MEDICAL IMAGING EQUIPMENT

THEORY

Richard Aston
y
r
s a to
ll i m
0 Co
be
Tu

m
B ea
n c il
Pe
s
c to r
D e te
 MEDICAL IMAGING EQUIPMENT
THEORY

Richard Aston

engineering
ABC research

33 Barney Street
Wilkes-Barre
Pennsylvania, U.S.A. 18702
(570)829-3163
E-mail: [email protected]
Web Site: <www.epix.net/~astonr>
or
<www.epix.net/~astonr/index.htm>

Copyright © 2004 by Richard Aston. All rights reserved. May not be copied or distributed without signed written
permission of the author.
DISCLAIMER: This material is derived from the literature as indicated in the references. It is meant for
educational purposes only. The author regrets any omissions, misunderstandings or errors and disclaims any liability
for them, and disclaims responsibility for any implications these ideas might have for existing equipment, its design,
use, or maintenance.
 PREFACE

This book is meant to serve as educational materials for clinical and biomedical engineer ing and
biomedical engineering technology students interested in understanding medical imaging
equipment. It is directed at the senior/graduate level.
One basic question at the center of the work is, how is it that one can take a picture of a
slice through the body with out opening the skin? A clear and systematic answer is given with
the help of Fourier Transform mathematics, along with discussions of the physical principles of
the hardware involved. This approach has great educational value because it is straight forward
logic and gives the student a basis for designing and maintaining this equipment. However the
Fourier Transform approach is not always used, and is often replaced with computer iteration
and statistical techniques. A study of these computer algorithms is beyond the scope of the
present work. Thus our focus here is more on the hardware than on the software of system
analysis.
Classroom presentation materials such a slide show of the figures in the text along with
concise explainations is provided on the author’s web site:<www.epix.net/~astonr> or
<www.epix.net/~astonr/index.htm>. Student end-of-chapter exercises and a solutions manual for
instructors are available at ABC engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A.
18702. ([email protected]).
 MEDICAL IMAGING EQUIPMENT THEORY

Richard Aston
June, 2005

TABLE OF CONTENTS

Chapter 1 Magnetic Resonance Imaging (MRI) Equipment Theory 1

Introduction 1
Nuclear Magnetic Resonance 2
RF Radiation 3
Inducing NMR 4
FID 6
Spin Echo 6
Magnetic Resonance Imaging 7
The DC Magnetic Coil 8
Field Shaping Coil in the Z-direction 9
Field Shaping in the X- and Y-directions 9
The Gantry 12
The Fourier Transform 13
The MRI Block Diagram 14
Signal Processing 18
Receiver Processing 19
Image Resolution 19
Signal-to-noise Ratio 20
Mathematical Model for Image Calculation 21
Hardware Data 22
RF Coils 25
Fast-Scan Techniques 26
References 29
Engineering Data from Vlaardingerbroek 31
Further Information on MRI from the WEB 31
Exercises

Chapter 2 X-ray: Basic Theory 34

X-ray Production 35
X-ray Tube Current 35
X-ray Characteristics 37
 Planar X-ray 39
Image Production 40
Mass Attenuation Coefficient 41
X-ray Hardware 43
X-ray Tube 44
Beam Shaping Parts 47
X-ray Detection 48
The Power Supply 49
References 50
Exercises 50

Chapter 3 CT Scanner Theory 51

CT Scanner 51
CT Image Calculation 51
Improving Scan Speed 55
X-ray Detectors 56
Gantry Hardware 57
Fifth Generation CT Scanner 58
Image Properties 58
CT Scanner Calibration 60
Resolution 60
Patient Dose 61
Spiral Scanners 63
References 64
More Information on CT Scanners from the WEB 64
Exercises 66

Chapter 4 Nuclear Medical Equipment: Part 1. Planar Imaging 67

Hydrogen and its Isotopes 67
The Cyclotron 68
Radioactivity 70
Radio Nuclides 71
Tracking Radio Pharmaceuticals 72
Well Counter 72
Anger Camera 73
References 74
Exercises 75

Chapter 5 SPECT and PET Scanners 76

The SPECT Scanner 77
The SPECT Algorithm 78
 Advances in SPECT Scanning 80
Scintillating Crystals for Gamma Radiation 82
Attenuation Correction 83
The Positron Emission Tomography (PET) Scanner 83
Electronic Collimation 85
Collimation Errors 87
Axial Collimation 88
Detectors 89
Overview, Advances and Comparison of SPECT and PET Scanners 90
SPECT Scanner Advances 91
PET Scanner Advances 93
Troubleshooting 94
References 95
Exercises 96

Chapter 6 Ultrasonic Imaging Equipment Theory 97

Ultrasonic Waves 97
The Wave Equation 97
Solutions to the Wave Equation 100
Characteristic Impedance 103
Ultrasonic Wave Reflections 105
Power Reflection Coefficient 107
Power Transmission Coefficient 108
Power Attenuation 109
Power Relations at a Boundary 111
Ultrasonic Transducers 114
Impedance Matching 116
A-Scope Ultrasound Imaging 117
B-Scope Imaging 119
M-mode Display 119
Phased Array 120
References 121
Exercise 121

Appendix
Appendix A Magnetic Field of a Current Loop 122
Magnetic Field of a Solenoid 123
References 124
Appendix B Fourier Transform 125
Fourier Transform of a Pulse 126
 Inverse Fourier Transform of a Pulse 127
Three Dimensional Fourier Transform 128
References 129
Appendix C Phase Shift Circuit 130
Appendix D MRI fast scan technique 131
References 134
Appendix E Fan Beam Projection 135
About the Author 138
 1

CHAPTER 1
MAGNETIC RESONANCE IMAGING
EQUIPMENT THEORY

NOTE: This material is for educational purposes only. See the disclaimer above.

Introduction

Magnetic Resonance Imaging (MRI) equipment provides high

quality two or three dimensional pictures of internal anatomical
structures of the body without the need for performing surgery. It
is comparable with Computer Tomography (CT) x-ray scans. It
differs fundamentally from x-ray in its physics; it does not
require ionizing radiation that presents a cancer risk, as does
x-ray.
The MRI is based on the nuclear magnetic resonance
(NMR) procedure, invented in 1946 by Felix Bloch, Edward
Purcell, and coworkers, and used in the second half of the
twentieth century to chemically analyze substances. In this
procedure, a substance is placed in a strong DC magnetic field,
then radiated with radio waves, often in the television band. The
chemist can then identify the substance by measuring the
frequency at which it absorbs the radio waves. In 1972, Raymond
Damadian was issued a US Patent for showing how to detect
cancer with NMR. The next year, Paul C. Lauterbur, a chemistry
professor at the State University of New York at Stony Brook
demonstrated that one could make a cross sectional image of the
body using NMR (1.1). It was immediately obvious that this
 2

technique was potentially valuable as a diagnostic tool in

medicine. This no doubt produced the funding necessary to rapidly
do the engineering development necessary to make this into a
clinical tool.

Nuclear Magnetic Resonance

Most atoms may be

visualized as a spinning
magnet. These magnets
often have their north and
south poles oriented
randomly, yielding zero net
magnetization. However
when placed under a DC
magnetic field, the magnets
tend to orient so that a
vector leaving their north
pole lines up with the
applied field lines, B0, as
shown in Figure 1.1,
spinning on their axes,
producing a magnetization,
M, in a volume element
(voxel).
Nuclear magnetic
resonance occurs in each atom when radio frequency (RF)
radiation, polarized perpendicular to B0, is applied to the aligned
atoms. Under the influence of both B0 and RF, the axis of the
magnetic moments of the aligned atoms shifts at an angle with B0.
It precesses around the direction of Bo like a gyroscope, as shown
in Figure 1.2.
The frequency of the applied radiation, f, must be the value given
by the Larmor equation

f = B0 (1.1)

in order to cause precession of the atoms in the Figure 1.2 voxel,

stimulated by RF. Here , the gyromagnetic ratio, is a measured
quantity characterizing each atom and has units of hertz per Tesla
 3

(Hz/T), f is in hertz, and B0 in Tesla.

The frequency of the RF depends on
the type of atom. For the one most
important in clinical MRI, hydrogen,
the RF must be 42.58 MHz/Tesla.
Thus if B0 is 1.5 Tesla (T), the RF
must be 63.87 MHz to cause
precession of hydrogen atoms. Other
gyromagnetic ratio values of clinical
interest are (1.2) given in Table 1.1.
In practice, the RF is radiated by an
antenna and polarized at right angles
(orthogonal) to the direction of the DC magnetic field which is
taken to be in the z-direction in a Cartesian
coordinate system (x, y, z). The duration and
magnitude of the RF is usually sufficient to
orient the magnetic dipoles of the precessing
atoms into the x-y plane, reducing the Mz
component to zero.

RF Radiation

One of the simplest antennas for radiating the

RF is a single loop wire shown in Figure 1.3.
Notice in the plane of the loop, the x - y plane
in this case, the magnetic field generated by
the loop is in the z-direction, designated Bz.
The z-component of the field along a line
through the center of the coil is derived in
Appendix A as Equation (1.A4), and is given
by, (1.3)
2
0R I
Bz = 2(R2 +Z 2 ) 3/2
(1.2)

Here 0 is the permeability of free space (1.257 X 10-6

henry/meter), I is the current in amps in the wire, R, the radius in
meters of the wire, and Z the distance in meters from the center of
the loop in the z-direction. The magnetic field also has components
in the x-direction and the y-direction. In the plane of the wire, Z
becomes zero and Equation (1.2) reduces to
 4

0I
Bz = 2R (1.3)

and in this case the other components of the magnetic field are
zero. Here it is obvious that when the loop is smaller, the magnetic
field is stronger for a given current in the coil. The operator should
choose an antenna as small as possible, but large enough to radiate
the entire sample of interest.

Inducing NMR

In order to create an MRI image it is necessary to stimulate a

nuclear magnetic resonance in the target. The body is placed in the
DC magnetic field. The RF pulse is polarized at 90 degrees with
the DC magnetic field. The angle of precession that occurs
depends upon the magnitude and duration of the pulse shown at
the top of Figure 1.4. The magnetization angle of precession with
the DC magnetic field, , may be calculated from,(1.4)

= 2 B1 (1.4)
where is the gyromagnetic ratio, B1, the magnetic field from the
RF pulse and is the duration of the pulse. Thus, by adjusting the
duration and/or the intensity of the RF pulse one can set to
different angles. In practice the angles /2 and radians are often
used. When /2 is chosen, we produce a 90 degree precession, and
the magnetization vector lodges in the x - y plane. After the RF
pulse is turned off, the magnetization vector continues to precess,
but it falls off exponentially with a time constant T1. This
produces a
continuation
of the RF
radiation,
illustrated in
Figure 1.4, as
a free
induction
decay (FID)
signal. That
is, the
 5

magnetization MZ returns to its original value

M0 according to the equation

M Z = M 0 (1 − e −t/T1 )
(1.5)
where MZ is the atomic magnetization in the
z-direction, M0 is the value after the effect of
the RF pulse has transpired, and T1 is the time constant. This
equation is plotted in Figure 1.5. At t = 0 MZ is zero, and after 5
time constants it rises to M0, and the magnetic dipoles are
realigned with the applied DC magnetic field.
The value of the time constant T1 varies from .2 to 2 seconds in
the brain. Some example values appear in Table 1.2 (1.5). The time
constant of decay for the Thalmus is 703 ms. This means the
magnetization will recover 63.21% of its original value in 703 ms
after the RF is shut off, and in five time constants it will be
essentially zero. The difference in the T1 values is used in the MRI
to distinguish between these tissue types.
The magnetization, MXY, in the x-y plane is influenced by a
dephasing process. The individual atoms in a particular pixel are
oscillating at slightly different frequencies. Therefore as time
passes they fall out of phase with each other, and their
magnetizations sum to a value less than that if they remained in
phase. This causes MXY to decrease. Dephasing due to frequency
differences is illustrated in Figure 1.6. Notice the two waves
begin in phase, but by the end of the cycle, the higher frequency
wave is leading in phase.
Magnetization in the x-y plane after the RF stimulus is
turned off follows the formula
M XY = M XY0 e −t/T2 (1.6)
where MXY is the atomic magnetization in the x-y plane, MXY0 is
the value the instant after the RF pulse has
transpired (t = 0), and T2 is the time constant.
This equation is plotted in Figure 1.7. After 5 time
constants the value of MXY diminishes nearly to
zero. Notice in Table 1.2 that the time constant T2
is much less than T1. Therefore the MXY fades
 6

away before MZ has re-aligned. This fact is important in the

Spin-echo RF stimulus technique discussed below.

FID

During the time T1, the magnetic dipoles are precessing around
their respective atoms, and they radiate RF energy, shown in the
lower part of Figure 1.4. Its magnitude diminishes exponentially
with the time constant T1, and the dephasing process. The signal,
called the field induction decay (FID), can be picked up by the
same antenna, as shown in Figure 1.2, that transmitted the
stimulus RF pulse. This time the antenna is switched to the
receiver. In the nuclear magnetic resonance (NMR) procedure, the
sample is radiated by a band of frequencies; the one which equals
the Larmor frequency, given by Equation (1.1), will stimulate the
sample. It precesses at its Larmor frequency when the RF is shut
off. By measuring the frequency of the received FID and using
Equation (1.1), one can calculate the gyromagnetic ratio , of the
sample, provided the DC magnetic field strength is known.
____________________________________________________
EXAMPLE 1.1. A sample is placed in a DC magnetic field of 1.5
T. It is radiated by a short RF pulse, whose magnetic field is
perpendicular to the DC field. An antenna connected to a radio
receiver and oriented as in Figure 1.1 is used to pick up the field
induction decay. The frequency of the FID is measured as 63.87
Hz.
a. Compute the gyromagnetic ratio of the sample.

b. Use Table 1.1 to identify the sample.

SOLUTION
Part a. From Equation (1) we compute

= Bf0 = 63.887Hz
1.5T = 42.58 Hz
T
Part b. The atom with this geomagnetic ratio is seen in Table 1 to
be hydrogen (1H).
_____________________________________________________

Spin Echo
 7

Spin Echo is a technique for stimulating NMR that helps

compensate for signal loss due to slight differences of the DC
magnetic field as a function of position. Such differences, or
inhomgenaities, cause atoms to precess at different frequencies
according to the Larmor equation. They therefore spread further
and further in phase as time progresses. The individual atom
magnetic vectors will therefore add up to less than they would if
they were all in phase, resulting in a loss of signal. To counter this
the spin-echo RF pulse sequence is applied as shown in Figure 1.8.
At a time t = 0, when the 90o pulse ends, and FID begins as shown
in Figure 1.8, some magnetization vectors lead others in phase. At
time TE/2, a 180o pulse of RF is applied which causes the leading
magnetization vectors to lag. But since they are at a higher
frequency, at time TE, they catch up and all magnetization vectors
are in phase again. (This process can be compared to a
noncompetitive race in which the runners start at the gun together.
The faster ones lead, but the gun is
fired again, signaling all the
runners to turn around and head
back to the starting line. They all
finish at the starting line together
because the faster ones run further.)
This gives an increased signal,
called the echo. Thus the spin-echo
technique improves the FID signal.

Magnetic Resonance Imaging

NMR evolved into a medical

imaging technique when Paul
Lauterbur investigated the Larmor
Equation (1.1) and realized that a
change in applied magnetic field
changes the frequency of
precession of any particular atom.
Thus if the applied magnetic field
intensity is shaped as a function of
position, a measurement of the
frequency of resonance of known
atom types can be correlated to their position. The human body is
 8

more than 60 percent water, therefore it is convenient to detect

hydrogen, 1H, listed in Table 1.1. The other important factor is the
relaxation times T1 and T2, some of which are listed in Table 1.2.
These values differ among the tissues the hydrogen appears in and
can be used to identify the type of tissue. In summary:
yFrequency of the FID measures
tissue position
y Relaxation times of the FID
identify tissue type
The first alternation of the NMR apparatus
needed to make it into an MRI is to shape
the magnetic field as a function of position
under the DC magnet. This is done with
coils which cause a variation of the field in
the x-direction, Gx, the y-direction, Gy ,
and the z-direction, Gz , written as:
ØB x
Gx = Øx (1.7a)
ØB y
Gy = Øy (1.7b)
ØB z
Gz = Øz (1.7c)
where Bx, By, and BZ are the magnetic fields created by the
x-direction, y-direction and z-direction coils respectively.

The DC Magnetic Coil

The DC magnetic field may be produced

by a permanent magnet, illustrated in
Figure 1.1 when small samples are being
imaged. But for most medical applications
an electromagnetic magnet is used, as
illustrated in Figure 1.9. The DC magnetic
field in this case is produced by a
solenoid, a wire wrapped around a
cylindrical core, large enough to allow one
to insert a prone human body for imaging.
The magnetic field in core, whose radius is
much smaller than its length, at its center,
is given approximately by (1.6)
 9

NI
B0 = l (1. 8)
Where is the permeability in Henry per meter (H/m) (1.257x10-6
H/m in free space), N is the number of turns, I is the current in
amperes, and l is the length in meters, and B0 is the magnetic field
in the direction of the axis of the coil in Tesla (T) units.
______________________________________________________
EXAMPLE 1.2. How many amps are required to produce a 1.5 T
magnetic field at the center of a solenoid which
is 4 meters long, having 10,000 turns of wire?
SOLUTION. Solving Equation 1.7 for I, and
substituting the numbers gives,

I= 1.5Tx4m
1.25x10 −6 (H/m)x10000 = 480 amps
_______________________________________

Field Shaping Coil in the Z-direction

The field shaping coil which produces GZ,

Equation (1.7c), the variation in the field in the
z-direction, is illustrated in Figure 1.10. Notice
at the top end of the coil, a current entering in the direction of the
arrow produces a positive magnetic field gradient, GZ. Then as
you proceed down the coil, the spacing between turns increases,
causing a smaller magnetic field. In the middle of the coil, the
magnetic field reduces to zero. Going further, the turns direction is
reversed, GZ become negative, and continues such to the end of the
coil. The spacing between the coils is varied so that the magnetic
field varies linearly from +GZ to -GZ.

Field Shaping Coil in the X- and Y-directions

To understand how field is caused to vary in the x-direction,

consider two coils side-by-side as shown in the top-view of
Figure 1.11. The magnetic fields caused by these currents are
shown in the front view. Notice the currents are going in opposite
directions, and therefore, by the “right hand rule” produce
opposing magnetic fields at the isocenter, which cancel to zero.
 10

Now four coils are arranged as shown at the top of Figure

1.12. Looking at section A-A’ we see that the magnetic fields
produced by the currents in the directions
shown, cancel at the center of the four-coil
structure. Thus an isocenter appears as shown
in the lower part of the figure. As you travel in
the positive x-direction, the magnetic field
increases positive, and in the negative
x-direction goes negative.
A set of coils turned at right angles to the
ones in Figure 1.12, would likewise produce a
magnetic field which varies in the same way in
the y-direction.
In the MRI, these coils must fit in the
cylindrical hole in the gantry, that the body of
subject being imaged must fit. Therefore the
coils are bent into a saddle shape as illustrated
in Figure 1.13. Here coils 1 and 3 are on the
top, and coils 2 and 4 are on the bottom.
The coils are arranged so that the fields
vary approximately linearly in the space of
interest, as shown in Figure 1.14.
 11

In this figure, empirical data from

saddle coils oriented to produce a linear
field variation in the x-direction is shown
(1.7)
. The range of linearity is
approximately 10 cm in the z-direction,
and at least 20 cm in the x-direction. The
steeper the gradient of the magnetic field
the better the resolution of the image,
because differences in field strength
translate into differences in frequency,
which in turn measure position in the
image. The coil sensitivity, C, is
measured in terms of the amount of
current required to produce a given
magnetic field gradient. A typical value
for C is 30 A/mT/m, according to
(1.8)
Vlaardingerbroek. In operation, this current must be delivered
by a short pulse. The pulse rise time, R, is an important design
parameter. In terms of the inductance L and resistance R of the
coil this is given as,

R = RL (1.9)
Vlaardingerbroek gives a typical value of L as 200 H. He also
cites the field gradient as 12 mT/m, rise time 0.6 ms, and the peak
power 43 kW .

____________________________________________________
EXAMPLE 1.3. A saddle coil used to shape the magnetic field in
the y-direction has an inductance of 200 H and a rise time of 0.6
ms. What is the resistance of the coil?
SOLUTION:
From Equation (1.9) we calculate
200x10 −6
R= L
R = 0.6x10 −3 = 0.33
____________________________________________________
EXAMPLE 1.4. A saddle coil draws 30 A peak current and 43 kW
of power. What is the peak voltage, V?
 12

SOLUTION:

V = 43000
30 = 1433.3 V
____________________________________________________
EXAMPLE 1.5. If the magnetic field gradient in the y-direction is
12 mT/m in a DC magnetic field of 1.5 Tesla, what is the
corresponding frequency gradient of the precessing hydrogen
atoms?

SOLUTION: From Table 1.1, the gyromagnetic ratio of 1H in the

1.5 T field is 42.58. Thus

f = 1.5x42.58 MHz x12x10 −3 m

T
T
= 0.766 MHz
m
____________________________________________________

The Gantry

The Gantry, shown in the

bottom of Figure 1.15, is a
structure that insures a
proper orientation of the
subject being imaged, the
magnetic coils and RF
antenna coil. In top of the
figure the coils described
above are shown, giving the
direction of the currents. The
plus sign, +, indicates the
current is going into the
page, and the dot, *, indicates
a current coming out. The
patient rests on a pallet and is
positioned in the hole in the
gantry, as shown.
 13

The DC coil in Figure 1.15 causes a constant magnetic

field, BZ, in the z-direction which is always on. The effect of the
other coils is to shape the magnetic field so that the atoms being
imaged are under different magnetic fields, and according to the
Lamor equation, will emit different frequencies, according to their
position. Energizing the GZ coil causes variation of the magnetic
field along the z-axis, GY along the y-axis, and GX along the x-axis.
Energizing the RF coil makes the atoms precess. We will now
discuss how and when this magnetic field shaping takes place, and
how an MRI image is developed. A knowledge of the Fourier
Transform is basic to our understanding this process.

The Fourier Transform

As we discuss in Appendix B, the Fourier Transform is an

extension of the fourier series. A fourier transform pair is given by
two integral equations, (B2) and (B3) in Appendix B, as:
∞
f( t) = 2
1
¶ F( )e j t
d
−∞
(1.10)
∞
F( ) = ¶ f( t) e − j t
dt
−∞

(1.11)
Here F( ) is the fourier transform of
f(t). Equation (1.10) is a function of
time alone, since the radian frequency,
, is a running variable. We therefore
say the function f(t) is in the time
domain, while F( ) is in the frequency
domain. Equation (1.11) expresses that
function in the frequency domain, since
it is a function of frequency alone.
Equations (1.10) and (1.11) form a
fourier transform pair, (1.11) being the fourier transform, and
(1.10) being the inverse fourier transform.
 14

An example fourier transform pair, shown in Figure 1.16

and calculated in Appendix B, is very important to the operation
of the MRI. A voltage pulse shown in the top of Figure 1.16 is a
function of time, v(t). Its fourier transform, calculated from
Equation (1.11) has the form V(f) shown in the lower part of the
figure, as calculated in Example B2 in Appendix B. The practical
significance of this in the MRI is that a pulse of RF applied to the
antenna coil of the shape shown in the top of the figure has
frequency components given in the lower part of the figure.
Furthermore the value of the frequency components is adjusted by
changing the voltage waveform in the time domain. Because each
of these frequency components causes atoms to precess, or be
illuminated, at different positions in the subject, this gives the
operator a means of selectively illuminating different parts of the
body with the RF antenna coil.

The MRI Block Diagram

The block diagram of the MRI appears in Figure 1.17 which is

used to show how the parts described above are interrelated to
produce an image.
15
 16

The subject being imaged is under the RF loop antenna and

through the gradient coils as shown in Figure 1.15. The transmitter
in Figure 1.17 energizes the RF loop antenna in order to cause the
hydrogen atoms in the subject to precess. The x, y and z-gradient
amplifiers create the magnetic field gradients by driving currents
into the appropriate gradient coils. When the transmitted RF is
turned off, the receiver is turned on to pick up the signal from the
subject and process it into an image. The details of this process
follow:
The signals controlled by the MRI
electronics are illustrated in Figure 1.18 which
follows Wehrli.(1.9) Here the time sequence of
the pulses which excite the RF antenna coil
and the magnetic field shaping coils is shown.
The sequence of pulses shown is repeated, for
example, 256 times to acquire the data needed
to make one slice through the subject imaged,
shown in Figure 1.15. The data from each
voxel in the slice is recorded in a matrix
representing the x-y plane as illustrated in
Figure 1.19. One way to understand how the
pulses in Figure 1.18 create an image is to
imagine yourself on one of the voxels in the
slice. The RF pulse creates a spin-echo
sequence as described above. The waveshape
is “sine x over x” (sinc x), the fourier
transform of which is a band of frequencies.
Simultaneous with the first 90o RF pulse in
time period A, the z-direction coils are
energized so that the magnetic field in one of the slices would
cause spins in its voxels corresponding to those in the RF pulse, as
demanded by the Lamor equation. Thus the RF pulse and magnetic
field gradient GZ work together to select or illuminate one slice.
The voxels in that slice precess at frequencies determined by the
Lamor equation and the gyromagnetic ratio of the tissue in the
voxel after the 900 pulse is turned off. The phase of the precession
in the slice differs from what it would be if it were at the isocenter
of the magnetic field because the frequency is different, as
illustrated in Figure 1.6. To compensate for this the next pulse in
time-period B from the z-direction gradient is in the opposite
direction, making a frequency shift of opposite sign from the
 17

isocenter. Thus all slices will have the same phase due to the
z-direction magnetic field as that at the isocenter.
In time-period B, GX and GY gradients are energized so that
the magnetic fields across the slice of Figure 1.19 vary linearly in
the x and y directions. Thus the voxels in the slice each take on
different frequencies as determined by the Lamor equation and the
differing magnetic fields, and by the end of the time-period, B
acquire different phases.
In time-period C, all magnetic field gradients are shut down
so the Lamor equation and the magnetic field B0 alone determine
the frequency of resonance, but the phase differences will remain
as set in time-period B. The 180o RF pulse is applied to cause
cancellation to dephasing according to the spin-echo method
described above.
In time-period D, the gradient GX is applied. This causes
frequencies of precession to vary linearly as a function of x in the
slice, but the phases in the y-direction will remain as set in
time-period B. In period D, the receiver in the block diagram,
Figure 1.17, is turned on. The summation of all the radiations from
the voxels in the slice are picked up by the antenna coil. The
received signal is in the time domain. The signal contains a set of
frequencies, magnitudes and phases of each of the 2562 voxels in
the slice which can be analyzed by fourier transform techniques as
discussed below.
The time periods discussed are in the order of 10 to 30 ms.
The voxels in the slice will continue to resonate up to 1 or 2
seconds, as shown in Table 1.2. After these resonances
have ceased, the time TR, another sequence of pulses in
figure is applied. They are all identical to the first set,
except the magnetic field gradient GY is set to another
value. Thus the 2562 voxels will precess as before, except
the phase shifts in the y direction will differ. This process
is repeated 256 times, in order to give enough data
characterize all 2562 voxels in the slice. Typical times
required to image one slice by this method is 256/60 = 4.3
minutes to 8 minutes. Since the echo time TE is in the
order of 10 milliseconds, computer time-sharing can be
used to make many slices in this 4 to 8 minutes.

Signal Processing
 18

To understand how the signal emitted from the voxels in the

patient produces an image in an MRI, it is useful to consider an
analogy of what happens when
we hear a cord struck on a
piano. In both cases fourier
transforms operate.
When we strike a cord on the
keys of a piano, the individual
strings vibrate at their own
resonant frequencies, much as
voxels do in the slice of the
patient in the MRI. The strings
are spatially arranged in the
piano as are the voxels in the
slice. The set of vibrating piano
strings form a fourier
transform. The sounds from the
individual strings add in the air
and become a pressure wave
that varies as a function of time
which is applied to the ear as
shown in Figure 1.20.
Likewise, the RF vibrations of
the voxels in the slice in the
MRI add in air and become a
function of time applied to the
receiver antenna loop.
In Figure 1.20, the sound pressure wave enters the inner ear
through the stapes bones that set up a standing pressure wave in
the cochlea. Low frequency sounds cause stronger vibrations deep
in the ear, to the right, and high frequency sounds cause stronger
vibrations to the left. The hair cells on the basilar membrane pick
up these vibrations and transmit the signal to particular locations in
the brain by way of the auditory nerve. Thus the ear serves to
transform a sound pressure wave shown in Figure 1.20 into
frequency components located as a function of position along the
cochlea. This transformation in mathematics is called a Fourier
Transform. Furthermore, because the process occurs fast enough
that one speech sound after the other can be individually analyzed
 19

into its frequency components and presented to the brain in real

time, the process is called a Fast Fourier Transform(FFT).
The frequency components of the sound are then
transmitted along individual nerve fibers to spatially distinct places
in the brain, and the listener interprets the sound as music. In the
MRI the radio wave picked up by the antenna loop is operated on
by an FFT and the individual frequency components are placed in
spatially distinct memory locations in the computer which create
an image on a display unit. Thus the computer in the MRI
functions in an analogous manner to the brain.

Receiver Processing

The output of a slice on the patient from the sequence of

excitations depicted in Figure 1.18 is an electromagnetic wave
having a frequency usually between 40 and 100 MHz. It is picked
up by the RF loop antenna and applied to two mixers (1.10) which
amplitude modulate it with a carrier, one being 90 degrees phase
shifted from the other. That is, one mixes with cosine functions
and the other with sine functions (1.11). After analog to digital
conversion (ADC) on the block diagram, a fast fourier transform
(FFT) is performed on these outputs to produce the magnitudes
and phases of one row of a matrix for each presentation of the
excitation sequence of Figure 1.18. The real part of the complex
number comes out of the Mixer (cos) and the imaginary part from
the Mixer (sin) in Figure 1.17. After 256 repeats of this excitation,
in the example we are considering, two matrices are formed, one
for the real part and one for the imaginary part, from which the
magnitudes of the radiations of one slice of voxels can be
calculated by algorithms in the computer and placed in their
correct positions to create a visual image.

Image Resolution

The image resolution, x, in the x-direction and y in the

y-direction, is the minimum size object that can be distinguished.
A measure of this can be derived from the Lamor Equation (1.1).
Differentiating this equation,

= = Gx =
df dB x f
(1.12)
dx dx x
 20

where we replace the differentials df and dx with discrete

differences f and x. So
x= 1
Gx f (1.13)
Here f represents the difference in frequencies between a voxel
and its adjacent one in a slice shown in Figure 1.19. The number of
voxels in the x-direction we call Nx. The duration of the output
pulse we call TD in Figure 1.18. Thus the bandwidth, BW, of the
output RF applied to the antenna coil is

BW = N x f (1.14)
Equation (1.13) shows that to achieve better spatial resolution it is
necessary to select between closer frequency spacings. One can
visualize tuning across the bandwidth with a more narrow band
receiver for better resolution. The time to scan the entire band
should be longer for better resolution. . S Wright (1.12) shows that

x= 2
G x TD (1.15)
Thus the best resolution occurs for a large magnetic field gradient,
Gx, and for a long duration RF output pulse.(1.13)
____________________________________________________
EXAMPLE 1.6 A typical MRI has a magnetic field gradient, GX =
12 mT/m, the DC magnetic Field is 1.5 T and an RF output
duration, TD = 20 ms . What is the resolution in the x-direction.
SOLUTION: The gyromagnetic factor, = 1.5x42.6 from Table
1.1. Thus Equation (1.14) yields,

x= 2
1.5x42.6x10 6 x12x10 −3 x20x10 −3 = 0.41mm
______________________________________________________
Thus we can theoretically resolve an object larger than 0.41 mm in
width in this case.

Signal to Noise Ratio

The signal to noise ratio, SNR, is increased by factors that make

the RF signal larger. These are the DC magnetic field, the voxel
volume, VV, and the output duration, TD. Conolly (1.14) gives the
value as proportional to
 21

V
SNR } B 0 TD V Vn (1.16)
where Vn is the noise volume. Vn increases with the total volume
being imaged, and the SNR is degraded when larger volumes are
imaged. Thus the SNR is degraded when one attempts to use a
smaller DC magnet, improve the resolution with smaller voxel
volumes, and to create images faster by decreasing TD.

Mathematical Model for Image Creation

A Mathematical model that can be programed on the computer to

create an image begins with the individual voxel. During the
readout of the magnetization resonance, TD, in Figure 1.18, each
voxel emits a magnitude due to its tissue type and depending on its
position, (x,y,z), and a phase that is generated by the magnetic
field gradients given by
t
−i ¶ ⎯→ →
G( )$ r d
e 0
where is the magnetization coefficient in units of MHz/T, G, the
magnetization gradient is a vector

G = xG x + yG y + zG z (1.16)
which forms a dot product with a position vector

r = xx + yy + zz (1.17)

All of these radiations reach the receiver antenna and add to form a
signal, s(t) as a function of time given as(1.5)
t
∞ ∞ ∞ −i ¶ ⎯→ →
G ( )$ r d
s ( t) = ¶ ¶ ¶ (x , y , z)e 0 dx dy dz
−∞ −∞ −∞

(1.18)

k-space is defined by the quantity

 22

→ t
→
k ( t) = 2 ¶ G ( )d (1.19)
0
which has the units per meter (m-1), and is a quantity depending on
the magnetic field gradients chosen in the MRI as well as the
values of the different tissues in the slice. This forms the k-space
vector

k = xk x + yk y + zk z (1.20)
so Equation (1.18) becomes this
∞ ∞ ∞ →→
s(t) = ¶¶¶ (x, y, z)e −i 2 k $ r
dx dy dz
−∞ −∞ −∞
→
= S( k )
(1.21)
This has the form of a three dimensional Fourier transform. As
stated in Appendix B, Fourier has shown that it has an inverse
equality given by
∞ ∞ ∞
→ → → →
(x, y, z) = ¶ ¶ ¶ S ( k )e + i 2 k $ r
d k
−∞ −∞ −∞

(1.22)
This equation is calculated by the computer for each of
approximately 256 times the pulse sequence of Figure 1.18 is
applied, and a gray-scale image is formed.

Hardware Data
 23

Progress in the development of the MRI is limited, in large

measure, to the size of the DC magnetic field. The higher this field
is, the better the image is. It has a higher signal to noise ratio and
more elements can be imaged other than hydrogen, such as
phosphorous and sodium. However the magnetic field is limited
by the physical size of the magnet. Three types of magnets have
been used: permanent, electromagnets, and superconductor
electromagnets. A Permanent magnet is illustrated in Figure 1.21.
The DC magnetic field must be as uniform as possible to prevent
signal loss due to dephasing, as discussed above. The soft iron
shims shown in Figure 1.21 provide a standard way of calibrating
the field. The metal alters the magnetic field pattern in a controlled
fashion, and can be conveniently adjusted . MRIs use fields
between .02 to 8 Tesla (1 Tesla = 10,000 Gauss) and more. The
higher the DC magnetic field, in many cases, the better because
y Larger fields yield better contrast and intensity of images
y Larger fields glean scans of more types of elements
It is easy to achieve a multitesla field if the gap in magnet is small.
But it becomes difficult if the gap must be large enough to pass a
person. Smaller gap MRIs may be used to image the head, or knee
alone for example. Permanent magnets may be used in low field
applications of less than 0.3Tesla ( T). They have the advantage of
low maintenance and operating costs. However they are heavy, in
the order of 20 tons at the 0.2 T level if iron is used in a magnet
capable of imaging a patient.
With neodymium alloy, the
weight reduces to 4½ tons.
A second alternative,
resistive magnets are
electromagnets in which the
magnetic field is produced
by a coil of ordinary
conductive wire, usually
copper, wrapped around an
iron core, such as illustrated
in Figure 1.9. They are
usually used only in small
field applications of 0.02 to
0.06 T. For high field
applications they require
high current and present
 24

cooling problems. For example, a 0.3 T magnet capable of imaging

a patient requires 100 kilowatts (kW) of power. (1.16) A third
alternative, superconductor magnets, were developed in the 1980s,
and have made 8 Tesla DC magnets practical. Superconducting
wire must be cooled to a very low temperature. For example
niobium-titanium wire becomes a superconductor below 10
degrees Kelvin (K) above absolute zero. Figure 1.22 shows a
superconducting coil magnet for an MRI. It has chambers for both
liquid nitrogen which is a low cost barrier to room temperature and
liquid helium which brings coils to superconductivity. The
superconducting coils are maintained at liquid helium
temperatures. As the temperature of the helium rises from heat
absorption, new liquid is added to the chamber through the
cryogenic transfer neck shown in the figure. The coolant is usually
replaced about every five weeks. A vacuum jacket is used to
provide insulation for maintaining cooling . The magnetic field
in the bore, or tube, in the gantry is made homogenous by
adjustment of the shim coils. If the temperature is maintained
below the
superconducting
threshold, the DC
magnetic field may be
maintained for weeks
without power being
delivered to the coils, as
the magnetic loss is less
than 0.001 gauss per hour.
For example, 200 Amps
of current produces a 1.5
T magnetic field in a 1.3
meter diameter tube.
Since the current is
flowing in a
superconductor, it
produces essentially no
power loss. A larger
diameter tube would
require more current to
produce this field
strength.
 25

Occasionally the magnet will suddenly loose its field in a

fault process called “quenching”. Vibrations in the coils may
produce enough heat to cause a local loss of super conductive
temperature and the current in the coils generate more heat. The
process avalanches and the magnetic field is rapidly lost,
sometimes with a loud noise. (1.17)
When an MRI scan is taken, the patient is placed on a
computer-controlled pallet and moved into the tube in the gantry.
In closed versions of the MRI, the tube is rather long and fits
closely around the patient. A small, but significant percentage of
the patients experience claustrophobia in this situation. That along
with the noise of the scanning equipment can make some patients
uncomfortable. One source of sound is from the current pulses on
the magnetic field shaping coils. Lorentz forces are set up which
cause vibrations heard as noise. (This is the same process that
causes hum in power transformers.) One does not feel the
electromagnetic fields, and they have no measured biological
effects other than those of heat generated by high power. However
the magnetic field is very strong, and if a metal object gets picked
up it will accelerate as it approaches the magnetic coils. Patients
have been severely injured, even fatally, when hit by a metal object
falling through the magnetic field.

RF Coils

An improvement on the single loop RF coil antenna, shown in

Figure 1.3, was introduced by C.E. Hayes and his group in 1985,
 26

the Bird Cage Coil in Figure

1.23. (1.18) This device improves
the sensitivity of the antenna
and the signal-to-noise (SNR)
ratio. The lines on the bird cage
antenna in the figure are
connected conductors. If one of
the rings is energized with RF at
two points, one stimulated 90o
out of phase with the other, and
displaced along the circle by 90o
it will resonate such that the RF
radiation is polarized in the x-y
plane, or in a patient slice, and
perpendicular to the DC
magnetic field. The 90o phase
shift may be introduced with a
¼ wave RF transmission line.
Or an RC network, shown in
Figure 1.23, introduces a 90o
phase shift under the conditions:

2 f r RC p 1
(1.23)
A V = −2 f r RC
(1.24)
This result is shown in
Appendix C. Here Av is the
buffer amplifier voltage gain,
and fr is the RF frequency, as
calculated by the Lamor
equation (1.1).
The bird cage antenna coil can be made large enough that a
patient could fit through the bore. Smaller coils are used to image
the head alone, and smaller to image a joint in the arm or leg.

Fast-Scan Techniques
 27

The spin-echo scan sequence

illustrated in Figure 1.18
produces a single MRI scan
in 4 or more minutes in some
cases. This is longer than a
breath-hold by a patient, and
certainly longer than the
period of pulsation of the
heart and other pulsating
organs. Thus to eliminate
organ motion and breathing
artifacts it is important to
find faster excitation
sequences. One such
method, the gradient-echo
imaging, called FLASH
(fast-low angle shot) by its
originators (19) uses a low flip
angle, , (as defined in
Equation 1.4) and eliminates
the 180o echo pulse, and does not require that the T1 and T2
transients pass before the following excitation pulse is applied. The
timing diagram for the gradient-echo sequence, Figure 1.24,
produces an echo by adjusting the gradient Gx so that the spins
have zero phase at the center of the sampling output window. (20)
 28

A derivation of the magnetization in the x-y plane resulting from

the FLASH pulse sequence is given in Appendix D as Equation
(D14),

− TE&
= M0 1 − e − TR sin e T2
M 0xy T1
TR
1−cos e − T1 (1.25)
where M0 is the initial longitudinal (z-direction) magnetization,
and
M 0xy is the component of the magnetization in the x-y plane. T2*
is the x-y plane magnetization time constant modified by machine
induced factors.
A plot of Equation (1.25) for the case TE = 30 ms,
T2 = 300 ms, ( here T2* is taken as equal to T2) and the tip
angles are taken successively as = 20, 30 and 90 degrees
is given in Figure 1.25. The plot is made by programing
Equation (1.25) on a Texas Instruments TI-89 graphics
calculator. The abscissa , TR/T1, is carried over a range
from 0 to 5.
Figure 1.25 shows that the magnetization in the x-y
plane holds its value at TR/T1 ratios about at 0.1. Thus the
scan time is significantly reduced. For example, if the
tissue imaged has a T1 of 0.5 s, a TR of 50 ms could be
practical. At 256 lines per scan, this would imply at scan
time of 0.05 X 256 = 12.5 seconds, enough so that the scan
could be done in one breath hold. In this method, TR
should be kept greater than TE, otherwise a focusing echo would
not develop.
The effect of the FLASH sequence in Figure 1.24 on the
signal-to-noise ratio (SNR) helps guide the user in selecting values
for TR, TE and the tip angle. The signal voltage will increase as
the number of exications (NEX) increases, thus increasing the
 29

SNR. Since noise is electrical power and signal is voltage, this

increase is proportional to NEX . Also, the TR is the time period
of these excitations, so signal increase proportional to [1/TR ] .
The signal is also proportional to Mxy in Equation
(1.25). So we write, (1.10)
TR
1 1− e − T1
SNR ~ TR 1− cos e − TR
sin
T1 (1.26)

This Equation shows the effect of repetition time, TR

and tip angle on the signal-to-noise ratio, and is plotted
using the TI-89 graphics calculator in Figure 1.26 for the
case TE = 30 ms, T2 = 300 ms, ( here T2* is taken as
equal to T2) and the tip angles are taken successively as
= 20, 30 and 50 degrees.

References

1.1. Lauterbur, P. C. “Image formation by induced local

interactions: Examples employing nuclear magnetic resonance.”
Nature London: vol 242 190-191, (1973).

1.2. Hornak, J.P. The Basics of the MRI Rochester: Rochester

Institute of Technology (www.cis.rit.edu/htbooks/mri/) Ch 3, p 2 .

1.3.Vlaardingerbroek, Marinus T., Jacques A. Boer. Magnetic

Resonance Imaging: Theory and Practice. Berlin: Springer, pp
32-33 (©1996).

1.4.Wehrli, Felix W. “Principles of Magnetic Resonance, Chapter

1” in David D. Stark and William G. Bradley, editors. Magnetic
Resonance Imaging St. Louis: Mosby Year Book, p 5 (© 1992).

1.5. Wehrli, F. W., Chao P. W., Yousem, D. M.: “Parameter

dependence of susceptibility-induced signal losses in gradient-echo
imaging. Magnetic Resonance Imaging 7 (Supplement 1): 139
(1989).

1.6. Kraus, John D., Keith R. Carver. Electromagnetics 2nd

Edition. New York: McGraw Hill, Inc. (1973) pp 155-158
 30

1.7. Vlaardingerbroek, op.cit. p 27.

1.8. ibid. p 28.

1.9. Wehril, “Principles of Magnetic Resonance, Chapter 1”, op.

cit. p11.

1.10 Schwartz, Mischa Information Transmission Modulation and

Noise. New York: McGraw Hill Book Company, Inc. (1959) pp
103-106.

1.11. Vlaardingerbroek, op. cit. p 38.

1.12. Wright, Steven. “Nuclear Magnetic Resonance and Magnetic

resonance Imaging”. Chapter 16 in John Enderle, Susan Blanchard,
and Joseph Bronzino, editors of Introduction to Biomedical
Engineering, San Diego: Academic Press (2000) p 821.

1.13. Conolly, Steven, A. Macovski, and J. Pauly. “Magnetic

Resonance and Imaging: Acquisition and Processing” in Chapter
65.1 in The Biomedical Engineering Handbook ed. by Joseph D.
Bronzino, Boca Raton, FL: CRC Press (1995) pp 1010.

1.14 Conolly, ibid. p1010.

1.15 Suetens, Paul. Fundamentals of Medical Imaging, Cambridge,

UK: Cambridge University Press (2002) p 112.

1.16. Schneck, John. “Magnetic Resonance and Imaging:

Hardware/Instrumetation” in Chapter 65.2 in The Biomedical
Engineering Handbook ed. by Joseph D. Bronzino, Boca Raton,
FL: CRC Press (1995) pp 1014-1020.

1.17. Matwiyoff, N. A. “Instrumentation, Chapter 3” in David D.

Stark and William G. Bradley, editors. Magnetic Resonance
Imaging St. Louis: Mosby Year Book, p 71. (© 1992).

1.18. Hayes, C.E., W.A. Edelstein, J. F. Schenck, et. al. “An

efficient highly homogeneous radiofrequency coil for whole body
 31

NMR imaging at 1.5 T. Journal of Magnetic Resonance, 63:622,

(1985).

1.19. Haase A., Frah J, Matthaei D, et al. FLASH imaging: rapid

NMR imaging using low flip angle pulses. J Magn Reson
1986;67:258-66.

1.20. Wehrli, Felix W. Fast-Scan Magnetic Resonance: Principles

and Applications. New York: Raven Press Ltd. (1991) pp. 10-21.

Engineering Data from Vlaardingerbroek:

y Ball park figures for saddle coils: L = 200 H; coil

sensitivity G/I = C = 30 A(mT/m)-1. Field gradient 12
mT/m. Rise time 0.6 ms, peak power 43 kW (p 28).
y DC magnetic field homogeneity: better than 5ppm. (p
12)
y Superconductor wire: niobium titanium alloy
superconductive below 12 K. Liquid Helium 4.3K. (p
18)
y Prism Magnet: 0.15 T, gap .7 m, length 1.05 m, total
weight 3400 kg. Permanent magnetic material NdBFe
(33%) and iron yoke (60%). (p 22)
y Magnetic homogeneity of less than 5 ppm within a 50
cm sphere (p 24)
y Spine may require 50 cm maximum length FOV and
penetration depth 16 cm. (p. 38)
y MRI signals have a dynamic range of about 8 X 1015 at
1.5 T and 1015 at .5 T. (p 40)
y Echo time 10.8 ms. (p 68)

______________________________________________________
_____________________________________________________

FURTHER INFORMATION ON MRI FROM THE WEB

______________________________________________________
 32

http://www.cis.rit.edu/htbooks/mri/
The Basics of MRI, J. P. Hornak

This is a book of 14 chapters by J. P. Hornak of Rochester

Institute of Technology on the MRI. Quite comprehensive and
widely referenced.
____________________________________________________

Http://www.t2star.com/modules/MR_Image_Contrast.pdf

“MR Image Contrast”, William Faulkner, 1997

Excellent 12 page text on T1 and T2 relaxation and its effect on the

image, from an operator’s prespective. He shows how the operator
can varry TE and TR to make T1 or T2 weighted images.
Spin-echo pulsing sequence discussed.
____________________________________________________

Http://www.mri.jhu.edu/~sreeder/pdf_papers/mse_t~va.pdf

“Development of High Speed, High Resolution Tagging

Techniques”. By Scott Reeder, Advisor Elliot McVeigh

A graduate thesis out of Johns Hopkins University, Baltimore,

Maryland; a 198 page document in full text.

This thesis now requires permission to access.

____________________________________________________
Http://www4.ncsu.edu/eos/users/w/wes/homepage/MRI/mrilesso.h
tml
“Magnetic Resonance Imaging: A Short introduction for
clinicians” by Wesley E. Snyder.

This is an MRI clinician’s lecture in complete text with simple

diagrams and superficial equipment ideas. He explains T1 and T2
weighting as a user method; shows how TE and TR choices result
in T1 or T2 weighting.
____________________________________________________
Http://www.gehealthcare.com/gecommunity/mri/ortho/ortho_apps-
techniques.html
 33

MRI Techniques for Orthopedic Imaging, Cynthia F. Maier, Lewis

Lauzon, Ramish Venkatesan
GE application note for FSE TIMING SEQUENCE

A seven-chapter industrial application note directed to

orthodontics, relevant to General Electric MRI equipment. Various
chapters discuss the CHESS timing sequence, and features of
interest to the clinician.
____________________________________________________
http://mrel.usc.edu/class/articles/Riederer-MRI2000.pdf
“Current Technical Development of Magnetic Resonance
Imaging”
Stephen J. Riederer. IEEE Transactions Engineering in Medicine
in Biology. (September/October, 2000).

This is an IEEE review article in complete text about the status of

MRI in 2000 by Riederer of the Mayo Clinic, a prolific expert in
this field.

____________________________________________________
Http://www.uib.no/med/avd/miapr/arvid/bfy-361/howseman_bowt
ell1999.pdf

“Fuctional Magnetic Resonance Imaging:Imaging Techniques and

Contrast Mechanisms” Alistair M. Howseman & Richard W.
Bowtell. Philosophical Transactions of the Royal Society of
London, vol 354, pp 1179 - 94.

The paper is given in full text.

____________________________________________________
____________________________________________________

Exercises
Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).
 34

CHAPTER 2

X-RAY, BASIC THEORY

NOTE: This material is for educational purposes only. See the disclaimer above.

Various particles and waves emerge from the atom. Early evidence
of this was discovered by Thomas Edison when he put a plate of
metal in a light bulb and placed it at a positive voltage with respect
to the tungsten filament. It drew a current, called the Edison Effect.
It was not known then that these were electrons, but was
investigated as cathode rays. One of these investigators, The
German Physicist, Wilhelm Conrad Röntgen investigating the
florescence cause by these rays, on November 8, 1895, noticed a
screen covered with barium platinocyanide glowed when brought
near the cathode ray tube, but not in the path of the cathode rays.
Serendipitously, he changed the focus of his investigation and
studied these mysterious rays, he named x-rays. Most importantly,
and amazingly, he also noticed that when he put his hand in the
path of the rays, an outline of his bones appeared on the barium
plantinocynide screen. He quickly published this and other
properties: x-rays could pass through a thousand-page book and
three and one half millimeters of aluminum. It was soon
discovered these rays could be collimated by pin holes in lead
sheets and develop photographic film. This led to techniques so
that x-rays could be used to image bone without disturbing the
skin, and in fact were used to diagnose broken bones and find
foreign objects in the body such as swallowed metal and bullets.
Thus radiography was invented.
 35

The modern x-ray vacuum tube, illustrated in Figure 2.1,

was invented by W. D. Coolidge in 1913, making possible the
production of x-rays at higher energies, produced by anode
voltages in excess of 50,000 volts.

X-ray Production

X-rays are produced as a result of the bombardment of an atom by

an electron or other atomic sized particles or photons. They are
produced in nuclear reactors, and are emitted from stars. But
almost all medical x-rays are produced by vacuum tubes.
The voltage, VF , in Figure 2.1 drives a current into a
filament, which, in turn heats the cathode of a vacuum tube.
Electrons boil off the cathode, and are attracted by the voltage, VA.
They accelerate to a high velocity and penetrate the anode. The
interaction of the electron with the atoms of the anode, often
tungsten, transfers the kinetic energy of the electron to an x-ray
photon, which emits from the anode.
The energy of the x-ray photon cannot exceed the kinetic
energy that the electron acquires as it falls through voltage, VA.
The intensity of the x-ray is measured by the number of photons
per second. Since each photon is produced by an electron, the
intensity of the x-ray is controlled by adjusting the electron beam
current.

X-ray Tube Current

The cathode in Figure 2.1 is made of a

metal such as tungsten. The heat from the
filament agitates the electrons so they
acquire enough energy to boil off the metal
into the air. An electron leaving the surface
will produce a positive charge there which
pulls the electron back in. However the
voltage VA, if large enough, will pull it out
and accelerate it toward the anode, creating
what is called a thermonic current. The
higher the temperature of the metal
cathode, the greater the current. The
 36

formula for the beam currrent, IB, has been derived using quantum
mechanics as (2.1)
−11600 E W
I =C A T e
B 0 C
2 T (2.1)
Here T is the absolute temperature in degrees kelvin, C0 is the
cathode material coefficient, AC is the area of the cathode in square
meters, and EW is the work function in electron volts
for the metal in the cathode. This is the current that
would flow if VA is so high that every electron that
boils off the cathode is attracted to the anode. X-ray
tubes are designed so this is the case, thus Equation
(2.1) gives the current in an x-ray tube. Equation (2.1)
shows that the current increases with filament
temperature and the area of the cathode, and decreases
for larger work functions.
C0 varies for different cathode materials. The work
function, EW, is a measure of the electrostatic force
required to pull an electron from the surface of the
 37

metal. Seely gives a table of values for these constants for

commonly used cathode materials, as listed in Table 2.1. (2.2)
These values can be used to generate current versus
temperature curves for x-ray tubes having either tungsten,
thoriated-tungsten, or Oxide-coated cathodes. The calculation in
Figure 2.2 was done by programing Equation (2.1) on a graphics
calculator, showing the cathode current for a tungsten versus a
thoriated tungsten cathode, of 1 cm2 area and using the values in
Table 2.1. Clearly, the thoriated tungsten cathode produces a beam
current at a lower temperature.

X-ray Characteristics

The electron beam in the tube of Figure 2.1 impacts the anode.
Most of the electrons enter the anode metal and are stopped,
creating heat. About one percent, however, create x-rays when the
incoming electrons encounter the atoms. In Figure 2.3, we
illustrate two types of x-ray radiation, characteristic and
bremsstrahlung.
 38

To produce characteristic radiation, the incoming electron collides

with an orbital electron in the anode atom. The orbital atom gets
knocked to a higher orbit, having absorbed energy from the
incoming electron, as illustrated in the top of Figure 2.3. It remains
unstable, until it returns to its original orbit, releasing its energy in
x-ray characteristic radiation. Characteristic radiation emits at a
specific energy level, depending upon the atom type in the anode.
The energy level is determined by the energy gap between electron
orbits in the atom. Bremsstrahlung radiation is more like noise,
and appears in a band of energy levels. It is
caused when the incoming electron is
deflected by the atom and slowed down,
transferring its kinetic energy into x-ray
radiation.
The difference between
characteristic and bremsstralung radiation
is illustrated in Figure 2.4. Because
bremsstralung radiation is caused by
collisions of various types, the energy of
the x-ray photons released follows the
distribution indicated by the curve. The
spike on the curve indicates the
characteristic radiation, and its specific
value is determined by the energy gap
between electron orbits. It is caused by very
few photons, compared to bremsstrahlung radiation, which
contains the majority of the energy in the x-ray. Bremsstrahlung
radiation is most important in medical x-ray, and characteristic
radiation can be neglected - it being more important in
crystallography. Since the bremsstrahlung radiation picks up its
energy from the individual electrons in the x-ray tube beam, the
energy of each photon cannot exceed the that of those electrons.
The energy in an electron in the beam, EE, is determined by the
applied voltage, VA, in Figure 2.1 as:

EE = e VA (2.2)
-19
where e is the charge on an electron (e = 1.602 x 10 coulomb),
and VA is the voltage on the anode of the x-ray tube. The units in
Equation (2.2) are
joules
coulomb x coulomb = joules
 39

The units may also be referred to as electron volts, where one

electron volt is the energy acquired by an electron as it falls
through one volt (1.602 x 10-19 joules), as is clear from Equation
(2.2).
The energy of an x-ray photon, EP, is proportional to its
frequency, f, and was given by Max Planck as

EP = h f (2.3)
-34
where h is plancks constant (h = 6.626 x 10 joule-seconds).
____________________________________________________
EXAMPLE 2.1. In Figure 2.4, The maximum energy of an x-ray
photon is given as 135 keV. What is the maximum frequency of an
x-ray photon. f
SOLUTION
The maximum energy in joules is

E Max = 135 keV x 1.602 x 10 −19 coul

= 2.162 x 10 −14 joules
Thus Equation (2.3) can be used to find the maximum frequency

EP 2.162 x 10 −14
f= h = 6.626 x 10 −34 = 3.26 x 10 19 Hz
____________________________________________________

Notice here that the maximum energy of a photon is equal to the

anode voltage in units of electron volts.
Since an x-ray photon is electromagnetic radiation, its
wavelength, , is given by
c
= f (2.4)
8
where c is the velocity of light (3 x 10 m/s), and f is the frequency
in hertz. Thus in Example 2.1, the minimum wavelength of the
radiation is 3 x 108/3.261 x 1019 = 9.20 x 10-12 m.

Planar X-ray

Throughout the twentieth century, planar x-ray images were

created by passing x-rays through the patient as shown in Figure
 40

2.5. X-rays from the tube pass through

the patient and are absorbed at different
rates depending upon what tissue they
pass. Bone for example absorbs far more
x-ray that soft tissue, and so would
appear as a shadow on the film. In order
to be precise about how the image is
formed we need to analyze how x-ray is
absorbed by biological tissue.

Image Production

The planar image is produced because

tissue absorbs x-ray at different rates, just
as light is absorbed differentially by
different materials. Glass is transparent to
light, whereas, metal is opaque to light.
The differences are made precise by
Lambert’s Law. He stated that light is
absorbed in proportion to the thickness of the material. For x-ray,
Lambert’s law states that the fractional change in x-ray intensity as
it passes through a tissue is proportional to the thickness of the
material. In mathematical terms this is:

=− x
where is the x-ray intensity in watts per square meter, x is the
thickness of the tissue and the proportionality constant is - . A
detailed discussion of this proportionality constant will follow.
Taken to the differential limit, this equation becomes

d
dx =− (2.5 a)

The solution to this equation,

= 0 e− x (2.5 b)

is verified by substituting it into Equation (2.5a), and showing the

equality.

Mass Attenuation Coefficient

 41

As x-ray photons pass through a tissue, they are reduced in

intensity, by being absorbed according to Equation (2.5b). The
absorption increases with the density of the tissue, , because the
greater number of atoms per unit volume increases the chance of a
collision and subsequent absorption of an x-ray photon.
Approximate values of this density for common tissue types are
given in Table 2.2.

The absorption of an x-ray photon is also dependent upon its

energy level as expressed by the mass attenuation coefficient, .
Values of for several common biological tissues is given in
Table 2.3. (2.3) There most often decreases with increasing
photon energy. Above 200 keV, the values do not differ greatly

TABLE 2.2.
Density of common
biological substances

SUBSTANCE DENSITY (g/cm 3 )

Air 0.0013

Water 1.0

Muscle 1.06

Fat 0.91

Bone 1.85

among the tissues listed. At these energies the absorption is

determined mostly by the tissue density.
 42

TABLE 2.3. Mass Attenuation Coefficient Versus Photon Energy

2
X-ray Photon Mass Attenuation Coefficient , (cm /g)
Energy
Water Compact Muscle
(keV) Air
Bone
10 4.89 4.66 19.0 4.96
15 1.32 1.29 5.89 1.36
20 0.523 0.516 2.51 0.544
30 0.147 0.147 0.743 0.154
40 0.0647 0.640 0.305 0.0677
50 0.0394 0.0384 0.158 0.0409
60 0.0304 0.0292 0.0979 0.0312
80 0.0253 0.0236 0.0520 0.0255
100 0.0252 0.0231 0.0386 0.0252
150 0.0278 0.0251 0.0304 0.0276
200 0.0300 0.0268 0.0302 0.0297
300 0.0320 0.0288 0.0311 0.0317

_____________________________________________________
EXAMPLE 2.1. An x-ray intensity, produced by a tube having 40
kV on the anode, of 1 W/m2 is incident on two adjacent tissues,
muscle and bone. Plot the intensity of the x-ray in both as they
progress through the tissue in the x-direction.

SOLUTION
 43

The x-ray photons have a maximum energy of 40 keV, so we take

for bone as 0.305 and for muscle as 0.0677 in Table 2.3. From
Table 2.2 the densities are 1.85 and 1.06 respectively. A plot of
Equation (2.5), on your graphics calculator, with these values is
given in Figure 2.6:

_____________________________________ _______________
It is clear in Figure 2.6 that x-ray is absorbed more in bone than in
muscle. This is due to the large difference in their densities and
mass attenuation coefficients. The x-ray which emerges from the
x-ray on to the film is very different in intensity from bone an
muscle. This makes for good contrast in x-ray images of bone
imbedded in muscle. Blood clots and fluids would make images
with good contrast because of the wide difference between the
density of lung tissue and muscle or blood, since lung contains a
lot of air, having a low density as indicated in Table 2.2. Medical
x-rays usually use tube voltages ranging from 20 kV to 200 kV, in
part because the values of mass attenuation coefficients get low
and undifferentiated at higher photon energy levels, as shown in
Table 2.3.

X-Ray Hardware

The x-ray tube is the heart of any x-ray unit. It is where the x-rays
are generated, and is an expensive part that wears out and must be
maintained and replaced periodically. Difficulties associated with
it are due to the fact it uses high voltages and must dissipate a lot
of excess heat. The heat arises from the inefficiency of the process
by which x-rays are generated.
 44

The efficiency of an x-ray tube, , is equal to the ratio of

the power in the x-ray beam to that in the beam of electrons. The
power of the x-ray beam, PX, increases as the square of voltage on
the anode, VA ,because the higher voltage means the electrons are
going faster and penetrate deeper into the anode, increasing their
chance of colliding with an atom. It also increases with the beam
current, IB, since this means more electrons are available to collide.
A large the atomic number, Z, also increases the probability of
collision (2.4). The net result is:

P X =k I B Z V A2
where k is an empirical constant, measured as k = 1.4 x 10 -9.
The power in the electron beam, PB, is the product of the voltage
and current as:

PB = I B V A
The x-ray tube efficiency is then the ratio of these two quantities

= k Z VA (2.6)

For the case of a tungsten cathode, (Z = 74) in a tube operated at

VA = 100 kV, this efficiency calculates to 1.04%.
In general the efficiency of x-ray production is about one
percent, and as a result, most of the energy of the electron beam is
converted to heat which must be dissipated by the vacuum tube.
Several techniques for doing this are illustrated in Figure 2.7.

The X-Ray Tube

 45

Two mechanisms for dissipating heat are the rotating anode and
circulating oil around the tube. The anode rotates between 3600
and 10000 rpm, driven by a motor with field coils exterior to the
tube, and an armature interior to the tube. This spreads the
impacting electrons over a large area of the anode. The oil is
circulated with a pump and cooled with a fan in large tubes. The
high tube voltages between 20 and 200 kV are insulated with high
voltage terminals. Volume compensation is used to adjust the size
as the oil expands when heated. The metal casing around the tube
blocks any x-ray from spraying around the room. The x-ray is
directed through a glass window toward the patient being imaged.
Fabrication of an x-ray tube is very labor intensive making it cost
more than one would expect, into the thousands of dollars. The life
of the tube can be extended if the operator is careful not to over
heat it.
 46

To prevent thermal damage to tube parts such as the

cathode and anode, the exposure time for an image often must be
limited. A hypothetical characteristic curve for an x-ray tube,
Figure 2.8, shows that longer exposure times require lower beam
current, IB, for any particular anode voltage, VA. That is, the
higher the electron beam power, the hotter will becomes the tube.
To prevent overheating, the time the tube is turned on must be

limited. For the tube illustrated, if the current is 0.2 A and the
anode voltage is 100 kV, the maximum exposure time is
approximately 5 seconds, according to Figure 2.8.
 47

Beam Shaping Parts

The essential elements of a planar

x-ray unit are shown in Figure 2.9.
The power supply must provide up to
200 kV to the x-ray tube. X-rays
which emerge from the tube are first
passed through an aluminum sheet
which filters them, reducing the
number of low energy photons which
enter the patient. These photons add
to the dose of ionizing radiation, but
do not help the image much because
they are strongly absorbed by the
tissue. The collimator places vanes in
the way of the x-ray to absorb energy
that would miss the target being
imaged and only contribute to more
dose for the patient. The bucky grid
absorbs the x-ray that has been
scattered in the patient, and prevents
it from causing unsharpness in the
image. It is placed against a
luminescent screen which converts
the x-ray to light which can develop
the x-ray film to produce the image.
The collimator, illustrated in Figure
2.10, casts the x-ray beam on the
desired area of the patient’s anatomy.

The beam consists of both x-ray and

visible light from illuminator tubes,
so the operator can see where the beam is striking the patient. The
adjustable baffles block unwanted radiation, thus reducing the dose
of ionizing radiation to the patient.
 48

X-ray Detection

As the x-ray has passes through the patient it

becomes selectively attenuated, depending upon the
tissue. Thus the x-ray intensity emerging from the
patient varies across the plane perpendicular to the
rays, giving the information for the image. The
quality of the image is degraded by scattering of the
x-ray in the patient, interference from the Bucky
grid, and blurring in the detector. The process of
detection is illustrated in Figure 2.11.. The scattered
x-rays are caused when the direct x-ray photon is
deflected by an atom in the tissue. When it is
scattered it is knocked off its path so that it would
not hit the film in a place collinear with the direct
ray, and would cause an error in the image. The
purpose of the Bucky grid is to absorb this x-ray in
its heavy metal, lead or tungsten, vanes. The vanes of the grid then
block some direct x-rays as well, because they have a finite width.
This can be compensated for, to a degree, by mechanically
shifting the grid during the exposure.
After passing through the grid, the x-rays enter a heavy
metal screen containing a luminescent material such as phosphor.
These heavy metals have high valence, Z, atoms which capture
more x-rays, so the screens are called Hi Z. The x-ray is then
converted to light. The light then hits the film where it is detected.
The film is sensitive to both x-ray and light, but most x-ray would
pass through the film. The purpose of the
illuminating screen is to capture more
x-rays, both before they hit the film, and
after they pass through the film. This
increases the capture efficiency of the film
detector. The conversation of x-ray to light
causes some degradation in image
resolution due to the blurring illustrated in
Figure 2.11. Other solid state detectors are
available to detect x-rays and other
ionizing radiation. These are discussed in
the nuclear medical instrumentation
chapter, Table 5.1.
 49

The Power Supply

The power supply, illustrated in Figure 2.12, must supply high

voltage in range
from 20 to 200
kV in pulses of
varying
durations from
a few
milliseconds to
seconds, and in
some cases
continuously.

Adjustment of
the x-ray tube
beam current is
done with a low
voltage power
supply
consisting of
varactor VAR1
and
transformer, T1.
Changes in
VAR1 may cause the tube filament temperature to change, thus
causing a change in the electron beam current according to
Equation (2.1). A separate winding of T1 may also be used to
energize the rotation of the tube anode. A timing circuit is used to
turn the x-ray tube anode voltage on and off. It is located on the
primary side of T2 to avoid switching at a high voltage level, which
would draw arcs and damage switch contacts. Likewise the anode
voltage is varied with VAR2 on the primary side of the
high-voltage transformer T2. To achieve full-wave rectification, the
positive phase of the voltage on the secondary winding of T2
causes diodes D2 and D4 to conduct pushing a positive current from
the anode to the cathode of the tube. On the negative phase, diodes
D3 and D1 conduct. Thus the AC voltage of the source, VAC, is
converted to a DC beam current in the x-ray tube.
In summary, in this diagram, the x-ray pulse duration is
controlled by the timing circuit and the beam current is controlled
 50

with VAR1. These quantities are often referred to as the Mas in

units of milliamp-seconds. The x-ray photon energy level is
controlled by VAR2 and is referred to as the kV.
People working with the power supply, such as biomedical
equipment technologists and clinical engineers need to take high
voltage precautions. The voltage breakdown in air at sea level is 75
kV/in. Thus this voltage can jump inches in air and electrocute
those who get near the high voltage electrodes in the power supply.
Thermal damage to the x-ray tube is prevented by thermal sensors
which shut the power supply down when the tube gets overheated.

References

2.1. Seely, Samuel. Electron-Tube Circuits. New York:

McGraw-Hill Book Company, (1958) p 4.
2.2. ibid.
2.3. U. S. Government Printing Office. Physical Aspects of
Irradiation (NBS Handbook No. 85). Washington DC: (March
1964), p 3.
2.4. Feinberg, Barry N. Applied Clinical Engineering. Englewood
Cliffs, NJ: Prentice Hall, Inc. (1986) p 245.
2.5 Aston, Richard. Principles of Biomedical Instrumentaton and
Measurement. Columbus, Ohio: Merrill Publishing Company.
(1990).

Exercises

Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).
 51

CHAPTER 3
CT SCANNER THEORY
For educational purposes only. See disclaimer above.

Conventional x-ray records the shadow of anatomical structures on

a film. In the 1970’s a breakthrough in imaging was achieved with
the invention of the Computer Tomography (CT) Scanner, one of a
series of devices capable of imaging a transverse slice through the
body. These slices can be combined to make a three dimensional
picture of the anatomy. The word “tomograph” means “a picture of
a slice”.
Early forms of tomography were used to make an image of
objects hidden behind others. One method of doing this is
illustrated in Figure 3.1. The x-ray tube is rotated around one
side of the circle and the film follows at the opposite end of a
diameter line. A series of pictures is taken on the same film. Each
picture will have the object in the center, but the objects off-center
will be blurred or averaged away. In some cases this procedure
could yield useful diagnostic data, but compared with modern CT
scanners it is extremely
crude.

The CT Scanner
 52

The CT Scanner, in one form, consists of an x-ray tube that

rotates around the patient in a cylindrical opening supported by a
gantry, as shown in the top of Figure 3.2. A pencil beam of x-ray,
collimated into a single line, passes through the patient to a
detector diametrically opposed as illustrated on the lower part of
the figure. The detector measures the attenuation of the x-ray beam
as it passes through the patient. This data is recorded in a
computer. The tube is then rotated a degree or so and the
attenuation along another line is recorded.
The data is then used by the computer to
calculate the x-ray attenuation of each
square millimeter of a transverse slice
through the patients body, as shown on the
lower part of the figure. The attenuation
numbers are converted to shades of gray
and presented on a computer monitor
screen as a picture of a slice of the
patient’s anatomy.
The process by which the physical
data is collected is easy to understand by
anyone familiar with conventional x-ray.
But the calculation procedure involves
advanced mathematics, and would not be
practical without modern dedicated
computers.

CT Image Calculation

First the intensity of the x-ray of a pencil

beam after passing through a slice is

measured as, (r), a

 53

function of r in Figure 3.3. The x-ray

intensity emerging from the tube is a
known quantity, 0. Thus we can
form a function
(r)
p (r) = − ln 0
(3.1)
which is called the attenuation
profile, p (r), as a function or r.
To collect this data, one
orients the tube, collimators, and
detectors as shown in the figure at a
specific angle 1. The tube is then
moved to the right along the
collimator, and the detector measures
for each of the pencil beams.
These data can be used to form the
function p 1(r). Now the whole
apparatus is rotated to another angle
2 and another set of measurements is
taken. This process is continued so
that the number of measurements
taken would typically be 256 x 256 =
65,536.
As an example calculation of (r),
consider the medium inside the
circle, Figure 3.3, to be uniform of
0- (r)
attenuation coefficient = 1 m-1, and
0 the radius of the circle 1 m. Then in
Exercise 3.1 we compute the graph in
Figure 3.4. Part (a) of the figure
(a)
0 shows the x-ray intensity as a
r r
max function of r on the detectors. And
p 0- ( r )
part (b) shows p (r) for this case. For
each value of as we rotate the
tube-detector pair we have a p (r)
(b)
which we can take the polar form of
0 the one-dimensional Fourier
r r
max transform (1DFT) of, according to the
formula
FIGURE 3.4. ATTENUATION PROFILE
 54

∞
P (k ) = ¶ p (r)e −2 i( k $r )
dr (3.2)
−∞

Here i is the imaginary number −1 and 2 k has the

dimensions radians per meter. k is a vector. This formula
transforms the spatial quantity p into k-space. This
operation can be repeated for each value as the
tube-detector is rotated through 0 to radians to yield a
two-dimensional function in k-space, P(k, ). k-space is
similar to the frequency domain of the fourier series. The
values of k are collected in polar coordinates, and can be
plotted is k-space, illustrated in
Figure 3.5. The two dimensional Fourier Transform in polar
coordinates is: (3.1)
∞
p (x , y ) = ¶ ¶ P ( k , )e 2 i( x k c o s + y k s in )
x k x dk d (3.3)
0 −∞
To create the set of real numbers p(x,y), the computer first
integrates on k to create a function of x, y, and . It then integrates
on to create the function p(x,y) which can be mapped as a
gray-scale image in rectangular coordinates. In practice we must
limit the range on k to a finite number, since the computer cannot
go on calculating for ever. A reasonable approximation is achieved
if Equation (3.3) is written so that the integration on k goes from -
k max to + kmax as (3.2)

k m ax
p (x , y ) = ¶ ¶ P (k , )e 2 i (x k c o s + y k s in )
x k x dk d
0 − k m ax

(3.4)
where

kmax = 21 r (3.5)
where =r is the distance between pencil beams in Figure 3.3,
which measures the spatial resolution of the image. The larger the
number of k frequencies integrated, the better the resolution. To
achieve a 1 mm resolution, kmax would equal 500 per meter, for
example. The image is an approximation to the ideal because a
 55

discrete, finite number of values of k and are chosen.

Mathematicians over the past quarter century have developed
many methods of improving the approximation and the speed
with which the calculation can be made by making
variations on the basic process of taking the Fourier
transform and its inverse as expressed in Equations (3.2) to
(3.5). (3.3). The process modeled by these equations is called
the filtered back-projection method. The magnitude k in
Equation (3.4) is called a ramp filter in k-space.

Improving Scan Speed

The first generation of CT scanners used a single detector

which followed the tube in a geometric arrangement shown
in Figure 3.3.
It took about 5 minutes to create an image of a slice.
Byconverting the beam into a fan shape, rather than parallel beams
as depicted, that time was reduced to twenty seconds per slice.
A third generation of scanners was developed which
eliminated the need to translate the tube, then rotate. In this case
the tube would rotate in coincidence with a bank of detectors,
around the patient in the center, illuminated by a fan beam
as shown in Figure 3.6. The tube and detectors are
supported by a gantry. The scan time is reduced to 0.5
seconds a slice with this arrangement, with a consequent
reduced x-ray dose to the patient, and it is widely used
today.
The formula for calculating the image from the fan
beam is similar to Equation (3.4). See Appendix E, Fan
Beam Projection. The derivation of an image formula in
terms of practical machine measured parameters is also
given there.
Another arrangement, called the fourth
generation is illustrated in Figure 3.7. It features a large
number of permanent detectors on the outer ring. The use
of up to 5000 detectors increases the cost of the
equipment, but the mechanical stress on these detectors is
reduced. The scan time remains about the same as in the
third generation.
 56

X-ray Detectors

The bank of detectors is placed between one half and

one meter from the center of the gantry opening. Slow
acting film is replaced by instantaneous detectors in the
CT scanner, most often of gas ionization chamber or
solid state scintillation detector design.
An ionization chamber design is
shown in Figure 3.8. The ionization
chambers (3.4) are formed between
tungsten electrodes. Purified xenon (Xe)
gas usually pressurized between 10 and
30 atmospheres absorbs about 60% of the
x-rays that enter the chamber. The
absorbed x-rays create positive ions and electrons
which are attracted to oppositely charged plates to
create an output signal current proportional to the
incident x-ray. The excitation voltage is chosen so
that the output current is proportional to the x-ray
incident. The tungsten in the electrodes tends to
absorb scattered x-rays, acting thereby as
collimators. The distance, d, between the signal
output electrodes is about 1.2 mm, insuring an
image resolution at that level. The response time of
Xe chambers is about 700 s. The excellent stability
of these detectors helps keep the calibration set.
Solid state detectors, such as illustrated in
Figure 3.9 can offer better detection efficiency. The
scintillating crystal cadmium tungstate (CdWO4)
absorbs almost all of the x-ray incident. Modern
solid state detectors reach 98 percent conversion
efficiency. Both solid state and gas detectors allow
scan times better than 2 seconds. The number of
detectors used in a CT scanner ranges from 500 to
5000.
 57

Crystal Diode

X-ray
Solid state detectors generally achieve higher quantum efficiency,
Scintillating Crystal
Photodiode and absorb more of the x-ray than xenon, thereby reducing dose
and allowing for smaller detectors. The fading characteristics of
d

log
I
0
Output Signal the detectors determine how quickly the signal reduces so that it
Io
-1 can be used again in the next scan. A comparison of CdWO4,
-2 Xenon gas, and a compound of Gadolinium Oxygen and Sulfur,
-3
FADING
CHARACTERISTICS
GdOS, ceramic (developed as Lightning UFC by Siemens
-4 CdWO4
Corporation) appears in Figure 3.9. (3.5) It is important that the solid
Xenon
state detector Lightning UFC has a faster fading time, as it also has
-5
Lightning UF
higher quantum efficiency and smaller size. This leads to faster
0 20 40 60 80 100
ms Time
and better resolution lower dose CT designs.
Figure 3.9. Solid State Detector

Gantry Hardware

The x-ray tube is mounted inside a gantry,

as shown in Figure 3.10. Control voltages
and signals are connected through slip
rings. The voltage is increased to
x-ray levels, near 100 kV, and applied to
the tube. X-rays are aimed in the
collimator to produce a fan beam, which is
detected by detectors mounted on a
circular strip around the patient in this
case. The cathode module controls the
filament current, and the anode module
keeps the anode spinning properly. All of
this hardware in s rotated with the tube,
and so presents a lot of inertia, which
limits the speed of scanning. The
patient lies on the table and can be
positioned in the hole in the gantry as
shown in the bottom of the figure.
 58

Fifth Generation CT Scanner

The fifth generation CT scanner, shown in Figure 3.11,

creates a focused electron beam
which is passed through deflection
coils that direct it toward a tungsten
target ring as shown.(3.6) The
resulting x-rays are directed out of
the tube, through the patient onto a
detector ring. The electron beam
can be scanned around a circle, at
electronic speeds, much faster than
a massive conventional x-ray tube.
In contrast to other approaches, this
has no moving parts other than the
electron beam. Commercial designs
are available which will produce an
image in 50 ms, making possible 17
frames per second. Thus, not only
breathing artifacts, but the effect of
the beating heart blurring the image
can be eliminated. One version of
electron beam tomography (EBT)
has a resolution better than .5 mm,
with a slice thickness of 1.5 to 10
mm.

Image Properties

Attempts at theoretical calculation

of x-ray contrast based on basic principles, as we did previously,
(3.7)
are limited by the fact that the attenuation constant, , is a
function of the energy of the x-ray photon, and the beam contains a
spread of energies. That calculation requires one to make a rather
arbitrary choice of what energy level to take the attenuation
constant. Thus it is common practice to resort to empirical
methods to obtain a measure of the contrast between, let’s say,
kidney and the viscera surrounding it.
A measure of contrast is given by the CT number, CTn, in
the formula (3.8)
 59

−
CT n = k ( )
w w
w w (3.6)
where w w is the attenuation constant of water and k is a
proportionality constant. If k is 100 for example, CTn would
represent the percentage difference between the attenuation
constant of the tissue in question, , as compared to that of water.
In practice, k is assigned the value 1000, and the CTn is called a
Hounsfield unit, HU, in honor of one of the inventors of CT.
− w w
HU = 1000 ( w w ) (3.7)
The least attenuating substance in the body is air, so we arbitrarily
assign the number 0 to its value. Thus the HU = -1000 for
air. Also, the HU number for water, where = w w, is 0. At
medical energies, around 70 keV, the for bone is approximately
2 times that of water. Thus the HU is taken as HU j 1000. All
other normal tissues in the body fall between these values.
However if contrast substances are injected, HU may exceed 1000.
Typical values of Hounsfield number for biological tissues
are given in Table 3.1.(3.9)

Table 3.1. Typical HU for Biological Tissues

Tissue Hounsfield
number
(HU)
Dense bone i1000
Red Blood Cell 93
Spleen 46
Muscle 44
Blood 42
Pancreas 40
Gray matter 38
White matter 30
Heart 24
Cerebral spinal fluid 0-20
Water 0
Fat -20 to -100
Bowel Gas (air) -1,000
 60

Notice how close the HU numbers are for spleen, muscle, blood,
pancreas are, portending low contrast for these tissues in a CT scan
unless absorbing substances are added as contrast agents.

CT Scanner Calibration

A phantom, such as is illustrated in Figure 3.12, is used to test the

ability of the CT scanner to distinguish objects of different size,
and having different attenuation coefficient, , values, expressed as
Hounsfield numbers. A set of rods of uniform size are made of
different plastics: Teflon (HU = 882), Bakelite (262), Plexiglas
(126), Nylon (89), Polystyrene (-28), Octane (-303). A CT image
of these rods will appear as different shades of gray, depending on
the HU number. Thus the rods can be used to calibrate the HU
number that corresponds to a particular shade of gray. They enable
one to set the range of HU values that are covered by the shades of
gray from white to black. If, for example, internal organs
are being imaged, the shades of gray would be set to range
from white to black as HU varied from say -100 to 100.
If, on the other hand bone was being imaged, white to
black might cover HU numbers from 500 to 1500. In
this way the contrast of the image can be adjusted.
The objects of diminishing size can be used to measure
the spatial resolution of the of the image. The phantom is a
known object that can be compared with the x-ray image,
so that any distortions can be clearly identified.

Resolution

The resolution of the image is the smallest distance

at which two objects can be distinguished. If two objects
are closer than the resolution distance, they will appear as
one object on the image. The resolution is limited, in a CT
scanner, by the size of the holes in the collimator. the width of the
detector.
On the imaging surface, the resolution is limited by the number
of pixels per meter in the image. A typical slice size may be 25
 61

centimeters in diameter. The display matrix may be 256 X 256. The

horizontal or vertical resolution in this case, RCT, cannot be better
than

RCT = 250256mm = 0.977 mm

To measure the resolution in x-ray, the concept of
modulation transfer function (MTF) pertains.
The MTF grid, shown in Figure 3.13, consists of lead
having holes of diminishing size as the fan beam rotates in
the counterclockwise direction. When the intensity of x-ray
that hits the detectors begins to decrease because of the
decreased spacing, the limit on the resolution has been
reached. In the case illustrated, the percentage of transfer
begins to diminish when the vanes and gaps reach 1 mm. If
the intensity has diminished significantly
at 2 cycles/mm, the resolution is
1
2/mm = .5mm

Patient Dose

The patient dose is the amount of x-ray

energy absorbed in the tissue per unit mass,
measured in grays (1 J/kg) or rads (.01 J/kg).
Each photon of x-ray gives up an amount of
energy as it passes through the tissue. Thus an
increase in the number of photons, or x-ray
intensity, increases the dose in a given mass. The
patient dose does not increase when one
increases the number of slices in a study, since
increasing the number of photons absorbed is
balanced by an increase in the mass. However if
x-ray scattering occurs, as it inevitably will, the
dose of the individual slices increases. The dose is increased if the
time to scan each slice is increased, or if the intensity of the x-rays
(mAs) is increased.
 62

The dose from a particular CT scanner depends on the

manufacturer. The scanners typically run at 120 kV and from 80 to
300 mA. Typical dose levels range from 1 to 8 rads.(3.10) The key
to reducing dose is to manufacture more sensitive x-ray detectors,
and generate computer algorithms that require shorter duration
x-ray exposures per scan. The demand for better resolution tends to
increase dose because thinner slices mean less mass per slice, and
longer tissue exposure times.

TUBE

FAN BEAM

FOV

DETECTOR ARRAY

FIGURE 3.14 SPIRAL SCANNER

 63

The Spiral Scanner

The Spiral Scanner was made possible by the introduction of slip

rings into the gantry so that the tube and detector assembly, as
oriented in the Generation 3 scanner (Figure 3.14) can rotate
continuously. Both the electrical power and the signal are
transmitted through the slip rings. The patient table is moved
slowly through the gantry hole so that the x-ray beam outlines a
spiral (helix) on the body being imaged. Each 360 degrees of
rotation constitutes a slice. Because of the rapid response time of
the detectors depicted in Figure 3.9, it is possible to go through one
slice in less than 0.5 seconds. The detectors can be made small
enough to allow 0.35 mm spatial resolution, and a slice resolution
of 0.4 mm. The detector array in Figure 3.14 may have up to 64
rows. In one study, as many as 800 slices are produced. A 64 slice
CT Scanner can scan the entire heart in 10 seconds. If the x-ray is
gated with an ECG waveform, an image of the heart in one phase
of its beat can be made. For example, one could trigger off the R
wave of the ECG on each beat to show the heart in contraction
(systole). Introducing a delay into the trigger could produce an
image of the heart in relaxation (diastole).
The Spiral Scanner may be used in angiography to view the
vessels of the heart. Temporal resolution of 185 milliseconds has
been reported. This compares to 20 milliseconds for conventional
angiography techniques (3.11). Although the CT resolution is not as
good in this instance, the procedure is much quicker with the CT
scanner, measured in the seconds, and it is noninvasive.
The large number of slices produced by the scanner poses
the problem of analyzing all of that data. It may produce
information overload for a human analyst. Machine data
processing may be developed. Combination of many slices into a
three-dimensional image also provides diagnostic data.
 64

References

3.1. Suetens, Paul. Fundamentals of Medical Imaging. New York:

Cambridge University Press, (2002) p 22.
3.2. ibid. p 78.
3.3 op.cit.
3.4. Alker, George, S. Rudin, D. Bednarek. “Computer
Tomography” in John Webster, Editor, Encyclopedia of Medical
Devices and Instrumentation, Volume 2. New York: John Wiley &
Sons, (1988). pp 826-848.
3.5 Günter Hahn, Rolf Hupke, Gerhard Kohl, Jürgen Leppert
“Developing an Ultrafast Radiation Detector for CT Scanning”
Siemens AG 2001, D-80312 München, (2001) accessed on line
May 2005 at <http://w4.siemens.de/FuI/en/archiv/
zeitschrift/heft1_97/artikel03/>.
3.6. Cunnigham, Ian A. & Philip F. Judy. “Computed Tomograpy,
Chapter 64” in Joseph D. Bronzino, Editor, The Biomedical
Engineering Handbook. Boca Raton, FL: CRC Press (1995) pp.
990-1001.
3.7. Aston, Richard. Principles of Biomedical Instrumentation and
Measurement. Columbus OH: Merrill Publishing Company (1990)
p456.
3.8. Alker, op.cit. p 833.
3.9. ibid. p834.
3.10. Cunnigham, op cit. p. 999.
3.11. Becker, “Tailoring contrast administration for cardiac CT”
University of Munich, Grosshadern, Germany. Accessed on line
May 2005 at<http://www.ctisus.org/cta_web/
12_04/12_04_Becker/12_04_CTA_Becker.html>.

______________________________________________________
More Information on CT Scanners from the WEB
______________________________________________________
http://ric.uthscsa.edu/personalpages/lancaste/DI2_Projects_2003/H
elicalCT_artifacts.pdf

“Artifacts in helical CT images”, Yu-Huei (Jessica) Huang.

Good article about the Helical CT scanner, with practical
observations and good references. A full text article.
______________________________________________________
 65

http://www.the-infoshop.com/study/go9670_tomography_scanners
_toc.html
A major marketing study of CT scanners available at a cost
of over 3000 dollars.
______________________________________________________
http://www.ispub.com/ostia/index.php?xmlFilePath=journals/ijmt/
vol1n1/pet.xml
Dario B. Crosetto: “A Modular VME Or IBM PC Based Data
Acquisition System For Multi-Modality PET/CT Scanners Of
Different Sizes And Detector Types.” The Internet Journal of
Medical Technology. 2003. Volume 1 Number 1.
This full text article presents a new design for a combination CT
and PET scanner, using a fifth generation CT.
________________________________________________________________
http://www.ctisus.org/cta_web/12_04/12_04_Becker/12_04_CTA_
Becker.html
Becker. “Tailoring contrast administration for cardiac CT”
This full text article presents clinical applications of the 64
slice CT Scanner.
________________________________________________________________
http://www.medical.toshiba.com/news/pressreleases/072704-484.h
tm
Advertising information from Toshiba about the 64 slice
CT.
Another such site form Siemens is:
Http://www.businesswire.com/webbox/bw.092804/242725134.htm
______________________________________________________

http://w4.siemens.de/FuI/en/archiv/zeitschrift/heft1_97/artikel03/
“Developing an Ultrafast Radiation Detector for CT Scanning”
Günter Hahn, Rolf Hupke, Gerhard Kohl, Jürgen Leppert.

An excellent report on developing new detector material

for the CT scanner in full text. This is an extensive article on CT
scanner detectors from Siemens is important to read. It reviews the
state of the art of CT scanner detectors as of 1997 by Siemens
researchers: Günter Hahn, Rolf Hupke, Gerhard Kohl, Jürgen
Leppert . They conclude lightening UFC is a promising material
for solid state detectors.
 66

______________________________________________________
http://www.devicelink.com/mddi/archive/99/05/004.html
A state-of-the art review of CT Scanners, 1999 .
_______________________________________________________________
http://www.isy.liu.se/cvl/ScOut/TechRep/Papers/LiTHISYR2682.
pdf
“The MATLAB/C program take - a program for simulation of
X-ray projections from 3D volume data. Demonstration of
beam-hardening artefacts in subsequent CT reconstruction”. Olle
Seger, Maria Magnusson Seger. Department of Electrical
Engineering
Linköping University, SE-581 83 Linköping, Sweden
Email: [email protected], [email protected]
March 15, 2005
A demonstration of how to use the mathematics program
MATLAB to construct CT images.
______________________________________________________
http://www.sv.vt.edu/xray_ct/parallel/Parallel_CT.html
Raman Rao, Ronald D. Kriz, et al “Parallel Implementation of the
Filtered Back Projection Algorithm for Tomographic Imaging”,
Department of Electrical Engineering, Virginia Polytechnic
Institute and State University, Blacksburg, Virginia 24061
In this project, they describe a parallel implementation of
the filtered back projection algorithm for finding 2D cross
sectional images. The results show that parallel processing is very
viable for CT reconstruction and results in significant run time
savings. They describe an implementation on two different
machines: the Intel Paragon and the Connection Machine.
______________________________________________________

Exercises
Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).
 67

CHAPTER 4
NUCLEAR MEDICAL EQUIPMENT: Part 1,
Planar Imaging
For educational purposes only. See disclaimer above.

Nuclear Physics found application in Nuclear Medicine before its

applications in electrical power production and in the military. Its
use is based on radio nuclides.
A nuclide is a species of atom characterized by the
contents of nucleus: its energy, number of protons and neutrons in
the classical model. If a nucleus is bombarded by a high energy
particle, it can be transmuted to a new element or an isotope of that
element. In in this case a new nuclide is formed. When the nuclide
is radioactive and emits particles, and is called a radio nuclide.
The concept of the
isotope is illustrated
Hydrogen with a discussion of
1
H 1 hydrogen.

Hydrogen and its

Electron Isotopes
Protron Deuterium
Neutron 2
1
H Hydrogen is found in
Nucleus nature in three forms,
hydrogen, deuterium
and tritium, illustrated
in Figure 4.1.
Tritium
3
H
The elements are
1
designated by
symbols in the form:
FIG URE 4.1. Hydrogen and Its Isotopes
 68

A
ZX
Where A is the mass number, Z is the atomic number, and X
represents the element symbol. A is the number of protons and
neutrons in the atom. Z is the number of protons in the nucleus.
Thus the number of neutrons, N, is given by,

N = A-Z (4.1)
The atom is orbited by electrons. In the classical model, the
nucleus consists of protons and neutrons, illustrated in Figure 4.1.
The mass of a proton is 1836 times that of an electron and the mass
of a neutron is nearly the same, 1840 times that of an electron. The
mass number A essentially allows you to calculate the mass of the
whole atom, since the mass of the electrons is small enough to be
neglected. The mass of the atom is approximately :
(1840 x N + 1836 x Z) x Mass of an electron. (4.2)
The mass of an electron is 0.000549 Atomic mass units (amu),
(9.1096 x 10-31 Kg, at rest).
Most hydrogen has the symbol shown in Figure 4.1. About
8:1000 parts of all hydrogen in nature appears as Deuterium, with
the symbol shown in the figure. Since it has the same number of
protons and orbiting electrons as hydrogen, it is an isotope of
hydrogen. It therefore has most of the same chemical properties of
hydrogen. Tritium is also an isotope of hydrogen. It does not
occur in nature but can be artificially produced in a cyclotron, as
discussed below.

The Cyclotron

There are several ways of producing nuclides artificially: as a

byproduct of nuclear reactions in a reactor such as in a nuclear
electrical power plant, in a linear accelerator, Van de Graaff
generator, synchrotron, or in a cyclotron. (4.1) The cyclotron will be
discussed here, since these are often used in the hospital
environment.
The cyclotron in Figure 4.2 consists of two hollow vacuum
D-chambers and a source, S, of protons generated from hydrogen
inserted in a chamber.
 69

The voltage VF heats a tungsten filament which emits electrons,

accelerated by voltage VA, and which in turn collide with the
hydrogen atoms, stripping their electrons. This results in a
positively charged
RF
Generator proton which is in
Deflection turn accelerated
Plate
B-Field by the RF
direction
out of the
generator, and
page launched into the
D chamber. There
s a magnetic field
B, perpendicular
to the path of the
proton, coming
D-Cham ber out of the page in
A v Figure 4.2, exerts
a force on it
v Proton
according to
F
Ampere’s law:
1
1
H

SOURCE, S
T

FIG URE 4.2. A Cyclotron

f = qvB (4.3)
where q is the charge in coulombs, v is the velocity in meters per
second and B is the magnetic field in Tesla (Webers/m2). The
force, f, is perpendicular to the direction of flow, and causes the
particle to travel in a circle. (To find the direction of f by the
right-hand rule, place your right-hand fingers in the direction of
the velocity of the positive charge. Roll your fingers in the
direction of the B-field, and your extended thumb will point in the
direction of the force on the charge.) The mass, m, of the proton
thus experiences a centripetal force (4.2) given by
m v2
f= r
where r is the radius of the circle. These two forces must be equal,
so:
 70

m v2
qvB = r
Solving for the radius r yields,
r= mv
qB (4.4)
The RF generator is of such a frequency that when the particle has
traversed a half-circle, it changes polarity so that as the particle
enters the gap between the two D-chambers it is accelerated in the
gap. In the second D-chamber, it follows a larger circle because it
is going faster. That causes it to follow a spiral path as shown in
Figure 4.2. The angular velocity, , of the particle, from Equation
(4.4) is,
qB
= v
r = m
which gives a frequency f = /(2 ) as
qB
f= 2 m (4.5)
where f is the frequency the RF generator needs to maintain in
order to keep in step with the spiraling proton in the cyclotron. The
spiral path keeps getting larger until the -particle exits the D
chamber on to a trajectory toward the target, T, as shown in Figure
4. 2. The trajectory is adjusted by a voltage on a deflection plate
along the path. The collision of the proton with the target creates a
radio nuclide.

Radioactivity

Nuclear bombardment, as in a cyclotron, can produce radioactive

substances which radiate particles spontaneously. Marie Curie
studied radium at the turn of the twentieth century and isolated
three different types of particles emitting from the atom, arbitrarily
called: , alpha, , beta and , gamma. Since then many other
particles have been discovered such as neutrinos, positrons, and
several quarks.
The -particle is a helium nucleus consisting of two protons
and two neutrons. Its orbiting electrons are stripped off, so it has a
net positive charge of +2. Because it is large and heavy it does not
penetrate matter very deeply, traveling only 4 centimeters in air,
and it cannot penetrate human skin. It therefore has limited use in
nuclear medicine.
 71

The -particle is either an electron, called -, or a positron,

+
, the magnitude of the charge is that of an electron. (The
-particle is assumed to be an electron unless specified otherwise.)
The -particle penetrates 0.42 cm in tissue at 1 Mev energy,
slightly better than an -particle.
The -ray is electromagnetic radiation, a photon. It has no
mass and no electrical charge. Depending upon its energy it
penetrates through most biological tissue with varying degrees of
attenuation. Because of its strong penetrating power it is very
useful in nuclear medicine.

Radio nuclides

Several example radio nuclides, used in nuclear medicine, radio

pharmaceuticals, shown in Table 1 illustrate some important
properties. The half life (4.3) indicates the time it takes for the
radioactivity to decrease by half, and indicates how long the
substance in the body will remain radioactive .
A longer half life will likely mean a larger dose of radiation
for the patient. On the other hand, long lasting radio nuclides are
easier to ship long distances. In the table, Chromium 51 has a
half-life of 27.8 days, while Technetium 99m emits half of its
radiation in 6 hours. Gamma radiation is the most common type
among these radionuclides. The higher the energy level of the
radiation the more its penetration power.
The nuclear medicine may be administered by injection, or
it may be ingested, or inhaled. Iodine 131, for example may be
administered either orally or intravenously, and it collects in the
thyroid. Xenon may be administered through a ventilation system
for lung imaging studies. Technetium 99m is a very versatile and
can be used to image the brain, liver, kidney, bone marrow, heart;
and check organ flow, function and location in most of these sites.

The radioactive substance may be monitored while in the

body, or it may be excreted for inspection in the blood, urine or
feces.

Tracking Radio Pharmaceuticals

Once a radio pharmaceutical is placed in the body, the fundamental

problem is that of determining its location at any point in time. The
 72

easiest case would be inspection of excreta or blood samples for

radioactivity. This may be done with a Well counter, an important
case because many nuclear medical imaging devices have evolved
from it.

Well Counter

A Well Counter consists of a sample, scintillating crystal, photo

cathode, photo multiplier tube (PMT) and electronic detecting
circuit, as illustrated in Figure 4. 3. (4.4)
The sample is a biological substance such as blood, urine, or tissue
which contains a
Sam ple radio
pharmaceutical.
Scintillation The purpose of
Crystal
the well counter
Photocathode
R is to measure the
Dynodes
radioactivity of
R 1
the sample. In
R 2 this case the
V
A emissions are
Photom ultiplier
R 3
Tube
gamma rays. The
sample is placed
R R 4
L in a scintillation
5
crystal.
A commonly
AMP COUNTER
used scintillation
DISPLAY
crystal consists of
FIG URE 4.3. W ell Counter sodium iodide
doped with
thallium (NaI(TI)). The thallium is necessary to make it scintillate,
or release light at room temperatures after a gamma ray impacts it.
The crystal is coated with aluminum on all sides except that next to
the photo cathode. The gamma ray passes through the crystal in
any direction, producing light rays which hit the photo cathode
directly, or they may reflect off the aluminum foil onto the photo
cathode.
The photo cathode consists of an opaque substrate coated
with a cesium lead compound, which emits an electron when a
photon of light strikes it. Typical sensitivities run from 3 to 150
microamps per lumen ( A/lm) of light.(4.5) The acceleration
 73

voltage, VA, causes equal voltage drops across successive dynodes,

or plates, numbered 1, 2, 3, 4, 5 in the diagram. These voltages
accelerate the electrons from the photo cathode, causing secondary
electrons to emit from each dynode, such that the electron beam
increases in intensity successively from dynode 1 to 5, as
illustrated in the diagram. The voltage, VA, may be adjusted by the
operator, since different pharmaceuticals emit different energy
levels, and the amplification required varies accordingly. Thus the
current through RL is amplified. The amplifier measuring the
voltage across RL increases the voltage of each pulse of radiation
and delivers it to a counter. The output of the counter may then be
displayed as counts per minute (cpm), which is the number of
nuclear disintegrations per minute.
If the sample has two different pharmaceuticals of
different energy levels, one may distinguish between them by
inserting a threshold detector in front of the counter in Figure 4. 3.
The threshold may then be set to eliminate the lower energy
pulses, thus counting the higher energy pulse. This number may
then be subtracted from the total count without any threshold to
yield the lower energy pharmaceutical count. This process may be
automated with a window detector that can select each count
individually. Typically the counter should handle more than 10
thousand counts per minute (cpm).

Anger Camera

ANGER
CAMERA A radio pharmaceutical in the body may be
PMTs
imaged on a plane with an Anger Camera, shown
NaI(TI)
cut in half in Figure 4. 4. In this device, a
Collim ator
NaI(TI) crystal, as large as 50 cm across is
Patient placed over a patient treated with a radio
(a) Section A-A' pharmaceutical which radiates gamma rays. The
rays emit in all directions. The collimator selects
those which are traveling perpendicular to the
plane of the crystal. Those light rays cause the
A'
crystal to scintillate, and emit light rays into the
PMTs
photo multiplier tubes (PMTs), which are
Collim ator NaI(TI) arranged in the plane of the crystal. Thus the
location of the radiating source can be mapped in
A
that plane by monitoring the output of the PMTs
(b) Bottom View

FIGURE 4.4. Anger Cam era

 74

fixed in known locations. The details of the process are illustrated

in Figure 4. 5.
Of the gamma rays that emit from the patient, the direct ray
passes through a lead lined hole in the collimator directly into the
scintillating crystal. Those rays that are traveling at the wrong
angle hit the lead or tungsten walls of the collimator hole and are
absorbed. Another ray may be scattered, or deflected by atoms in
the body so that it goes into a collimator hole not directly over it.
This could be confused with a direct ray and cause an error in the
image. However the scattered ray is of lower intensity than the
direct ray, having given some of its energy to the deflecting atom.
Thus the pulse height detector can distinguish between direct and
scattered rays on the basis of their intensity. The pulses then
Pulse height Signal
are processed and produce a dot on the display, mapping the
Display
detector Processing position of the PMT, resulting in a planar image of the
radioactive sites in the body. The pulse height detector can
also select among different pharmaceuticals radiating
different energy levels, as indicated in Table 4.1. Pulses
PMTs with higher energy levels create brighter scintillations.
Scintillating Those pulses not within the selection window would be
Light
Crystal rejected.
Collim ator It is assumed that each radiating gamma ray
Scattered Direct Absorbed emerges at a different time. Counts have been achieved with
Gam m a an Anger camera as high as 2.22 x 109 counts per minute
rays
(b) Signal Processing (cpm). (4.6) The time between counts is therefore 27
nanoseconds (ns). The rise times of the processing circuits
FIGURE 4.5. Planar Nuclear Im aging
need to be shorter than this to distinguish between counts.
Since Tc-99m is one of the most popular radio
pharmaceuticals in single photon nuclear medicine, many
clinical Anger cameras are optimized for use at its radiating
energy, 140 keV. This can be altered by changing the pulse height
detector window.

References:

4.1. Blanchard, C. H., C. R. Burnett, R. G. Stoner, R. L. Weber.

Introduction to Modern Physics. Englewood Cliffs, NJ:
Prentice-Hall, Inc. (1958) pp 364 - 379.

4.2. Sears, F. W., M. W. Zemansky. College Physics. Reading,

MA: Addison Wesley Publishing Co. (1957) pp. 152-156.
 75

4.3. Weast, Robert. Handbook of Chemistry and Physics, 52nd

Edition. Cleveland, OH: The Chemical Rubber Company (1971) p.
B245.

4.4. Early, Paul J., D. B. Sodee. Principles and Practice of Nuclear

Medicine. St. Louis, MI: C.V. Mosby Company (1985) p 232.

4.5. Aston, Richard. Principles of Biomedical Instrumentation and

Measurement. Columbus, OH: Merrill Publishing Company (1990)
p 478.

4.6. Croft, Barbara Y. “Nuclear Medicine: Instrumention” Chapter

66.1 in in Joseph D. Bronzino, Editor, The Biomedical Engineering
Handbook. Boca Raton, FL: CRC Press (1995) pp. 1049.

Exercises

Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).
 76

CHAPTER 5
SPECT AND PET SCANNERS
For educational purposes only. See disclaimer above.

The single photon emission computer tomography (SPECT) device

and positron emission tomography (PET) device are both nuclear
medicine imaging instruments that create three dimensional
pictures of internal anatomical sites. The SPECT scanner is a
direct extension of the nuclear medicine imaging device discussed
previously in Chapter 4. It can use the same Anger camera and the
same nuclear pharmaceuticals. It uses the same principle, namely
that the pharmaceutical is inserted into the patient and accumulates
in the tissue of interest. The nuclear radiation is detected by the
Anger camera and processed into a visual image, useful in
diagnosis. Each nuclear event creates a singe photon, hence the
name single photon emission computer tomography. The PET
scanner differs in that it uses different pharmaceuticals, emitting
positrons, and a different detecting camera. Fundamentally, the
physics of the nuclear emission is different in the PET, as will be
discussed below, in that two photons of gamma radiation result
from each nuclear reaction.
 77

The SPECT Scanner

A SPECT scanner using same detecting components as a nuclear

medical imaging device that produces a two-dimensional image is
illustrated in Figure 5.1.
The anger camera is fixed to a c-arm and is rotated around the
patient. At each position, a scan of the radiation from the nuclear
pharmaceutical is taken and stored in a computer. The same
algorithm as is used in the CT scanner can then be made to
generate an image of a slice through patient, as explained in the
next section.
 78

The SPECT Algorithm

The geometric relation between the

radioactive tissue being imaged and
the anger camera is shown in
Figure 5.2. The situation is similar
to that of the CT scanner (Figure
3.3 in the CT Scanner chapter). In
that case, attenuation, or energy
absorption, varies along the line
collinear with the ray being
detected. In this case it is energy
generation that varies along that
line. Mathematically, this is just a
change in the sign; absorption
negative is replaced by generative
positive. Thus the form of the
mathematical equations describing
the process remains the same.
The Anger camera in Figure
5.2 measures rays of radiation that
get through the collimator. The
collimator is ideally designed to
absorb any rays not collinear with
the s axis. Thus the collimator
insures that the measured nuclear
event occurs on a cord of the circle
formed by the rotating Anger
camera. At a given angle 1, this
forms a function of the axis r,
p 1(r). We may take the
one-dimensional Fourier transform of this as

∞
P (k) = ¶ p 1 (r)e −2 i(k$r) dr (5.1)
−∞
Here i is the imaginary number −1 and 2 k has the dimensions
radians per meter, where k is a vector. This formula transforms the
 79

spatial quantity p 1 into k-space. This operation can be repeated

for each value as the Anger cammera is rotated through 0 to
radians to yield a two-dimensional function in k-space, P(k, ).
k-space is a mathematically created region containing imaginary
numbers.
Thus by scanning linearly along the r-direction in Figure
5.2, rotating to another value of , and scanning again we have
created a two-dimensional Fourier transform P(k, ). It follows that
the inverse Fourier Transform can be calculated using the two
dimensional Fourier Transform in polar coordinates (5.1)
∞
p (x , y ) = ¶ ¶ P ( k , )e 2 i( x k c o s + y k s in )
x k x dk d
0 −∞

(5.2)
To create the set of real numbers p(x,y), the computer first
integrates on k to create a function of x, y, and . It then integrates
on to create the function p(x,y) which can be mapped as a
gray-scale image in rectangular coordinates. In practice we must
limit the range on k to a finite number, since the computer cannot
go on calculating for ever. A reasonable approximation is achieved
if Equation (5.3) is written so that the integration on k goes from -
k max to + kmax as (5.2)

k m ax
p (x , y ) = ¶ ¶ P (k , )e 2 i (x k c o s + y k s in )
x k x dk d
0 − k m ax

(5.3)
where

kmax = 21 r (5.4)
where r is the distance between radiation beams in Figure 5.2,
which measures the spatial resolution of the image. The larger the
number of k frequencies integrated, the better the resolution. In
practice, the resolution is in the order of 1 cm for the SPECT
scanner, as compared to 1 mm for an x-ray CT scanner. The image
is an approximation to the ideal because a discrete, finite number
of values of k and are chosen. Mathematicians over the past
quarter century have developed many methods of improving the
 80

approximation and the speed with which the calculation can be

made by variations the basic process of taking the Fourier
transform and its inverse as expressed in Equations (5.1) to (5.4)
(5.3)
. This process is called the filtered back-projection method. The
magnitude k in Equation (5.4) is called a ramp filter in k-space.

Advances in SPECT Scanning

Engineering development of the SPECT scanner should have the

objective of improving the scan time, count detection efficiency,
resolution and accuracy.
The scan time can be shortened by adding more anger
cameras, as shown in Figure 5.3, where two are shown.
 81

The radiation from each nuclear event goes in all directions. Thus
with more cameras surrounding the patient one would pick up
more events and the count capability would be higher, up into the
150K count per second range. Also the patient can be scanned with
less rotation of the cameras, decreasing therefore the scan time.
This logic would indicate adding more cameras around the
circle, at the cost of more complexity. With a full size anger
camera, four would fill the circle. However if the camera is made
smaller, ten or more could fit around the circle. (5.4) Often in
clinical practice, complexity and cost considerations keep the
number of cameras to two. The resolution of the image is
determined by the size of the collimator holes. A brain SPECT
image using about 500K counts per slice has a spatial resolution
reported at 8 mm, whereas a myocardial study using 150K counts
per image produces a resolution of 15 mm. (5.5)
Most clinical nuclear medical instruments use
thallium-activated sodium iodide (NaI(Tl)) scintillators. These
have some disadvantages: they require PTMs; they are
hygroscopic and thus yellow as they absorb moisture; and their
density is relatively low, so their stopping power for high energy
radiation is relatively low. Thus the search for better crystals is
ongoing. Table 1(5.6) lists a number of candidates.
 82

Scintillating Crystals for Gamma Radiation

Table 5.1 Properties of

Scintillators (5.6)
Detector Density Effective Decay Light Hygro-
g/cm 3 Z time(ns) Output scopic
(%)

NaI(TI) 3.67 50.6 230 100 Y

CsI(TI) 4.51 54.1 1000 45 N
CsI(Na) 4.51 54.1 630 85 Y
BGO 7.13 74.2 350 15 N
LSO 7.4 65.5 42 75 N

As the density and effective atomic number Z of the crystals in

Table 5.1 increase the stopping power increases, which means
higher energy gamma rays are detected. Thus bismuth germanate
(BGO) crystals, and lutetium oxyorthosilicate (LSO) crystals have
higher stopping power than NaI(TI).
The decay time limits the number of counts per second that
can be detected. LSO has the fastest response time, and
thallium-activated cesium iodide crystal (CsI(TI)) the slowest.
LSO with a 42 ns decay time could detect as much as 1/(42(10-9))
= 23.8 Million counts per second. That number for NaI(TI) is
1/(230(10-9)) = 4.35 Million counts per second.
The light output for NaI(TI) is best of the set of crystals
considered, and BGO is the worst. Overall LSO has the best
properties, having superior stopping power, higher count
capability, relatively high light output and no hygroscopic
yellowing. E.V. Garcia reports (5.7):
LSO can stop a 511 keV photon in about 1.16 cm, BGO in
1.05 cm, and NaI(TI) in 2.88 cm. Yet, LSO has five times
the light output of BGO and about 1/10th the decay time,
 83

making it an excellent scintillator. One drawback of the LSO

crystal is that it produces its own natural radiation.

Attenuation Correction

As in the case of x-ray,

gamma rays are
attenuated as they pass
through tissue. The
propagation of the gamma
ray is governed by
Lambert’s law as
discussed in the x-ray
Chapter 3. Although this
is nearly the same for
most viscera tissue, lung
and bone have widely
different attenuation
coefficients. Thus large
errors can be introduced
in to the SPECT scan. To
compensate for this a CT
absorption scan is taken.
A x-ray tube may be used
as a source. In Figure 5.4,
a radioactive source is illustrated. The source is translated along its
holder, and rotated around the patient so that the gamma rays emit
from the source and pass through the patient’s body to be collected
by the Anger Camera. The CT algorithm is then used to map the
attenuation coefficients. These are used to correct the SPECT scan
resulting from the radio pharmaceutical in the patient.
The increased accuracy of the scan is achieved at the cost
of an extra radiation dose in the patient, and more radioactive
material to be contended with in the hardware.

The PET Scanner

 84

The Positron emission tomography (PET) scanner is similar to the

SPECT scanner in that it detects nuclear pharmaceuticals in the
body tissue of a patient. It differs in the fact the the
pharmaceuticals it uses emit positrons which in turn emit gamma
rays.
Commonly used positron emitting with their half-lives and
example uses are listed in Table 5.2. The half-lives listed are very
short, from 2 to 110 minutes. Therefore the cyclotron used to
produce 11C, 13 N, and 15O isotopes, for example has to be close by
the use facility. This adds significantly to the cost. The short life
times minimize the radiation dose to the patient, but make the
pharmaceuticals highly perishable.
The PET scanner is useful in heart perfusion tests through
the use of 11C and in measurement of brain properties through
glucose metabolism, oxygen utilization, and neuroreceptor studies.
The physics of positron interaction in tissue imposes
differences in PET and SPECT hardware. After the positron is
emitted from the isotope, it travels 1 to 4 mm before it encounters
an electron with which it annihilates and produces a gamma ray as
illustrated in Figure 5.5. This imposes a limit on the resolution of
the PET scanner, because the position of the gamma ray is 1 to 4
 85

mm away from the pharmaceutical containing the isotope, which is

the point desired.
The gamma ray travels in opposite directions within a
tolerance of plus or minus
0.25o (0.004363 rad) . Each
gamma ray is at the 511 keV
level. Since the rays are
assumed to be collinear in the
image reconstruction
mathematics, this imposes
another error in measurement.
The gamma ray photon is
electromagnetic radiation and
travels at the speed of light, 3 x
108 m/s. This means to travel
one meter the photon requires
3.33 ns. To travel 1 mm, the
order of resolution desired,
would require 0.0033 ns.
Electronic circuitry capable of
resolving the 0.0033 ns would
be very difficult to design.

Electronic Collimation

The fact that the PET scanner can use electronic collimation,
illustrated in Figure 5.6, and the SPECT scanner cannot, creates a
major difference between the two devices.
 86

A radio pharmaceutical is depicted as being in the right shoulder of

a patient. At the top of the figure, positron-electron annihilation,
 87

event (a), creates a collinear pair of gamma ray photons that strike
detectors 12 and 32 on the detector ring. Thus it is known the event
occurred on a cord of the circle of the detector ring. Event (b)
depicted below strikes detectors 8 and 27, telling us the event is on
the cord going between those two detectors. Since millions of these
isolated events occur one after another, going randomly in all
directions, in a short time all detector pairs will have received a
pair of gamma rays. Therefore each event, and the cord on which it
occurred, can be stored a computer memory site. This is exactly
the information given by the SPECT scanner as illustrated in
Figure 5.2 above. This information can be used in Equations (5.1)
to (5.6) to form an image of the location of the radio
pharmaceutical in a cross sectional slice through the patient.

Collimation Errors

The path of the gamma rays in both the SPECT, Figure 5.2, and
PET scanners, Figure 5.6 are shown to be straight lines. Some
photons scatter, that is they collide with atoms and interact with
electrons. This may cause their path to deviate from a straight line,
and their intensity level to diminish because they loose energy in
the collision. Thus they may hit the wrong detector, or enter the
wrong collimation hole, and cause an error. As many as 30% of the
gamma rays to enter the detectors may be scattered. To reduce the
effect of scattered photons, a threshold detector is used to block
low energy counts.
Errors may result if two nuclear events occur almost
simultaneously, and seem to merge into one event. In this case in
the PET, two gamma rays, one from each event, may hit detectors
and two may be lost. Statistical methods are used classify events as
(5.9)
random.
 88

Axial Collimation
Septa Detector

The collimation in Figure 5.6 is

in the x-y plane, a transverse
slice through the body. The
nuclear event may send gamma
rays in any direction. To reject
gamma rays from outside the
slice desired, tungsten septa may
Sintillator be used to absorb them as
illustrated in Figure 5.7 (5.10). In
modern PET scanners, Budinger
Z reports the septa separate 15 to 47
levels. This collimator works as
FIGURE 5.7. Axial Collim ation described in Figure 4.5.
 89

Detectors

In order to improve the

resolution limit imposed by the
size of a photo multiplier tube
(PMT), a scheme illustrated in
Figure 5.8 is used.(5.11) The PMT
is typically 16 mm square. A
bismuth germanate (BGO)
crystal is attached to an array of
four such PMTs, cut into squares
as illustrated in the figure. The
squares may be as small 4 by 4
mm, thus improving the
resolution. The depth of the cut
into the crystal changes the
character of the light that
impinges on the PMT. The deep
cuts tend to focus the light on
one PMT, while the shallow cuts
spread the light over several
PMTs.
In the case illustrated, if the gamma ray were to fall on
BGO crystal 1, it might illuminate only PTM 1. If it fell on BGO 2,
it might illuminate both PMT 1 and 2. If it fell on BGO 3, it would
illuminate both PMT 1 and 2, but at a lesser intensity. Continuing
this coding down to BGO 36 which would illuminate only PMT 4,
we have a scheme such that the 36 BGO locations can be coded
into combinations of the 4 PMTs. PET scanners are used which
extends this to 64 BGO crystals coded down to 4 PMTs.
Further improvement in resolution would result from the
use of solid state detectors that create an electrical signal directly
from the gamma ray, eliminating the need for a scintillating crystal
and PMT. One such crystal is cadmium zinc telluride (CZT) (5.12). It
can be made small enough that the resolution would be limited
only to the 1 to 4 mm determined by the basic physical process of
conversion of a positron to gamma rays, as illustrated in Figure
5.5.
 90

OVERVIEW, ADVANCES AND COMPARISON OF SPECT

AND PET SCANNERS (5.13)

The intent of this section is to give you an overall review of

nuclear imaging which emphasizes the needs of people who
manage, maintain, and design such equipment. The information
here is expanded greatly by the inclusion of many web sites
immediately and fully available to all readers.

Nuclear imaging differs from the others such as X-ray, CT

Scan, MRI and Ultrasound in that it images primarily the
biological function or process of organs rather than their
anatomical structure or morphology alone. This is not a totally
exclusive characteristic, since the MRI can also do functional
imaging, and anatomical data is sometimes gleaned from nuclear
imaging. Common to all nuclear imaging, whether planar imaging,
or tomography as in single photon emission computer tomography
(SPECT) or positron emission tomography (PET), is the nuclide in
a radiopharmaceutical which is injected into the body being
imaged. The insertion is either by needle injection, oral, or by
inhalation. The radiopharmaceutical is usually specific to a
particular organ which absorbs it and becomes radioactive. The
radiation from radioactive decay is almost always of gamma rays,
which are electromagnetic waves of higher frequency and energy
than either visible light or x-rays.
The planar imaging device which is designed to make a
two-dimensional image of the radioactive organ or substance can,
fortunately, be extended into and become an element of the
tomography device, SPECT, and does not have to be redesigned.
Thus if you can solve all the problems regarding SPECT scanners,
you can deal with those with planar nuclear imaging as well - a
nice double whammy, which justifies our emphasis on the SPECT
Scanner here.

The PET scanner differs fundamentally from the SPECT in

that the nuclear decay process is in two steps: first a positron (the
anti-body of an electron) emerges and travels 1 to 4 millimeters
until it hits an electron and annihilates, converting the masses into
two gamma rays traveling in essentially opposite directions. The
 91

travel of the positron creates a blurring effect, such that the

resolution of the PET scanner cannot be better than that 1 to 4
millimeter travel distance. The resolution of nuclear imaging
devices is more course than MRI or CT Scan because of this,
compton scattering, and the need for collimators or equivalent
gamma ray selective software. If a commercial imager has a
resolution between 3 and 6 millimeters, it is probably operating
within specification.

The other major design problem with nuclear imaging is in

attenuation of the gamma ray as it travels from the target organ to
the imaging camera. This is variable depending upon the
interfering tissue. Often this attenuation is accepted as clinically
tolerable error, but higher-end equipment is available which
calculates the attenuation the gamma ray encountered and thereby
improves the image. This correction may be built into a SPECT
scanner, or one may take advantage of the fact that the
fundamental process by which a CT scanner makes an image is to
calculate a map of the attenuation of each element (voxel) in a
slice. Thus we see on the market now combinations such as the
PET/CT Scanner. This happy combination both improves the PET
image by calculating the attenuation map of the slice, and by
making an anatomical image of the slice as well as a biofunctional
image. The MRI is not combined on a large scale yet because of
the interference of its large magnetic field with the PET process.
Work is also being done to combine PET and SPECT Scanners and
SPECT and CT.

SPECT Scanner Advances

The basics of SPECT Scanner is available in full text on line at the

web site < http://venda.uku.fi/~aoseppan/SPECTthesis.pdf > 5.14
The SPECT scanner depicted in Figure 5.3 consists of two anger
cameras at right angles with each other. The cameras rotate around
the patient and collect gamma rays emitted from the patient. The
information thus obtained is of the same form as that in a
Computer Tomography X-ray scanner. Thus the same kind of
signal processing can be used to create a two dimensional (slice)
 92

image of the organ containing the pharmaceutical. The basic data

collection hardware is shown in Figure 4.5. Suppose the
pharmaceutical is concentrated as shown in the patient. The
gamma rays emitted may go in any direction. Those which go
directly through the collimator enter a scintillating crystal, often
thallium doped sodium iodide (NaI(TI)) and as a result of
collisions with atoms produce a visible light photon. The light
enters a photomultiplier tube where it is converted into an electron
beam which is amplified. These gamma rays are attenuated by a
factor depending on the patient’s body tissue. Other gamma rays
are scattered by collisions with atoms along the path. These rays
tend to be of lower amplitude than the direct rays and may be
identified by the signal processing equipment, most simply by a
pulse height detector, or better with software. The collimator has
septa, often of lead which absorbs gamma rays that enter it. The
two major figures of merit of the SPECT scanner: sensitivity, the
percentage of nuclear events which send rays to the detector
PMTs, and the resolution which is a measure of the smallest
particles that can be distinguished, are affected by the parts in
Figure 4.5. Over the years resolution has been improved by
developments in the collimator, the reduced size of PMTs, and in
signal processing. Sensitivity has been improved by developments
in the type of scintillating crystal and collimation process.

If the holes on the collimator are made smaller in diameter,

the resolution improves, but at the cost of less sensitivity, because
the gamma ray is more likely to be absorbed by the septa material
when the hole is small. Smaller size PMTs also improve
resolution. A position-sensitive photomultiplier tube (PSPMT) is
available from Hamamatsu (see their web site <
http://nedelec.home.cern.ch/nedelec/UHE/hardware/R5900-C8.pdf
>) which has an array of 64 anodes in an approximately 1 square
inch area package. Experiments with PET scanners using PSPMT
show the possibility of a 2.8 mm resolution. (see the web site <
http://www.bic.mni.mcgill.ca/users/chris/ANIPET_2000.PDF >5.15
for a complete paper) Another approach to SPECT scanning is to
employ solid state diode detectors instead of a PMT. Such a device
is available from Digirad Corporation. Their website describes
their Thallium doped Cesium Iodide (CsI(TI)) detector based
equipment. Extensive description is offered there along with
photographs. (see < http://www.digirad.com/default.asp >)
 93

Improvements in sensitivity of nuclear imaging equipment are

being sought by investigating other crystals other than the
traditional thallium doped sodium iodide NaI(TI) as scintillators.
Bismuth germanate (BGO) has higher stopping power but less
light output per unit volume than NaI(TI), and is used in the PET
scanner which emits high energy, 511 keV gamma rays. E.V.
Garcia reports 5.16 that lutetium oxyorthosilicate (LSO) can stop
such a photon in about 1.16 cm, BGO in 1.05 cm, and NaI(TI) in
2.88 cm. Although LSO has less light output than NaI(TI) it has
five times the light output of BGO and about 1/10th the decay
time, making it an excellent scintillator. One drawback of the LSO
crystal is that it produces its own natural radiation, which could
raise the ionization dose to the patient, and pose a hazard to the
users. Gadolinium oxyorthosilicate (GSO) also has excellent
stopping power and is used in PET scanners.

An excellent article on how to assess and control the

quality of the SPECT image is available in full text on line at the
web site: <http://tech.snmjournals.org/cgi/reprint/29/1/12.pdf >5.17

PET Scanner Advances

One of the major advantages of the positron emission tomography

(PET) scanner is that it improves sensitivity over the SPECT
scanner by utilizing electronic collimation and getting rid of the
mechanical collimator. Two dimensional (2 D) PET scanners use
electronic scanning in the cross-sectional plane, and use
mechanical colliminators in the axial direction. Three dimensional
(3D) scanners eliminate the mechanical device in all planes, and
use software to colliminate in the axial direction.
In the positron emission tomography (PET) electronic
collimator illustrated in Figure 5.6. The gamma rays are detected
in the PET much as in the SPECT, except the PET rays are higher
energy (511 KeV) so BGO crystals are often used with PMTs
because of their higher stopping power, and a coincident detector
is added to select those gamma rays which hit opposing detectors
at the same time.
 94

Since the SPECT and PET are primarily used to assess

biological function rather than anatomical structure, it is useful to
use the CT scanner in conjunction with either of these to add
anatomical data. The CT scanner also measures the attenuation
map so that both the PET and SPECT scans can be made more
accurate. That is, the computer can take into account the
attenuation the gamma ray experiences as it travels through the
body tissue to the detector, using the CT data.
A PET/CT scanner has been evaluated in a full text
document available on the web as the site:<
http://www.petmadrid.com/documentos/Performance%20evaluatio
n%20of%20the%20new%20whole-body%20PET_CT%20scanner
_Discovery%20ST.pdf > 5.18 Written in 2004, this comprehensive
article gives a description of state of the art PET scanners,
describes the phantoms used for test and gives complete resolution
and sensitivity data for the General Electric Corporation,
Discovery PET-CT Scanner. Excellent pictures of the Philips
Corporation PET/CT open Gemini scanner appear on the corporate
website: <
http://www.medical.philips.com/main/products/pet/assets/docs/GE
MINI_Brochure_040903.pdf >.
Some experimental work is being done to combine SPECT
and CT,5.19 but to avoid excessive amounts of hardware,
attenuation correction can be done by using a radioactive
substance as a known source of radiation, instead of an x-ray tube,
which is rotated around the patient to make a CT scan from which
an attenuation map can be made to correct the SPECT scan without
adding a separate CT scanner, as shown in Figure 5.4.

Troubleshooting

The quality of nuclear imaging equipment must be frequently

tested in order to insure accurate data for disease diagnosis 5.20.
Phantoms containing radio active substances such as tectnetium
(99mTc) as a known source of radiation are used to measure
sensitivity and resolution of clinical hardware. Low sensitivity
could be caused by a cloudy scintillating crystal, damaged
colliminator, or weak PMTs. A failed PMT for example might
produce a blank spot in the image. Good troubleshooting tips about
replacing PMTs are found in the web site <
http://www.atral.com/pmtwrap1.html >. Cracks in the scintillating
 95

crystal could produce artifacts. Both resolution and sensitivity are

improved if the camera is placed closer to the patient. SPECT
scanners are available that have the camera follow the contour of
the body for this reason.

REFERENCES

5.1. Suetens, Paul. Fundamentals of Medical Imaging. New York:

Cambridge University Press, (2002) p 22.
5.2. ibid. p 78.
5.3 ibid.
5.4. Milster, T.D., Aarsvold, J.N., et.al., “A Full-field Modular
Gamma Camera.” J. Nuclear Medicine 31:632.
5.5. Tsui, B. M. W., “SPECT (Single-Photon Emission Computer
Tomography)” in Joseph D. Bronzino, Editor, The Biomedical
Engineering Handbook. Boca Raton, FL: CRC Press (1995) pp.
1055.
5.6. Garcia, E.V., T.L. Faber, et.al. “Advances in Nuclear emission
PET and SPECT imaging.” IEEE Engineering in Medicine and
Biology, vol 19, No 5, September/October 2000, pp 22-33.
5.7. ibid. p 25.
5.8. Graham, M. G. “Positron-emission Tomography” in
Encyclopedia of Medical Devices and Instrumentation, Volume 4,
Ed. by John Webster, New York: John Wiley and Sons (1988), p
2343.
5.9.Hoffman, E.J., S.C. Huang, et.al. “Quantitation in positron
emission computed tomography: 4. Effect of accidential
coincidences.” J. Computer Assisted Tomography. 5:391, 1981.
5.10. Budinger, T.F., “Positron Emission Tomography
Instrumentation” in Joseph D. Bronzino, Editor, The Biomedical
Engineering Handbook. Boca Raton, FL: CRC Press (1995) pp.
1140-1150.
5.11.Suetens, Paul op.cit. p. 203.
5.12. Garcia, op.cit. p 22.
5.13 Aston, R. “The Fundamentals of Nuclear Imaging: SPECT
and PET Scanners” Biomedical Instrumentation and Technology.
vol 38, no 5, pp. 363-366 (Oct/Nov, 2004). This section is taken
from the article cited here.
 96

5.14. Seppänen, Aku O., “Correction of Collimator Blurring and

Attenuation in Single Photon Emission Computed Tomography”
University of Kuopio, M.S. Thesis, on line <
http://venda.uku.fi/~aoseppan/SPECTthesis.pdf
>
5.15. Nguyen, K. O. “Operational and Performance Report on
ANPET, a versatile PET Scanner for Laboratory Animals”.
Montreal Neurological Institute, on line <
http://www.bic.mni.mcgill.ca/users/chris/ANIP
ET_2000.PDF >
5.16. Garcia, E.V., T.L. Faber, et.al. “Advances in Nuclear
emission PET and SPECT imaging.” IEEE Engineering in
Medicine and Biology, vol 19, No 5, September/October 2000, pp
22-33.
5.17. Groch, M. W., W. D. Erwin, Journal of Nuclear Medicine
Technology 2001; 29:9-15. Text on line at <
http://tech.snmjournals.org/cgi/reprint/29/1/12.pdf >
5.18. Bettinardi, V. et al. “Performance evaluation of the new
whole body PET/CT scanner: Discovery ST,” European Journal of
Nuclear Medicine and Molecular Imaging, 10.1007/
500259-003-1444-2 (10 February, 2004). Full text is on line at
http://www.petmadrid.com/documentos/Performance%20evaluatio
n%20of%20the%20new%20whole-body%20PET_CT%20scanner
_Discovery%20ST.pdf
5.19. Kalki K, et al. "Myocardial perfusion imaging with a
combined x-ray CT and SPECT system. "J Nucl Med. 1997
Oct;38(10):1535-40.
5.20. Early, P. J., D. B. Sodee. Principles and practice of nuclear
medicine. Saint Louis:C.V. Mosby Company (1985) Ch 13, pp
372 - 425.

Exercises

Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).
 97

CHAPTER 6
ULTRASONIC IMAGING EQUIPMENT
THEORY

For educational purposes only. See the disclaimer above.

Ultrasonic Waves

Ultrasonic waves are similar to the pressure and flow waves seen
in fluid flow. A pressure difference, p, across two points in matter,
whether air, tissue, or metal, causes a displacement of the atoms,
giving them a velocity, v. The atoms do not move very far because
they are bound by elastic forces. However, the energy of one atom
is transferred to the other atoms, and it propagates through matter
at its own velocity.
There exists an analogy of ultrasonic waves to voltage
waves: ultrasonic pressure, p is analogous to voltage, and the
particle velocity, v, of ultrasonic waves is analogous to current.
Furthermore, the acoustic impedance is analogous to the
impedance of an electrical circuit. Therefore, your knowledge of
electrical circuits will help you understanding of ultrasonics.
We must begin with a fundamental understanding of
ultrasound itself. Most biological tissue of interest in ultrasound
imaging, soft tissue, supports ultrasonic waves according to the
following model: A solid surface is caused to vibrate and is put in
contact with the tissue to launch a pressure wave in the
longitudinal direction. That is, the tissue displacement is in the
same direction as the wave travel.

The Wave Equation

To analyze how sound propagates in matter, we treat a fluid

medium first. Most soft tissues, the primary object of medical
ultrasonic imaging, would fit this model. Consider an element of
 98

fluid with a surface area S (m2 ) as shown in Figure 6.1. An

external sinusoidal force, F is applied over that surface. The
density of the fluid, , has units of kilograms per meter cubed
(Kg/m3). MB represents the bulk modulus of the element which is
defined in the Handbook of Chemistry and Physics by the equation
p 2 −p 1
M = B v 1 −v 2 (6.1)
v1
where p1, p2 are the pressures on either side of the volume, and the
denominator is the fractional change in volume due to the pressure
differential. The units on Equation (6.1) are pressure in newtons
per meter squared (n/m2). MB is measured by applying a pressure
across the material under test, and measuring the fractional volume
change.
In equilibrium, the element in Figure 6.1(b) extends
from a position of x to x + < x. A force is applied at the
Area, S
MB surface x which displaces the left surface by a distance u
x and the right surface by a distance u + <u, as shown in part
(a) c of the figure. The incremental displacement u arises
from spring action or elasticity in the element. The energy
inserted into the surface at x is transferred into the element,
setting it, as we see from the resulting equations, into
x x+ x harmonic motion. Assuming first a lossless case, the
(b) energy inserted is equal to the potential energy stored in the
spring at its maximum displacement. (At other points in the
F
oscillation the energy would be distributed between
potential and kinetic energies). Thus we can form the
u u+ u equation

F Øx = M B S Øu
(c)
FIGURE 6.1. An elem ent of m atter in ultrasound

The units on both sides here are newton-meters (n-m). Thus we

can write

F = M B S Øu
Øx (6.2)
Then differentiating this with respect to x
ØF Ø2u
Øx = MB S Øx 2 (6.3)
The displacement <u shown in Figure 6.1(c) is caused by the
difference in forces on the left and right surfaces of the element,
 99

<F. Since we have a linear system we can say the ratio of <F to
<x extends down to infinitesimal dimensions so that Equation
(6.3) becomes
¿F ØF Ø2u
¿x = Øx = MB S Øx 2 (6.4)
Now we remember Newton’s law that force equals mass times
acceleration and realize that the differential force <F in Figure
6.1(c) accelerates the mass S <x, (where is the element density
in kilograms per meter squared, and S is the area in meters
squared) according to the equation
Ø2u
¿F = S ¿x Øt 2 (6.5)
where u is the displacement in meters of the element from its
equilibrium position and t is time. Combining Equations (6.4)
and (6.5) gives
Ø2u Ø2u
M B S ¿x Øx 2 = S ¿x Øt 2
Hence
Ø2u Ø2u
Øx 2 = MB Øt 2
(6.6)
We now define a quantity, c,
MB
c= (6.7)

where c has the dimensions meters per second (m/s) as shown in

Exercise (6.1), and, as will shown below, is the velocity of sound
in the medium. Thus Equation (6.6) becomes
Ø2u 1 Ø2u
Øx 2 = c 2 Øt 2 (6.8)
This is the wave equation for a longitudinal sound wave. It is
longitudinal because the force that displaces the matter travels in
the same direction as the wave, as illustrated in Figure 6.1. A
Transverse wave is the kind commonly seen on the string of a
guitar. The displacement of the matter is perpendicular to the
direction of sound wave flow. It can be shown that the wave
equation for transverse waves is identical to that for longitudinal
waves.(6.1)

Solutions to the Wave Equation

 100

The wave equation (6.8) describes the displacement u as a function

of time and position. In a linear system, the displacement is
proportional to the force, or pressure which creates it. Thus we can
write u = a p, where a is a constant and p is the pressure in newtons
per square meter (n/m2). Substituting this into Equation (6.8) gives
Ø2a p 1 Ø ap
2

Øx 2 = c 2 Øt 2
Canceling the constant a on both sides we have the wave equation
for the pressure, p,

Ø2p 1 Ø p
2

Øx 2 = c 2 Øt 2
(6.9)

The solution to this equation is shown in Exercise (6.2) to be

x−
p = P 0 e j( t) (6.10)
where is a phase constant with units (m-1) , is the radian
frequency, in radians per second (rad/s), both of which are
measurable physical quantities, and
c= (6.11)
having units meter per second (m/s). In this solution,
the sign before both the or may be either plus (+)
or minus (-). Equation (6.10) is a complex number,
having the imaginary number designator

−1 = j
One solution of Equation (6.10) is the real part

p = P 0 cos( x − t)
(6.12)
This is the expression for a forward traveling wave, as illustrated
in Figure 6.2. where an example calculation is worked out as done
in Example 6.1.
______________________________________________________
 101

EXAMPLE 6.1. Program Equation (6.12) on a graphics calculator

and plot p(x) for several instants in time to illustrate that this is a
forward traveling wave. Take = 1 rad/m, = 1 rad/s, and P0 = 10
N/m2, and plot for successive times.
SOLUTION
Equation (6.12) is programed into a graphics calculator for two
cases, t = t1, and t = t2. The resulting printout for t1 = 0 and t2 =
1.57. The result plotted in Figure 6.2 shows that from the time t = 0
to t = 15.7 the crest of the sound pressure wave has moved from x
= 0 to x = p/2. Thus as time increases the
Period
T wave travels in a positive direction. Thus it
p is called a forward traveling wave.
tim e, t ___________________________________
(a) Sound pressure versus tim e for a Example 6.1 shows that Equation (6.12)
particular position in space describes a forward traveling sound
pressure wave. Notice the minus (-) sign in
Com pression front of the wt term. If one changes that to
Expansion
W avelength a plus (+) sign, Exercise 6.3 shows that the
sound pressure wave travels in the negative
x direction, and is called a backward
p position, x traveling wave, with the expression

p = P 0 cos( x +
(a) Sound pressure versus position for a
particular point in tim e
t)
FIGURE 6.3. A logitudinal sound pressure (6.13)
wave Since the sound wave is longitudinal, it
travels through the air, or other physical
medium, causing alternate compressions and expansions as
illustrated in Figure 6.3. Part (a) of the figure is a plot of Equation
(6.13) when x is fixed at a particular value, and is simply a
sinusoidal wave form. The period is the time between
crests,

T= 1
f (6.14)
Where = 2 f.
The wavelength, , is the distance between compression
crests when Equation (6.13) is plotted at a particular point in time,
as illustrated in part (b) of the figure. When x passes through a
distance , the quantity x equals 2 radians. Thus

=2
 102

or
= 2
(6.15)
The velocity of the sound wave is the speed at which a crest
travels through space. Mathematically this is derived by noting that
if we would travel along a crest in the wave the pressure would
remain constant. Thus in Equation (6.12) we would have

x− t = cont
The crest velocity, c, is dx/dt. Differentiating both sides,
dx
dt − =0
and
dx
dt = =c (6.16)
This shows that the velocity c in Equation (6.11) is indeed the
speed of sound in the medium. The c is also the same as in
Equation (6.7), which means that c is a physical property of the
medium, dependent on its bulk modulus and density.
Also, combining Equations (6.15) and (6.16) gives another
relationship for the speed of sound as

c= f (6.17)

Characteristic Impedance

A sound pressure, p, propagating through a medium causes a

displacement of its particles as illustrated in Figure 6.3 at a
velocity, v. Materials with higher characteristic impedances cause
less velocity for a given pressure. Therefore it makes sense to
define characteristic impedance, Z0, as the ratio of the pressure to
the velocity, as
−p (when a one− directional traveling wave)
Z0 = v (when a one− directional traveling wave)
(6.18)
where the negative sign arises because a positive pressure gradient
causes a velocity in the negative x-direction. (In an electrical
system, this impedance is defined as the ratio of voltage to current,
 103

where voltage is analogous to sound pressure and current is

analogous to particle velocity.)
To derive relationships for Z0 we begin with an equation
for the displacement, u, of the particles. Since in a linear system,
the displacement is proportional to pressure, p, Equation (6.13)
yields

u = U 0 cos( x − t) (6.19)
where U0 is the amplitude of the displacement. The velocity, v, of
the particle is du/dt, so

v= U 0 sin( x − t) (6.20)
We can also differentiate u with respect to x, yielding
du
dx = − U 0 sin( x − t) (6.21)
Then from Equation (6.2) we have
du
dx = MFB S
and the pressure p is equal to F/S, so this becomes
p
du
dx = MB (6.22)
This into Equation (6.21) gives

p = − U 0 M B sin( x − t)
Using Equation (6.20) gives
p − U0 MB sin( x− t) − MB
v = U0 sin( x− t) =
This into Equation (6.18) and (6.7) yields the formula for
characteristic impedance as
c2
Z0 =
Then from Equation (6.11) we have

Z0 = c (6.23)
 104

So we see the characteristic

impedance of a material is
equal to the product of its
density and the speed of
sound in the material, both
measurable physical
properties of the medium.
The characteristic impedance
of several materials appears
in Table 6.1(6.2).

Ultrasonic Wave
Reflections

Ultrasonic imaging
equipment works by
detecting the reflections of
incident ultrasound waves,
called echoes, on the tissue
being investigated. Tissues
are distinguished from one
another by their relative
reflected wave intensities. It is important to understand what
property of the tissue influences reflection so that the proper uses
and limitations of the equipment may be appreciated. We shall see
that tissue density is important in distinguishing different organs
under test. The exact analysis of ultrasound wave reflections is
very complex because the boundaries in biological tissues are
pretty much randomly distributed so that incident waves are
largely scattered back. However in order to understand what
factors affect reflection, and to make approximations of the
intensity levels of reflections one may analyze the simple case of
plane wave incident on a boundary between two tissues extending
infinitely in either direction from the boundary as illustrated in
 105

Boundary
Tissue 1 Tissue 2

c Z = c2
Z = 1 1 02 2
01

Incident
wave

Transm itted
Reflected wave

+x-axis 0 - x-axis

FIGURE 6.4. Ultrasound reflection from a boundary

between two substances
Figure 6.4. Here the x-axis is drawn positive to the left. The
expression for the sound wave traveling in the negative x-direction
is given by Equation (6.13), it having a plus sign in the argument

p 1 = P 01 cos( x + t)
This is called the incident wave in the figure. As this wave hits the
boundary between two tissues with different characteristic
impedances, Z01 or Z02, a portion will be reflected from it, p2,
which will travel in the positive x-direction, and will be referred to
as the reflected wave

p 2 = R P 01 cos( x − t)
where R is the fraction of the incident wave which bounces off the
boundary. R, the pressure reflection coefficient is defined as

pressure magnitude refected at the boundary at x = 0

R= pressure magnitude incident on the boundary at x = 0
(6.24)
Both the incident and reflected waves are traveling in tissue 1, and
so the total pressure there is the sum of these, p = p 1 + p2. Thus
 106

p = P 01 [cos( x + t) + R cos( x − t) ] (6.25)

The pressure in tissue 1 causes the molecules to move with a
velocity, v, proportional to the pressure in a linear system. The
reflected wave will travel in the opposite direction of the incident
wave, and will therefore have a minus sign, a 180o phase shift, so
that

v =V 01 [cos( x + t) − R cos( x − t) ] (6.26)

The ratio of pressure to velocity is defined as the wave impedance
in the tissue, z,
p
z(x) = v (6.27)
That is, the wave impedance equals the pressure divided by the
velocity of the wave composed of the sum of all incident and
reflected waves, and is a function of position, x. Notice z is not the
characteristic impedance as defined in Equation (6.18) because it
contains waves going in multiple directions..
A means of measuring the reflection coefficient is derived
by considering the wave impedance at the boundary. First set x = 0
in Equations (6.25) and (6.26) and insert these in to Equation
(6.27) to yield
P [cos( t)+R cos(− t)]
z(0) = V0101[cos( t)−R cos(− t)]
Here the cosines cancel because cos(wt) = cos(-wt). Also the
characteristic impedance in tissue 1 is given by
P 01
Z 01 = V 01
since this is the ratio of the magnitudes of a wave traveling in one
direction only in the tissue. Thus these two equations become

z(0) = Z01 1+R

1−R
Also at the pressure on one side of the boundary at x = 0 equals
that on the other side, since both sides are infinitesimally close.
Also on the right side of the boundary there are no reflected waves,
so the impedance z(0) is equal to the characteristic impedance of
that medium, Z02. So this equation becomes

= 11 +R
Z02
Z01 −R (6.28)
This is solved for the reflection coefficient as
 107

Z 02 − Z 01
R= Z 02 + Z 01 (6.29)

2 c2 − 1 c 1
R= 2 c2 + 1 c 1 (6.30)
These equations show that the sound pressure reflection
coefficient, R, can be calculated from the physical properties of the
tissue, namely its density, r, and the speed of sound in the
material, c, or the characteristic impedance of the tissue.
Appropriate data for several common biological materials are
given in Table 6.1.

Power Reflection Coefficient

The quantity ultrasound pressure times velocity, pv, has the

physical units
n m
m2 s = n−m 1
s m2
Here we see on the right side of the equation, the newton-meters
per second represent energy flow,or power. Thus the right hand
side represents power per meter squared, or power density, PWR.
Since p and v are sinusoidal steady state waveforms, we know by
analogy with electrical power theory that the power density is

PWR = P rms V rms (6.31)

where Prms and Vrms are the root mean square magnitudes of the
sinusoidal pressure and velocity waveforms. In the case of a single
traveling wave of pressure in the medium we note that
Z0 =
P rms
V rms
which inserted into the above equation gives
P 2rms
PWR = Z0 (6.32)
where PWR is the power density in watts per meter squared
(W/m2).
The power reflected from a boundary such as is illustrated
in Figure 6.4 is calculated as follows: The incident pressure wave
in medium carries a power given by
 108

P 2rms(inc)
PWR inc = Z 01
and in the reflected wave, the pressure is reduced by a factor of the
reflection coefficient, R such that
(RPrms(inc) ) 2
PWRref = Z01
The ratio of these two equations, where we define the ratio of
reflected to incident power as GR, then yields
PWR ref
R = PWR inc = R2 (6.33a)
Expressed in decibels, the decibel ratio of reflected to incident
power, GdBR, is
PWR
dBR = 10 log PWR inc
ref
= 20 log R (6.33b)
____________________________________________________
EXAMPLE 6.2. A power density of 1.5 Watt per meter squared is
incident on a boundary of muscle and bone. The physical constants
for the bone are B = 1.85 g/cm3 and cB = 3360 m/s, and those for
muscle are M = 1.06 g/cm3 and cM = 1570 m/s. How much power
is reflected from the boundary?
SOLUTION
From Equation (6.30) we calculate

R = 1.85x3360−1.06x1570
1.85x3360+1.06x1570 = 0.578
Then Equation (6.33) gives GR = 0.5782 = 0.333. Thus the reflected
power is PWRR = .333x1.5 = 0.5 W.
____________________________________________________

Power Transmission Coefficient

The power transmitted, PWRtrans through the boundary illustrated

in Figure 6.4 is equal to the difference between the incident power,
PWRinc, and that reflected, PWRref . That is

PWR trans = PWR inc − PWR ref

(6.34)
 109

since the power into the boundary must equal that out. The power
transmission coefficient is define as

trans =
PWRtrans
(6.35)
PWRinc
From Equation (6.35) and (6.33) then we have

trans =1− ref = 1 − R2 (6.36a)

Expressed in decibels,

dBtrans = 10 log trans = 10 log(1 − R 2 ) ( 6.36b)

Power Attenuation

As ultrasound propagates through matter, the magnitude of the

wave attenuates due to several factors including friction between
moving particles, dispersion or spreading of the beam, scattering,
and spurious oscillations it may induce. To keep the mathematical
model simple enough to be useful in practical calculations, we
consider only friction. In a linear system this is proportional to
velocity, which in the case of traveling sinusoidal waves means it
is proportional to the rate of phase change. Thus the traveling wave
in Equation (6.10) becomes

p = P 0 e (− n x+j( x− t))
or

p = P0e− nx
e j( x− t)
The actual wave is the real part of this, or

p = P0e− nx
cos( x − t) (6.37)
where an is the attenuation in nepers per meter.
______________________________________________________
EXAMPLE 6.3. Program Equation (6.37) on a graphics calculator
and plot p(x) to illustrate how it attenuates. Take an = 0.1
neper/m, = 1 rad/m, = 1 rad/s, and P0 = 10 N/m2, and plot as a
function of x for t = 1s.
 110

SOLUTION Programing Equation (6.37) on a graphics calculator

results in the graph of Figure (6.5). Notice the loss in amplitude as
x increases.
____________________________________________________

According to Equation (6.32) the power at x = 0 in Figure 6.6

10
P 2rms
6.2
5 1.52 4.7
2.85 Z0
pressure, p (n/m2 )

where Z01 is the characteristic impedance of the medium. an1 is the

position, x (m)
neper attenuation of the pressure wave. So the power at point x1 is
FIGURE 6.5. Attenuation as a function of position
given by

2 e- 2an1x1
2
P rm P rm s
s
P 2rms −2
Z
n1 x 1
Z0 e
Z 0 0

Z 0 a n1

x1
The power attenuation at x1 relative to that at x = 0 is the ratio
0
FIGURE 6.6
of these two power levels,

att = e −2 n1 x 1
By the definition of decibel, this attenuation in decibels, GDB is
given by

= 10 log DB att=
− 2 n1 x 1 log e = 0.8686 n1 x 1

We may now define the decibel unit attenuation coefficient as dB

with units decibel per centimeter as,

dB = dB
x1 = 0.8686 n dB/cm

(6.38)
 111

where the units on a are neper per centimeter. Values for dB for
several common materials are given in Table 6.2.
Table 6.2. Typical ultrasonic dB
attenuation coefficient parameters

Material adB at 1 MHz Frequency

(dB/cm) dependence

Air 1.2 f 2

1.3
Blood 0.1 f

1.5
Bone 10 f

Fat 0.6 f

Lung 40 f 0.6

Notice the great variation in attenuation in different media:

Muscle 1.5 - 2.5 f
10 dB/cm in bone, 1.5 to 2.5 dB/cm in muscle at 1 MHz.
Water 0.002 f2 The frequency dependency of attenuation also varies. In
Source: Bronzino, Joseph D, Biomedical Engineering and Instrumentation, p
351.
fat, as you double frequency, attenuation doubles. In water
since it goes as f2 , so doubling
Tissue 1 Tissue 2
frequency quadruples the
GdB 2-1 =10 log ( PW R2 )
attenuation in dB/cm. As a rule of
PW R 1 PW R 2
PW R1 thumb approximation, the
attenuation of ultrasound in the
GdB4-1 =10 log ( PW R 4 )
whole body is taken as directly
PW R 1
PW R 3
PW R 5 proportional to frequency.
PW R4
GdB 4-3 =10 log ( PW R4 ) Power Relations at a Boundary
PW R3 GdB 3-2 =10 log ( PW R 3 )
PW R 2
0 It is important to be able to x1
x-axis
estimate power levels as an
ultrasonic wave travels through
GdB 4-1 = GdB 2-1 + GdB 3-2 + GdB 4-3 the body because the amount of
reflection allows us to distinguish
between tissue masses so that we
FIGURE 6.7. Power relations in a lossy m edium
can see the boundaries. To
analyze the situation we follow
Figure 6.7.
In this figure two lossy materials form a boundary at x1. A
pulse of power PWR1 at x =0 is directed at the boundary, and
experiences a loss in magnitude as it traverses tissue 1, and
acquires the value PWR2 incident on the boundary. A portion is
reflected, depending on power reflection coefficient, producing a
power level PWR3 . The pulse then travels back through tissue 1
and is attenuated to produce a power PWR4 . We need to know the
 112

PWR4/PWR1 in order to estimate the signal to noise ratio of a

receiver which detects PWR4 for example. The process of
calculation is as follows. Clearly,
PWR4 PWR2 PWR3 PWR4
PWR1 = PWR1 PWR2 PWR3
since canceling like terms on the right gives the ratio on the left.
To convert to decibels, we take 10 log of both sides as
PWR4 PWR2 PWR3 PWR4
10 log PWR1 = 10 log PWR1 PWR2 PWR3
Since the log of the products is equal to the sum of the logs we
write this as

PWR 4 PWR 2 PWR 3 PWR 4

10 log PWR 1 = 10 log PWR 1 + 10 log PWR 2 + 10 log PWR 3 (6.39)

In words, the total dB loss around a path is equal to the sum of the
dB losses over each segment of that path. Using the symbolism of
Equation (6.33b) we have
dB4−1 = dB2−1 + dB3−2 + dB4−3 (6.40)

_____________________________________________________
EXAMPLE 6.4. In Figure 6.7, a pulse of power, very short
compared to the transit time, heading toward the boundary is 2
W/m2 at x = 0. It travels to the boundary at x = 2 cm, a portion is
reflected, which travels back to x = 0. Tissue 1 is muscle (Z01 = 1.7
x 106 kg/m2-s , db = 2 dB/cm ) and Tissue 2 is bone (Z02 = 4 x
106 kg/m2-s, dB = 10 dB/cm ). How much power arrives back at x
= 0 after making the round trip.
SOLUTION: First we calculate the dB losses experienced by the
power pulse on its round trip.
Loss from x =0 to 2 cm in muscle
= -2dB/cm x 2 cm = -4 dB
To find the reflection loss at the boundary we compute the
reflection coefficient, R from Equation 6.29 as
(4x106 kg/m2 −s) − (1.7x106 kg/m2 −s)
R= (4x106 kg/m2 −s) + (1.7x106 kg/m2 −s) = 0.404
Thus the power reflection coefficient is by Equation (6.33a) =
0.4042 = 0.163. Then by Equation (6.33b), GdBR = 20 log 0.163 = -
15.75 dB. Then
Loss from x=2 to 0 cm in muscle is as above = -4 dB.
 113

Thus the total round trip loss is additive as shown in Equation

(6.40). So
Total round trip loss = -4 dB - 15.75 dB - 4 dB = -23.75 dB

To find the power we note that decibels are equivalent to 10 x the

log of the power ratio, so in Figure 6.7 we have
PWR 4
−23.75 dB=10 log PWR 1
Taking the antilog to find the power ration gives
PWR 4
PWR 1 = 10 −2.375 = 0.00422
Thus the return power, PWR4 = 0.00422 x 2 W/cm2 = 8.44
mW/cm2
_____________________________________________________
EXAMPLE 6.5. For the situation described in Example 6.4,
calculate the power launched into tissue 2 at the boundary x = 2
cm.
SOLUTION: First we calculate the dB losses experienced by the
power pulse on its trip to the boundary:.
Loss from x =0 to 2 cm in muscle = 2dB/cm x 2 cm = -4
dB.
To find the reflection loss at the boundary we compute the
reflection coefficient, R from Equation 6.29 as
(4x106 kg/m2 −s) − (1.7x106 kg/m2 −s)
R= (4x106 kg/m2 −s) + (1.7x106 kg/m2 −s) = 0.404
To find the loss the power experiences passing through the
boundary, we use Equation (6.36b)

dBtrans = 10 log(1 − R 2 ) = 10log(1 − 0.404 2 )

= −0.774 dB
The total loss from x = 0 to the right side of the boundary is
therefore -4 - 0.774 = -4.774 dB.
The to find the power level,
Figure 6.7 we have
PWR 5
−4.774 dB=10 log PWR 1
Taking the antilog to find the power ration gives
 114

Ultrasonic W ave

= 10 −0.477 = 0.333
PWR 5
PWR 1
Metal
Thus the transmitted power,
Electrode
PWR5 = 0.333 x 2 W/cm2 =
Mechanical 0.666 W/cm2.
Displacem ent ________________________
________________________
Electrical Piezoelectric
Exicitation Crystal ______

Ultrasonic Transducers
Metal
Electrode Ultrasound is usually
generated with piezoelectric
FIG URE 6.8. Ultrasound generated with sinusoidal exication
transducers. The typical
situation is illustrated in
Figure 6.8.

The transducer converts the electrical excitation applied to

the two metal plates into an ultrasonic wave setup by the resultant
vibrations in the crystal. A piezoelectric crystal has the property
that when a voltage is placed across it, the electrical forces on the
atoms are disturbed and cause the atoms to change their position.
The net result, in the case shown, is the surface of the crystal in
contact with the metal vibrates. If the excitation is a sinusoidal
wave, the mechanical displacement will also be sinusoidal. If the
metal electrode is in contact with a tissue for example, this
vibration will set up a propagating, longitudinal ultrasound wave
perpendicular to the electrode surface.
Piezoelectricity was first discovered in natural crystals like
quartz and tourmaline, quartz (silicon dioxide) being the best of
the two. Better piezoelectric materials such as barium titanate and
amonium dihydrogen phosphate can be fabricated. One which
produces relatively large mechanical displacements for a given
voltage, lead titanate zirconate, called PZT (a registered trade
name of Vernitron Ltd.) normally consists of grains in a random
crystallographic orientation. Their orientation can be made
systematic by application of an applied electric field, in a process
called poling. In the process, the temperature of the material is
raised above the Curie point, then slowly lowered so that the
 115

piezoelectric effect remains at room temperature.(6.3) Some

physical constants associated with common piezoelectric materials
are given in Table 6.3(6.4).

TABLE 6.3. Properties of some piezoelectric materials

Quartz, x-cut PZT-4 PZT-5A
(see note i) (see note ii) (see note ii)

Tensor Subscript 11 33 33

Transmitting Constant -12 -12 -12

d (m/V) 2.31x10 289x10 374x10

Receiving constant -2
-2 -2
g (v m/N) 5.78x10 2.61x10 2.48x10

Coupling coefficient
ke 0.1 0.70 0.71
Dielectric Constant -11 -11 -11
e T (F/m) 4x10 1150x10 1500x10
Wave velocity
c (m/s) 5740 4000 3780
Density, r
kg/m 2650 7500 7750
Characteristic Imped. 7 7 7
Z (kg/s m2 ) 1.52x10 3.0x10 2.93x10
Mechanical Q
> 25000 > 500 75
Curie Temp.
( deg C) 573 328 365
Table from P.N.T. Wells, p52. note(i) Data quoted from Cady, 1946) note (ii) Data extracted from Bulletin 66011/E,
Vernitron Ltd., Southamptonn, UK. PZT is a registered trade name of Vernitron Ltd. Southampton, UK.

We note that PZT has a hundred fold better transmitting constant

than quartz, and transmits more power per volt. It can also be
molded into almost any shape and can be polarized in what ever
direction desired, giving them extensive practical advantages over
quartz in transducer production. The lower mechanical Q means
that PZT will produce a broader bandwidth of ultrasound waves.

Impedance Matching

In order to transfer the maximum amount of energy from the

piezoelectric transducer into the tissue, the crystal or ceramic must
 116

Metal be placed in an
Gel
electrodes appropriate structure that
provides optimum
oscillation and impedance
matching. The basic
structure is illustrated in
Electrical
Figure 6.9.
l l To set up an
source 2 4
ultrasonic wave in the
Tissue piezoelectric transducer, a
sinusoidal voltage source
may be applied between
Backing Piezoelectric two metal electrodes. A
m aterial m aterial longitudinal oscillation
will be set up such that
Im pedance the transducer is one half
m atching wavelength ( /2) long.
transform er (An analogy is, suppose a
FIGURE 6.9. An ultrasound transducer guitarist plucks a string
and puts a finger on it. A
node appears there and the string will vibrate between that point
and the point where it is fixed to the guitar, one half wave length
away. In general resonators vibrate at frequencies such that their
length is an integer multiple of half wavelengths.) The transducer
is coupled to the tissue through a quarter wave transformer ( /4).
The quarter wavelength is chosen because such a transformer will
deliver maximum power to the tissue if its impedance is chosen
according to the formula.
ZT = ZINZOUT (6.41)
where ZT is the characteristic impedance of the transformer, ZIN, is
the characteristic impedance of the material at the input to the
transformer (in this case the transducer), and ZOUT is the
characteristic impedance of the material at the output of the
transformer (the tissue). This relationship for a quarter wave
transformer is derived from electrical transmission line theory, and
is valid because the voltage on the line is analogous to the sound
pressure, and the current on the line is analogous to the velocity of
the ultrasound. An example derivation is found in Kraus.(6.5)
A gel is needed between the transducer and the tissue to fill
up the air gaps that inevitably form. As you can see, from Table
 117

6.1, the characteristic impedance of air is very much smaller than

tissue, and so will introduce huge reflections that make for a bad
connection. The characteristic impedance of the gel is close to that
of tissue, making it suppress these reflections. The characteristic
impedance of the /4 transformer can be adjusted by using
different viscosities of oil encased in a rubber container. Thus a
transducer coupled to the skull bone would use oil of a different
viscosity from that coupled to muscle for maximum power
transfer. In general, notice from Tables 6.1 and 6.3 that the
characteristic impedance of the transducer is more that ten times
that of most body tissues, making impedance matching very
important for adequate power transfer. Other polymers are used to
make transducers for a variety of applications. In Exercise 6.7 we
show that a PZT-4 transducer can produce almost 10 mW/cm2 with
1 volt rms excitation, meaning it can be operated from low voltage
transistor circuits.
____________________________________________________
EXAMPLE 6. 6. Calculate the thickness of a PZT-4 transducer
operating at 1 megahertz.
SOLUTION: From Equation 6.17 and Table 6.3 we have

4 = 4cf = 4000
4x10 6 /s = 1 mm
m/s
which is the thickness of the transducer.
____________________________________________________
In practice, ultrasound transducers may be as small as 1 mm3.
The backing material in Figure 6.9, ideally, would reflect
all of the energy from the
Sweep piezoelectric device back toward
Gen the /4 transformer. However if
Pulse it does, ringing occurs, so an
Transm it Pulse Gen empirical compromise is reached
Incident
Return
X-ducer
in the choice of the backing
10 20 30 40 50 60
Depth (m m ) Reflected
material to suppress those
pulse oscillations without absorbing
too much energy.
A-Scope Display Lim iter

A-Scope Ultrasound Imaging

Rcvr
A simple block diagram of an
ultrasound imaging system is
given in Figure 6.10. Its
FIGURE 6.10. Basic ultrasound im aging block diagram
 118

operation begins with the production of a pulse of voltage in the

pulse generator, triggered by a timer which also begins the trace
across the A-scope. The transmit pulse appears as a deflection on
the left edge of the A-scope. The voltage across the transducer
sets up a pulse of ultrasound power which is delivered to the
subject, creating an incident power pulse that travels toward an
organ to be imaged. The reflected power returns to the transducer
where it is fed to the the limiter. The limiter limits the voltage level
so that the high level transmitter pulse (perhaps 100 dB above the
return signal) does not overload or burn out the receiver. The
receiver accepts signals in the 1 to 10 MHz range in typical
ultrasound equipment, and then feeds the return pulse to the
A-scope. This return pulse appears as a deflection on the A-scope
trace seen as the return pulse in Figure 6.10. The trace measures
the time elapsed from the transmit pulse to the return, which is the
time the ultrasound takes to travel from the transducer to the object
imaged and back.
The trace on the A-scope is calibrated in millimeters of
depth by converting the time of round trip travel to distance. The
whole body speed of ultrasound is taken as 1540 m/s. Thus 1
microsecond measured on the trace is equivalent to 1.540 mm
round trip, or half this in depth to the object imaged. (ie. 1
microsecond is equivalent to 0.770 mm depth). This type of
display may be used to find the depth of an object in the body. It is
similar to a range finder in a sonar on ships that determine the
distance to obstacles under water, or the depth to the ocean floor.
Controls on the receiver may be adjusted so that the receiver gain
increases in proportion the distance along the trace. This
compensates for attenuation the power pulse experiences in travel
and keeps the size of the reflection pips about the same for similar
boundaries. The A-scope gives information about the depth of
objects under the skin. To obtain a planar view of the object,
showing both the depth and breath of the object, a B-scope is used.
 119

B-Scope Imaging

A planar view of an organ

Depth
can be obtained by causing
the transducer to sweep
2 through an angle, as
1 3
illustrated in Figure 6.11. In
position 1, the ultrasound
3
pulse travels to the object
Organ and sends a reflection at the
outline 2
two boundary points where it
enters and leaves. These are
Origin 1
recorded on a screen as an
Organ increase in intensity of the
B-m ode display beam which makes a brighter
X-ducer
R Aston, 12/03 mark. The beam then rotates
FIGURE 6.11. A B-m ode im age toward position 2, giving
reflections which trace out
Vertical
trace
more of the boundary on the screen, proceeding
to postion 3, etc. This type of scope is widely
Back used in producing ultrasound images of fetuses.
wall
The transducer may be rotated mechanically
Depth (m m )

Heart
m otion to achieve this display, and such transducers are
X-ducer in use in the clinic. To achieve faster display
Front
sweeps, phased array transducers, explained
wall 0 1 2 3
sec below, are used. These rotate the sweep
Heart
electronically, so that one is not slowed down by
the inertia of a mechanical system, making it
R A ston, 12/03

possible achieve clear images of moving objects

FIGURE 6.12. An M -m ode im age such as the heart.

M-mode Display

Motion of an organ such as the heart can be imaged using the

M-mode on an ultrasound imager, as shown in Figure 6.12.
The vertical trace is proportional to the depth of the object imaged,
as in the A-scope, and the height of the echo is displayed by a
brightening of trace. The trace then moves at a speed much slower
than the transit time of the ultrasound pulse, so that it records
changes in depth as a function of time. In the case shown, heart
 120

valve motion appears as a deflection, showing its variation in the

depth as the heart beats.

Phased Array

The phased array approach , illustrated in Figure 6.13, allows one

to sweep a B-mode image at electronic speeds, and to focus the
ultrasound deep under the skin. The piezoelectric transducer
elements, which can be as small as one millimeter across, are
aligned in a row as illustrated, with a distance, D, separating them.
Consider the ultrasound at the point P at the end of a line OP,
which is very much longer than the length of the transducer.
(Figure 6.13 is not in scale.). The difference in the OP path length,
l, between adjacent transducer elements is related to D and the
angle, J, between the transducer array and OP. The expanded
view at the two elements shows a right triangle, for which the
following relationship holds,

cos = l
D (6.42)
Of the two paths shown, the lower one in this case is longer, but its
pulse leaves its element TD seconds sooner than the upper one.
Thus the two pulses arrive at point P at the same time when

l = c TD (6.43)
where c is the speed of sound in the medium. Combining
Equations (6.43) and (6.42) yields,
c TD
cos = D (6.44)
This has the effect of causing constructive interference, or pulse
addition at the point P. Similarly, the pulses from the other
transducer elements will arrive in phase at point P and
Transducer
add. Now if the time delay is changed to a new value, a
elem ents new angle J will be made. This process can be carried
TD
on at electronic speeds, making it possible to sweep a
P
B-mode display more rapidly than is possible by
physically rotating the transducer.
q

Delay line O

References
D
D
Input pulse

q
l

l
cos q =
D
FIGURE 6.13. Ultrasound Phased Array
 121

6.1. Wells, P.N.T. Biomedical Ultrasonics. New York: Academic

Press (1977) p 1.16.
6.2. Bronzino, Joseph D. Biomedical Engineering and
Instrumentation: Basic Concepts and Applications. Boston, MA:
PWS Engineering (1986) p 351.
6.3. Hennig, Ewald D. “Piezoelectric Sensors” in Encyclopedia of
Medical Devices and Instrumentation, ed. by John Webster, New
York: John Wiley & Sons, vol 4. (1988) pp2310-2319.
6.4. Wells, P.N.T. op.cit. (p 52).
6.5. Kraus, John D and K. R. Carver. Electromagnetics, Second
Edition. New York: McGraw Hill Book Co. (1973) p 509.

Exercises

Student end-of-chapter exercises and a solutions manual for instructors are available at ABC
engineering research, 33 Barney St. Wilkes-Barre, PA, U.S.A. 18702. ([email protected]).

.
 122

APPENDIX A

Magnetic Field of a Current Loop (1A)

According to the Biot-Savart Law, the magnetic field density

caused by a current, I, in an element of wire, dl, in a circular
current loop, shown in Figure A1 is
Idl sin
dB= 4 r2
(A1)
where dB is the magnetic field
element, is the angle between the
element of wire and the radian, r,
and is the permeability of the
medium. The direction of dB is
shown perpendicular to the radian.
Because of the symmetry of the
loop, the magnetic field exists only
as the component BZ along the axis
of the loop. BZ is given by

dBZ =dBcos =dB Rr

(A2)

The angle between the direction of

dl and the radian r is 90 o , making
the sin equal to 1 in Equation
(A1). The arc length dl is given by,
dl = R d
and because it in a right triangle, r
is,

r = R2 + z2 (A3)
Substituting these into Equation
(A2), yields
 123

IR 2
dB Z = 4 r3 d
Then substituting Equation (A3) yields,
IR 2
dB Z = 3 d
4 (R 2 +z 2 ) 2
To get the total field along the z-axis, we integrate with respect to

IR 2
B = ¶dB Z = ¶ 20 3 d
4 (R 2 +z 2 ) 2
Integrating we have

IR2 IR2
B= 3 [ ] 20 = 3
4 (R2 +z 2 ) 2 2(R2 +z 2 ) 2
(A4)
Which is the magnetic field along the z-axis of the loop. Off that
axis the field is more complicated.

Magnetic Field in a Solenoid (2A)

The magnetic field along the axis of a solenoid can be derived

from Equation (A4). If the coil has a large number of closely
packed turns, N, each coil can be considered a single element
containing a current
NIdz
L amps (A5)
as illustrated in Figure 1.9.
Equation (A4) then expresses an element of the magnetic field dB
made by the current in Statement (A5), which replaces the “I” of
Equation (A4). Thus we have
NIR 2
dB = 3 dz
2L(R 2 +z 2 ) 2
The total field along the axis of the solenoid is then found by
integrating the contributions of all elemental loops from -L/2 to
L/2 along the length of the coil, L. That is,
 124

B = − L2 ¶ dB
L
2

Substituting Equation (1.5) gives

NIR2 L/2
B= 2L ¶ −L/2
dz
3 (A6)
(R2+z2) 2
From standard integral tables we can find the relationship

¶ dz
3 =R2 R2+z2
z
(R +z ) 2
2 2
Evaluating this between the limits -L/2 to +L/2 yields,

L
− L2
2
2
− 2
= R2 2L
4R2+L2
R2 R2+ L4 R2 R2 + L4
This result substituted into Equation (A6) gives
NI
B= 4R 2 +L 2
(A7)

where R is the radius of the coil, L the length, N the number of

turns, I the current and is the permeability. Equation (A7)
calculates the magnetic field flux at the center of the solenoid.

Reference Appendix A

A1. Kraus, John D., Keith R. Carver. Electromagnetics 2nd

Edition. New York: McGraw Hill, Inc. (1973) pp 147-148.

A2. ibid. pp 155-157.

 125

APPENDIX B FOURIER TRANSFORM

The fourier transform arises from the Fourier Theorem. This

theorem states that a periodic function can be expressed as a series
of harmonically related sinusoidal signals. Mathematically this is
stated as:

N N
f( t) = a 0 + a n cos n 0t + b n s in n 0t
n=1 n=1
(B1)
Here n is an integer, N is the upper limit on the integers n, w0 is
the fundamental radian frequency, t is time and a n and bn are
unknown constants given by the following relationships.
T
a0 = 1
T ¶ f( t) d t
0

T
an = 2
T ¶ f( t) c o s ( n 0 t) d t
0

T
bn = 2
T ¶ f( t) s in ( n 0 t) d t
0
where T is the period of the periodic signal. Thus at any particular
time, t, f is a function of
w = nw0
which are harmonically related radian frequencies.
If the period of T is taken to be infinite ( ∞) one can show
the Fourier Theorem implies that (B1)
∞
f( t) = 2
1
¶ F( )e j t
d (B2)
−∞
∞
F( ) = ¶ f( t) e − j t
dt (B3)
−∞
Here F( ) is the fourier transform of f(t). Equation (B2) is a
function of time alone, since the frequency is a running variable.
We therefore say the function is in the time domain. Equation (B3)
 126

expresses that function in the frequency domain, since it is a

function of frequency alone. Equations (B2) and (B3) form a
fourier transform pair, (B3) being the fourier transform, and (B2)
being the inverse fourier transform.

The Fourier Transform of a Pulse

The pulse in Figure B1 is described mathematically as:

TD
v(t) = A when t <
2
TD
= 0 when t >
2
This inserted in Equation (B3) gives,
TD
2

V( ) = A ¶
−T D
e −j t dt (B4)

2
To solve this we need to recall two standard mathematical
identities:

¶e ax
dx = e ax
a (B5)
and
e jx −e −jx
2j = sin x (B6)
Using Equation (B5) in (B4) we have
e −j t
V( ) = A −j evaluated between limits -T D/2
and TD/2 gives,
TD TD
−j −j
V( ) = A( e −j − e
)
2 2
−j
Equation (1.7) into this gives
TD
sin
V( ) = 2A 2

And in terms of frequency f = /2 , we this becomes,

 127

A sin( f TD)
V(f ) = f (B7)
______________________________________________________
EXAMPLE B1
A voltage pulse of amplitude, A, 3.14 volts, and duration, T D = 2
sec. Use a graphics calculator to plot the fourier transform of this
pulse.
SOLUTION
On the TI-85 calculator set the Equation (B7) in the GRAPH
mode, keystroke

y1 = (A*sin(p*x*TD))/(p*x)

where x is the running variable for frequency on the calculator.

Store the values:
3.14 STO A
2 STO TD
A resulting plot of the fourier transform V(f) is shown at the
bottom of Figure B1.
______________________________________________________

The Inverse Fourier Transform of a Pulse

It is important to know what function of time will produce a band

of frequencies of constant amplitude in the frequency domain in
the operation of an MRI. This question is answered if we know
what function in the time domain corresponds to the following in
the frequency domain:

V(f ) = 0 when f > f0

= B when f < f0
(B8)
where f is the frequency in Hertz (Hz) and B is a real constant. To
find the answer we use the inverse Fourier transform, Equation
(B2), and integrating as,
 128

v(t) = 2
1
¶ Be j t
d = B ej t
t 2j − 0
0

− 0
Here 0 is finite, so v(t) only approximates the inverse
Fourier transform, since in Equation (B2) the limits are
infinite. The larger 0 the better the approximation will
be. Evaluating the limits gives,
B ej 0 t −e −j 0 t B sin( 0 t)
v(t) = t 2j = t
(B9)
This written in terms of frequency f 0 = 0/2 is
sin(2 f 0 t)
v(t) = B t (B10)
____________________________________
EXAMPLE B2. A radio frequency (RF) voltage is
modulated by a pulse of shape given in Equation (B10)
of 10 Hz = f0 and B = . Find the corresponding fourier
transform, V(f) in the frequency domain, and plot the
two functions.
SOLUTION
To plot v(t) of Equation (B10) on a graphics calculator such as the
TI-85, 86 or 89 operated in the y = editor, keystroke
Y2=(B*sin(2 *f0*x))/( *x) (B11)
where x is a running variable representing time in the calculator.
In this case store the following quantities:
STO B
10 STO f0
A plot of Equation (B11) is given at the top of Figure B2.

Three Dimensional Fourier Transform(B2)

The Fourier transform extends into three dimensions as

follows:
∞ ∞ ∞
→ → − 2 i →k $ →r →
S( k) = ¶ ¶ ¶ s( r ) e d r
−∞ −∞ −∞
(B12)
The inverse Fourier transform is:
 129

∞ ∞ ∞
→ → → → →
s( r) = ¶ ¶ ¶ S( k ) e + 2 ik $ r
d k
−∞ −∞ −∞
(B13)
where i is the imaginary number operator and the vector

→
multiplication is a dot product.

r = x$x+y$y+z $z (B14)

→
is a spatial vector, and

k = x $ kx + y $ ky + z $ kz (B15)
is a frequency vector, as expressed in rectangular coordinates. In
one case r may have the dimension meters (m), and k the
dimension m-1. In another case r may represent time (t) and k have
the dimension t-1.

REFERENCES

B1. Papoulis, Athanasios. The Fourier Integral and its

Applications. New York: McGraw-Hill Book Company, Inc.
(1962) p7.

B2. Bracewell, R.N. The Fourier Integral and Applications, First

Edition. New York: McGraw Hill (1978) p2
 130

APPENDIX C 90O PHASE SHIFT CIRCUIT

Analyzing the circuit in Figure C1,

1
VOUT
=A = AV 1+j21 fRC
j2 fC
VIN V
R+ j2 1fC
Under the condition

2 fRC p 1 (C1)
this becomes
V OUT
V IN = AV 2 fRC ø − 90
1 o
Then if

A V = −2 fRC
(C2)

V OUT = V IN ø90 o (C3)

Thus the output leads the input voltage by 90 o.

 131

APPENDIX D

MRI FAST SCAN TECHNIQUE

One method of decreasing the scan time over the Spin Echo
sequence is to eliminate the 180 degree pulse and replace the 90
degree excitation pulse with one that causes the tilt angle of the
magnetization, , to be less to be less than 90 degrees. That this
process works is indicated mathematically as follows: The Bloch
equation, which gives the relationship between tissue
magnetization, M and the applied magnetic field strength, B is
modified for the existence of an inital magnetization caused when
TR is less than the time constant of the FID as follows:

→ → → Mx a x +My a y (M z −M0z )a z
dM
dt = M% B − T2 − T1 (D1)
Where M0z is the magnetization in the z-direction at t = 0. The
solution to this equation is (D1)
M z (t) = M 0 (1 − e −t/T 1 ) + M z0 e −t/T1
(D2a)
Mx (t) = M0xy cos 0t e −t/T2
(D2b)
My (t) = −M0xy sin t e −t/T2 (D2c)
The z-component is called the longitudinal component of the
magnitization because it follows the MRI core axis, and the x and
y-components are the transverse component. For the fast scan, the
z-component, Mz (t), is measured by the RF loop and the x and y
components are ignored. Following the derivation given by Scott
Reeder(D2), M bn represents the magnetization just before the nth RF
excitation pulse , and M an represents the magnitization just after.
After the first pulse the longitudinal magnetization is:
M a1 = M b1 cos = M 0 cos (D3)
Where is the tip angle of the induced magnitization, as set by the
size of the magnitization pulse given by Equation (1.4), and M 0 is
the initial logitudinal magnitization. The next pulse occurs after a
time TR and Equation (D2a) becomes:
 132

M b2 = M0 − TR
1 − e T1 a − TR
+ M 1 e T1
(D4)
Substituting the above equation yields,
M b2 = M 0 1 − e − T1 + M b1 cos e − T1
TR TR

(D5)
Likewise after the third pulse,
M b3 = M 0 1 − e − T1 + M b2 cos e − T1
TR TR

(D6)
Again after the fourth pulse,
M b4 = M0 − TR
1 − e T1 + M b3 cos − TR
e T1
(D7)
Substutiting Equations (D6) and (D5) into (D7) gives,

M b4 = M 0 (1 − e − T1 ) + M 0 (1 − e − T1 ) + M 0 (1 − e − T1 ) + M 0 cos e − T1 cos e − T1 cos e − T1

TR TR TR TR TR TR

(D8)
This rearranged is,

M b4 = M 0 (1 − e − T1 ) + M 0 (1 − e − T1 ) cos e − T1 + M 0 (1 − e − T1 ) cos 2 e − + M 0 cos 3 e −

TR TR TR TR 2TR 4TR
T1 T1

Rearranging again gives,

= M 0 (1 − e − TR
) 1 + cos e − T1 + cos 2 e − T1 + M 0 cos 3 e − T1
TR 2TR 4TR
M b4 T1 (D9)
st
Extending this in general, the magnitization before the n+1 pulse, where n is a
summing integer, becomes,
n+1 i n
M bn+1 = M 0 1 − e − T1 cos e − T1 + M 0 cos e − T1
TR TR TR
(D10)
i=0
Where i is a summing integer. A closed form for this is given by
the geometric series equation,(D3)
N2 N 1 − N 2 +1
k = 1− , N2 m N1, >1
k=N 1
(D11)
Applying this to the summation in Equation (D10) gives,
TR n
1− cos e − T1 n
M bn+1 = M 0 1 − e − TR
+ M 0 cos e − T1
TR
T1
TR D12)
1−cos e − T1
For large values of n, the second term of this equation can be
neglected as
 133

(cos e − T1 ) n approaches zero, and we have,

TR

M bn+1 = M 0 1 − e − T1
TR 1
TR (D13)
1−cos e − T1

Thus we have the magnitude of the magnitization vector, which is

tilted at degrees. The magnitization projected into the x-y plane,
M0xy is proportional to the sin , and it will decay with a time
constant T2 up to the time it is measured TE. So from Equation
(D13) we have

− TE&
= M0 1 − e − TR sin e T2
M 0xy T1
TR (D14)
1−cos e − T1

Where T2* includes modifications of T2 due to equipment induced

factors.
A plot of Equation (D14) for various values of TR << T1 shows a
considerable measurable magnitization, thus yielding faster scans.
 134

REFERENCES APPENDIX D

D1. Wright, Steven, “Nuclear Magnetic Resonance and Magnetic

Resonance Imaging”, Chapter 16 in Introduction to Biomedical
Engineering Ed. By J. Enderle, S. Blanchard & J. Bronzino. San
Diego: Academic Press, p. 801 (2000).

D2 Reeder, Scott R., Development of High Speed, High

Resollution Magnetic Resonance Imaging and Tagging Techniques
Baltimore, MD: Johns Hopkins University, MS Thesis pp31-34
(1993). Accessed on the WEB, October 2004 at:
Http://www.mri.jhu.edu/~sreeder/pdf_papers/mse_t~va.pdf

D3 Oppenheim, Alan V. & Ronald W. Shafer. Discrete Time

Signal Processing. Englewood, NJ: Prentice Hall (1982).
 135

Appendix E Fan Beam Projection

R

TUBE
y

FOV

r
-F

x
(x,y)
AN
AN
2
G
LE

TORS
DETEC
+F
AN
ANG
2 LE
FIGURE E1

Given, the constructed, empirically measured

Fourier Transform, P(k, ), we can form the 2D inverse Fourier
Transform (FT) as,
∞
f(x, y) = ¶ ¶ P(k, ) k e i2 k(x cos + y sin )
dk d (E1)
0 −∞
The Fan Beam Projection formula given in Equation (E1) is
modified to be calculable from easily measurable machine
parameters shown in Figure E1: , the angle between the center
line of the field of view FOV and the ray from the tube to a
particular detector; , the angle between the center line of the FOV
and the y-axis; and R, the distance from the tube to the center of
the FOV. The derivation of this modified formula is as follows:
We break Equation (E1) into two parts as (E1)

f(x, y) = ¶ p & (r, ) d (E2)

0
where
r = x cos + y sin (E3)
And where
∞
p & (r, ) = ¶ P(k, ) k e i2 k(r) dk (E4)
−∞
Equation (E4) is the one dimensional transform of the product of
two FTs; P(k, ) and ækæ.
By the projection theorem, (E2)
 136

∞
P(k, ) = ¶ p(r, )e −2 ikr dr (E5)
−∞
This says P(k, ) is the one dimensional FT of p(r, ). Thus, p(r, ) is
the inverse FT of P(k, ). We define q(r) as
∞
q(r) = ¶ k e i2 kr dk (E6)
−∞
q(r) is the inverse FT of xkx.
The Convolution Theorem states(E3)
Æ −1 [Æ 1 p(r, * k ] = p(r, ) W Æ −1 k (E7)
Which reads in words: The inverse Fourier Transform of the
Fourier Transform of p times the magnitude of k equals the inverse
Fourier Transform of the Fourier Transform of p convoluted with
the inverse Fourier Transform of the magnitude of k.
Thus we may form
∞
p & (r, ) = ¶ p(r ∏ , ) q(r − r ∏ ) dr ∏ (E8)
−∞
where r is given by Equation (E3).
Substitution of (E8) into Equation (E2) gives
∞
f(x, y) = ¶ ¶ p(r ∏ , ) q(r − r ∏ ) dr ∏ d (E9)
0 −∞
Now we change the limits on to be from 0 to 2 and take half the
integral. Then we limit the integration on r∏ to the field of view
(FOV). So Equation (E9) becomes
FOV
2 2
f(x, y) = 1
2
¶ ¶ p(r ∏ , ) q(x cos + y sin − r ∏ ) dr ∏ d (E10)
0 −FOV
2
Also, from Figure E1, we observe new coordinates
= +
And r = R sin (E11)
Substituting this into Equation (E10) gives
Fan angle
2 2
f(x, y) = 1
2
¶ ¶ p( , ) q(x cos( + ) + y sin( + ) − R sin ) R cos d d (E12)
0 −Fan angle
2
Where q is given by Equation (E6). Finite limits are set on
Equation (E6) by the Ram-Lax filter,(E4)
 137

k MAX = 1
ßs
Equation (E6) becomes
k MAX
q(r) = ¶ k e i2 kr
dk (E13)
−k MAX
where r = x cos + y sin . Calculation of Equations (E13) and
(E12) produce f(x,y) which can be coded in a gray scale to make
the image.(E5)

References Appendix E

E1. Swindell, W., & S. Webb. “X-ray Transmission Computed

Tomography, Chapter 4” in The Physics of Medical Imaging ed.
By S. Webb, Philadelphia: Institute of Physics Publishing (1988)
p. 113.
E2. Suetens, Paul. Fundamentals of Medical Imaging Cambridge:
Cambridge University Press (2002) pp 75-76.
E3. Macovski, A., Medical Imaging Systems, Englewood Cliffs,
NJ: Prentice Hall (1983) p 127.
E4. Ramachandran, G.N., & Lakshminarayanan, A. V., (1971)
Proc. National Academy of Science, USA, vol 68, pp 2236-2240.
E5. Suetens, op cit p80.
 138

ABOUT THE AUTHOR

Richard Aston has published three text books, two in medical

instrumentation: Principles of Biomedical Instrumentation and
Measurement published by Merrill Publishing Company in 1990,
and Medical Instrumentation for Nurses and Allied Health-Care
Professionals with Katherine K. Brown, published by Jones and
Bartlett, Inc. in 1994. He taught medical instrumentation in several
colleges over the past 20 years, most recently at East Tennessee
State University in the biomedical engineering technology program
for 10 years. He taught a medical imaging equipment course to
undergraduate seniors about ten times. He was granted a Ph.D. by
The Ohio State University in 1969, and is a registered professional
engineer (P.E.) in Pennsylvania.

The book Principles of Biomedical Instrumentation and

Measurement by Richard Aston is available from Prentice Hall at
their web site:

Http://vig.prenhall.com/catalog/academic/product/0,1144,0675209439,00.html

ISBN: 0-675-20943-9