HET408 Medical Imaging

An Introduction to Projection Radiography

The X-ray Spectrum

X-rays form part of the electromagnetic spectrum extending from

wavelength() : 109 6 1012 m frequency(f ) : 3 1017 5 1019 Hz The principle physical processes upon which X-ray generation rest are Thermionic Emission Bremsstrahlung and Characteristic Emission When an electromagnetic wave interacts with a charged particle the amounts of energy and momentum which are exchanged in the process are those corresponding to a photon.
Frequency Hz
10 22 10 10
21

Photon Energy eV
gamma rays 10
7

Wavelength m

10 -13 10 -12 10 -11 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 0 10 1 10

2

10 20
19

10 6 X-rays 10 5 10 4 10 3 ultraviolet visible infrared 10 2 10 1 10 0 10 -1 10 -2 10 -3 microwave 10 -4 10 -5 TV/FM Standard broadcast radiofrequency 10

-6

10 18 10 17 10 16 10 15 10 14 10 13 10 12 10 10 10 10 10 10 10 10 10 10
11 10 9 8 7 6 5 4 3 2

10 -7 10 -8 10 -9 10 -10 10 -11 10
-12

10 3 10 4 10 5 10 6

Figure 1: The Electromagnetic Spectrum

Thermionic Emission

The energy band corresponding to the uppermost atomic shell, occupied by the valence electrons, primarily determines the bulk electrical properties. If the uppermost band is not completely full it is called the conduction band, if it is full it is called the valence band and the empty band just above it is called the conduction band. If the total number of electrons in the conduction band is less than the total number of energy levels available in the band, then electrons will occupy all energy states up to a maximum energy called the Fermi Energy, F , if the metal is in its ground state. The uppermost electrons in the conduction band can be thermally excited. The states occupied by thermally excited electrons fall in an energy region of the order of 20 kT above F . At high temperatures, T , the occupation of electronic states extends over energies well above F . To extract an electron from a metal it is necessary to give a conduction band electron at least the energy e where e = elementary electronic charge = work function and is the energy required to extract an electron from the highest occupied energy level. Thus at high temperatures some electrons will escape from the metal to form an electron gas around the solids (metal/lament) surface. Based on the Fermi-Dirac distribution it can be shown that the thermo-electric current density, j, coming form the surface of a solid is given approximately by the RichardsonDushman equation, 4me 2 2 k T exp[e/kT ] h3 (1)

j= where me = electron rest mass T = temperature in Kelvin

(2)

k = Boltzmanns constant ( 1.38 1023 JK 1 ) 3

PSfrag replacements 20 kT

dn dE 0

Emax

Figure 2: Diagrammatic illustration of the energy states of free electrons in a solid at absolute zero (dashed line) and at 20 kT above absolute zero (solid line).

Thermionic emission can be implemented by passing an electric current through a piece of wire (lament) with a very high melting point e.g tungsten. In the production of X-rays this resultant electron gas is accelerated through an electric eld, E. The kinetic energy acquired after an electron has traveled a distance x is 1 me v 2 = E e x 2 or Energy = V e (4)

(3)

approx. 20kT

E max

Figure 3: The distribution of free electrons between energy levels in the conduction band.

This last expression motivates the denition of the electron volt as a unit of energy. 1 electron volt is equal to the work done on a particle of charge e when it moves through a potential dierence of 1 V. Thus 1 eV is

eV

= (1.6021 1019 C)(1 V ) = 1.6021 1019 J

Bremsstrahlung and Characteristic X-ray emission

X-rays are produced by the impact of fast electrons against the anode material of an X-ray tube 1 . The emitted X-ray spectrum will be composed of two main components dependent on the energy of the incident photons; Bremsstrahlung or deceleration radiation Characteristic X-ray emission For large anode potentials the emitted X-ray spectrum will be dominated by energy peaks corresponding to characteristic emission.

Bremsstrahlung

It is well known that an accelerated charge radiates electro-magnetic energy. The rate of energy radiation by a charge, q, moving with a velocity v and acceleration a, when v c is dE q 2 a2 = dt 6 0 c3 where = permittivity of free space

(5)

c = velocity of light If a particle is decelerated instead of being accelerated the previous equation still holds. Thus when a fast charge such as an electron hits a target and is stopped a substantial part of its
Most of the energy absorbed from the electrons will appear in the form of heat. Only a small proportion, less than 1%, will appear in the form of X-rays
1

6 X-ray intensity (arbitrary units) 5 4 3 2

25 kV continuous radiation (bremsstrahlung) characteristic radiation

1 0 0 10 20 30 wavelength (nm)

20 kV 15 kV 10 kV 5 kV

Figure 4: The intensity distribution for the X-ray spectrum of Molybdenum as a function of the applied voltage. Note the appearance of the K-series as excitation potential is increased. The K-series excitation potential is 20.1 kV .

total energy is released as electro-magnetic radiation. This is called deceleration radiation or bremsstrahlung. The energy of an electron may be radiated o as the result of successive collisions such that several photons are produced. the result of a single collision such that just one photon is produced. Photons emitted by just one electron-anode collision will be the most energetic and thus have the shortest wave-length. Thus the wavelengths of the X-rays so produced will be equal to or longer than a threshold wavelength satisfying the relationship

E = h hc = Thus

hc Ve 6

(6.6 1034 Js)(2.99 108 m s1 ) V 1.6021 1019 C 1.24 106 m = V = Example If electrons are accelerated by a potential dierence of 18 kV , the minimum wavelength of X-rays produced is

min =

1.24 106 1.8 104 = 6.9 1011 m

Characteristic X-ray Emission

Electrons lling the inner complete electron shells constitute the core or kernel. The binding energy of the kernel electrons is much higher than that of the valence or conduction electrons. These kernel electrons remain practically undisturbed in most of the processes in which the atom participates. However if the energy of the incident electrons is suciently large it is possible to knock one of the electrons out of the kernel. For example when the energy of incident electrons is increased hitting a Molybdenum (Mo) target certain well dened peaks appear in the emitted X-ray spectrum. These peaks are called characteristic X-rays. The wavelengths of these peaks are independent of the applied voltage, whereas the relative amplitude of these peaks is a function of the applied voltage. The dynamical state of each electron can be described by four quantum numbers n = energy l = angular momentum ml = quantization of orientation of angular momentum ms = electron spin States with n = 1, 2, 3, 4 constitute the K,L,M,N,.. shells. Consider that an ejected electron is from the K-shell. When such a K-electron is removed and empty state (or hole) is left in the K-shell. Another electron in a higher energy level of the kernel (or even a valence or free electron) may fall into the vacant state in the K-shell. The radiation emitted by the electron falling into this vacant state lies in the X-ray region of the 7

spectrum. Such an electron may have proceeded from the L,M,N,... etc shells and thus a series of X-ray lines may be produced designated as K , K , K ... etc. If the vacant state produced by electron ejection is in the L-shell then transitions may only arise from the M, N, ... etc shells.
n infinity 4 (N-shell)

3 (M-shell)

2 (L-shell)

L L
L-series

L absorption edge

1 (K-shell)

K K K

K-series

Figure 5: X-ray transitions in an atom with atomic number Z 36.

In many cases the photon emitted in an X-ray transition is absorbed by another electron within the same atom, which is therefore ejected as a result of an internal photo-electric eect. This process of the internal conversion of X-rays into photo-electrons is called the Auger eect and the emitted photo-electrons are called auger electrons.

Interactions of EM radiation with tissue

When a beam of electro-magnetic radiation passes through an object the energy of this beam is gradually absorbed by various processes photo-chemical reactions atomic photo-electric eect (or photo-ionization) Compton scattering pair production 8

K absorption edge

 photo-nuclear reactions In the energy range used in diagnostic radiology (17 - 150 keV ) the important photon interactions with tissue are the photo-electric eect and Compton scattering.

Photoionization

When an incident photon has enough energy, its absorption by an molecule or an atom may result in the ejection of an electron i.e A + h = A+ + e (6)

This process is called photoionization or the atomic photo-electric eect. Thus when a beam of ultraviolet, X- or - radiation passes through matter, it produces ionization along its path. Th energy required to extract an electron from an atom or a molecule is called the ionization potential, I. The kinetic energy of the ejected electron is given by Ek = h I Listed below are some typical values for the ionization potentials for some molecules Molecule H2 O CO2 O2 C9 H10 O3 Ionization Potential eV 12.6 13.8 12.1 8.4 (7)

Table 1: Ionization potentials for some common organic molecules.

A photo-electric interaction is followed by the ejection of a photo-electron, one or more characteristic X-rays and Auger electrons. The highest energy electron that will be emitted in this process will be of the order of 150 keV and would have a range in water typically of 0.03 cm. Thus for most purposes the electrons can be considered as being locally absorbed. These emitted electrons will greatly contribute to the patient dose and to the energy absorbed in the image receptor.

Compton Scattering

When the energy of a photon is much larger than the binding energy of the electron in an atom or a molecule, the electron can be considered free. In this case Compton scattering is a more probable process than the photo-electric eect. Analysis of electromagnetic radiation that has passed through a region in which free electrons are present shows in addition to the incident radiation another radiation of dierent frequency. This new radiation is interpreted as the radiation scattered by free electrons. The frequency of the scattered radiation is smaller than the frequency of the incident radiation. Accordingly the wavelength of the scattered radiation is longer than the wavelength of the incident radiation. This interesting eect is called the Compton eect.

The Attenuation of X-rays

Given that I0 is the intensity of radiation before it enters the substance, its intensity after it has traversed a thickness x of the substance is given by the familiar equation I = I0 e x (8)

where (in say cm1 ) is a quantity characteristic of each substance and each process and is variously called the coecient of linear absorption, coecient of linear attenuation or the macroscopic cross-section. For each substance there is one macroscopic cross-section for each possible process that we have outlined. The total cross-section of a substance is the sum of all partial cross-sections. These cross-sections will also be a function of the energy of the incident photons. For the two important attenuating processes occurring in diagnostic X-ray we have Attenuation Process Attenuation incident energy(E) atomic number(Z) E 3 > E 3 Z4 Z

Photoionization Compton Scattering

Table 2: Attenuation processes as a function of incident electromagnetic energy and atomic number

This table illustrates that scattering will will be less important in providing contrast between tissues with diering average Z. For soft tissue the photo-electric cross-section is larger than the scatter cross-section for energies 25 keV . 10

10 4
K absorption edge

10 3
I(Pb)

(m-1 )

II(Pb)

10 2
Al II(Al)

Pb III(Pb)

10 1

I(Al) III(Al)

10 0 10 -2 10 -1 10 0 10 1 10 2

photon energy (MeV)

Figure 6: Macroscopic cross sections for the absorption of photons in aluminium (Al) and lead (Pb). Curves labelled I are the partial cross sections due to photoelectric absorption. Curves labelled II are the partial cross sections due to Compton scattering. Those curves labelled III are partial cross sections due to pair production (not important for the energies used in diagnostic radiography).

10

Image Formation

The projected radiographic image can be considered as the distribution of absorbed energy. In what follows we will assume the following a mono-chromatic X-ray source emits photons of energy E. the X-ray source is suciently far away from the patient such that the photon beam can be considered parallel to the normal of the the surface of the image receptor. each photon reacts with the image substrate locally, the response of this imaging substrate being linear such that the image so formed can be considered as the distribution of the absorbed energy. If N photons are incident at the surface of the image receptor (or substrate) and I(x, y) dx dy is the energy absorbed in the image receptor plane dx dy then we can write the following expression 11

1.0

incident

transmitted

relative number of photons

0.8

0.6

0.4

0.2

20

40

60

80

100

20

40

60

80

100

(a)

Energy (keV)

(b)

Figure 7: X-ray spectra for a tungsten anode (100 kV accelerating potential with 2.5 mm aluminium added). (a) X-ray spectra before and (b) after attenuation by 18.5 cm soft tissue plus 1.5 cm bone. Note that the lower energy components have been attenuated more heavily than the higher energy components after passage through the human tissue. Thus the average photon energy increases and the beam becomes harder. Such beam hardening can be troublesome in the context of tomographic image reconstruction as the exponential law of linear attenuation no longer holds.(data based on Birch R, Marshall M and Andran G M (1979). Catalog of Spectral Data for Diagnostic X-rays. Hospital Physicists Association, London.).

I(x, y) =

(E, 0)N E exp( +

(x, y, z) dz)

(Es , )Es S(x, y, Es , ) d dEs

= primary + scatter where S(x, y, Es , ) is the number of scattered photons in the energy range Es to Es + dEs in the solid angle range to + d. This is known as the scatter function. (E, ) is the energy absorption eciency coecient of the image substrate as a function of photon energy and angle . (x, y, z) is the linear attenuation (made up of all the partial cross-sections). The scatter function will in general be a complicated function of position and distribution of attenuation within the tissue. For this reason it is often considered as a slowly varying function of x and y. It can be shown that 12

10 linear attenuation coefficient (cm -1 )

Compton scattering

10

-1 photoelectric

10

-2

50

100

150

photon energy (keV)

Figure 8: The variation of the partial cross sections (photo-electric and Compton) with incident photon energy for soft tissue.

I(x, y) = N (E, 0)E exp( = N (E, 0)E exp( where

(x, y, z) dz) + S (E)E (x, y, z) dz)(1 + R)

(9) (10)

R=

scattered primary

(11)

The rst equation essentially says that we consider the scatter proportional to the energy of the emitted X-rays. i.e as we increase the energy of the source X-rays so does the amount of scatter.

10.1

Contrast and Unsharpness

Referring to the above gure 9 we can dene contrast as (I1 I2 ) I1

C=

(12)

where I1 and I2 give the energy absorbed per unit area. Using the previous denitions for I(x, y) it can be shown for the simple model of contrast that the contrast, C, is given by

13

2 PSfrag replacements

I2

I1

Figure 9: A simple model for the denition and estimation of image contrast in planar radiography.

C = {1 exp[(2 1 )x]}/(1 + R) Thus from the above equation we can see that the factors aecting contrast will be the thickness of the target (i.e x).

(13)

the dierence in the linear attenuation coecients between dierent structures that we would like to visualize. the scatter-to-primary ratio R i.e a greater proportion of scatter will act to degrade image contrast. The linear attenuation coecients will be a function of the incident photon energy. The scatterto-primary ratio will also be a function of the incident photon energy. Thus we expect contrast to vary with the mono-chromatic energy of the incident X-rays. Increasing the energy of the emitted X-ray photons will result in reduced linear attenuation coecients and hence reduced patient dose. However increased source photon energy will result in reduced contrast. Thus the energy of the X-ray spectrum chosen for radiographical procedures will be a compromise between patient does and contrast 2
Problem 1: By using equations (12) and (9) (why not equation 10 ?) and referring to gure 9 derive equation (13)
an aluminium plate interposed between the X-ray source and the patient is often used to lter out the low-energy component of the emitted X-ray spectrum. As the low energy components of the spectrum will be heavily attenuated and thus contribute little to image formation the eect of such a lter is to remove the low energy component that would otherwise unacceptably contribute to patient dose.
2

14

10.2

Relationships between noise and dose

Even during a simple plain-lm radiogram there will be sources of noise that will conspire to degrade contrast and image quality. The contributions to this noise will in general arise from two major sources quantum noise due to statistical uctuations in the number of photons detected per unit area by the imaging substrate. uctuations due to spatial and temporal inhomogeneities in the properties of the image receptor (substrate) and display system Quantum noise, because it is a statistical process, can be reduced by increasing the number of incident photons. However this will be at the expense of an increased dose to the patient. Therefore a central consideration is What surface dose is required to achieve a contrast C over an area A against a background noise arising from the eects of quantum noise ? It can be shown that the minimum surface dose, Sd , required to dierentiate two areas with dierent linear attenuation coecients is Sd = N E t (14)

where t is the mass energy absorption coecient for tissue ( 0.004 m2 kg 1 ). t = En / where En is the tissue mass energy cross-section or linear attenuation coecient and is the tissue density. E the photon energy and N , the number of incident photons per unit area
2 N = kt (1 + R) exp(1 X)/[ (E, 0)(2 1 )2 x4 ]

(15)

The constant kt is the signal-to-noise ratio for which an object just becomes detectable. Empirical evidence suggests this value is 5. From this equation note the following the minimum dose required to visualize an object (against a background of 1 ) increases as the inverse fourth power of the size of the object and as the inverse second power of the dierence in the linear attenuation coecients. i.e larger doses will be required to visualize thinner objects in the image plane thus ... for a xed dose and contrast there will be a minimum object size that can be visualized. low contrast resolution will vary with object size.

15

Example 1: Calculate the number of incident photons per unit area, N , and the surface dose, Sd for imaging 1 mm3 of tissue with 1% contrast assuming the following values E = 50 keV = 0.3 x = 1 mm kt = 5 1 = 22.6 m1 t = 0.004 m2 kg 1 R = 2 C(1 + R)/x (2 1 ) power series expansion of equation (13) X = 0.2 m

2 = kt (1 + R) exp(1 X)/[ (E, 0)(2 1 )2 x4 ] 52 (1 + 2) exp(22.6 m1 0.2 m) = 0.3 9 102 m2 1012 m4 = 2.55 1013 photons m2

thus the surface dose Sd is Sd = N t E = 22.55 1013 m2 0.004 m2 kg 1 50 keV = 5.1 1012 keV kg 1 = 0.817 mGy

where 1 Gy = 6.242 1015 keV kg 1 . Problem 2: By assuming that the projected area of the object having a linear attenuation coecient of 2 (gure 9) is A calculate the ratio of the total energy exiting the region A with respect to the expected variation in the total energy exiting a region of area A of thickness X and linear attenuation coecient 1 . This is the signal-to-noise ratio (SNR). By assuming that the SNR must be of at least kt before an object is visible solve the expression obtained for the SNR to obtain the number of photons per unit area that must be incident in order to obtain a contrast of C. Hint: Assume signal = (I1 I2 )A = CI1 A, noise = I1 A (i.e the number of photons detected per unit area is a Poisson process) and C is given by equation (13). Also see pp 124-127 Guy and ytche

16

11

Image Receptors and Substrates

The image receptor or substrate is what converts the transmitted X-rays into a visible image. All image receptors form an image by the absorption of energy from the transmitted X-ray beam. In general the transmitted spectra will consist of a distribution of photon energies and numbers. Part of the performance of image receptors is measured by its response to particular photon energies. The more restricted is its response, in general the better the resolution. Below is an incomplete list of the major methods for visualizing the energy of the transmitted X-ray beam. Direct exposure X-ray lm Screen-lm combinations Image-intensiers Xero-radiography Ionography Stimulated Luminescence (scintillation detector)

11.1

Direct Exposure X-ray Film

The direct exposure of X-ray lm is not commonly used in radiography because it has a low absorption eciency ( ). However the basic principles of operation are used in screen-lm combinations. In general there are two lm emulsions consisting of silver bromide suspended in gelatin. This is illustrated below
protective layer

film emulsion (20 micrometers)

subbing layer

film base (200 micrometers)

Figure 10: Schematic outline of the construction of direct exposure X-ray lm.

17

When using X-ray lm it is particularly important to know the relationship between the optical density of the lm as a function of exposure. Dene the optical density of the lm, D, as D = log10 (I0 /I) (16)

where I0 is the intensity of a light beam (eg a light box) before passage through the lm and I is the intensity after passage through the lm. It can be shown that the relationship between optical density of the direct exposure lm and the X-ray dose received by the lm is given by D = Dmax [1 exp(kB)] (17)

where k is a constant characterizing the response of the lm and B is the dose applied to the lm from transmitted X-rays. i.e the greater the lm exposure the blacker the lm. For a mono-energetic beam of X-rays the dose B is related to N by B = EN (18)

where is the photon absorption eciency of the lm. Curves that relate optical density to lm exposure are known as characteristic curves or H and D curves (after Hurter and Drield). It is common practice to plot lm dose on a logarithmic scale. It is found that the lm characteristic curve will have a central portion for which the following linear relationship is valid

D = 0.847 Dmax log10 (1.883 k B) B = log10 ( ) B0 where D is the optical density, B/B0 is the relative lm exposure and is known as the lm gamma (typically between 2 and 3). For the case of an homogeneous solid of thickness x we expect ln(B/B0 ) = x and it can be shown that we can calculate according to log10 (I) = x log10 (e) + log10
I0 B 0 ( E N0 (ic )t)

(19)

where E N0 (ic ) is the emitted X-ray intensity as a function of the lament current. Thus the lm can be estimated as the slope of log10 (I) versus x log10 (e). Direct exposure X-ray lm is excellent for high resolution studies where high resolution is more important than dosing considerations eg hands and teeth. However in conventional plain lm radiographic practice Screen-Film combinations are preferable as they are faster and 18

thus minimize patient dose. The main drawback is that screen-lm combinations do not have as good inherent resolution as direct exposure lm and for this reason are used when dose minimization is more important than the preservation of ne detail.

3.0 optical density

2.0

1.0 A B

0.5

1.0

1.5

2.0

2.5

log (relative exposure)

Figure 11: Film characteristic curves for a direct-exposure X-ray lm (curve A) and for a lm-screen combination (curve B). Note that the useful dynamic range for the latter is less than the former.

12

References

Aird EGA (1988). Basic Physics for Medical Imaging. Heinemann, Oxford. Alonso M and Finn EJ (1968). Fundamental University Physics, Volume III: Quantum and Statistical Physics. Addison-Wesley, Reading MA. pp 176-178,243-251. Guy G and Dominic tyche (2000). An Introduction to the Principles of Medical Imaging. Imperial College Press, London. pp 89-112,113-168, Webb S(ed) (1992). The Physics of Medical Imaging. IOP, Bristol. pp 20-54.

19

A
A.1

International SI Units of radiation exposure and dose

Exposure

The index of exposure, iexposure , is dened by iexposure = q/m (20)

where q is the total electric charge liberated by photoemmission in a mass m of air. There is no name for the SI unit of exposure ( 1 C kg 1 of air). Thus the intensity of a beam of X-rays can be specied in terms of the ionization per unit mass of air. However in general strengths of X-ray sources are specied in terms of the energy deposited per unit mass as it is the energy deposited in tissue that determines the biological hazard. The kinetic energy released per unit mass (kerma) of air is measured in units of Grays (Gy) such that 1Gy = 1 J kg1 Sources are also described in terms of air kerma rates i.e diexposure /dt.

A.2

Absorbed Dose

The absorbed dose will depend on the energy of incident radiation and the atomic composition of the absorber. In general this is linearly related to iexposure by an empirical dimensionless factor that depends on the absorber and the incident radiation energy, fabsorber (E), i.e doseabsorbed = fabsorber (E) iexposure Note that doseabsorbed will be in units of Gy 3 . Absorbed dose X-ray dose can also measured in terms of its relative biological eectiveness (equivalent absorbed dose) such that it can be quantitatively compared to other forms of radiation (e.g electron, proton or neutron radiation). The modern SI unit of equivalent absorbed dose is the Sievert (Sv) and for X-rays has the following units 1 Sv 1 J kg1 = 1 Gy
3 4

(21)

The older CGS unit of absorbed dose was the rad (1 erg s1 ). 1 Gy = 100 rad. The old unit of absorbed equivalent dose was the rem (radiation equivalent man) with 1 Sv = 100 rem.

20

A.3

Eective dose

The amount of energy deposited in a particular organ will depend on the what fraction of the total surface dose (contained in a weighting factor wR ) is absorbed which can only be inferred from theoretical models and phantom studies. In general the wR is estimated assuming a 70 kg standard man and will also incorporate information regarding the relative susceptibility of an organ to radiation damage. Thus the eective organ dose will be given as doseorgan = wR doseabsorbed wR will range from 0.01 (dental, spine) to 0.2 (gonads).

21