slightly different rates, increasing
phase dispersion.
 Unlike T₂, T₂* is not a true relaxation process because it is not
random—it’s caused by predictable field variations.
 T₂* typically dominates over T₂ in signal decay, leading to faster
loss of coherence.

CHAPTER 7 46
 Spin Echoes and T2*

 To recover signal lost due to T₂*, MRI uses a technique

called a spin echo:
• After a 90° excitation pulse, spins begin to dephase.
• A 180° refocusing pulse is applied around the y-axis.
• This reverses the relative positions of faster and slower spins.
• After an equal time interval, spins re-align, forming an echo.
 The spin echo restores signal lost due to T₂*, but cannot
recover T₂ decay, which is truly random.
 Spin echo sequences are essential for:
• Improving image quality
• Reducing artifacts
• Enhancing contrast based on true T₂ values

CHAPTER 7 47
 Spin Echoes and T2*
Formation of a spin echo. Top row: spins start to dephase
due to different rates of precession after they have been
flipped into the transverse plane. Bottom row: A 180°
inversion pulse is applied then the spins start to come back
into phase, but perfect recovery of signal is not achieved
due to T2 effects.

B1

B1

CHAPTER 7 48
 Spin Echoes and T2*
 Rephasing of spins by gradient reversal:
• The fast hare represents spins precessing
rapidly (and accumulating phase) by virtue
of their location in a stronger portion of
the gradient; the tortoise represents more
slowly precessing spins in a weaker part of
the gradient.
• The fast hare travels much farther initially
(corresponding to a larger phase
accumulation). The reversal of direction
halfway through the race corresponds to
the gradients being applied with opposite
polarities. The hare again runs faster but in
the opposite direction, having more
distance to make up. Finally both return to
the starting line at the same time
(equivalent to net phase shift = 0).

CHAPTER 7 49
 Spin Echoes and T2*
Formation of a spin echo. Formation of the Spin Echo

Conventional (single echo) SE pulse

sequence.
• TE: Echo Time
• TR: Repetition Time

CHAPTER 7 50
 Fourier
Reconstruction—MRI

CHAPTER 7 51
 Overview
 We will consider Fourier-based reconstruction as is found
in magnetic resonance imaging (MRI).
 We will see how gradient fields are used to encode the
signals with spatial information and how a Fourier
transform can then be used to recover a 2D or 3D image.

CHAPTER 7 52
 Introduction
 Main Magnet: Generates a strong, uniform magnetic field (B0​)
essential for aligning hydrogen nuclei in the body, forming the basis
for MRI signal generation.
 Switchable Gradient Coils: Superimpose spatially varying magnetic
fields on top of the main field. These gradients are crucial for spatial
encoding, allowing the MRI system to distinguish signals from
different locations in the body.
 RF Coils and Power Amplifiers: Transmit radiofrequency (RF) pulses
to excite hydrogen nuclei and receive the emitted signals. These coils
are tuned to the Larmor frequency of hydrogen and are essential for
signal detection.
 Pulse Sequence and Receive Electronics: Coordinate the timing and
control of RF pulses and gradient switching and process the received
signals for image reconstruction.
 Role of Gradient Coils in Imaging: While the magnet and RF coils are
necessary for basic NMR, gradient coils are what enable spatial
resolution—they encode positional information into the signal,
making image generation possible.

CHAPTER 7 53
 Gradients
 Uniform Main Magnetic Field (B0​): Initially, the magnetic
field B0 is assumed to be uniform and aligned along the z-
axis, which is essential for basic NMR signal generation.
 Role of Gradient Coils: Gradient coils modify the
magnitude (not the direction) of B0 in a spatially
dependent way, enabling spatial encoding of the MRI
signal.
 Three Orthogonal Gradients: There are three gradient
coils aligned along the x, y, and z axes. Each can add or
subtract a magnetic field component depending on
position:
𝐵𝐵 = 𝐵𝐵0 + 𝐺𝐺𝑥𝑥 𝑥𝑥 + 𝐺𝐺𝑦𝑦 𝑦𝑦 + 𝐺𝐺𝑧𝑧 𝑧𝑧 𝒛𝒛� = 𝐵𝐵0 + 𝑮𝑮 ⋅ 𝒓𝒓 𝒛𝒛�
 The direction of the field along z, the main scanner axis,
is maintained. Only the magnitude is changed
CHAPTER 7 54
 Gradients
 Gradient Strength and Units: The gradient vector applied G =
𝐺𝐺𝑥𝑥 , 𝐺𝐺𝑦𝑦 , 𝐺𝐺𝑧𝑧 is measured in millitesla per meter (mT/m), with
clinical systems typically reaching up to 40 mT/m.
 Rapid Switching and Slew Rate: Gradient coils must switch on
and off very quickly (within 0.1–1 ms). The slew rate (rate of
change of gradient strength) ranges from 5 to 250 mT/(m·s)
and is limited to reduce eddy currents in the patient.
 Spatial Encoding Mechanism: By altering the local magnetic
field, gradient coils excite spins at the Larmor frequency
depending on their position. This variation is used to encode
spatial location into the MR signal.
 Image Formation Principle: The spatially varying Larmor
frequency and the phase of the transverse magnetization are
key to reconstructing an image from the received NMR signals.
CHAPTER 7 55
 Slice Selection

 Purpose of Slice Selection: The goal is to isolate a specific

slice of tissue so that the MRI signal originates only from
that region, enabling targeted imaging.
 Gradient Application for Spatial Encoding: Applying a
gradient field G = 𝐺𝐺𝑥𝑥 , 𝐺𝐺𝑦𝑦 , 𝐺𝐺𝑧𝑧 causes the Larmor
frequency to vary with position:
ω 𝒓𝒓 = γ 𝐵𝐵0 + 𝑮𝑮 ⋅ 𝒓𝒓
 Selecting a Slice Along the z-Axis: To focus on a slice at a
specific z-position, a gradient G = (0, 0, Gz​) is applied,
resulting in:
ω 𝑧𝑧 = γ 𝐵𝐵0 + 𝐺𝐺𝑧𝑧 · 𝑧𝑧

CHAPTER 7 56
 Slice Selection

 Ideal vs. Practical Slice Thickness:

• In theory, exciting only the frequency corresponding
to a specific z-value would select an infinitesimally
thin slice.
• In practice, this is not feasible due to:
o Finite RF pulse bandwidth: Real RF pulses excite a range of
frequencies.
o Signal strength limitations: A very thin slice contains few
protons, resulting in weak signals.
 Trade-off in Slice Thickness: A thicker slice improves
signal-to-noise ratio (SNR) but reduces resolution in the
slice direction. The slice thickness is controlled by the
gradient strength and RF bandwidth.

CHAPTER 7 57
 Slice Selection
 RF Pulse Excitation Over a Frequency Range: Instead of using a
single-frequency RF pulse, MRI applies a pulse that spans a range
of frequencies [w1​, w2​], which excites a corresponding range of
tissue along the z-axis—resulting in a thicker slice or "slab."
 Gradient Strength Affects Slice Thickness:
• A smaller z-gradient (G1​) results in a thicker slice, as the frequency
range covers a larger spatial region.
• A larger z-gradient (G2​) results in a thinner slice, as the same frequency
range corresponds to a smaller spatial region.
 Slice Selection Depends on Three Parameters:
• Gradient strength in the z-direction: (Gz​)
𝑤𝑤1 +𝑤𝑤2
• RF centre frequency: 𝑤𝑤
�= 2
• RF bandwidth: Δω = ∣ω2​−ω1∣
 Controlling Slice Properties: By adjusting these three parameters,
MRI systems can precisely control both the position
and thickness of the selected slice.

CHAPTER 7 58
 Slice Selection

Example of the use of a gradient (in z) and selective

excitation to achieve slice selection in MRI

CHAPTER 7 59
 Slice Selection
 Sequential Slice Acquisition for Volume Imaging: To image a
full volume, multiple slices are acquired one after another. This
increases total scan time proportionally to the number of
slices.
 Slice Thickness and Signal Quality:
• Thicker slices contain more tissue (i.e., more water/protons), which
improves signal-to-noise ratio (SNR) and contrast-to-noise ratio
(CNR).
• However, thicker slices reduce resolution in the slice direction.
 Resolution Independence Across Dimensions:
• Resolution in the slice direction is determined by slice thickness.
• Resolution within the slice (in-plane) is governed by frequency and
phase encoding.
• It’s common to have voxel dimensions that differ between slice
thickness and in-plane resolution to balance SNR and image clarity.

CHAPTER 7 60
 Slice Selection

 Imperfect Slice Edges Due to RF Pulse Limitations:

• RF pulses have finite bandwidth and duration, making
it impossible to excite only the exact frequencies
needed for perfectly sharp slice boundaries.
• Pulse shape and duration introduce extra frequency
components, leading to blurred slice edges.
 Practical Adjustments in Multi-Slice Imaging:
• Small gaps are often left between slices to reduce
overlap and edge artifacts.
• RF pulse shapes are optimized to improve slice
definition and overall image quality.

CHAPTER 7 61
 Frequency Encoding

 Slice Selection alone is Not Enough: After

selecting a slice using gradient fields, further
spatial encoding is needed to distinguish
different locations within that slice for image
construction.
 Transverse Magnetization after RF Excitation:
• The transverse magnetization Mxy​(t) evolves over
time as:
𝑀𝑀𝑥𝑥𝑥𝑥 𝑡𝑡 = 𝑀𝑀𝑥𝑥𝑥𝑥 (0+ ) 𝑒𝑒 −𝑗𝑗 2πω0 𝑡𝑡−ϕ 𝑒𝑒 −𝑡𝑡/𝑇𝑇2
o Where 𝑀𝑀𝑥𝑥𝑥𝑥 (0+ ) = 𝑀𝑀𝑧𝑧 (0− ) · sin(𝛼𝛼) represents the
magnetization immediately after applying a flip
angle α.

CHAPTER 7 62
 Frequency Encoding

Received Signal as an Integral Over the Slice:

 The MRI signal s(t) is the sum of contributions from all
excited tissue within the slice:
∞
𝑠𝑠 𝑡𝑡 = � 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ · 𝑒𝑒 −𝑗𝑗𝑗πω0𝑡𝑡 · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞
• A is a scanner-specific gain factor,
• both Mxy and T2 vary spatially depending on tissue properties.

Simplifying Assumptions:
 The phase ϕ is assumed to be zero for simplicity.
 The spatial variation of T2 and Mxy reflects differences in
tissue composition and relaxation properties.

CHAPTER 7 63
 Frequency Encoding

 Implication for Imaging: This formulation sets the stage

for applying spatial encoding techniques (like frequency
and phase encoding) to extract positional information
from the received signal and reconstruct an image.
 Object-Related Terms Grouped into Effective Spin
Density:
• To simplify the signal expression, all object-specific
terms (including scanner gain and tissue properties)
are grouped into a function called effective spin
density:
𝑓𝑓 𝑥𝑥, 𝑦𝑦 = 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦

CHAPTER 7 64
 Frequency Encoding

Received Signal Expression:

 The received signal s(t) becomes:
∞
𝑠𝑠 𝑡𝑡 = 𝑒𝑒 −𝑗𝑗𝑗πω0 𝑡𝑡 � 𝑓𝑓 𝑥𝑥, 𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞

Demodulated (Baseband) Signal:

 After demodulation (removing the carrier frequency ω0​),
the signal becomes:
∞
𝑠𝑠0 𝑡𝑡 = 𝑒𝑒 +𝑗𝑗𝑗πω0 𝑡𝑡 · 𝑠𝑠 𝑡𝑡 = � 𝑓𝑓 𝑥𝑥, 𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞

CHAPTER 7 65
 Frequency Encoding

Loss of Spatial Information:

 This demodulated signal is a constant value, as the spatial
variation has been integrated out. No positional
information remains in the signal.
Need for Spatial Encoding:
 To reconstruct an image, spatial encoding must be
introduced—typically by applying a gradient during
the Free Induction Decay (FID) period, which causes
spatially dependent frequency shifts.

CHAPTER 7 66
 Frequency Encoding

Need for Spatial Encoding Within a Slice:

 After selecting a slice, we must distinguish different
positions within it to construct an image.
 This is achieved by applying a readout gradient
orthogonal to the slice selection gradient—commonly in
the x-direction.
Position-Dependent Larmor Frequency:
 Applying a gradient in the x-direction modifies the Larmor
frequency spatially:
ω 𝑥𝑥 = γ 𝐵𝐵0 + 𝐺𝐺𝑥𝑥 𝑥𝑥 = ω0 + γ𝐺𝐺𝑥𝑥 𝑥𝑥

CHAPTER 7 67
 Frequency Encoding

Received Signal with Spatial Encoding:

 The signal becomes an integral over all spins in the slice, now
with spatially varying frequency:
∞
𝑠𝑠 𝑡𝑡 = � 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ · 𝑒𝑒 −𝑗𝑗𝑗π·ω 𝑥𝑥,𝑦𝑦 𝑡𝑡 · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞
 Substituting the frequency expression:
∞
𝑠𝑠 𝑡𝑡 = � 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ 𝑒𝑒 −𝑗𝑗𝑗π ω0 +γ𝐺𝐺𝑥𝑥 𝑥𝑥 𝑡𝑡 · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞
 Factoring out the constant frequency term:
∞
𝑠𝑠 𝑡𝑡 = 𝑒𝑒 −𝑗𝑗𝑗πω0𝑡𝑡 � 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ 𝑒𝑒 −𝑗𝑗𝑗πγ𝐺𝐺𝑥𝑥 𝑡𝑡𝑡𝑡 · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞

CHAPTER 7 68
 Frequency Encoding
Demodulated Signal Using Effective Spin Density:
 Using the definition:
𝑓𝑓 𝑥𝑥, 𝑦𝑦 = 𝐴𝐴 · 𝑀𝑀𝑥𝑥𝑥𝑥 𝑥𝑥, 𝑦𝑦, 0+ · 𝑒𝑒 −𝑡𝑡/𝑇𝑇2 𝑥𝑥,𝑦𝑦
 The baseband signal becomes:
∞
𝑠𝑠0 𝑡𝑡 = � 𝑓𝑓 𝑥𝑥, 𝑦𝑦 · 𝑒𝑒 −𝑗𝑗𝑗π·γ𝐺𝐺𝑥𝑥 𝑡𝑡𝑡𝑡 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞
Interpretation as a Fourier Transform:
 This signal resembles a 2D Fourier transform of the spin
density f(x,y), where:
• Spatial frequency in x-direction: u = γ·Gx·t
• Spatial frequency in y-direction: v = 0
• In MRI, this frequency domain is called k-space, with:
• kx​=u, ky​=v
CHAPTER 7 69
 Scanning k-Space
 MRI and k-Space Scanning:
• MRI imaging involves scanning 2D Fourier space, known as k-
space, where spatial frequency data is collected to reconstruct
images.
 Basic Frequency-Encoding Process:
• The simplest form of scanning involves frequency encoding
along the u-axis (typically the x-direction in physical space).
• This is achieved by applying a readout gradient during signal
acquisition.
 Pulse Sequence Overview:
• The process begins with RF excitation combined with a slice
selection gradient, which isolates a specific slice of tissue.
• Following excitation, an x-gradient is applied during the readout
phase, indicated by the ADC (Analog-to-Digital Converter) being
active.

CHAPTER 7 70
 Scanning k-Space

 Gradient Duration and k-Space Coverage:

• The longer the Gₓ gradient is applied, the further the sampling
progresses along the u-axis in k-space.
• This creates a trajectory through k-space, collecting frequency
data needed for image reconstruction.
 Bidirectional Gradient Application:
• For more complete coverage, the gradient is first applied in
the −Gₓ direction, then in the +Gₓ direction.
• This allows sampling from −u to +u, effectively covering the full
range of spatial frequencies along the x-axis.

CHAPTER 7 71
 Scanning k-Space

Using a Gx readout gradient to acquire samples

along u

Gradient refocusing

readout

CHAPTER 7 72
 Scanning k-Space

Reading out the whole of the u-axis

readout

CHAPTER 7 73
 Scanning k-Space

Beyond Frequency Encoding:

 While frequency encoding allows sampling along the u-
axis (typically the x-direction), it alone is insufficient for
full 2D image reconstruction.
 To capture spatial information in the v-direction (typically
the y-direction), phase encoding is introduced.
Phase Encoding Mechanism:
 A gradient in the y-direction (Gy​) is applied before the
readout phase, for a fixed duration Tp​.
 This gradient induces a position-dependent phase shift in
the spins along the y-axis.

CHAPTER 7 74
 Scanning k-Space

 Mathematical Representation:
• The resulting signal, incorporating both frequency and phase
encoding, is:
∞
𝑠𝑠0 𝑡𝑡 = � 𝑓𝑓 𝑥𝑥, 𝑦𝑦 · 𝑒𝑒 −𝑗𝑗𝑗πγ𝐺𝐺𝑥𝑥𝑥𝑥𝑥𝑥 · 𝑒𝑒 −𝑗𝑗𝑗πγ𝐺𝐺𝑦𝑦 𝑦𝑦𝑇𝑇𝑝𝑝 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
−∞
 Phase Accumulation:
• The phase shift introduced by the Gy gradient is:
ϕ𝑦𝑦 = −γ𝐺𝐺𝑦𝑦 𝑦𝑦𝑇𝑇𝑝𝑝
• This phase shift encodes the y-position of spins into the signal.
 Timing of the Gradient:
• The Gy gradient is not active during readout; it is applied before
the ADC is turned on.
• This allows the readout to occur along a specific line in k-space,
offset in the v-direction.
CHAPTER 7 75
 Scanning k-Space

Using a phase encoding gradient

Tp

readout

CHAPTER 7 76
 Scanning k-Space

k-Space Trajectory:
 By varying the strength of the Gy gradient across multiple
acquisitions, different lines in k-space are sampled.
 This process enables full 2D coverage of k-space.
Combined Encoding Strategy:
 The combination of frequency encoding (x-direction)
and phase encoding (y-direction) allows MRI to reconstruct
a complete 2D image from the acquired data.
frequency encoding phase encoding

∞
𝑠𝑠0 𝑡𝑡 = ∬−∞ 𝑓𝑓 𝑥𝑥, 𝑦𝑦 · 𝑒𝑒 −𝑗𝑗𝑗π(γ𝐺𝐺𝑥𝑥 𝑡𝑡)𝑥𝑥 · 𝑒𝑒 −𝑗𝑗𝑗π(γ𝐺𝐺𝑦𝑦 𝑇𝑇𝑝𝑝 )𝑦𝑦 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

CHAPTER 7 77
 Scanning k-Space

A k-space trajectory to
sample the whole of 2D
k-space from one
excitation

CHAPTER 7 78
 MRI Reconstruction
Image Reconstruction via Inverse Fourier Transform:
 After acquiring data in k-space (the frequency domain), the
spatial image is reconstructed by applying an inverse Fourier
transform to the function F(u,v), which represents the spatial
frequency content.
Efficiency with Discrete Sampling:
 In practice, k-space is discretely sampled, meaning data are
collected at specific intervals.
 This allows the use of the Fast Fourier Transform (FFT), a highly
efficient algorithm for computing the inverse transform.
MRI Acquires Frequency Domain Data:
 The MRI scanner does not directly capture spatial images;
instead, it collects data in the frequency domain.
 The transformation from frequency to spatial domain is what
produces the final image.

CHAPTER 7 79
 MRI Reconstruction

 Raw data in k-space array (a) and corresponding

image data in image space (b). In both cases, the
magnitude of the data is presented.

CHAPTER 7 80
 MRI Reconstruction

Implications for k-Space Sampling:

 Because the image is reconstructed from frequency data,
the way we sample k-space has a direct impact on image
quality.
 Choices in sampling density, coverage, and pattern affect:
• Field of view (FoV).
• Image resolution.
• Presence of artifacts (e.g., aliasing).
Importance of Sampling Strategy:
 Proper sampling ensures that the reconstructed image
accurately represents the object.
 Undersampling or improper spacing can lead to
distortions or loss of detail.

CHAPTER 7 81
 Field of View

 Sampling Theory in MRI:

• MRI reconstruction relies on principles of sampling theory, similar
to those used in signal processing.
 Sampling as Multiplication with Delta Functions:
• Sampling a signal is mathematically equivalent to multiplying it by
a train of delta functions spaced at regular intervals.
• This operation replicates the signal’s frequency spectrum at
intervals determined by the sampling rate.
 Effect of Sampling Rate on Frequency Replicas:
• A higher sampling rate results in greater spacing between replicas
in the frequency domain, avoiding overlap.
• Conversely, a lower sampling rate brings replicas closer together,
increasing the risk of overlap.

CHAPTER 7 82
 Field of View
 Extension to MRI: Sampling in 2D Frequency Domain:
• In MRI, sampling occurs in 2D (or 3D) k-space, which is the
spatial frequency domain of the object.
• This is analogous to multiplying the object’s true frequency
spectrum by a 2D grid of delta functions.
 Impact on Image Domain:
• The result of this multiplication is the creation of replicas of the
image in the spatial domain.
• The spacing between these replicas depends on how densely k-
space is sampled.
 Risk of Wrap-Around Artifacts:
• If k-space is not sampled densely enough, these image replicas
can overlap, causing wrap-around artifacts.
• These artifacts manifest as parts of the image appearing in
incorrect locations, degrading image quality.

CHAPTER 7 83
 Field of View
 The effect of under-sampling (aliasing) in MRI. The top row is ‘fully’
sampled, in the bottom row every other line in k-space has been
removed (factor of 2 under-sampling), resulting in overlapping in the
reconstructed image

2D FT

CHAPTER 7 84
 Field of View

 Under-sampling has caused out of view anatomy

CHAPTER 7 85
 Field of View

Nyquist Theorem and Sampling Requirements:

 According to the Nyquist theorem, to avoid aliasing, a
signal must be sampled at a rate at least twice the
highest frequency present in the signal.
 In MRI, this highest frequency is related to the bandwidth
of the signal.
Bandwidth and Field of View (FoV):
 In MRI, the concept of bandwidth translates to the Field
of View (FoV) in the reconstructed image.
 The FoV determines the spatial extent of the image and is
directly linked to how k-space is sampled.

CHAPTER 7 86
 Field of View

FoV and k-Space Sampling Interval:

 The FoV in each spatial direction is inversely proportional
to the sampling interval in k-space:
1 1
𝐹𝐹𝐹𝐹𝑉𝑉𝑥𝑥 = , 𝐹𝐹𝐹𝐹𝑉𝑉𝑦𝑦 =
Δ𝑘𝑘𝑥𝑥 Δ𝑘𝑘𝑦𝑦
 Smaller sampling intervals (i.e., denser sampling) result in
larger FoVs and vice versa.
Avoiding Wrap-Around Artifacts:
 To prevent wrap-around artifacts, the FoV must be large
enough to encompass all physical tissue being imaged.
 If tissue lies outside the defined FoV, it will be aliased—
appearing incorrectly within the image volume.

CHAPTER 7 87
 Field of View
Strategies to Prevent Aliasing:
 Ensure dense sampling of k-space (reduce Δ𝑘𝑘𝑥𝑥 𝑦𝑦Δ𝑘𝑘𝑦𝑦 ).
 Use gradient-based excitation to restrict the imaging volume
to within the FoV.
 These strategies are analogous to applying an anti-aliasing
filter in signal processing to remove high frequencies that
could cause aliasing.

CHAPTER 7 88
 Field of View

Strategies to Prevent Aliasing:

 Increase field of view (FOV): This can be accomplished by
expanding either the imaging FOV or the FOV in the phase-
encoding direction.

CHAPTER 7 89
 Field of View

Strategies to Prevent Aliasing:

 Adjust the phase-encoding direction: It can help mitigate
aliasing artifacts. Choosing a phase-encoding direction that is
less susceptible to aliasing, such as selecting the right-left
direction instead of the anterior-posterior.

Phase encoding direction A-P

CHAPTER 7 90
 Field of View
Strategies to Prevent Aliasing:
 Oversampling in the phase-encoding direction: Acquiring
additional phase-encoding steps beyond what is necessary for the
desired FOV can help reduce aliasing artifacts. This oversampling
allows for a better estimation and reconstruction of the image
without aliasing.

CHAPTER 7 91
 k-Space Coverage and Resolution
Choice of k-Space Coverage:
 In MRI, we can choose how much of k-space to acquire during
imaging.
 This decision directly affects the image resolution and scan
duration.
Importance of High-Frequency Data:
 The highest spatial frequencies in k-space correspond to fine
details in the image.
 Acquiring higher frequencies requires more time during the
readout phase of the pulse sequence.
Trade -off Between Time and Detail:
 Limiting k-space acquisition to lower frequencies speeds up the
scan but results in loss of fine image detail.
 This is analogous to 1D Fourier transforms: discarding high-
frequency components reduces resolution.

CHAPTER 7 92
 k-Space Coverage and Resolution

Central k-Space Dominance:

 The centre of k-space contains most of the image’s overall
structure and contrast information.
 It is essential to sample this region thoroughly to preserve
the core features of the image.
Resolution Formula:
 The effective spatial resolution of the image is inversely
related to the maximum frequency acquired:
1
Δ𝑥𝑥 =
2 · 𝑘𝑘𝑚𝑚𝑚𝑚𝑚𝑚
 Higher kmax values yield finer resolution.

CHAPTER 7 93
 k-Space Coverage and Resolution
k-Space vs Image Space

CHAPTER 7 94
 k-Space Coverage and Resolution

CHAPTER 7 95
 k-Space Coverage and Resolution

CHAPTER 7 96
 k-Space Coverage and Resolution

CHAPTER 7 97
 The Effect of T2 Decay

Time-Dependent Nature of k-Space Sampling:

 MRI acquires k-space data after RF excitation, during a
period when the signal is actively decaying.
 This acquisition takes time, and during this time,
relaxation processes occur.
T₁ vs. T₂ Decay:
 Both T₁ and T₂ relaxation processes are present, but:
• T₂ decay is most relevant during readout because it affects
the transverse magnetization, which is what MRI measures.
• T₂ decay causes the signal amplitude to decrease over time,
especially impacting later-acquired k-space data.

CHAPTER 7 98
 The Effect of T2 Decay

Impact on Image Quality:

 T₂ decay can lead to signal loss and blurring, particularly
in high-frequency regions of k-space.
 This results in a less accurate reconstruction of the true
object.
Minimizing T₂ Effects:
 To reduce the impact of T₂ decay, the acquisition time
should be kept as short as possible.
 Faster acquisition helps preserve signal strength and
image fidelity.

CHAPTER 7 99
 The Effect of T2 Decay

Segmented Acquisition Strategy:

 Instead of acquiring the entire k-space in one excitation,
MRI can use multiple excitations.
 Each excitation captures only a portion of k-space (e.g., a
single line).
 This approach reduces the time per readout but extends
the total scan duration.
Trade-offs:
 Single-shot acquisition is faster but more vulnerable to T₂
decay.
 Segmented acquisition improves signal quality (reduces
effect of T2 decay) but increases overall scan time.

CHAPTER 5 100
 More Advanced k-Space Trajectories: Partial Fourier

Total Acquisition Time and k-Space Sampling:

 The total time to acquire an MRI image depends on how
long it takes to sufficiently sample all of k-space.
 To minimize the effects of T₂ decay, it’s important to limit
the time spent sampling after each excitation.
Strategy: Partial k-Space Acquisition:
 One effective strategy to reduce acquisition time is to not
collect the entire k-space.
 This is based on the fact that real-valued images exhibit
symmetry in k-space, making some data redundant.

CHAPTER 7 101
 More Advanced k-Space Trajectories: Partial Fourier

Theoretical Basis:
 In theory, due to conjugate symmetry, it is sufficient to acquire
only half of k-space (e.g., only the positive ky lines).
 The missing half can be inferred mathematically if the image is
assumed to be real-valued and noise-free.
Practical Considerations:
 In real-world imaging, imperfections and noise make it risky to rely
solely on symmetry.
 Therefore, a slightly larger portion of k-space is typically acquired—
especially around the centre, which contains most of the image’s
energy and contrast information.
Goal of Partial Acquisition:
 The aim is to balance scan time and image quality by reducing the
number of samples while still enabling accurate reconstruction.
 This approach helps shorten acquisition time without significantly
compromising image fidelity.

CHAPTER 7 102
 More Advanced k-Space Trajectories: Non-Cartesian

Flexibility in k-Space Sampling:

 MRI is not limited to traditional Cartesian (grid-based)
sampling of k-space.
 Alternative trajectories such as spirals or radial spokes
can be used.
Advantages of Non-Cartesian Trajectories:
 These methods can offer lower sensitivity to motion
artifacts, making them beneficial for imaging moving
tissues (e.g., heart, lungs).
 They may also allow for faster acquisition or more
efficient coverage of k-space.

CHAPTER 7 103
 More Advanced k-Space Trajectories: Non-Cartesian

Challenges with Reconstruction:

 Most efficient Fourier transform algorithms (like FFT) require
data to be sampled on a uniform grid.
 Non-Cartesian data must be resampled or interpolated onto a
Cartesian grid before image reconstruction.
Preprocessing Requirement:
 This resampling step adds computational complexity and may
introduce interpolation errors if not handled carefully.
Use Cases:
 Non-Cartesian trajectories are especially useful in real-time
imaging, angiography, and functional MRI, where motion
robustness and speed are critical.

CHAPTER 7 104
 More Advanced k-Space Trajectories: Non-Cartesian

 Alternative k-space trajectories of cartesian sampling (a). The

spiral (c) can be completed after a single excitation. The spokes
(b) might be done from separate excitations; otherwise, it
would be necessary to traverse a trajectory from the end of one
spoke to the start of the next

CHAPTER 7 105
 More Advanced k-Space Trajectories: 3D k-Space

Slice-by-Slice 2D Acquisition:
 A common method for 3D imaging is to acquire 2D slices sequentially.
 Each slice is excited individually, followed by a 2D readout within that
slice.
 This approach is straightforward but only excites a small portion of
tissue at a time.
Limitation: Lower Signal-to-Noise Ratio (SNR):
 Since only a small volume is excited per slice, fewer hydrogen nuclei
contribute to the signal.
 This results in a lower SNR, which can affect image quality.
Alternative: Full 3D Volume Excitation:
 A more advanced method excites a larger volume of tissue in a single
excitation.
 It uses frequency encoding in one direction and phase encoding in two
directions (e.g., y and z) to fill a 3D k-space.

CHAPTER 7 106
 More Advanced k-Space Trajectories: 3D k-Space

Advantage: Higher SNR:

 More hydrogen nuclei are involved in the signal, leading to
stronger and clearer signals.
 This improves the overall image quality, especially in low-signal
regions.
Disadvantage: Longer Acquisition Time:
 Traversing the full 3D k-space takes significantly more time
than acquiring a single 2D slice.
 This extended duration increases the risk of motion artifacts
and signal decay.
Resolution Trade-off:
 Due to T₂ decay during the longer acquisition, the point spread
function (PSF) in 3D imaging is often poorer than in multi-slice
2D imaging.
 This can lead to reduced spatial resolution in the final image.

CHAPTER 7 107
 Compressed Sensing
Traditional Sampling Assumptions:
 Conventional MRI techniques assume that k-space is fully
sampled or that any missing data can be efficiently
interpolated.
 This approach is based on the Nyquist sampling theorem,
which ensures accurate reconstruction if sampling is sufficient.
Consequences of Undersampling:
 If k-space is not adequately sampled, the reconstructed image
will contain artefacts, such as distortions or aliasing.
Redundancy in Real-World Images:
 Most medical images contain redundant information, meaning
not every pixel or frequency component is essential.
 This is why image and video compression techniques (e.g.,
JPEG, MPEG) are highly effective.

CHAPTER 7 108
 Compressed Sensing

Implication for MRI:

 In theory, it may be possible to reconstruct an image
without acquiring all of k-space, by leveraging this
redundancy.
 However, this requires moving beyond traditional Fourier-
based reconstruction algorithms.
Need for Advanced Techniques:
 To reconstruct images from incomplete k-space data, MRI
must use alternative methods, such as compressed
sensing, which exploits known properties of the image
(e.g., sparsity).

CHAPTER 7 109
 Compressed Sensing

Compressed Sensing Overview:

 Compressed sensing is a technique that allows image
reconstruction from under-sampled k-space data,
challenging the traditional need for full sampling based
on the Nyquist theorem.
Random Sampling and Noise-Like Artefacts:
 In an ideal compressed sensing setup, k-space is sampled
randomly.
 This leads to incoherent artefacts in the image, which
resemble noise rather than structured distortions, making
them easier to suppress during reconstruction.

CHAPTER 7 110
 Compressed Sensing

Traditional vs compressed sensing

CHAPTER 7 111
 Compressed Sensing

Reconstruction with Constraints:

 Compressed sensing relies on additional constraints
during image reconstruction to extract meaningful data
from noisy signals.
 These constraints are based on known properties of the
image, such as sparsity.
Sparsity as a Key Constraint:
 The most common constraint is sparsity: the idea that the
image can be represented compactly in a transformed
domain.
 For example, angiographic images of blood vessels are
naturally sparse, with most voxels being either
background or vessel.

CHAPTER 7 112
 Compressed Sensing
Sparsifying Transforms:
 Images can be transformed using techniques like finite differences,
wavelets, or total variation to highlight sparse features.
 These transforms help guide the reconstruction algorithm to recover
the image accurately from limited data.
Practical Limitations:
 True random sampling of k-space is not feasible in practice due to
hardware constraints and the need to follow a trajectory through k-
space using frequency and phase encoding.
 Nonetheless, compressed sensing principles can still be applied
to structured under-sampling schemes.
Integration into Reconstruction Algorithms:
 Compressed sensing is implemented as part of advanced
reconstruction algorithms.

CHAPTER 7 113
 Compressed Sensing

 Wavelet transform yields a sparse representation of the image.

CHAPTER 7 114
 Compressed Sensing

Compressed Sensing is based on:

• Image has a sparse
representation in a known
transform domain (i.e., is
compressible).
• The need for non-linear
reconstruction that enforces
sparsity.
• Incoherence of the under-
sampling artifacts in the
transform domain (noise like).

CHAPTER 7 115
 Compressed Sensing
a) Image compression: first acquires a fully sampled image and then compresses it in the second
step.
b) Compressed sensing: builds the compression into the encoding process, thus acquiring only a
subset of the encoding steps in a random pattern. The image is subsequently reconstructed from
undersampled data with a suitable nonlinear algorithm.
F: Fourier transform; T: sparsifying transform (wavelet); CS: compressed sensing.

CHAPTER 7 116
 Compressed Sensing

 (a) Regular undersampling

generates coherent replicas of
the signal structure.
 (b) Random undersampling
generates incoherent artifacts
that appear like added noise.
 (c) radial sampling permits
undersampling along both
spatial dimensions and thus
enables a higher level of
incoherence.

CHAPTER 7 117
 Compressed Sensing

 The process of nonlinear iterative reconstruction for CS MRI.

CHAPTER 7 118
 MR Image Acquisition
T1 CONTRAST
T2 CONTRAST
PD CONTRAST

CHAPTER 7 119
 MR Image Acquisition
The acquisition process in MRI involves repeating an imaging cycle
multiple times.
 Each cycle includes:
• Excitation of tissue using RF pulses.
• Relaxation of magnetization.
• Collection of RF signals for image formation.
 The total acquisition time is determined by:
• The duration of each cycle.
• The number of cycles performed.
 The cycle duration is defined by the TR (Time of Repetition):
• TR is an adjustable protocol factor.
• It is used to select the desired image contrast (e.g., T1, T2, PD).
 The number of cycles is also adjustable:
• It depends on the required image quality.
• More cycles can improve image resolution and reduce noise but increase
scan time.

CHAPTER 7 120
 MR Image Acquisition

Reconstruction
 The image reconstruction process is typically:
• Much faster than the acquisition process.
• Automated, requiring no operator intervention or manual
adjustments.
 It involves mathematical processing of the acquired RF signals
to generate the final image.
 Once the acquisition is complete, reconstruction is performed
quickly and efficiently.
Imaging Protocol
 Each MRI procedure is governed by a protocol programmed
into the system.
 The protocol defines how the imaging process is executed and
what characteristics the resulting image will have.

CHAPTER 7 121
 MR Image Acquisition
 Key factors to consider when selecting, modifying, or designing a
protocol for a clinical procedure include:
• Imaging method:
o Choice between spin echo, gradient echo, or hybrid methods.
• Image contrast type:
o Selection of contrast weighting: PD (Proton Density), T1, T2, etc.
• Spatial characteristics:
o Slice thickness.
o Number of slices.
o Slice orientation and positioning.
• Detail and noise requirements:
o Desired image resolution.
o Acceptable levels of visual noise.
• Selective signal suppression techniques:
o Techniques to suppress unwanted signals (e.g., fat suppression).
• Artifact reduction techniques:
o Methods to minimize motion artifacts, magnetic field inhomogeneities, etc.
CHAPTER 7 122
 MR Image Acquisition

Imaging Methods
 MRI offers multiple imaging methods to generate diagnostic
images.
 The main difference between these methods lies in the
sequence and timing of:
• RF (radiofrequency) pulses.
• Magnetic field gradients.
 These methods are commonly referred to as pulse
sequences.
 Pulse sequences define how the magnetization is
manipulated during the acquisition process.
 Each imaging method requires the user to adjust specific
parameters to achieve desired image characteristics.

CHAPTER 7 123
 MR Image Acquisition

Selection of Imaging Methods and Their

Characteristics
 The choice of imaging method and its parameter settings is
primarily based on:
o The desired contrast sensitivity to specific tissue characteristics:
‒ Proton Density (PD)
‒ T1 relaxation
‒ T2 relaxation
o The required acquisition speed for the clinical application.
• Additional considerations include:
o Visual noise levels in the resulting image.
o Sensitivity to artifacts, which may vary depending on the method used.

CHAPTER 7 124
 MR Image Acquisition

Spin echo and gradient echo imaging methods.

TR: Time of Repetition

TE: Time to Echo
TI: Time of Inversion
TS: Time of Saturation

CHAPTER 7 125
 MR Image Acquisition
Spin Echo Methods
 The echo event is generated by applying a 180° RF pulse.
 These methods are typically used to:
• Minimize T2* effects.
• Produce high-quality T1 and T2-weighted images.
 More robust against magnetic field inhomogeneities.
Gradient Echo Methods
 The echo event is produced by applying an inverse
magnetic field gradient.
 These methods are:
• Faster.
• More sensitive to T2* effects.
• Useful for dynamic imaging and certain functional studies.

CHAPTER 7 126
 MR Image Acquisition

Magnetization Phases and Contrast Development

 Each MRI imaging cycle consists of two distinct magnetization
phases:
• Longitudinal magnetization phase
• Transverse magnetization phase

Contrast Determination
 The dominant contrast in the final image depends on:
• The duration of the longitudinal and transverse phases.
• The transfer of magnetization from the longitudinal to the
transverse phase.
 Proper selection of TR (Time of Repetition) and TE (Time to
Echo) controls which contrast (T1, T2, or PD) is emphasized.

CHAPTER 7 127
 MR Image Acquisition
Longitudinal Phase (T1)
 Associated with T1 contrast.
 During this phase, tissues recover their magnetization at different rates
depending on their T1 relaxation times.
 The differences in recovery rates create T1-weighted contrast.
Transverse Phase (T2)
 Associated with T2 contrast.
 In this phase, tissues lose transverse magnetization at different rates
based on their T2 relaxation times.
 These differences result in T2-weighted contrast.
Proton Density (PD) Contrast
 PD contrast is always present, as it reflects the concentration of protons
in each tissue.
 It becomes most visible when neither T1 nor T2 contrast dominates the
image.
 PD contrast is often revealed when acquisition parameters minimize T1
and T2 effects.
CHAPTER 7 128
 MR Image Acquisition

 A common characteristic of all methods is that there are

two distinct phases of the image acquisition cycle:

 The longitudinal and

transverse
magnetization phases
of an imaging cycle.
 T1 and PD contrast
are produced during
the longitudinal phase
 T2 contrast is
produced during the
transverse phase.

CHAPTER 7 129
 MR Image Acquisition

TR and TE – Timing Parameters in MRI

 The duration of the longitudinal and transverse
magnetization phases is controlled by two key protocol
parameters:
• TR (Time of Repetition)
• TE (Time to Echo)
 These parameters are essential for determining the type
of contrast (T1, T2, PD) that will dominate in the final
image.

CHAPTER 7 130
 MR Image Acquisition

TR – Time of Repetition
 TR is the time interval between:
• The start of longitudinal relaxation (after saturation).
• The moment when longitudinal magnetization is converted into
transverse magnetization via an excitation pulse.
 It defines when the image is captured relative to the
longitudinal magnetization phase.
 Since longitudinal relaxation is relatively slow, TR also
represents:
• The duration of the entire imaging cycle.
• The repetition time between cycles.

CHAPTER 7 131
 MR Image Acquisition

TE – Time to Echo
 TE is the time interval between:
• The start of transverse relaxation (after excitation).
• The moment when magnetization is measured to generate
image contrast.
 This measurement occurs at the echo event, which is
when the image is captured relative to the transverse
magnetization.
 TE determines how much T2 contrast is present in the
image.

CHAPTER 7 132
 MR Image Acquisition
Excitation in MRI
 The excitation process marks the transition from:
• Longitudinal magnetization (stable state).
• To transverse magnetization (excited, unstable state).
 This transition is achieved by applying an RF
(radiofrequency) pulse.
 The RF pulse causes the
magnetic vectors of
protons to "flip" into the
transverse plane,
initiating the signal
acquisition phase.

CHAPTER 7 133
 MR Image Acquisition

Flip Angle and Its Role

 The flip angle (α) defines how much longitudinal
magnetization is converted into transverse magnetization.
 A 90° flip angle:
• Converts all longitudinal magnetization into transverse
magnetization.
• Is typically used in spin echo sequences.
 Flip angles < 90°:
• Convert only a fraction of the longitudinal magnetization.
• Are used in gradient echo methods.
• Help to reduce acquisition time, making these methods faster
but more sensitive to field inhomogeneities.

CHAPTER 7 134
 MR Image Acquisition

The Echo Event and Signals

 The transverse magnetization phase concludes
with the echo event.
 This event generates the RF signal that is emitted
by the tissue.
 The RF signal is the fundamental data used to
construct the MRI image.
 The echo event can be triggered by:
• An RF pulse (used in spin echo methods).
• A magnetic field gradient (used in gradient echo
methods).

CHAPTER 7 135
 MR Image Acquisition

Contrast Sensitivity in MRI

 In MRI, the imaging process is typically optimized to
highlight one specific tissue characteristic:
• T1 relaxation
• Proton Density (PD)
• T2 relaxation
 The goal is to achieve maximum or sufficient
contrast sensitivity for the selected characteristic.
 This results in an image that is heavily weighted by
that characteristic, enhancing diagnostic clarity.

CHAPTER 7 136
 MR Image Acquisition

Factors Influencing Contrast Sensitivity

 Contrast sensitivity is determined by:
• The imaging method used (e.g., spin echo, gradient
echo).
• The combination of protocol parameters, especially:
oTR (Time of Repetition)
oTE (Time to Echo)
 These parameters control:
• The duration of magnetization phases.
• The timing of signal acquisition, which affects contrast.

CHAPTER 7 137
 MR Image Acquisition
1- Formation of T1 Contrast
 T1 contrast develops during the longitudinal relaxation phase
of the imaging cycle.
 After an RF pulse, the longitudinal magnetization is reduced to
zero (saturation).
 Tissues then begin to regrow their magnetization at different
rates, depending on their T1 values.
Tissue Behavior
 Tissues with shorter T1 values:
• Regrow magnetization faster.
• Reach higher magnetization levels earlier.
• Produce stronger RF signals.
• Appear brighter in T1-weighted images.
 Tissues with longer T1 values:
• Regrow more slowly.
• Appear darker in T1-weighted images.

CHAPTER 7 138
 MR Image Acquisition
Image Acquisition Timing
 The TR (Time of Repetition) determines when the image is
captured during the regrowth process.
 At the selected TR time:
• The longitudinal magnetization is converted to transverse
magnetization.
• The resulting signal is measured and displayed as pixel brightness.
• This produces a T1-weighted image.

Initial State of Tissues

 At the start of each imaging cycle, all tissues appear dark due
to zero longitudinal magnetization (saturation).
 As tissues regain longitudinal magnetization, they become
brighter in the image.

CHAPTER 7 139
 MR Image Acquisition

Effect of TR on Contrast
 A short TR:
• Interrupts the regrowth process before full recovery.
• Results in lower signal intensity and darker tissues.
• Enhances T1 contrast, as differences in regrowth rates
are more pronounced early on.
 A long TR:
• Allows tissues to fully recover their magnetization.
• Increases signal intensity and image brightness.
• Reduces T1 contrast and emphasizes Proton Density
(PD) contrast.

CHAPTER 7 140
 MR Image Acquisition

Practical Considerations
 Full magnetization recovery typically occurs when TR
exceeds ~3× the T1 value of the tissue.
 Although multiple cycles are needed to form a complete
image, the magnetization is always measured at the
same point in each cycle, as defined by the TR setting.

CHAPTER 7 141
 MR Image Acquisition

Optimal TR Selection
 TR should be chosen based on the T1 values of the
tissues being imaged.
 If TR ≈ T1, the tissue has regained about 63% of its
magnetization.
• This timing provides maximum contrast between tissues with
small differences in T1.
 Selecting the right TR is a trade-off between:
• Contrast sensitivity.
• Signal intensity.
• Clinical practicality.

CHAPTER 7 142
 MR Image Acquisition
2- Proton Density (PD) Contrast in MRI
 Proton Density (PD) refers to the concentration of hydrogen protons
in each tissue voxel.
 PD determines the maximum level of magnetization a tissue can
achieve during the longitudinal relaxation phase.
Contrast Formation
 Differences in PD between tissues result in variations in signal
intensity, which can be used to generate image contrast.
 Tissues with higher PD:
• Achieve greater magnetization.
• Produce stronger RF signals.
• Appear brighter in PD-weighted images.
 Tissues with lower PD:
• Reach lower magnetization levels.
• Emit weaker signals.
• Appear darker in the image.

CHAPTER 7 143
 MR Image Acquisition

 PD contrast is most
visible when T1 and T2
effects are minimized.
 This is typically achieved
by using:
• Long TR (to reduce T1
weighting).
• Short TE (to reduce T2
weighting).

CHAPTER 7 144
 MR Image Acquisition

PD Contrast from Magnetization Levels

 PD contrast arises from differences in the maximum
magnetization levels that tissues can reach.
 Example: Two tissues with the same T1 values but different
PDs will reach different magnetization plateaus.
• A tissue with 80% PD reaches only 80% of the magnetization level of
a tissue with higher PD.
• This difference in magnetization is the source of PD contrast.
Timing and Visibility of PD Contrast
 Early in the imaging cycle, PD contrast is present but relatively
weak compared to T1 contrast.
 As the cycle progresses:
• T1 contrast fades.
• PD contrast becomes more dominant, especially as tissues
approach their maximum magnetization.

CHAPTER 7 145
 MR Image Acquisition

Key Differences Between T1 and PD Contrast

 T1 contrast is based on the rate of magnetization growth (relaxation
speed).
 PD contrast is based on the final magnetization level (determined by
proton concentration).
 T1 contrast is strongest early in the relaxation phase.
 PD contrast is strongest later, as magnetization approaches its
maximum.
Producing PD-Weighted Images
 To emphasize PD contrast:
• Use a relatively long TR to allow tissues to reach near-maximum
magnetization.
• “Snap the picture” during the later portion of the relaxation phase.
 For almost pure PD contrast, select a TR value that is at least three
times the T1 of the tissues being imaged.
• At this point, tissues reach 95% of their magnetization, minimizing T1
influence.

CHAPTER 7 146
 MR Image Acquisition

3- T2 Contrast in MRI. Transverse Magnetization Phase

 T2 contrast develops during the transverse relaxation phase of
the imaging cycle.
 After excitation, tissues begin to lose transverse magnetization
at different rates depending on their T2 values.
Tissue Behaviour
 Tissues with longer T2 values:
• Retain transverse magnetization longer.
• Emit stronger RF signals.
• Appear brighter in T2-weighted images.
 Tissues with shorter T2 values:
• Decay faster.
• Emit weaker signals.
• Appear darker in the image.

CHAPTER 7 147
 MR Image Acquisition

Contrast Formation
 The difference in decay
rates between tissues
creates T2 contrast.
 This contrast becomes
more pronounced as time
progresses during the
transverse phase.
Visualization
 The contrast at the time
of the echo event (TE)
reflects these differences
in magnetization levels.

CHAPTER 7 148
 MR Image Acquisition

Echo Event and TE (Time to Echo)

 The echo event is when magnetization is measured and
converted into RF signals.
 The TE value determines when during the transverse
phase the image is captured.
 Longer TE values:
• Enhance T2 contrast.
• Allow more time for differences in decay to become visible.
 However, very long TE values may result in low signal
intensity, making images less useful.
 A balance must be struck between maximizing T2
contrast and maintaining adequate signal strength.

CHAPTER 7 149
 MR Image Acquisition

Interaction with T1 and PD

 T2 contrast is added to the PD contrast present at the
start of the transverse phase.
 T1 and T2 contrasts oppose each other:
• Tissues bright in T1 images tend to be dark in T2 images.
• Mixing both contrasts can cancel out useful information.
 To produce a pure T2-weighted image, it is necessary to:
• Use a long TR to minimize T1 influence.
• Use a long TE to enhance T2 contrast.

CHAPTER 7 150
 MR Image Acquisition

Mind Map Summary

CHAPTER 7 151
 MR Image Acquisition
Summary of TR and TE for image type

Contrast type TR TE Dominant feature

T1 Short Short Longitudinal relaxation
PD Long Short Proton density
T2 Long Long Transverse relaxation

CHAPTER 7 152
 MR Image Acquisition

T1-Weighted Imaging
 Best for:
• Fatty tissues (e.g., subcutaneous fat, bone marrow) (short T1)
• Anatomical detail (e.g., brain structure, spinal cord anatomy)
• Post-contrast imaging (e.g., gadolinium-enhanced scans)
• Clinical uses:
o Detecting tumours (with contrast agents)
o Evaluating brain anatomy
o Visualizing white matter in the brain
 Appearance:
• Fat: Bright
• Water/CSF: Dark
• Gray matter: Intermediate
• White matter: Brighter than grey

CHAPTER 7 153
 MR Image Acquisition

T2-Weighted Imaging
 Best for:
• Fluid-rich tissues (e.g., oedema, inflammation, CSF) (long T2)
• Pathological changes (e.g., tumours, infections, cysts)
• Clinical uses:
o Identifying oedema, inflammation, or fluid accumulation
o Detecting lesions in the brain or spine
o Evaluating joint and soft tissue injuries
 Appearance:
• Water/CSF: Bright
• Fat: Intermediate to dark
• Pathologies with high water content: Bright

CHAPTER 7 154
 MR Image Acquisition

Proton Density (PD)-Weighted Imaging

 Best for:
• Tissues with subtle differences in proton concentration
• Joint imaging (e.g., menisci, cartilage)
• Brain imaging when both T1 and T2 effects are minimized
• Clinical uses:
o Assessing cartilage integrity
o Visualizing ligaments and tendons
o Providing high anatomical detail with low contrast
 Appearance:
• Tissues with high PD: Bright
• Tissues with low PD: Dark
• Less contrast than T1 or T2, but better spatial resolution

CHAPTER 7 155
 MR Image Acquisition

MR Images

CHAPTER 7 156
 MR Image acquisition in a nutshell

CHAPTER 7 157
 MR Image acquisition in a nutshell

CHAPTER 7 158
 Summary
 NMR (now MRI) uses strong magnetic fields and radiofrequency waves to probe
atomic nuclei, especially hydrogen (protons).
 Protons possess spin, acting like tiny magnets with two energy states: parallel (low
energy) and anti-parallel (high energy).
 In a magnetic field, protons precess at the Larmor frequency, generating measurable
signals.
 Macroscopic magnetization arises from slight excess of protons in the low-energy
state, enabling tissue contrast.
 Transverse magnetization (signal) is induced by RF pulses that tip magnetization into
the xy-plane.
 RF excitation must match the Larmor frequency to effectively manipulate
magnetization.
 Relaxation processes:
• T₂ (transverse): dephasing of spins, causing signal decay.
• T₁ (longitudinal): recovery of magnetization along the z-axis.
 Bloch equations model magnetization dynamics including relaxation and RF
excitation.
 Spin echoes (via 180° pulses) help recover signal lost due to T₂* effects
(inhomogeneities).
 MRI rooms use Faraday cages to block external RF interference.

CHAPTER 7 159
 Summary
 MRI Spatial Encoding with Gradients: Gradient coils modify the
magnetic field to encode spatial information, enabling slice selection
and pixel-level resolution by manipulating the Larmor frequency across
space.
 Slice Selection and Frequency Encoding: Specific slices are selected
using RF pulses and gradients, while frequency and phase encoding
allow differentiation of positions within the slice, forming the basis for
image construction.
 k-Space Sampling and Image Reconstruction: MRI collects data in k-
space (frequency domain), and images are reconstructed using inverse
Fourier transforms. Sampling density affects field of view (FoV),
resolution, and potential aliasing artifacts.
 Advanced Acquisition Techniques: Strategies like partial Fourier
sampling, non-Cartesian trajectories (e.g., spiral or radial), and 3D k-
space acquisition improve efficiency, signal-to-noise ratio, and reduce
motion sensitivity.
 Compressed Sensing and Practical Considerations: Compressed
sensing allows under-sampling of k-space by exploiting image sparsity,
enabling faster scans. Practical challenges include T₂ decay during
readout and the need for optimized pulse sequences.
CHAPTER 7 160