Professional Documents
Culture Documents
November 2017
Norges Teknisk-Naturvitenskapelige Universitet, Trondheim, Norway.
1
Contact: anderhja@stud.ntnu.no
2
Contact: thanhqn@stud.ntnu.no
Contents
Introduction v
Todo list v
2 Signal Processing 4
2.1 What is a signal? . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Looking at things . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 What is a filter? . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Ultrasound 6
3.1 What is ultrasound? . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.1 How to make an image? . . . . . . . . . . . . . . . . . . . 6
3.1.2 Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.3 Sound wave in medium . . . . . . . . . . . . . . . . . . . 7
3.1.4 Beam steering and focusing . . . . . . . . . . . . . . . . . 7
3.1.5 Array system . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1.6 Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Limitations and artefacts . . . . . . . . . . . . . . . . . . . . . . 9
3.2.1 Shadows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.2 Reverberations . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.3 Ring down artefacts . . . . . . . . . . . . . . . . . . . . . 9
3.2.4 Side lobes . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.5 Angle dependency . . . . . . . . . . . . . . . . . . . . . . 9
3.2.6 Foreshortening . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.7 Out of plane motion . . . . . . . . . . . . . . . . . . . . . 9
3.3 Clinical applications . . . . . . . . . . . . . . . . . . . . . . . . . 9
i
CONTENTS ii
5 Magnetic Resonance 11
5.1 Clinical applications . . . . . . . . . . . . . . . . . . . . . . . . . 11
II Ultrasound 15
7 On Wavy Things 16
7.1 Developing the wave equations . . . . . . . . . . . . . . . . . . . 16
7.2 Transmission and reflection . . . . . . . . . . . . . . . . . . . . . 19
7.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
11 Scattering 64
11.1 The Born Approximation . . . . . . . . . . . . . . . . . . . . . . 64
11.2 Scattering and absorption . . . . . . . . . . . . . . . . . . . . . . 65
A MATLAB examples 77
III CT 84
13 Some electromagnetics 85
13.1 Maxwell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.2 Make it hetero . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
13.3 Quick and dirty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
14 Make it smile 90
14.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
14.2 Getting an image back . . . . . . . . . . . . . . . . . . . . . . . . 92
14.2.1 Inverse Radon Transform . . . . . . . . . . . . . . . . . . 93
14.2.2 Inverse Fourier Transform . . . . . . . . . . . . . . . . . . 93
14.2.3 Simple Back Projection . . . . . . . . . . . . . . . . . . . 93
14.2.4 Filtered Back Projection . . . . . . . . . . . . . . . . . . . 94
B MATLAB examples 97
IV MR 100
15 The Easy Parts 101
15.1 Love and attraction . . . . . . . . . . . . . . . . . . . . . . . . . 101
15.1.1 Electrons and protons . . . . . . . . . . . . . . . . . . . . 101
15.1.2 Some mechanics . . . . . . . . . . . . . . . . . . . . . . . 102
15.2 Hit Me Baby One More Time . . . . . . . . . . . . . . . . . . . . 104
CONTENTS iv
16 Gradients 111
16.1 Adding a gradient field . . . . . . . . . . . . . . . . . . . . . . . . 111
18 NMR 117
18.1 NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
References 119
Introduction
v
Todo list
vi
Part I
1
Chapter 1
2
CHAPTER 1. LET’S START EASY 3
Signal Processing
4
CHAPTER 2. SIGNAL PROCESSING 5
Ultrasound
3.1.2 Modes
As mention there are many technique that can be used to get an image and
depend on the image quality and functionality, we define different modes:
• A-mode (Amplitude): measure the the arrive time of the echoes relative
to the time it was transmitted for a single pulse.
• B-mode (Brightness): measure a large number of pulse-echo, send in dif-
ferent angular direction, which can be used to generate 2D or 3D image.
6
CHAPTER 3. ULTRASOUND 7
3.1.6 Resolution
The spatial resolution of ultrasound imaging is determine by a number of factors:
• The axial resolution is determine by the pulse length.
Both axial and lateral resolution will decrease with the depth of the tissue due
to the absorption and scattering.
CW PW
+ Range resolution No Yes
− Frequency aliasing No Yes
− Range ambiguity No Yes
− Limited max. c measure No Yes
− Smaller transducer aperture (reduced resolution) Yes No
What are advantages and drawbacks of pulse waves and continuous waves
in relation to Doppler measurements? Some are listed in Table 3.1.
CHAPTER 3. ULTRASOUND 9
3.2.1 Shadows
3.2.2 Reverberations
3.2.3 Ring down artefacts
3.2.4 Side lobes
3.2.5 Angle dependency
3.2.6 Foreshortening
3.2.7 Out of plane motion
Missing
An example of medical image made by the imag-
figure
ing method
Chapter 4
Missing
An example of medical image made by the imag-
figure
ing method
10
Chapter 5
Magnetic Resonance
Missing
An example of medical image made by the imag-
figure
ing method
11
Chapter 6
Optical Coherence
Tomography
6.1 Introduction
From the technique’s name, you probably know a lot already. We are dealing
with light, its coherence, and we are doing scanning. Add refer-
ence.
12
CHAPTER 6. OPTICAL COHERENCE TOMOGRAPHY 13
For the moment, assume both the sample and the reflector act as mirrors (that
is, no depth dependcence on backscattering in sample). Then we have to new
mirrors with reflectivities rs and rr ,
−rrs 0
Mrs = . (6.5)
0 −rrs
Also the beams travel some distance lrs ,resulting in the transfer matrix
Ml = ∗ ∗ ∗ ∗ ∗ (6.6)
Missing
figure A sketch of the system set up
Finally, the electrical field entering the detector will look like
i 1 i 1
ED = rr √ √ Ei e−i2klr + rs √ √ Ei e−i2kls ≡ ER e−i2klr + ES e−i2kls (6.7)
2 2 2 2
We define the autocorrelation function as
Γ12 (τ )
γ12 (τ ) = √ (6.9)
I1 I2
often also referred to as g (1) (τ ) in the literature. This function will take values
such as
Laser light g (1) (τ ) = e−iω0 τ
2
Gaussian chaotic light g (1) (τ ) = e−iω0 τ −π(τ /τc ) /2
for central frequency ω0 and coherence time τc . For a fully coherent system,
|g (1) (τ )| = 1.
The intensity detected will be
from which you easily see the effect of constructive (∆l = nλ/2) and destructive
(∆l = (2n + 1)λ/4) interference as you change ∆l (integer n).
Another approach is by using an electrical field of the form Add refer-
ence.
Ei (ω, t) = s(ω)e−i(ωt+kz) , (6.10)
that is, a wave packet where s(ω) is the amplitude spectrum. The beam is split
as before, but we assign a spatial response function H to the sample, so that
the field reflected from the sample is
√
Es = tr tt Ei H (6.11)
Missing
An example of medical image made by the imag-
figure
ing method
Part II
Ultrasound
15
Chapter 7
On Wavy Things
∂ψ
u(z, t) = (Vibration velocity), (7.1)
∂t
∂u
a(z, t) = (Acceleration of plane). (7.2)
∂t
1 ∂ψ
p≈− (Pressure), (7.3)
κ ∂z
∂p 1 ∂ ∂ψ 1 ∂ ∂ψ 1 ∂u
=− =− =− . (7.4)
∂t κ ∂t ∂z κ ∂z ∂t κ ∂z
The mass has to be conserved as the wave passes, keeping the mass ∆m
between two neighbouring planes, say at equilibrium positions z and z + dz,
constant. Define ρ0 as the mass density in the unstrained material in the volume
16
CHAPTER 7. ON WAVY THINGS 17
We now turn to Newton’s laws. Pressure is force per unit area. The area
of each plane is A, and thus the force at the volume element ∆V equals the
pressure difference between the two planes at z and z dz,
∆F ∂p
= p(z, t) − p(z + dz , t) ≈ p(z, t) − [p(z, t) + dz ]
A ∂z
∂p
= − dz .
∂z
As ∆V0 = A dz, we get
∂p
∆F = − ∆V0 (Net force on ∆V ). (7.6)
∂z
From (7.2) we know the acceleration of ∆V , hence
∂u ∂p
∆F = ∆ma = ∆m = − ∆V0 .
∂t ∂z
We remember that the mass density of unstrained material is ρ0 = ∆m/∆V0 ,
giving
Important 7.1.3: Newtons acceleration equation
∂u ∂p
ρ0 = − . (Newtons acceleration equation) (7.7)
∂t ∂z
This equation represents kinetic energy, as the traditional ma = F.
CHAPTER 7. ON WAVY THINGS 18
The wave equations are made by combining (7.4) and (7.7) by differentiating
(7.7) with respect of t and z. Starting with t,
∂ ∂u ∂ ∂p
ρ0 =−
∂t ∂t ∂t ∂z
∂2u ∂2p
ρ0 2 = −
∂t ∂t∂z
∂2u 1 ∂2u
ρ0 2 = .
∂t κ ∂z 2
Similarly, by differentiation with respect to z,
∂ ∂u ∂ ∂p
ρ0 =−
∂z ∂t ∂z ∂z
∂2u ∂2p
ρ0 =− 2
∂z∂t ∂z
2 2
∂ p ∂ p
ρ0 κ 2 = 2 .
∂t ∂z
We define the wave propagation velocity c as
Important 7.1.4: Wave Propagation Velocity
1
c= √ , (7.8)
ρ0 κ
∂2u 1 ∂2u
− = 0, (7.9)
∂z 2 c2 ∂t2
2 2
∂ p 1 ∂ p
2
− 2 2 = 0. (7.10)
∂z c ∂t
In three dimensions you have to introduce some vectors and be more careful
when developing the two equations. However, the results are very similar to
(7.9) and (7.10), namely
Important 7.1.5: Homogeneous Wave Equation in 3D
1 ∂2u
∇2 u − = 0, (7.11)
c2 ∂t2
2
1 ∂ p
∇2 p − 2 2 = 0, (7.12)
c ∂t
Write two equations of physical phenomena that are the basis for ultra-
sound wave propagation in tissue and explain the material parameters
and the field variables that are found in the equations. What energy
forms do the two equations represent.
Answer
Newtons acceleration equation:
∂u(r, t)
ρ(r) = −∇p(r, t), (7.7 revisited)
∂t
representing kinetic energy. Linear elasticity equation:
∂p(r, t) ∂p(r, t)
κ(r) + hab (r, t) ∗t κ(r) = −∇ · u(r, t),
∂t ∂t
first term representing elastic potential energy, the second representing
absorption.
Arguments, variables and parameters are
r Spatial coordinate
t Time coordinate
p(r, t) Acoustic pressure
u(r, t) Vibration velocity
ρ(r) Mass density
κ(r) Elastic compressibility
hab (r, t) Acoustic power absorption kernel
zj = ∗ ∗ ∗, j = 1, 2. (7.13)
CHAPTER 7. ON WAVY THINGS 20
7.3 Energy
Example 7.3.1: Exam 2012.2b
Answer
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
21
Chapter 8
1 ∂2u
∇2 u − = −s(r, t), (8.2)
c2 ∂t2
1 ∂2p
∇2 p − 2 2 = −s(r, t). (8.3)
c ∂t
22
CHAPTER 8. GIVE ME SOME HELMHOLTZ 23
representing fluid-rigid surface interfaces with fluid being able to slip along the
surface.
We introduce power loss through the Sommerfeld radiation condition
∂p 1 ∂p
lim r − = 0. (Sommerfeld radiation condition) (8.9)
r→∞ ∂r c ∂t
This removes solutions entering from the infinity, making sure s is a source and
not a sink of energy.
Let us assume that the source s is confined to some volume in space and that
there is no reflections or refractions of the wave. Now we start to transform,
and we denote the Fourier transform of a function as capital letters. (Fourier
who? See Example 8.1.3 Fourier transform) Thus (8.3) takes the form
ω2
(k 2 − )P (k, ω) = S(k, ω)
c2
S(k, ω)
P (k, ω) = 2 ≡ G(k, ω)S(k, ω), (8.10)
(k − ω 2 /c2 )
hence
1
G(k, ω) = . (Fourier of Green’s Function) (8.11)
(k 2 − ω 2 /c2 )
In position and time space Green’s function, through transforming back (8.11),
gets the form
δ(t − r/c)
g(r, t) = F −1 {G} = . (Green’s Function) (8.12)
4πr
And as we know, multiplication in Fourier space is convolution elsewhere, so
t − t0 − |r − r0 |/c = 0
t0 = t − |r − r0 |/c
will will give any contribution to the integral. We call this the retarded time,
as is takes into account the time needed for the wave to propagate from r to r0
with speed c,
tr = t − |r − r0 |/c (Retarded time). (8.13)
CHAPTER 8. GIVE ME SOME HELMHOLTZ 24
So, in short,
ˆ
s(r0 , tr )
p(r, t) = d3 r0 . (Retarded potential) (8.14)
4π|r − r0 |
But the general principle is so important, that we will put it in red:
Important 8.1.2: Pressure
creating a pressure
ˆ ˆ
3 δ(t − t1 − |r − r1 |/c)
p(r, t) = d r0 dt0 g(r − r0 , t − t0 )s(r0 , t0 ) = .
4π|r − r1 |
as a retarded time, taking into account the time the wave needs to prop-
agate from r0 to r. So a bit more general, assuming a source s,
ˆ
s(r0 , tr )
p(r, t) = d3 r0 , (8.14 revisited)
4π|r − r0 |
s(r, t) = f (t)δ(r),
f (tr )
p(r, t) = ,
4πr
that is, the pressure is the signal from the point source taking the delay
into account.
CHAPTER 8. GIVE ME SOME HELMHOLTZ 25
Finish exam-
Example 8.1.2: Dipole source ple
F −1 {F G} = F −1 {F } ∗ F −1 {G} = f ∗ g,
∂f (t)
→ iωF (ω),
∂t
∇f (r) → ikF (k).
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
0 2 4 6 8 10 12
What this means is illustrated in Figure 8.1 and 8.2. We apply the wave equa-
tions, and it can easily be argued that the one for pressure achieve the form,
0.8
0.6
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-1
0 2 4 6 8 10 12
Figure 8.2: Illustration of pe and p = Re pe eiωt , for f = 0.7 and pe as shown.
This is just an example of the relation between p and pe , it may not solve the
wave equation. For code, see A.
CHAPTER 8. GIVE ME SOME HELMHOLTZ 29
So, we have now got three forms of the Green’s function, namely
δ(tr )
g(r, t) = , (Pulse) (8.12 revisited)
4πr
e−ik1 r
ge (r, ω) = , (Harmonic source (complex)) (8.24 revisited)
4πr
cos ωtr
g(r, t; ω) = . (Harmonic source (real part)) (8.25 revisited)
4πr
The latter is illustrated in Figure 8.3.
Figure 8.3: Illustration of (8.25), for ω = 60π and c = 1500 m s−1 For code,
see A.
CHAPTER 8. GIVE ME SOME HELMHOLTZ 31
to us,
1 ∂ 2 p(r0 , t0 )
∇20 p(r0 , t0 ) − = −s(r0 , t0 ), (8.3 revisited)
c2 ∂t20
1 ∂ 2 g(r − r0 , t − t0 )
∇20 g(r − r0 , t − t0 ) − 2 = −δ(r − r0 )δ(t − t0 )
c ∂t20
(8.26 revisited)
δ(tr − t0 )
g(r − r0 , t − t0 ) = , (8.12 revisited)
4π|r − r0 |
tr = t − |r − r0 |/c, (Retarded time) (8.27)
g ∂2p
g∇20 p − = −gs
c2 ∂t20
the second with p,
p ∂2g
p∇20 g − = −pδ(r − r0 )δ(t − t0 )
c2 ∂t20
subtract the second from the first,
" # " # " # " #
2 g ∂2p 2 p ∂2g
g∇0 p − 2 2 − p∇0 g − 2 2 = − gs − − pδ(r − r0 )δ(t − t0 )
c ∂t0 c ∂t0
g ∂2p p ∂2g
g∇20 p − − p∇ 2
0 g + = pδ(r − r0 )δ(t − t0 ) − gs
c2 ∂t20 c2 ∂t20
and integrate over the volume V and all time, that is
ˆ ˆ ∞ " #
3 2 g ∂2p 2 p ∂2g
d r0 dt0 g∇0 p − 2 2 − p∇0 g + 2 2
V −∞ c ∂t0 c ∂t0
ˆ ˆ ∞ " #
3
= d r0 dt0 pδ(r − r0 )δ(t − t0 ) − gs . (8.28)
V −∞
Next trick assumes that p → 0 as t → ±∞. Notice that, a bit similar to the
one above,
∂2p ∂2g
∂ ∂p ∂g ∂g ∂p ∂p ∂g
g −p = +g 2 − −p 2
∂t0 ∂t0 ∂t0 ∂t0 ∂t0 ∂t0 ∂t0 ∂t0 ∂t0
2 2
∂ p ∂ g
=g 2 −p 2.
∂t0 ∂t0
we see that
ˆ ∞ ˆ ∞
∂2p ∂2g
∂ ∂p ∂g
dt0 g 2 −p 2 = dt0 g −p
−∞ ∂t0 ∂t0 −∞ ∂t0 ∂t0 ∂t0
∞
∂p ∂g
= (g −p ) =0
∂t0 ∂t0 −∞
CHAPTER 8. GIVE ME SOME HELMHOLTZ 33
Add the two tricks together, and we are done. In all its beauty, the Helmholtz-
Kirchhoff’s formula takes the form
Important 8.3.1: Helmholtz-Kirchhoff ’s formula
ˆ ˆ ∞
3
α(r)p(r, t) = d r0 dt0 g(r − r0 , t − t0 )s(r0 , t0 )
V −∞
ˆ ˆ ∞
d2 r0
+ dt0 n̂0 · g(r − r0 , t − t0 )∇0 p(r0 , t0 ) (8.31)
S −∞
− p(r0 , t0 )∇0 g(r − r0 , t − t0 ) .
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
s(r, t) Source
S(k, t) Spatial Fourier transform of s(r, t)
S(r, ω) Temporal Fourier transform of s(r, t)
S(k, ω) Spatial and temporal Fourier transform of s(r, t)
se (r, ω) Complex envelope for source field
pe (r, ω) Complex envelope for pressure field
k1 = ω/c(= kr ) Propagation factor in the Helmholtz equation
m
k1 = ω/c − ia|ω| Propagation factor with damping
α(r) Describing the δ-function inside V , on S and outside
V
k Fourier space coordinate of r
tr = t − |r − r0 |/c Retarded time
∇0 Differentiation (as ∇) with respect to r0
34
Chapter 9
I know what you are thinking: Why on Earth do we want to do more approxi-
mations?
There is basically three regions we will look at. Now, assume a circular
transducer of radius a and diameter D = 2a centered at r0 , then the three
regions will be (r = |r − r0 |)
1. Extreme nearfield r < 0.25D2 /2λ
9.1 Fraunhofer
So, there is basically two approximations we should remember from this section,
namely
35
CHAPTER 9. OH NO! THE APPROXIMATIONS ARE COMING! 36
|r − r0 | ≈ r − r̂ · r0 , (9.1)
1
where r = |r| and r0 = |r0 | (as usual). When we are looking at |r−r0 | ,
we even skip the last term,
1 1
≈ . (9.2)
|r − r0 | r
D2
r> . (Rayleigh condition) (9.3)
2λ
|r − r0 | ≈ r − r̂ · r0 , (Fresnel approximation)
for tr = t − |r − r0 |/c and some source s. We quickly see that our two approxi-
mation can both be used, yielding
CHAPTER 9. OH NO! THE APPROXIMATIONS ARE COMING! 37
ˆ
s(r0 , tr ) r − r̂ · r0
p(r, t) ≈ d3 r0 , tr ≈ t − , (Far-field approximation)
4πr c
(9.4)
9.2 Tissue
So
∂u
ρ = −∇p
∂t
and thus
∂∇u
−∇2 p = −ρ
∂t
∇2 ψ + k12 (ω)ψ = 0
k 2 = k12 (ω) = ω 2 ρκ∞ (1 + H∞ (ω)) = ω 2 ρκ(ω)
so that
ψ = Ae−kd ·r ei(ωt−kr ·r) ,
CHAPTER 9. OH NO! THE APPROXIMATIONS ARE COMING! 38
which makes it obvious that kd is a damping factor and that propagation velocity
only involves kr , c = ω/kr !
We often use a wave vector of the form
Important 9.3.1: Wave vector with damping factor
ω
− ia|ω|m
k1 (ω) ≈ (9.5)
c
where m will depend on which material we are looking at.
This will affect our Green’s function. But in fact, only by adding a loss
function
L(r, ω) = e−kd (ω)r (Loss function) (9.6)
so that
e−ikr (ω)r e−ikr (ω)r
ge (r, ω) = e−kd (ω)r = L(r, ω)
4πr 4πr
We can also transform back to time space,
ˆ ∞
1 e−ikr (ω)r l(r, tr )
g(r, t) = dω L(r, ω) ≈
2π −∞ 4πr 4πr
If we use
ω
k1 (ω) ≈ − ia|ω|m (9.5 revisited)
c
we get for a → 0, by definition,
If we use
ω
k1 (ω) ≈ − ia|ω|m (9.5 revisited)
c
we get for m = 1, according to Angelsen [2000, eq. 4.375],
1 ar c acr
l(r, tr ) = = ,
π (ar)2 + t2r π (acr)2 + (r − ct)2
If we use
ω
k1 (ω) ≈ − ia|ω|m (9.5 revisited)
c
we get for m = 2, according to Angelsen [2000, eq. 4.374],
2
e−tr /4ar
l(r, tr ) = √ ,
4πar
for l as defined in (9.7). This wave is shown in Figure 9.1.
We further define
1
c= √ (9.10)
ρa κa
κf (r)
β(r) = , (9.11)
κa
ρf (r)
γ(r) = , (9.12)
ρ(r)
please note that the definitions are not equal, for β the denominator is a con-
stant, whilst for γ it is the total, spatial function.
CHAPTER 9. OH NO! THE APPROXIMATIONS ARE COMING! 40
Starting from
∂ρu
= −∇p,
∂t
1
p = − ∇ψ,
κ
we want to deduce a wave equation for φ. Remember that ∇φ = −ρu. Thus
∇φ = −ρu
∂∇φ ∂ρu
=− = ∇p
∂t ∂t
∂φ
p=
∂t
∂p ∂2φ
= 2
∂t ∂t
∂2φ ∂ 1∇ · ψ
2
=− κ
∂t ∂t
∂2φ 1 ∂ψ
2
=− ∇·
∂t κ ∂t
∂2φ 1
=− ∇·u
∂t2 κ
∂2φ 1 1
= ∇ · ∇φ
∂t2 κ ρ
Now,
1 1 1 ρa 1 ρ − ρf 1 ρ − ρf
= = = =
ρ ρa + ρf ρa ρa + ρf ρa ρ − ρf + ρf ρa ρ
1 ρf 1
= 1− = 1−γ
ρa ρ ρa
so that
" #
1 1 1 2 1 1
∇· ∇φ = ∇ ∇φ + ∇ φ = ∇ 1 − γ · ∇φ + 1 − γ ∇2 φ
ρ ρ ρ ρa ρa
" #
1 γ 1
=∇ − · ∇φ + 1 − γ ∇2 φ
ρa ρa ρa
1 1
= − ∇γ · ∇φ + 1 − γ ∇2 φ
ρa ρa
CHAPTER 9. OH NO! THE APPROXIMATIONS ARE COMING! 42
∂2φ 1 1
κ =− ∇γ · ∇φ + 1 − γ ∇2 φ
∂t2 ρa ρa
∂2φ
ρa (κa + κf ) 2 = −∇γ · ∇φ + ∇2 φ − γ∇2 φ
∂t
1 ∂2φ ∂2φ
+ ρa κf 2 = −∇γ · ∇φ + ∇2 φ − γ∇2 φ
c2 ∂t2 ∂t
1 ∂2φ κf ∂ 2 φ
+ ρa κa = −∇γ · ∇φ + ∇2 φ − γ∇2 φ
c2 ∂t2 κa ∂t2
2
1 ∂ φ β ∂2φ
2 2
+ 2 2 = −∇γ · ∇φ + ∇2 φ − γ∇2 φ
c ∂t c ∂t
Finally, we notice that
∇ · (γ∇φ) = ∇γ · ∇φ + γ∇2 φ,
yielding
1 ∂2φ β ∂2φ
2 2
+ 2 2 = −(∇γ · ∇φ + γ∇2 φ) + ∇2 φ
c ∂t c ∂t
1 ∂2φ β ∂2φ
+ = −∇ · (γ∇φ) + ∇2 φ.
c2 ∂t2 c2 ∂t2
Or by flipping things around,
Use the Fourier version of (9.13) for φ → p and the Greens function
e−ik|r−r0 |
ge (r − r0 , ω) = ,
4π|r − r0 |
with k = ω/c, to make an integral equation for the sum of incident and
scattered field. Also discuss how this new integral equation cna be solved
through iterations.
Answer:
The last part of the question will be answered later, after we have looked
at Born approximations. Remember that the temporal Fourier trans-
∂
forms ∂t → iω. Also, for temporal transfoms, we often do not use
capital letters. But here we go,
+ ∇ · (γ(r)∇φ(r, ω))
∇2 φ(r, ω) + k 2 φ(r, ω) ≡ −s(r, ω)
s(r, ω) = β(r)k 2 φ(r, ω) − ∇ · (γ(r)∇φ(r, ω))
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
s(r, t) Source
S(k, t) Spatial Fourier transform of s(r, t)
S(r, ω) Temporal Fourier transform of s(r, t)
S(k, ω) Spatial and temporal Fourier transform of s(r, t)
se (r, ω) Complex envelope for source field
pe (r, ω) Complex envelope for pressure field
k1 = ω/c(= kr ) Propagation factor in the Helmholtz equation
m
k1 = ω/c − ia|ω| Propagation factor with damping
α(r) Describing the δ-function inside V , on S and outside
V
k Fourier space coordinate of r
tr = t − |r − r0 |/c Retarded time
∇0 Differentiation (as ∇) with respect to r0
D Opening diameter of some sort through which a wave
is passing
44
Chapter 10
You may have noticed that the word ’Ultrasound’ until now only has been
mentioned in the title. Of course it may function as click-bait, but since you
haven’t stopped reading yet, I think you deserve your moneysworth as well.
The ultrasound equipment consists of mainly two parts,
1. Transducer
2. Receiver
which are quite self-describing names. The transducer sends out a pulse or
harmonic wave, and the receiver detects what is reflected from things inside
the body. The transducer is important and can be made several ways. You
could have one vibrating element, or you could have several, and you can also
experiment with the shape. We will look into several kinds of transducers and
waves, like
• Transducers
– Single element
– Multi element
– Circular
– Elliptical
– Rectangular
– Flat (planar)
– Bent (concave)
• Waves
– Continuous wave
– Pulse
45
CHAPTER 10. TURN UP THE MUSIC 46
0 0
0.005 0.005
0.01 0.01
0.015 0.015
0.02 0.02
0.025 0.025
0.03 0.03
-0.01 -0.005 0 0.005 0.01 -0.01 -0.005 0 0.005 0.01
If the transducer is just one single element, all parts of the surface moves
at the same time and there is no spatial variation in the normal velocity,
thus
Un = 1.
This we will use several times in the sections to come.
If ρ is the mass density (and it is), we define the surface momentum distri-
bution as
p = ρu = −∇φ
and then
1
u = − ∇φ. (10.3)
ρ
CHAPTER 10. TURN UP THE MUSIC 47
It should be noted that one usually refers to the velocity potential, which is
closely related, only missing the factor 1/ρ. Also, it is worth noting that since
∂u
ρ = −∇p, (7.7 revisited)
∂t
it follows that
∂φ
p= . (10.4)
∂t
Example 10.0.2: Exam 2010.2a
How is the relationship between the impulse potential φ(r, t), the
vibration velocity u(r, t) and the pressure p(r, t)?
Answer
In short
1
u(r, t) = − ∇φ(r, t) (10.3 revisited)
ρ(r)
∂φ(r, t)
p(r, t) = (10.4 revisited)
∂t
ˆ
h(r, t) = d2 r0 gp (r − r0 , t − t0 )Mn (r0 )
St
h may also function as a filter, which is easiest to see in the frequency space.
There H may be close to zero for some frequency and thus the signal in the
point we are looking, will have low amplitudes of this signal. Usually there is
also a damping effect as a function of distance from the source, as the signal
will lose energy on its way.
F
D = 2a
P
z
a Radius of transducer
D = 2a Diameter of transducer
F Radius of curvature
FN = F/D F-number
λ Wavelength of signal
A sketch is shown in Figure 10.2. At the focus point F = (0, 0, F ) the intensity
will be at its highest. In a plane normal to z containing the focus point, we may
move some distance aF , and the intensity will decrease. These is things we will
look into.
Important 10.1.1: Main idea
To see what happens, we need to grasp that the wave sent out by the
transducer will depart from the whole surface at the same time. Thus
there will be interference with itself along its way. In the focus point,
the whole thing will add up as all parts of the transducer has the same
distance and there is no phase shift.
proofread
section
10.1.1 On the central axis
So first, let us look at a general point P = (0, 0, zP ) located at the z-axis as
shown in Figure 10.2 and even better in Figure 10.3 and try to understand how
the wave will look like as it interfere with itself at P . Let Q(α) be a point on the
transducer so that the angle between z-axis and QF is α. What is the distance
CHAPTER 10. TURN UP THE MUSIC 49
Q
D = 2a
P α F
z
αmax
rQP from Q to P ?
rQP = rP − rQ
rQ = rF − rQF
rQP = rP − rF + rQF
rQP = −rP F + rQF
q q
|rQP | = (rQF − rP F )2 = rQF 2 + rP2 F − 2rQF rP F cos(α)
p
= F 2 + (F − zP )2 − 2F (F − zP ) cos(α)
q
= zP2 + 2F (F − zP )(1 − cos(α))
There exists an αmax which denotes the boundary of the transducer. The
depth d of the transducer then becomes
d = F − F cos(αmax ) = F (1 − cos(αmax )). (10.6)
Also,
F 2 = (F − d)2 + a2
F 2 = F 2 − 2dF + d2 + a2
0 = d2 − 2dF + a2
√
2F ± 4F 2 − 4a2 p
d= = F ± F 2 − a2
2
p a2
= F 1 − 1 − (a/F )2 ≈ F 1 − (1 − )
2F 2
a2
= .
2F
√
where we have used the approximation 1 + x ≈ 1 + x/2 and only looked at the
negative sign as d < F . The distance from P to the boundary of the transducer
CHAPTER 10. TURN UP THE MUSIC 50
is
q
rbP = |rQP | = zP2 + 2F (F − zP )(1 − cos(αmax ))
α=αmax
q
= zP2 + 2d(F − zP )
by using (10.6).
Now, all points on the transducer between an angle α and α + dα will have
the same distance to P . Remember that we are in 3D, so this is a circular strip
on the transducer. The area of this strip will be
dA = 2πrA dr = 2πF sin(α)F dα = 2πF 2 sin(α) dα
where dr = F dα and rA = F sin(α) follows from the geometry.
Do you remember the spatial frequency response for nonabsorbing materials?
Well, here it is
ˆ
e−ikr |r−r0 |
H(r, ω) = d2 r0 Mn (r0 )
St 2π|r − r0 |
kr = ω/c = 2π/λ
Mn = ρUn
where Mn is the surface momentum distribution. We choose Un = 1. This
integral we evaluate with d2 r0 = dA = 2πF 2 sin(α) dα and r = rP so that
|r − r0 | = |rQP |, which we recently found an expression for, and for all circular
strips, i.e. from 0 to αmax . Put it all together,
ˆ
e−ikr |rQP |
H(rp , ω) = dA ρ
St 2π|rQP |
ˆ αmax √ 2
2 e−ikr zP +2F (F −zP )(1−cos(α))
= ρF dα sin(α) p 2
0 zP + 2F (F − zP )(1 − cos(α))
which according to Angelsen can be evaluated to
H(rp ) = ρde−ikr (rbP −zP )/2 G(zP ),
F rbP − zP sin(kr (rbP − zP )/2)
G(zP ) = . (10.7)
d F − zP kr (rbP − zP )/2
This is exciting because sin(x) = 0 for x = nπ for any non-zero inte-
ger n, leaving H = 0 for (but NB: for rbP − zP = 0 we need to use that
limx→0 sin(x)/x = 1)
kr (rbP − zP )/2) = nπ
2π rbP − zP
= nπ
λ 2
rbP − zP = nλ
q
zP2 + 2d(F − zP ) = nλ + zP (Condition for minimum)
CHAPTER 10. TURN UP THE MUSIC 51
For zP close to√F , (F − zP )/zP2 will be small, and we can use our familiar
approximation 1 + x ≈ 1 + x/2,
d(F − zP ) d(F − zP )
rbP ≈ zP (1 + ) = zP + ,
zP2 zP
which yields for (10.7)
F sin(kr d(F − zP )/2zP )
H(zP ) ≈ ρde−ikr (zP +d(F −zP )/2zP ) .
zP kr d(F − zP )/2zP
We will find the intensity on the axis for zP close to F . Introduce the Fresnel
parameter
λ
S= . (Fresnel parameter)
2d
If you remember well, d ≈ a2 /2F and FN = F/D, D = 2a, so that
λ Fλ Fλ 4FN
S≈ = 2 = 2 =
a2 /F a D /4 D/λ
which you can put into
!2
ωkr π F sin(π(F − zP )/2SzP )
I(zP ) = |H(zP )|2 ≈ I0 ,
2ρ S zP π(F − zP )/2SzP
r=F+w
CHAPTER 10. TURN UP THE MUSIC 52
40
FN = 0.5, m = 4
35 FN = 1, m = 3
FN = 3, m = 2
FN = 3, m= 1
30 FN = 4, m = 0.5
25
20
15
10
0
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
|r − rQ | = r − rQ · w/r.
θ
F z
a2 e−ikF
H(F, w = 0) = ρ .
2F
What about the depth? Obviously, the maximum occurs for zP = F .
*****Add more
CHAPTER 10. TURN UP THE MUSIC 55
0.5
0.4
0.3
0.2
0.1
-0.1
-100 -50 0 50 100
leaving
Figure 10.7: ***. |H| for a = 5 cm, c = 1500 m s−1 , f = 1 MHz, F = 0.5 m,
ρ = 1 kg m−3 . For code, see A.
CHAPTER 10. TURN UP THE MUSIC 58
e−ikr r J1 (θL )
HL (r, ω) = ρAL ,
πr θL
q
θL = kr a x2P + yP2 /r
AL = πa2
e−ikr r sin(θR (a)) sin(θR (b))
HR (r, ω) = ρAR
2πr θR (a) θR (b)
θR (x) = xkr yP /r
AR = 2a2b
We note that limθ→0 sin(θ)/θ = 1 and limθ→0 J1 (θ)/θ = 1/2, so that the limiting
values for HL and HR are equal.
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
s(r, t) Source
S(k, t) Spatial Fourier transform of s(r, t)
S(r, ω) Temporal Fourier transform of s(r, t)
S(k, ω) Spatial and temporal Fourier transform of s(r, t)
se (r, ω) Complex envelope for source field
pe (r, ω) Complex envelope for pressure field
k1 = ω/c(= kr ) Propagation factor in the Helmholtz equation
m
k1 = ω/c − ia|ω| Propagation factor with damping
α(r) Describing the δ-function inside V , on S and outside
V
k Fourier space coordinate of r
tr = t − |r − r0 |/c Retarded time
∇0 Differentiation (as ∇) with respect to r0
D Opening diameter of some sort through which a wave
is passing
φ Momentum potential
a Radius of circular transducer
D = 2a Diameter of circular transducer
62
CHAPTER 10. TURN UP THE MUSIC 63
Scattering
64
CHAPTER 11. SCATTERING 65
φk−1 = φi + φk−1
s ,
φ0 = φi .
where h is the spatial impulse response. The exact form of h depends on the
situation, as we have looked into before. Scatterers in the medium will scatter
φt and create a scattered wave φs . The total wave received will be
ˆ ˆ ∞
1
φ(t) = d3 r0 dt0 hr (r0 , t − t0 )s(r0 , t0 ) (11.3)
ρ −∞
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
s(r, t) Source
S(k, t) Spatial Fourier transform of s(r, t)
S(r, ω) Temporal Fourier transform of s(r, t)
S(k, ω) Spatial and temporal Fourier transform of s(r, t)
se (r, ω) Complex envelope for source field
pe (r, ω) Complex envelope for pressure field
k1 = ω/c(= kr ) Propagation factor in the Helmholtz equation
m
k1 = ω/c − ia|ω| Propagation factor with damping
α(r) Describing the δ-function inside V , on S and outside
V
k Fourier space coordinate of r
tr = t − |r − r0 |/c Retarded time
∇0 Differentiation (as ∇) with respect to r0
D Opening diameter of some sort through which a wave
is passing
φ Momentum potential
a Radius of circular transducer
D = 2a Diameter of circular transducer
66
CHAPTER 11. SCATTERING 67
where rs,0 is the position of the source at t = 0. The distance between the
source and the observer is
68
CHAPTER 12. IT’S ALIVE! ON DOPPLER ULTRASOUND 69
rs (t) = rs,0 + vt
and
yielding
c
fo = f
c − v cos θ
and the Doppler shift is
c v cos θ v
fD = fo − f = f −1 =f ≈ f cos θ (12.3)
c − v cos θ c − v cos θ c
when c v. Let f → ω, then
ω
ωD (r̂) = r̂ · v (12.4)
c
is a nice formulation, and r = ro − rs,0 .
CHAPTER 12. IT’S ALIVE! ON DOPPLER ULTRASOUND 70
Table 12.1: Blood velocities in cm s−1 in inferior venae cavae (IVC) and superior
venae cavae (SVC) in rest and exercise during systole (S) and diastole (D). The
exercise was performed by D. B. for 150 s, L. W. for 90 s and J. M. for 165 s.
The peaks are averages for 10 cardiac cycles during shallow breathing. Data
from Wexler et al. [1968].
Subject D. B. I. G. L. W. J. M.
Age 35 39 29 30
Height [cm] 184 185 174 169
Weight [kg] 80 85 78 67
the original angular velocity from the transducer, and ω the one measured by
the receiver. We can then combine the earlier results, and get a total shift for
ωd = ω − ω0 ,
v
ωd = ωD (r̂t ) + ωD (r̂r ) = ω0 (r̂t + r̂r ) · = k(r̂t + r̂r ) · v = kv(cos θt + cos θr ),
c
The simplest case is when the transducer and receiver is very close in
space so that θt = θr . Then
ωd c
v= .
ω0 2 cos θ
400
200
-200
-400
0.8
0.6 0.02
0.4 0.01
0
0.2
-0.01
0 -0.02
Figure 12.1: Velocity as a function of Doppler shift and angle between velocity
of scatterer and transducer/velocity (approximately at the same position). c =
1500 m s−1 . For code, see A.
CHAPTER 12. IT’S ALIVE! ON DOPPLER ULTRASOUND 73
In reality, the situation is a bit more complex. We have used many pages to
analyse how the tissue affects the signals sent, and we will continue in the same
path.
ˆ
iω0 t
sc (t) = e ak02 υa d3 r1 Ht (r1 , ω0 )Hr (r1 , ω0 )n(r1 , t) (12.17)
ˆ
se (t) = ak02 υa d3 r1 Ht (r1 , ω0 )Hr (r1 , ω0 )n(r1 , t) (12.18)
ψ(z, t) Displacement
u(z, t) = ∂t ψ Vibration velocity
−1
p(z, t) = −κ ∂z ψ Pressure
κ Volume compressibility
ρ0 Mass density in unstrained material
ρ Mass density
c = (ρ0 κ)−1/2 Wave propagation velocity
s(r, t) Source
S(k, t) Spatial Fourier transform of s(r, t)
S(r, ω) Temporal Fourier transform of s(r, t)
S(k, ω) Spatial and temporal Fourier transform of s(r, t)
se (r, ω) Complex envelope for source field
pe (r, ω) Complex envelope for pressure field
k1 = ω/c(= kr ) Propagation factor in the Helmholtz equation
m
k1 = ω/c − ia|ω| Propagation factor with damping
α(r) Describing the δ-function inside V , on S and outside
V
k Fourier space coordinate of r
tr = t − |r − r0 |/c Retarded time
∇0 Differentiation (as ∇) with respect to r0
D Opening diameter of some sort through which a wave
is passing
φ Momentum potential
a Radius of circular transducer
D = 2a Diameter of circular transducer
75
CHAPTER 12. IT’S ALIVE! ON DOPPLER ULTRASOUND 76
MATLAB examples
A.1 Figure
Source for Figure 10.7.
x = -0.1:0.0005:0.1;
y = -0.1:0.0005:0.1;
rho = 1;
a = 0.05; % diameter 10 cm
omega = 2* pi *10^6; % f =1 MHz
c = 1500; % m / s
k_r = omega / c ;
F = 0.5;
[X , Y ] = meshgrid (x , y ) ;
w = sqrt ( X .^2+ Y .^2) ;
r = sqrt ( X .^2+ Y .^2 + F ^2) ;
theta = k_r .* a .* w ./ F ;
H = rho * a ^2 * ( exp ( -1 i * k_r .* r ) ./ r ) .* besselj (1 , theta )
,→ ./ theta ;
h = surf (X ,Y , abs ( H ) ) ;
set (h , ’ LineStyle ’ , ’ none ’)
colormap parula
xlabel ( ’ $x$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $y$ ’ , ’ Interpreter ’ , ’ latex ’)
zlabel ( ’ $ | H |$ ’ , ’ Interpreter ’ , ’ latex ’)
A.2 Figure
Source for Figure 10.8.
77
APPENDIX A. MATLAB EXAMPLES 78
x = -0.1:0.0005:0.1;
y = -0.1:0.0005:0.1;
rho = 1;
a = 0.05; %10 cm
b = 0.10; %20 cm
omega = 2* pi *10^6; % f =1 MHz
c = 1500; % m / s
k_r = omega / c ;
z_P = 0.5;
[X , Y ] = meshgrid (x , y ) ;
r = sqrt ( X .^2+ Y .^2 + z_P ^2) ;
theta1 = k_r .* a .* X ./ r ;
theta2 = k_r .* b .* Y ./ r ;
H = rho * ( exp ( -1 i * k_r .* r ) ./(2* pi * r ) ) .* (2* sin ( theta1
,→ ) ./ theta1 ) .* (2* sin ( theta2 ) ./ theta2 ) ;
h = surf (X ,Y , abs ( H ) ) ;
set (h , ’ LineStyle ’ , ’ none ’)
colormap parula
xlabel ( ’ $x$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $y$ ’ , ’ Interpreter ’ , ’ latex ’)
zlabel ( ’ $ | H |$ ’ , ’ Interpreter ’ , ’ latex ’)
A.3 Figure
Source for Figure 10.6.
x = -100:0.1:100;
y = besselj (1 , x ) ./ x ;
plot (x , y )
xlabel ( ’ $x$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $J_1 ( x ) / x$ ’ , ’ Interpreter ’ , ’ latex ’)
A.4 Figure
Source for Figure 10.4.
F = 0.3;
FN = [0.5 1 2 3 4];
m = [4 3 2 1 0.5];
c = 1500;
f = 10^6;
lambda = c / f ;
zP = 0.2:0.00005:0.5;
for i = 1: length ( FN )
APPENDIX A. MATLAB EXAMPLES 79
D = F / FN ( i ) ;
S = 4* FN ( i ) * lambda / D ;
theta = pi * (F - zP ) ./ (2* S * zP ) ;
I = (( pi / S ) * ( F ./ zP ) .* ( sin ( theta ) ./ theta ) )
,→ .^2;
plot ( F ./ zP , I *10^( - m ( i ) ) )
hold on ;
end
legend ({ ’ FN = 0.5 , m = 4 ’ , ’ FN = 1 , m = 3 ’ , ’ FN = 3 , m =
,→ 2 ’ , ’ FN = 3 , m = 1 ’ , ’ FN = 4 , m = 0.5 ’}) ;
xlabel ( ’ $F / z_P$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $ ( I / I_0 ) \ cdot 10^{ - m }$ ’ , ’ Interpreter ’ , ’ latex ’)
A.5 Figure
Source for Figure 9.1.
r = 0:0.001:0.5;
t = 0:0.001:0.5;
c = 1500;
a = 0.05;
[r , t ] = meshgrid (r , t ) ;
t_r = t - r / c ;
A.6 Figure
Source for Figure 8.1.
f = 3; % Hz
omega = 2* pi * f ;
t = 0:0.0001:12;
phi_e = exp ( - t ) .* abs (3* sin ( t ) .* t + 7* t .^2 - 30* sin
,→ (0.3* t ) .* t .^3) ;
phi_e = phi_e / max ( phi_e ) ;
phi = real ( phi_e .* cos ( omega * t ) ) ;
plot (t , phi )
APPENDIX A. MATLAB EXAMPLES 80
hold on ;
plot (t , phi_e , ’r - - ’)
xlabel ( ’ $t$ ’ , ’ Interpreter ’ , ’ latex ’)
legend ({ ’ $p$ ’ , ’ $p_e$ ’} , ’ Interpreter ’ , ’ latex ’)
A.7 Figure
Source for Figure 8.2.
f = 0.7; % Hz
omega = 2* pi * f ;
t = 0:0.00001:12;
phi_e = zeros (1 , length ( t ) ) ;
a = round ( length ( t ) /2) ;
phi_e ( round ( a *0.8) : round ( a *1.3) ) = 1;
phi_e ( round ( a *0.2) : round ( a *0.5) ) = 0.5;
phi = real ( phi_e .* cos ( omega * t ) ) ;
plot (t , phi )
hold on ;
plot (t , phi_e , ’r - - ’)
xlabel ( ’ $t$ ’ , ’ Interpreter ’ , ’ latex ’)
legend ({ ’ $p$ ’ , ’ $p_e$ ’} , ’ Interpreter ’ , ’ latex ’)
A.8 Figure
Source for Figure 8.3.
t = 0:0.0001:0.1;
r = 0. 01:0.000 1:0.11;
c = 1500;
t_r = t - r / c ;
[r , t_r ] = meshgrid (r , t_r ) ;
omega = 2* pi *30;
g = cos ( omega * t_r ) ./ (4* pi * r ) ;
h = surf (r , t , g ) ;
set (h , ’ LineStyle ’ , ’ none ’)
colormap parula
xlabel ( ’ $r$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $t$ ’ , ’ Interpreter ’ , ’ latex ’)
zlabel ( ’ $g$ ’ , ’ Interpreter ’ , ’ latex ’)
A.9 Figure
Source for Figure 12.1.
c = 1500;
APPENDIX A. MATLAB EXAMPLES 81
w = (0 .99:0.00 1:1.01) - 1;
th = 0.4:0.05:0.7* pi ;
[W , Th ] = meshgrid (w , th ) ;
v = W .* c ./(2* cos ( Th ) ) ;
h = surf (W , Th / pi , v ) ;
set (h , ’ LineStyle ’ , ’ none ’)
colormap parula
xlabel ( ’ $ \ omega_d / \ omega_0$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $ \ theta / \ pi$ ’ , ’ Interpreter ’ , ’ latex ’)
zlabel ( ’ $v$ [ m / s ] ’ , ’ Interpreter ’ , ’ latex ’)
A.10 Figure
Source for Figure ??.
% Continuous wave from point sources
%15.01.01 Hans Torp
% last update 11.01.2017
clf ;
f0 =2 e6 ;% frequency
c =1540;% speed of sound
axis ( ’ image ’) ;
[ x0 , z0 ]= ginput (1) ;
end ;
%% circular source
dx0 =0.3 e -3;
r0 =9 e -3;
x0 = -5e -3: dx0 :5 e -3;
z0 = 1.5* r0 - sqrt ( r0 ^2 - x0 .^2) ;
p =0* x ;
for j = 1: length ( x0 ) ,
r = sqrt (( x - x0 ( j ) ) .^2+( z - z0 ( j ) ) .^2) ;
p = p + exp ( i * k * r ) ./ r ;
image ( xaxis , zaxis ,0.05* real ( p ) ) ;
axis ( ’ image ’) ;
pause (0.01) ;
end ;
A.11 Figure
Source for Figure ??.
% Continuous wave from point sources
%15.01.01 Hans Torp
% last update 11.01.2017
clf ;
f0 =2 e6 ;% frequency
c =1540;% speed of sound
%% plane source
APPENDIX A. MATLAB EXAMPLES 83
z0 =5 e -3;
dx0 =0.3 e -3;
p =0* x ;
for x0 = -5e -3: dx0 :5 e -3 ,
r = sqrt (( x - x0 ) .^2+( z - z0 ) .^2) ;
p = p + exp ( i * k * r ) ./ r ;
image ( xaxis , zaxis ,0.05* real ( p ) ) ;
axis ( ’ image ’) ;
pause (0.01) ;
end ;
Part III
CT
84
Chapter 13
Some electromagnetics
13.1 Maxwell
Ok, I hope you know some electrodynamics already. Let us, however, do a quick
review.
85
CHAPTER 13. SOME ELECTROMAGNETICS 86
F = qE, (13.5)
F = qv × B, (13.6)
D = 0 r E = E, (13.7)
B = µ0 µr H = µH. (13.8)
If ∇r = 0, then we can develop some easy wave equations by combining the
Maxwell Equations. We start by finding the curl of (13.1) and inserting (13.2)
(through (13.8)),
∂∇ × B
∇ × (∇ × E) = ∇(∇ · E) − ∇2 E = −
∂t
∂j ∂2E
= −µ − µ 2
∂t ∂t
becoming
1 ∂2E ∂j ∇ρ
∇2 E − =µ + , (13.9)
c2 ∂t2 ∂t
√
where c = 1/ µ. This is indeed very similar to
1 ∂2u
∇2 u − = −s(r, t), (8.2 revisited)
c2 ∂t2
∂j
showing that ∂t and ∇ρ at the right hand side are sources.
In an equivalent way, we combine the two others,
∂∇ × D
∇ × (∇ × H) = ∇(∇ · H) − ∇2 H = ∇ × j +
∂t
∂2H
= ∇ × j − µ
∂t2
making
1 ∂2H
∇2 H − = −∇ × j, (13.10)
c2 ∂t2
which also resembles (8.2), making ∇ × j a source.
We know from **** that a vector field can be decomposed into a scalar
potential and a vector potential, and from (13.4) that ∇B = 0, leaving
B=∇×A (13.11)
for the magnetic vector potential A. Similarly things we can do for E, but here
we use that the curl of a gradient field is zero, yielding
∂A
E = −∇φ − . (13.12)
∂t
This allows us, by applying the Lorentz condition/gauge to A (which we are
free to do, just check),
∂φ
∇ · A + µ = 0, (Lorentz condition) (13.13)
∂t
to transform our wave equations into
1 ∂2A
∇2 A − = −µj (13.14)
c2 ∂t2
1 ∂2φ ρ
∇2 φ − 2 2 = − (13.15)
c ∂t
CHAPTER 13. SOME ELECTROMAGNETICS 88
√
where c = 1/ µ as before. This is even more similar to our acoustical wave
equations (8.2) and (8.3), but the sources are for A the µj (current density) and
for φ the free current ρ/.
We now need boundary conditions ****write about boundary conditions
∇ × E = −ωB, (13.16)
∇ × H = j + iωD, (13.17)
∇ · D = ρ, (13.18)
∇ · B = 0, (13.19)
look back to Maxwell Equations if you need to refresh things, but remember
that (r, t) → (r, ω). As we did for acoustics in ***, we split the permitivitty
and permeability into a slowly term representing an average over a larger area,
a , and one that varies more rapidly, f , representing local deviations. That is,
ρ ∇
∇·E= − E, (13.22)
∇µ
∇·H=− H. (13.23)
µ
√
We can also rewrite (13.16) and (13.17) as, for k1 = ω µ = ω/c
ρ ∇
∇2 E + k12 E = ∇ + iωµj − ∇ · E − iωH × ∇µ (13.24)
∇µ
∇2 H + k12 H = −∇ × j − ∇ · H + iωE × ∇. (13.25)
µ
These are Helmholtz type wave equations as we have looked at before, ****
meaning the first equation contains two active sources (first two terms) and two
scatterers (last two terms), whilst the latter equation only have one active term
(the first) but also two scatterers.
According to Angelsen we often have µ ≈ µ0 , so that it is variations in
which is causing scattering in tissues.
CHAPTER 13. SOME ELECTROMAGNETICS 89
Similarly to ***, we can make it more beautiful by writing (using the Lau-
rentz condition)
∇(µ) ∇µ
∇2 A + k12 A = −µj + ∇ · A + (∇ × A)∇ × , (13.26)
µ µ
ρ ∇
∇2 φ + k12 = − + · E, (13.27)
√
where k1 = ω µ = ω/c as before.
Make it smile
14.1 Basics
We introduce the coordinate systems given in Figure 14.1. proofread
As it is useful to be capable of transforming back and forth between the section
systems, we see that the unit vectors becomes
Look back to Figure 14.1. The source sends a ray which is detected at the
detector. This ray is parallel to the s-axis and for the moment we will call the
ray path for a given r and angle θ for L (r, θ). The detected signal will be the
sum of all attenuation f on this line L ,
ˆ
p(r, θ) = ds f (x, y) = R{f }(r, θ), (14.5)
L (r,θ)
and this is the Radon transform of f . Following Bolles [2011] and Beatty [2012],
90
CHAPTER 14. MAKE IT SMILE 91
y
Source
S r
θ L (r, θ)
x
r Detector
This relation is often called the central slice theorem, the Fourier slice theorem
or the slice projection theorem.
or by applying (14.10),
ˆ ∞ ˆ ∞
1
f (x, y) = dk x dk y F (kx , ky )e−i(kx x+ky y) , (14.14)
(2π)2 −∞ −∞
where we have found F (kx , ky ) through (14.10). The problem with doing this,
is that we must approximate the k-space values into a grid, which introduces
artefacts in our final image.
that is, fˆ is estimated by the sum of all projections crossing this point.
We will compare this method with the Fourier method later. Approximate
(14.15) with a continuous integral
ˆ
1 π
fˆ(x, y) = B{p} = dφ p(x cos φ + y sin φ, φ), (14.16)
π 0
CHAPTER 14. MAKE IT SMILE 94
|kr |F {R{f }(kr , θ)}(kr cos θ, kr sin θ)eikr (x cos θ+y sin θ)
ˆ π
1
= dθ 2πF −1 {|kr |F {R{f }(kr , θ)}(x cos θ + y sin θ, θ)}
(2π)2 0
1
= B{F −1 {|kr |F {R{f }(kr , θ)}(x, y)}}, (14.18)
2
where B is the back projection given in (14.16). Let us compare them closely,
1
f (x, y) = B{F −1 {|kr |F {R{f }(kr , θ)}(x, y)}} (14.18 revisited)
2
fˆ(x, y) = B{F −1 {F {R{f }}}} (14.17 revisited)
Now, this is not very useful as it contains a lot of sums and stuff, so we want
to find a function φ so that
F {φ(r)}(kr ) = |kr |,
p(d, θ) = p(d).
This means ( √
2 R 2 − d2 , d ≤ R
p(d) =
0, elsewhere
and
ˆ ˆ R p J1 (Rkr )
F (k) = dr p(r)e−ikr r = dr 2 R2 − r2 e−ikr r = 2πR2
−R Rkr
ˆ ˆ
q
2πR2 ∞ ∞ J1 R kx2 + ky2
f (x, y) = dk x dk y ei(kx x+ky y)
(2π)2
q
−∞ −∞ 2
R kx + ky2
CHAPTER 14. MAKE IT SMILE 96
Missing
figure An example of simple radon transform
MATLAB examples
B.1 Figure
Source for Figure ??.
len = 101;
% make things
f = zeros ( len ) ;
rV = - floor ( len /2) : floor ( len /2) ;
thetaV = [0 pi /2];
% project things
i = 0;
j = 0;
for r = rV
i = i + 1;
for theta = thetaV
j = j + 1;
temp = 0;
for x = 1: len
for y = 1: len
if ( checkCoord (r , theta , x , y , len ) )
temp = temp + f (x , y ) ;
end
end
end
p (i , j ) = temp ;
97
APPENDIX B. MATLAB EXAMPLES 98
end
j = 0;
end
% plot things
[X , Y ] = meshgrid ( rV , rV ) ;
h = surf (X ,Y , f ) ;
hold on
p = p / max ( max ( p ) ) ;
plot3 ( rV , rV ( end ) * ones (1 , length ( rV ) ) ,p (: ,2) )
plot3 ( rV ( end ) * ones (1 , length ( rV ) ) ,rV , p (: ,1) )
xlabel ( ’ $x$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $y$ ’ , ’ Interpreter ’ , ’ latex ’)
zlabel ( ’ $z$ ’ , ’ Interpreter ’ , ’ latex ’)
legend ({ ’ $f (x , y )$ ’ , ’ $ \ theta = \ pi /2 $ ’ , ’ $ \ theta = 0$
,→ ’} , ’ Interpreter ’ , ’ latex ’)
view (3)
B.2 Figure
Source for Figure 14.2.
len = 101;
% make things
f = zeros ( len ) ;
rV = - floor ( len /2) : floor ( len /2) ;
thetaV = 0:0.05: pi ;
% project things
i = 0;
j = 0;
for r = rV
i = i + 1;
for theta = thetaV
j = j + 1;
temp = 0;
for x = 1: len
for y = 1: len
if ( checkCoord (r , theta , x , y , len ) )
temp = temp + f (x , y ) ;
end
end
end
p (i , j ) = temp ;
end
j = 0;
end
% plot things
[r , th ] = meshgrid ( rV , thetaV ) ;
p = p / max ( max ( p ) ) ;
contourf (r , th , transpose ( p ) ,15)
xlabel ( ’ $r$ ’ , ’ Interpreter ’ , ’ latex ’)
ylabel ( ’ $ \ theta$ ’ , ’ Interpreter ’ , ’ latex ’)
colorbar ( ’ southoutside ’)
set ( gcf , ’ Renderer ’ , ’ opengl ’)
MR
100
Chapter 15
where Tp is the period, i.e. the time the electron uses for one rotation, which of
course is the inverse of the frequency,
1
f= (Frequency)
Tp
This rotation is de facto a current (defined as charge crossing a surface per time)
q q eω0
j= = =− (Current) (15.1)
Tp 2π/ω0 2π
eω0 r2
Mo = jπr2 = − . (Orbital magnetic dipole moment)
2
There is also an angular magnetic momentum
101
CHAPTER 15. THE EASY PARTS 102
Now, the electron also possesses a spin which give rise to a magnetic dipole
moment Ms and a mechanical one Is . Experimentally, one has found that
Ms
γs = = ge γo , (Spin gyromagnetic ratio)
Is
where ge is the g-factor or the spectroscopic splitting factor measured to be
and angular momentum I. Then we know that the torque T is, by definition
(and that γ = M/I)
dI 1 dM
T= = (Torque)
dt γ dt
but also
T=M×B
giving
1 dM
= M × B. (The Bloch Equation) (15.2)
γ dt
Solving (15.2) is best done using matrices as it contains three components
d
dt Mx 0 Bz −By Mx
d My = γ −Bz 0 Bx My
dt
d By −Bx 0 Mz
dt Mz
d
dt Mx = γ 0 Bz Mx
.
d
dt My
−Bz 0 My
Thus Mz stays unchanged (as its derivative is zero). The next can be shown to
have the solution
Mx = M⊥ cos(−γBz t + φ),
My = M⊥ sin(−γBz t + φ).
hence
Mx = M⊥ cos(ωL t + φ),
My = M⊥ sin(ωL t + φ),
Mz = Mz .
z
B
ωL
process made by all the particles in some area as a sum, the spin of each single
atom is not known.
Add refer-
Example 15.1.2: Some numerical values ence.
fL (1 H) = 42.57 MHz,
fL (31 P) = 17.24 MHz,
fL (23 Na) = 11.26 MHz.
What we see is that in the rotating coordinate system, the dipole experiences
an external magnetic field
ω
B0 = (Bz + )ẑ
γ
instead of simply B = Bz ẑ in the stationary coordinate system.
CHAPTER 15. THE EASY PARTS 106
z
B
M0
ω0
Brot
x0
y0
|M|
Ta
Tb
t
T1 (Tb ) T1 (Ta )
Let us take a short step back and look how M is behaving in the stationary
coordinate system,
Now we turn off the rotating field, and measure the magnetic field in the
xy-plane, sometimes using the same coils as are generating the RF-pulses. Let
us take a look at what we will detect.
dMx Mx
= γ(M × B)x − (15.5)
dt T2
dMy My
= γ(M × B)y − (15.6)
dt T2
dMz Mz − M0
= γ(M × B)z − (15.7)
dt T1
1 1 ∆Bz (r)
= +γ (15.8)
T2∗ (r) T2 (r) 2
m0 (r) = µ0 ρ(r)(1 − e−tm /T1 (r) ) (15.9)
∗
m(r, t) = m0 (r)e−t(1/T2 (r)+1/T2 (r)) ≡ m0 (r)e−t(1/Tt (r)) (15.10)
1 1 1
= + (Local transversal dephasing time constant) (15.11)
Tt (r) T2 (r) T2∗ (r)
t = 0 at the end of rf pulse
me Mass of electron
M Magnetic dipole moment
I Mechanical angular momentum
T Torque
γ = M/I Gyromagnetic ratio
γo Orbital gyromagnetic ratio
γs Spin gyromagnetic ratio
g Splitting factor
ge Splitting factor for electron
ωL Larmor Frequency (even if it is an angular velocity)
fL = ωL /2π Larmor Frequency
110
Chapter 16
Gradients
111
CHAPTER 16. GRADIENTS 112
Look closely, and imagine k = γGt, and you will see that this looks very much
CHAPTER 16. GRADIENTS 113
and thus,
m(r) = (γt)3 Se (t, γrt). (16.3)
Definitions
me Mass of electron
M Magnetic dipole moment
I Mechanical angular momentum
T Torque
γ = M/I Gyromagnetic ratio
γo Orbital gyromagnetic ratio
γs Spin gyromagnetic ratio
g Splitting factor
ge Splitting factor for electron
ωL Larmor Frequency (even if it is an angular velocity)
fL = ωL /2π Larmor Frequency
G(r) = ∇Bz Gradient field
PNS Peripheral Nerve Stimulation
s Measured, real signal
sc Complex representation of s
se Envelope in sc
114
Chapter 17
Show Me Something
So how do we turn all this data into an image? In the last chapter, we saw that
where Se (t, r) = F {se (t, G)} is the envelope in sc = se eiω0 t , which in turn is
the complex representation of our measured signal s = Re{sc }.
115
Definitions
me Mass of electron
M Magnetic dipole moment
I Mechanical angular momentum
T Torque
γ = M/I Gyromagnetic ratio
γo Orbital gyromagnetic ratio
γs Spin gyromagnetic ratio
g Splitting factor
ge Splitting factor for electron
ωL Larmor Frequency (even if it is an angular velocity)
fL = ωL /2π Larmor Frequency
G(r) = ∇Bz Gradient field
PNS Peripheral Nerve Stimulation
s Measured, real signal
sc Complex representation of s
se Envelope in sc
116
Chapter 18
NMR
18.1 NMR
117
Appendix C
MATLAB examples
118
References
119