Mri Gradient

3/4/2011
MRI Chap 1
FYS-KJM 4740 Bloch equations and main

principles
FYS-KJEM 4740 Frédéric Courivaud (PhD) 2
B0
The Bloch equation z
B0
Larmor equation ωL
z Mz
M
y
ωL
ωL
dM x
M Mz = γ (M × B )
dt
y ω0 (rad/s) = γ (rad/s/Tesla) x B0 (Tesla)
ωL
x
γhydrogen = 2.68 *108 rad/s/Tesla
/2π
Figure Error! No text of specified style in document.-1. The magnetic moment M rotates around
The magnetic moment M rotates around the static B-field at the Larmor frequency
Joseph Larmor
fL (MHz) = γ (MHz/Tesla) x B0 (Tesla)
γhydrogen = 42.58 MHz/Tesla
FYS-KJM 4740 3 FYS-KJM 4740 AtleBjørnerud 4
Flipping away the Magnetization

Rotating frame of reference
from its equilibrium
In MRI, short RF pulses are used to
B0
ω0 z B0 To change the direction of the magnetization
z` M z
M
To get M to rotate around x or y axis,
A linearly polarized magnetic field B1 is used
during short time (pulse) to get M to rotate
y around B1 axis
y` y`
x
B1
x` Rotating frame x`
Laboratory frame
FYS-KJEM 4740 5 FYS-KJEM 4740 6
1
3/4/2011
Additional B1-field induces phase

Baseline, no phase coherence in
coherence in transverse (X,Y)
transverse (X,Y) plane
plane
B0 B0
B1
Zero net Mx and My components Net Mx,y component > 0
Rotation z
The magnetic field due to

the RF pulse, B1, is generated
by two circularly polarized
fields with opposing direction B1
y
cos( −Ωt ) cos( Ωt )  cos( Ωt )
of rotation with angular B = B + B 1 1+ 1− = B1  sin( − Ωt )  + B1  sin( Ωt )  = 2 B1  0 
 0   0   0 
frequency +/- Ω x
2B1cos(Ωt)
B1 Ω
-Ω
time B1- B1+ The RF-coil generates a magnetic field B1 along the x-axis
FYS-KJEM 4740 12
2
3/4/2011
Bloch equation RF-eksitasjon med Larmor frekvens (rotating frame)
z’
dM
= γM × B eff z
dt
z’
Ω= γB0
y Ω
Beff= B1 y Ω
x’ Ω
x’ Ω
x y’
x y’
The ‘rotating frame’ (x’, y’, z’-coordinates)
 0 
dM
= γM × B eff
Beff= B0 + B1 + Ω/γ Ω =  0 
dt − Ω
FYS-KJEM 4740 Frédéric Courivaud (PhD) 13 FYS-KJEM 4740 14
z’ ω
:
Using Matrix formalism B0
α
1
RF pulse
1
M
ω1 = −γB1
0 0 0  M x 
dM   
= γ 0 0 B1x  M y  RF pulse duration is proportional to the
dt y’
0 − B 0  M z  B1 wanted flip angle, α.
 1x z’ ω1
x’
α1
M y ' = A sin( γB1x ' t ) + B cos(γB1x ' t )
ω1 = −γB1 tB1 = α / γ B1 M
Using the boundary conditions My’=My’(0) and Mz’=Mz(0) at t=0, we get

1 0 0  Mx (0) y’
   B1
M(t) = 0 cos(ω1t) sin(ω1t)  My (0) = R ⋅ M0
0 − sin(ω t) cos(ω t) M (0)  x’
 1 1  z 
FYS-KJEM 4740 15 FYS-KJEM 4740 Frédéric Courivaud (PhD) 16
Relaxation Relaxation
dM x M
=− x dM z M − M0
dt T2 =− z
dt T1
dM y My
=−
dt T2
1 1
= + γ∆B0
T2* T2
FYS-KJEM 4740 Frédéric Courivaud (PhD) 17 FYS-KJEM 4740 Frédéric Courivaud (PhD) 18
3
3/4/2011
Relaxation Condition: relaxation during RF excitation is neglected
dM dM
= γM × B eff − R (M − M 0 ) = − R (M − M 0 )
dt dt
M z (t ) = M 0 [1 − exp(− t / T1 )] + M z (0) exp(− t / T1 )
M xy (t ) = M xy ( 0) exp( −t / T2 )
1 
 0 0
 T2 
 0  M x 
R=0

1
0

M 0 =  0  M =  M y 
T2
 1  M 0   M z 
0 0 
 T1 
Summary with excitation and

relaxation
dM
= γM × B eff − R (M − M 0 )
dt
Chap. 2
− 1 / T2 0 0  M x'   0 
dM 
= 0 − 1 / T2 γB1x   M y '  +  0 
dt Slice-Selective RF excitation
 0 − γB1x − 1 / T1   M z '   M 0 / T1 
Image formation
Gradients coding in space

Z
sample volume voxel
Slice Selective RF pulse B0
Y
X
Use of field gradient pulses in the 3 directions
FYS-KJEM 4740 Frédéric Courivaud (PhD) 23 Frédéric Courivaud (PhD) KJEM/FYS 4740 24
4
3/4/2011
Magnetic Field Gradients Principles of Slice Selection
y
δBG,z (z) = Gzz
δBG,z (y) = Gyy
z
δBG,z (x) = Gxx
FYS-KJEM 4740 Frédéric Courivaud (PhD) 25 Frédéric Courivaud (PhD) KJEM/FYS 4740 26
Magnetfelt gradient RF pulse generation

FAVOURABLE SITUATION
Gradient styrke (mT/m)
Stigningstid (ms)
Gradient styrke
Slew rate (mT/m/ms) =
Stigningstid B1 applied for a certain duration (ms) B1 pulse frequency profile (HZ)
Effect of Magnetic Field Gradient Slice selective RF excitation

Example: use of z field gradient
B0  
 B1 
B0 ω = γ × (B0 + gz ) B eff = B 0 + G ⋅ r + B 1 +
Ω
= 0 
γ  Ω
 B0 + G z z1 − 
Z  γ
ω0
θ µ If we set: Ω=γ(B0 + Gzz1), then the effective field at z=z1 becomes:
Spatial information included in the
precession frequency  B1 
precession frequency ω : Applied field gradient (g) B eff =  0  Slice-selective excitation @ z=z1
ω0 = γ × B0  0 
What would happen if B0 + Gzz1 –Ω/γ >> B1 ??
Frédéric Courivaud (PhD) KJEM/FYS 4740 29 FYS-KJEM 4740 Frédéric Courivaud (PhD) 30
5
3/4/2011
All effects together: excitation, All effects together: excitation,

precession and relaxation precession and relaxation
Transverse (Mxy) relaxation
Ω=ωL ;Beff=B1x Precession around z-axis
 − 1 / T2 γG ⋅ r 0  M x '   0 
 − 1 / T2 γG ⋅ r 0  M x '   0  dM     
dM      =  − γG ⋅ r − 1 / T2 γB1x  M y '  +  0 
=  − γG ⋅ r − 1 / T2 γB1x  M y '  +  0  dt  
dt    0 − γB1 x − 1 / T1  M z '   M 0 / T1 
  
 0 − γB1 x − 1 / T1  M z '   M 0 / T1 
  
Excitation around x-axis Longitudinal (Mz) relaxation
Transverse magnetization, Mxy, Relaxation and precession Transverse Magnetization, Mxy, Excitation and Precession
MT = Mx + jMy B1x=0 MT = Mx + jMy T2=∞
derive! Condition: Mz≈M0 (How can this be achieved?)
 t
(
MT = MT (0)exp − jγr ⋅ ∫ G(t)dt)exp− T  dM T
2 = − jγ (G ⋅ r) M T + jγB1 M 0
dt
Longitudinal magnetization, Mz derive! General Solution:
  t   t 
M z (t ) = M 0 1 − exp −   + M z (0) exp −  As we have seen  t 
  T1    T1 
before M T = A(t ) exp − jγr ⋅ ∫ G (t ' )dt ' 
 
 t1 
Transversal magnetisering, Mxy, Eksitasjon og presesjon For a constant gradient along the z-axis: G(t) = Gz
T2=∞
Gz(t) t
Dersom vi sier at RF puls starter ved –T/2 og varer i T

sek: B1(t)
T /2
 T /2

M T (T / 2, r ) = jγM 0 ∫ B1 (t ) exp − jγr ⋅ ∫ G (t ' )dt ' dt
−T / 2  t  -T/2 0 T/2
t T /2
Gz(t) M T (T / 2, z ) = jγM 0 exp(− jγzG z T / 2 ) ∫ B (t ) exp( jγzG t )dt
−T / 2
1 z
-T/2 0 T/2 Phase dispertion Fourier transform of the B1 ”envelope”

(green shape)
6
3/4/2011
For a constant gradient along the z-axis: G(t) = Gz Elimination of the phase dispertion in x-y plane use of an extra gradient
of opposite polarity and half the length: -Gz
Gz(t)
T /2
M T (T / 2, z ) = jγM 0 exp(− jγzGz T / 2 ) ∫ B1 (t ) exp( jγzG z t )dt
−T / 2 B1(t)
Slice profile = Fourier transform of B1(t) -T/2 0 T/2

The phase of MT(z) in the x-y- plane is a function of z
This gives
kT
B1 ( k )
M T (T , z ) = jM 0 ∫
− kT
Gz
exp ( jkz )dk
k=γGzt and kT=γGzT/2.
We wish to have a ‘block’ excitation: MT(z) = M0sin(α)

between –d/2 og d/2 og MT=0 resten RF pulse generation
Find the B1(t) profile from the Fourier integral:
Develop at home!
d/2
sin(γG z t ⋅ d / 2)
B1 (t ) = G z ∫ exp( jγG t ⋅ z)dz = G d ⋅
−d / 2
z z
γG z t ⋅ d / 2
B1 pulse frequency profile (HZ) B1(t) applied for a certain duration (ms)
Slice selective RF excitation … in practice

Gz(t) Gz(t)
B1(t) B1(t)
-T/2 0 T/2 -T/2 0 T/2
Gradient can not be perfectly rectangular (Hardware limitations

Slew rate = Gradient strength / Time to reach the strength
Slew rate (mT/m/ms)
7
3/4/2011
B1(t) and truncation Summary selecting rephasing lobe

lobe
B0
B1(t) duration can not be
too long
The shape is
selected slice
“truncated” as a
compromise between
duration and achieved
RF pulse at the cost of
the “slice profile”
selecting rephasing
lobe lobe
time
Frédéric Courivaud (PhD) KJEM/FYS 4740 43 Frédéric Courivaud (PhD) KJEM/FYS 4740 44
Slice selective 180°inversion RF

Slice profile & “cross-talk”
pulse
coherence transfer
pathway
GZ
dephasing rephasing
Slice Selection Slice Selection
Frédéric Courivaud (PhD) KJEM-FYS 4740 47 Frédéric Courivaud (PhD) KJEM-FYS 4740 48
8
3/4/2011
Slice Selection Slice Selection
Frédéric Courivaud (PhD) KJEM-FYS 4740 49 Frédéric Courivaud (PhD) KJEM-FYS 4740 50
Slice Selection
Frédéric Courivaud (PhD) KJEM-FYS 4740 51 FYS-KJEM 4740 Frédéric Courivaud (PhD) 52
RF excitation RF inversion
9
3/4/2011
Exercise
• We want a slice excitation with a given slice
thickness of 3mm and a gradient strength of
10mT/m. We Assume the use of a sinc pulse
shape for B1(t). Image formation principles
• Find the half-width duration of the B1(t) and digital sampling

envelope (shape)
• Now, having this B1(t) envelope, find the
intensity of B1 (µT) to get a 90 degree RF
pulse.
Linear combination of gradients x

Gradients coding in space
and y
Z Gy
sample volume
B0
Gx Y
selected slice
Y
X
The precession of the spins depends on the field gradients
Slice selective excitation: done! Gx x and Gy y
the precession depends on the position
Frédéric Courivaud (PhD) KJEM/FYS 4740 57
Reminder: spin-echo experiment SE magnetization evolution

90º pulse 180º pulse
M0 z z z
t=0 t=τ
x x y
x y y
z z z
t > 2τ echo t = 2τ
x x y x y
y
10
3/4/2011
Use of gradients to make an image 1st 2D MR image

echo
slice selection Gz
Gx
Gy
Combination of Gx and Gy to “rotate” the total gradient orientation

reconstruction by back projection
Nature Editor letter to Paul C.

Zeugmatography
Lauterbur
“With regret I am returning your manuscript which we feel is not sufficiently wide significance for
inclusion in Nature. This action should not in any way be regarded as an adverse criticism of your
work, nor even an indication of editorial policies on studies in this field. A choice must inevitably be
made from the many contributions received; It is not even possible to accommodate all those
manuscripts which are recommended for publication by referees.”
Paul C. Lauterbur answered:

“Several of my colleagues have suggested that the style of the manuscript was too dry and spare, and
that the more exuberant prose style of the grant application would have been more appropriate. If you
should agree, after reconsideration, that the substance meets your standards,… I would be willing to
incorporate some of the material below in the revised manuscript…”
Nature answered short and positive:

“would it be possible to modify the manuscript so as to make the applications more clear?”
Relationship between a three-dimensional object, its two-dimensional projection along the Y-axis,
and four one-dimensional projections at 45° intervals in the XZ-plane. The arrows indicate the gradient directions.
Lauterbur PC. Image formation by induced local interactions: examples of employing nuclear magnetic resonance. Nature 1973; 242: 190-
191.
From last Chap.: definition of k

k notation
variable
d/2 Use of k notation is VERY IMPORTANT in MRI, we will see why…
sin(γG z t ⋅ d / 2)
−d / 2
z z
γG z t ⋅ d / 2
kx 
This equation can be simplified by introducing the following notation t
 
k = γ ∫ G(τ )dτ = ky 
kx 
t
 
0
 kz 
k=γGzt in general, this can be written k = γ ∫ G(τ )dτ = ky 
0
 kz 
sin(k ⋅ d /2)
B1 (t) = Gz d ⋅
k ⋅ d /2
11
3/4/2011
Phase effect given by a gradient y y
pulse (e.g. y direction) Gy
t
α (r, t ) = −γ ∫ G y (t ) ⋅ r ⋅ dτ 0
0
From Bloch’s equation, the Transverse magnetization is then given by (Assume
negligible relaxation here): α
a x b x
0 0
 t

M xy = M T (r, t ) = M T (r ,0) ⋅ exp − jγ ∫ G y (t )rdτ  The phase angle of the transverse magnetization vector before (a) and
 0  after (b) the application of a magnetic field gradient in the y-direction.
t
k x 
k = γ ∫ G (τ )dτ = k y  M T (t ) = ∫∫ ρ (r) ⋅ exp(− jk ⋅ r )dr
 
 k z 
0 slice
kx 
MR signal = integral of transverse t  
k-space k = γ ∫G(τ )dτ = ky 
magnetization over entire volume 0
 kz 
Limiting discussion to a slice (2D – xy plane), magnetization distribution is given by the

 t 
MT (t) ∝ S(t) ∝ ∫∫∫ ρ(r)exp− jγ ∫ Gy (t)rdτ dr
2-dimensional Fourier transform of the spin distribution across the slice
 0  M T (t) = ∫∫ ρ(r)⋅ exp(− jk⋅ r )dr

slice
NB: Fourier transform

ρ(r) is obtained from the inverse Fourier transform of MT(t) under the influence of a
known gradient configuration
F(k) = ∫∫∫ f (r)exp(− jkr)dr Fourier transform from k space to
geometric “space” of a continuous function f(r) 1
R ρ (x,y) =
2π ∫∫M
kx ky
T ( )
(kx ,ky )exp j(kx x + ky y) dkx dky
 t 
S(t) ∝ ∫∫∫ ρ(r)exp − jγ ∫ G(t)rdτ dr
  k-space = visualization of the distribution of spatial frequencies in the image.
R 0
k-space = Fourier transform of the MR image
Reminder: Fourier… and

k-space illustrations
frequencies
FT
MT(t)
{FT}-1
t
k x 
k = γ ∫ G (τ )dτ = k y 
{FT}-1
 
FT 0
 k z 
1
ρ ( x, y ) = ∫ ∫M T (k x , k y ) exp ( j (k x x + k y y ))dk x dk y
2π kx k y
12
3/4/2011
Low and high frequencies
Introduction to k-space
image sampling
What describes waves (signal) Spin Echo: freq. encoding

echo
Frequency information
slice selection Gz
Sum of waves with
different frequencies
Read-out direction
Gx
(frequency
encoding)
Difference of phases
between the 2 sums
Same frequencies as
above dephasing
sampling of the signal which

contains frequency information
(x-axis)
acquisition of a profile
Phase encoding The phase angle of a spin in a slice at a time t is pulsed field gradient along x
given by:
echo t′
∫ ω (x, y, t )dt = γG
0
yn ty + γG xtx Gx
Gradient y “on”
Definition of k:
slice selection Gz
∫
t′ tx
ki = γ Gi (t )dt (in the direction i) surface
Read-out direction 0
Gx
(frequency
encoding) The total transverse magnetisation is a function of kx, ky and the position in the
slice: MT(kx, ky)
1
phase encoding
Spatial information in
Gy Image reconstruction: m(x, y ) =
2π
∫ ∫ M (k ,k )exp[i(k x + k y )]dk dk
T x y x y x y
kx ky
“y” direction
2D Fourier Transform
13
3/4/2011
Acquisition of one “profile” ≡ k-

Repeated Phase encoding
space line acquisition of a profile
echo
180º rf pulse
slice
selection
ky slice selection Gz
read-out
Read-out direction
(frequency encoding) Gx
Acquisition of a profile
phase encode
phase encoding
Gy
Spatial information in “y”
direction
kx
Repeated acquisition of profiles Schematic of k-space acquisition

kx profiles Frequency information: x-axis
One ky line
1st ky
line
the field of view:
FOV ky
Phase
information
y-axis 2nd ky
FOV depends on the line
- gradient strengths
- sampling time of a profile
Image generation
Signal intensity
ky distribution in the
phase encoding selected slice IMAGE
kx 2D FT
frequency encoding
ωy y
1
m(x, y ) = ∫ ∫ M (k ,k )exp[i(k x + k y )]dk dk
2π k x ky
T x y x y x y
ωy
frequency ωx
frequency ωx x
14
3/4/2011
CONTINUE PPT FROM HERE Discrete sampling
MR-signal (MT)
U(t)
Tread
Sampling intervall: -Tread/2 - Tread/2
Schematic of k-space acquisition Discrete sampling and PSF

Frequency information: x-axis
One ky line
Discrete sampling
1st ky
line
Phase
information
y-axis 2nd ky
line
Pulse sequence K-space vs image space
ky
S(r) F(k)
∆y ∆ky
11
.ρ(x,y)
FT
0.5
RF-Excitation, detection
B1(t)
RF
FoVy
00
kx
0.5
{FT}-1
00 20 40 60 80 100
100 120 140
140
time (ms)
Slice-select G
z
Ny∆ky
Gy_max kx_max
Phase-encode G ∆x ∆kx ts
y
FoVx
Nx∆kx
Gx_r -ωmax = -γGxFoVx/2 ωmax = γGxFoVx/2
Ty
Read-out G
(freq. encode) x 1 k x 
(k x , k y ) exp ( j (k x x + k y y ))dk x dk y
t
Tread
ρ ( x, y ) =
2π ∫ ∫M
kx ky
T
k = γ ∫ G(τ )dτ = k y 
0
k z 
15
3/4/2011
K-space egenskaper
Resolution (x):
Maximum frequency in read-out (x) direction
2π
δx =
γG x N x t s ± ω max = ±γG x FoV x / 2
Image FoVy
Object FoVy
Field of view (x):
Min sampling rate (x):
2π 2π
λ x ,max = = = FoVx 1 / t s ≥ γG x FoV x / 2π
k x ,min γG x t s
‘Sampling rate ‘ (y):

Field of view (y):
πN y N y = γG y _ max T y FoV y
λ y ,max = FoV y =
γG y _ maxTy
Increase 1/ts, Nx FFT 1D: Truncation Artefact
Object FoV
Image FoV
discarded discarded
FFT 2D: Truncation artefact Truncation artifact

Ringing- (or truncation) artifacts in regions with high spatial frequencies (edges)
in a phantom. The artifacts are more evident in the right image due to a lower
matrix (N=112, vs N=256 in the left image).
16
3/4/2011
K-space sampling principles

echo echo
slice selection Gz slice selection Gz
Gx Gx
Gy
Combination of Gx and Gy to “rotate” the total gradient orientation

reconstruction by back projection
Gradient Echo
90º pulse echo
M z z z z z
0
x x y x y x
x y y y
TE
Gi
17
3/4/2011
B1(t) for slice selective excitation Definition of k variable

d/2 d/2
sin(γG z t ⋅ d / 2) sin(γG z t ⋅ d / 2)
−d / 2
z z
γG z t ⋅ d / 2
−d / 2
z z
γG z t ⋅ d / 2
This equation can be simplified by introducing the following notation
kx 
t
 
B1(t)
k=γGzt in general, this can be written k = γ ∫ G(τ )dτ = ky 
0
 kz 
sin(k ⋅ d /2)
B1 (t) = Gz d ⋅
k ⋅ d /2
k notation
Use of k notation is VERY IMPORTANT in MRI, we will see why…
kx 
t
 
k = γ ∫ G(τ )dτ = ky 
0
 kz 
18

Mri Gradient

Uploaded by

Copyright:

Available Formats

Mri Gradient

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mri Gradient

Uploaded by

Copyright:

Available Formats

3/4/2011

FYS-KJM 4740 Bloch equations and main

FYS-KJEM 4740 Frédéric Courivaud (PhD) 2

Flipping away the Magnetization

FYS-KJEM 4740 5 FYS-KJEM 4740 6

Additional B1-field induces phase

Zero net Mx and My components Net Mx,y component > 0

The magnetic field due to

Bloch equation RF-eksitasjon med Larmor frekvens (rotating frame)

FYS-KJEM 4740 Frédéric Courivaud (PhD) 13 FYS-KJEM 4740 14

Using the boundary conditions My’=My’(0) and Mz’=Mz(0) at t=0, we get

Relaxation Condition: relaxation during RF excitation is neglected

Summary with excitation and

Gradients coding in space

Slice Selective RF pulse B0

Use of field gradient pulses in the 3 directions

Magnetic Field Gradients Principles of Slice Selection

δBG,z (y) = Gyy

Magnetfelt gradient RF pulse generation

Gradient styrke (mT/m)

FYS-KJEM 4740 Frédéric Courivaud (PhD) 28

Effect of Magnetic Field Gradient Slice selective RF excitation

What would happen if B0 + Gzz1 –Ω/γ >> B1 ??

All effects together: excitation, All effects together: excitation,

Excitation around x-axis Longitudinal (Mz) relaxation

MT = Mx + jMy B1x=0 MT = Mx + jMy T2=∞

derive! Condition: Mz≈M0 (How can this be achieved?)

Dersom vi sier at RF puls starter ved –T/2 og varer i T

-T/2 0 T/2 Phase dispertion Fourier transform of the B1 ”envelope”

Slice profile = Fourier transform of B1(t) -T/2 0 T/2

We wish to have a ‘block’ excitation: MT(z) = M0sin(α)

Find the B1(t) profile from the Fourier integral:

Slice selective RF excitation … in practice

-T/2 0 T/2 -T/2 0 T/2

Gradient can not be perfectly rectangular (Hardware limitations

B1(t) and truncation Summary selecting rephasing lobe

Slice selective 180°inversion RF

Slice Selection Slice Selection

Slice Selection Slice Selection

• Find the half-width duration of the B1(t) and digital sampling

Linear combination of gradients x

Frédéric Courivaud (PhD) KJEM/FYS 4740 57

Reminder: spin-echo experiment SE magnetization evolution

FYS-KJEM 4740 Frédéric Courivaud (PhD) 60

Use of gradients to make an image 1st 2D MR image

Combination of Gx and Gy to “rotate” the total gradient orientation

Nature Editor letter to Paul C.

Paul C. Lauterbur answered:

Nature answered short and positive:

From last Chap.: definition of k

Phase effect given by a gradient y y

pulse (e.g. y direction) Gy

Limiting discussion to a slice (2D – xy plane), magnetization distribution is given by the

 0  M T (t) = ∫∫ ρ(r)⋅ exp(− jk⋅ r )dr

NB: Fourier transform

FYS-KJEM 4740 Frédéric Courivaud (PhD) 70

Reminder: Fourier… and

Frédéric Courivaud (PhD) KJEM/FYS 4740 72