Chapter 2

Electromagnetic waves
C=f λ Energy of 1 photon
C – 3x108 m/s2 hc
E=hv=
λ – wavelength λ
f – frequency h – 6.626x10-34 J.s or 4.135x10-15 eV.s

1D point impulse

Function value is 1 only at x=0

Integral of point impulse is 1

Integral of f(x) multiply by impulse

function gives f(0)

2D point impulse

Function value is 1 only at x and y = 0

Integral of function is 1

Integral of f(x,y) multiply by impulse

function gives f(0,0)

Point impulse demonstration

1D 2D
∞ ∞ ∞

∫ δ( x )dx = 1 ∫ ∫ δ ( x , y )dxdy = 1
−∞ −∞ −∞
 Sifting property
Time shift the impulse function to get a point of interest of the function

1D 2D

Any function f(y) can be expressed in terms of delta function

∞ ∞
f ( y ) =∫ f ( x ) δ ( x− y ) dx=¿ ∫ f ( x ) δ ( y−x ) dx ¿
−∞ −∞
This property uses the symmetrical nature of impulse function

Similarly for 2D
∞ ∞
f ( p , q )=∫ ∫ f ( x , y ) δ ( x−p , y −q ) dx dy
−∞ −∞

Scaling property
1
δ ( ax , by )= δ( x , y )
|ab|

Line impulse
δ ( x , y )=δ (xcosθ+ ysinθ−l)

When θ = 0, line becomes parallel to Y-axis

When θ = 90, line becomes parallel to X-axis
 Systems

If the system S is understood well, the output can be determined for any input
f(x,y)

A system is linear if the superposition principle holds

g ( x , y )=S [ a f 1 ( x , y )+ b f 2 ( x , y ) ]=a g 1 ( x , y ) +b g2 (x , y)

Shift invariant system

g ( x−x o , y− y o ) =S [ f (x −x o , y− y o)]

A system is called linear shift-invariant (LSI) is system is both linear and shift-
invariant

Impulse response
Impulse response is the system output when the input is an impulse function
Defined using h

h ( x , y ) =S [δ (x , y)]

If system is LSI, system can be represented using impulse response

If system is LSI, S [ δ ( x− p , y−q ) ]=h( x−p , y −q)

System output
Can find out system output of LSI system if impulse response and input are given

Linearity

Linearity
Impulse response

Convolution
 ∞ ∞
g ( x , y )= ∫ ∫ f ( p , q ) h ( x−p , y −q ) dpdq=h ( x , y )∗f ( x , y)
−∞ −∞

Contrast
Difference between object and background or between adjacent objects (local
contrast)

f t −f b
C=
fb
f t – target intensity
f b – background intensity

Resolution
Ability of a system to resolve distinct sources

Calculated by finding FWHM (full width half maximum)

1
1. Let the function = and let x=xo
2
2. Solve the equation to find xo
3. FWHM = 2xo

Chapter 3
X-ray beam attenuation

−μo L
I =I o e
I o – Input intensity of X-ray
I – Output intensity of X-ray
μ( x ) – Linear X-ray attenuation coefficient
As different material have different attenuation coefficient
 D

−∫ μ (x)dx
I =I o e A

−¿ ¿
¿ Ioe
Assume all lengths are equal
−(μ +μ + μ )∆ x
I =I o e 1 2 3

This is applied to x-rays as body has multiple mediums with different coefficients
Hounsfield units (HU)

μ−μ water
CT number = 1000 ×
μwater

Air has -1000 HU

Water has 0 HU (used as reference)
Bone has 1000 HU

Image formation: Line integral

I =I o e ∫
− μ (x , y )dxdy

I
=e ∫
− μ (x , y )dxdy
Io
I
ln ⁡( )=−∫ μ(x , y ) dxdy
Io
∞ ∞
I
ln ⁡( )=−∫ ∫ μ ( x , y ) δ ( xcosθ+ ysinθ−L ) dxdy
Io −∞ −∞
∞ ∞
I
ln ⁡( )=−∫ ∫ f ( x , y ) δ ( xcosθ+ ysinθ −L ) dxdy
Io −∞ −∞
I
Since projection = p ( L, θ ) =−ln ⁡( )
Io
 ∞ ∞
p ( L, θ ) =∫ ∫ f ( x , y ) δ ( xcosθ + ysinθ−L ) dxdy
−∞ −∞
This is known as Radon Transform
∞ ∞
R [ f ( x , y ) ] = ( L , θ )= ∫ ∫ f ( x , y ) δ ( xcosθ + ysinθ−L ) dxdy
−∞ −∞
Radon transform is used to plot a sinogram

To convert sinogram to image, use image reconstruction technique

Simple backprojection

For each projection that detects an object, assume that there is an object
along the entire line

The more backprojections done, the location of object can be found at the
intersection

Fourier reconstruction
 Fourier slice theorem
∞ Synthesis eqn from
P ( ρ ,θ )=∫ p ( L ,θ)e
− j 2 πρL
dL signals, let w=2 π
−∞
∞ ∞ ∞
P ( ρ ,θ )=∫ ∫ ∫ f ( x , y )δ ( xcosθ+ ysinθ−L)dxdy e− j2 πρL dL dxdy
−∞ −∞ −∞ Linearity
∞ ∞ ∞
P ( ρ ,θ )=∫ ∫ f (x , y ) ∫ δ (xcosθ+ ysinθ−L)e
− j 2 πρL
dL dxdy
−∞ −∞ −∞
Sub in eqn of
∞ ∞
line
P ( ρ ,θ )=∫ ∫ f (x , y ) e
− j 2 πρ(xcosθ + ysinθ)
dxdy
−∞ −∞
Let u= ρ cosθ , v= ρ sinθ
∞ ∞
P ( ρ ,θ )=∫ ∫ f ( x , y ) e
− j 2 π (xu + yv )
dxdy
−∞ −∞
P ( ρ ,θ )=F (u , v)
Filtered backprojection
Steps
 Get projection p ( L, θ )
 Do 1D Fourier transform to obtain P ( ρ ,θ )
 Multiply with ramp filter |ρ|, high pass filter: |ρ| P ( ρ ,θ )
 1D inverse Fourier transform
 Backprojection to get image
Point 2 to 4 is equivalent to doing filtering in time domain

Convolution backprojection
Steps
 Get projection p ( L, θ )
 Convolution of c(L) with p ( L, θ )
 Backprojection to get image

Chapter 4
Types of nuclear imaging
 Planer scintigraphy: Use anger scintillation camera

 Single photon emission computed tomography (SPECT): use rotating

camera to get projections. Image depend on radiotracer labelled with
radioactive atoms whose decay produce a single gamma photon directly

 Positron emission tomography (PET): Based on radiotracers labelled with

radioactive atoms whose decay produce a positron that is subsequently
annihilated, producing two gamma photons of 511KeV going in opposite
directions

Together, SPECT and PET are referred to as emission computed

tomography
Detection of positron decay

Positron is the opposite of an electron, denoted as e +

Positron are anti-matter, when collide with electron (matter), collision release
photon energy (511KeV) in opposite direction

Detector will detect photon pairs via time coincidence

Relative effect of attenuation

SPECT PET
 Lower energy photon  Higher energy
 Less penetrating  More penetrating
 Each photon only travel short  Photon pair travel entire body
distance
Assume d1=d2=15cm
Assume d = 15cm 0.153
μ511 KeV = (Attenuation
0.153 cm
μ140 KeV = (Attenuation
cm coefficient)
coefficient) I −μ
(d 1+ d 2)

=e =0.056
511 KeV

Io
 I −μ
d
Therefore, 5.6% transmitted
=e =0.1
140KeV

Io
Therefore, 10% transmitted
Example above shows SPECT and PET have high absorption rate, hence
absorption correction is needed to reconstruct images

Scattering is also involved in both, this cause miscalculation

For PET, if 2 positron-electron collision occur at the same time, but detector only
detect 1 photon from each event due to high absorption. This will cause false
reconstruction. This is miscalculation in coincidence detection.

PET/CT image
PET is not good for structure; PET scans reveal metabolic change on cellular level
CT scan can be used to identify structure
Use both together

Chapter 5
Ultrasound general information
 For good resolution and good penetrating ability, frequency range is 1MHz
to 20MHz
 Applied to body externally, non-invasive

Physics of ultrasound
 Ultrasound waves need a medium to travel
 Acoustic waves travel at speed of sound

1
C=
√ kρ
C – 1500m/s2
K – compressibility
ρ – density

 Typical measurements
Frequency – 1 to 20 MHz
C
Wavelength = =1.5 mm ¿ 0.075 mm
f

λ
Resolution = (estimate)
2
 Impedance formula
z=ρc
Increase in frequency, decrease in wavelength, decrease in resolution
Above 20MHz, there is high attenuation. Ultrasound wave must bounce back to
generate a signal, attenuation is not good.
 Reflection and refraction

Snell’s law
sin θi C1 speed ∈the first medium
= =
sin θt C2 Speed ∈second medium

Boundary conditions give reflected and transmitted pressure

I R Z 2 cos θ i−Z 1 cos θt 2
R I = =( )
Ii Z 2 cos θ i+ Z 1 cos θt
2
I 4 Z Z cos θ i
TI= t= 1 2
Ii ¿¿

Example calculation of reflected intensity

Ir
R I= =¿
Ii
¿¿

Transducer
Piezoelectric materials convert electric field to mechanical strain
 PZT (lead zirconate titanate) – more commonly used for transducer
 PVDF; a polymer film

These materials are characterised by 2 constants

strain
 Transmission constant D ( )
electric field
Potential
 Receiving constant G ( )
Unit stress
PZT
 D = 300x10-12 m/V
 G = 2.5x10-2 v/(N/m)

PVDF
 D = 15x10-12 m/V
 G = 14x10-2 V/(N/m)

PZT is more common is clinical applications due to good D and G. Can be used as
both transmitter and receiver

Matching layer and acoustic window purpose is to gradually change the

impedance of the transducer to match the impedance of the tissue. Acoustic gel
can be applied as well. This prevents attenuation.

Ultrasound wave

The pulse of ultrasonic energy travels into tissue. It is reflected from tissue
boundaries, causing echoes

By the time the echoes reach the transducer, the pulse generator has turned off,
and the echoes create an oscillation in the transducer again

The received ultrasound echo is in the radio frequency (RF) in the 1 to 20 MHz
range. It is processed by the signal processing unit for imaging and display
 Attenuation
Due to attenuation, the reflected echoes are weaker in magnitude and power

To overcome this problem, use time-gain compensation (TGC) unit

Vary the gain with time after pulse is transmitted to amplify magnitude

Rule of thumb in tissue: 1dB/cm/MHz

If a 2 MHz transducer is used to image abdomen up to a depth of 10 cm what will

be the total loss in the signal strength?

Total loss = 1dB/cm/MHz x 20cm x 2MHz = 40dB

Imaging mode
 A-mode (measure amplitude)
Transducer and organ is stationary
 B-mode (measure brightness)
Displays 2D image by translating transducer while organ is not moving

Either move transducer up and down or adjust the contact angle, making a
sweeping motion

Generating the image is like generation of sinogram

 M-mode (measure motion)

Keep transducer still and measure motion of moving organ

 C-mode (measure depth)

 Doppler mode (accounts for doppler effect)

Doppler effect: If a sound source or target is moving, the perceived

frequency of noise will be shifted

C
f t= f
C−vcosθ o
ft – received frequency
fo – original frequency
 v – velocity
c – 1500m/s2

Doppler frequency is given by

vcosθ vcosθ
f D =f t−f o =( )f o ≈ fo
c−vcosθ c

Pulse echo doppler is given by

2 vcosθ
f D =( f o)
c
The 2 is due to the shift occurring twice due to echoes

Chapter 6
Optical Coherence Tomography
 Similar to ultrasound, transmit light into tissue
 Different layer reflects light
 However in ultrasound, echo can be directly recorded as sound wave travel
very fast which makes the delay minimal
 Light wave do not travel as fast, hence will have delay
 In OCT, echo is detected by interferometry
 Maximum depth of OCT is 1-2 mm
 Spatial resolution: 1-10 microns
 Contrast: Backscattering and polarisation
 Eye: Optically transparent window until the retina
In ultrasound imaging sound waves are transmitted into the tissue and some of the
ultrasound energy gets reflected back to the transducer at tissue boundaries. This
reflected sound waves are recorded, and images are formed.
In optical coherence tomography the same principle is applied, but light waves are
used instead. Since the light travels much faster than the sound waves inside the
tissue, it is hard to capture the reflected lights using conventional electronics.
Therefore, instead of directly capturing the reflected light waves and calculating the
time of flight, an indirect method is used to calculate the depth from which they were
reflected. By using the low coherence light and interference the depth is measured.
Light vs sound
 OCT can be regarded as an optical analogue of ultrasound imager
 Difference between OCT and ultrasound imager comes from the difference in
wave speed
Application of OCT
 OCT used to image the retina of the eye
  Used to identify fovea and optic disc to observe morphology

Chapter 7
Digital Image processing (DIP)
Purpose of DIP
 To improve pictorial information
 Process image data for storage, transmission, and representation for
machine perception
Applications
 Noise filtering
 Content enhancement
o Contract enhancement
o Deblurring
 Cancer detection
Key stages in DIP
1. Image acquisition
2. Image enhancement
3. Image restoration
4. Morphological processing
5. Segmentation
6. Representation and description
7. Object recognition
8. Image compression
9. Colour image processing
Image compression
An image contains a lot of redundancy that can be removed to compress the file

Examples
 Interpixel redundancy
 Coding redundancy
 Psychovisual redundancy

Benefits of compression
 Reduced storage
 Reduced bandwidth

Image sampling and quantization

To digitize the image, sampling and quantization is needed

Sampling: Digitizing the coordinate values

Quantization: Digitizing the amplitude values
Quantisation Sampling
8
8-bit quantisation: 2 =256 gray levels Something like resolution
(0: black, 255:white) Digital still camera: 640x480, 1024x1024
Digital video camera: 640x480 at 30 fps
1 bit quantisation: 2 gray levels
(0: black, 1:white)
 Definition of an image
An image is a 2-dimensional function f(x,y) where x and y are spatial coordinates

Amplitude of f at any pair of coordinates (x,y) is called the intensity or gray level

An image is called a digital image when x, y and the amplitude values of f are all
finite and discrete

Pixel is the term used to denote an element of a digital image

Digital image can be represented in a compact matrix form

Each element in the matrix is a pixel and the value of the function is the gray level

M – rows of the matrix

N – column of the matrix

L=2 K
L – gray level
K – number of bits

Image size
B=M × X × K
Spatial and Gray Level Resolution
Spatial resolution is the smallest number of discernible line pairs per unit
Gray level resolution is the smallest discernible change in gray level
 Binary images

A binary image is where the gray level of each pixel is either 0 or 1

Binary images are produced as a result of some decision process on pixels of a

input image and is used to specify regions in the image where some property or
condition is true
Basic relationship between pixels

Connectivity

Connectivity between pixels is useful in

 Establishing object boundaries
 Defining image components/regions

To establish if 2 pixels are connected it must be determined

 If they are neighbours
 If their gray levels satisfy a specified criterion of similarity, usually defined in
a vector V

3 types of adjacencies
4-adjacency: 2 pixels P and Q with values from V are 4-adjacent if Q is a 4-
neighbour of P
8-adjacency: 2 pixels P and Q with values from V are 8-adjacent if Q is a 8-
neighbour of P
M-adjacency(mixed adjacency): 2 pixels P and Q with values from V are m-
adjacent if
 Q is a 4-neighbour of P, or
 Q is a diagonal-neighbour of P and the overlapping 4-neighbour of Q and P
have no pixels whose values are from V

Path
A digital path from pixel P with coordinates (x,y) to pixel Q with coordinates (S,T) is
a sequence of distinct pixels with coordinates

N is length of path and pixels are connected based on the different types of
adjacencies
 Region
A region of an image is a set of connected pixels

Boundary
Boundary of a region is a set of pixels in the region that have one or more
neighbours that are not in the region
Distance measures

Euclidean distance between Q and P

De ( p , q )=¿

D4 distance (also called city-block distance)

D 4 ( p , q ) =|x−s|+| y−t|

D8 distance (also called chessboard distance)

D 8 ( p ,q )=max ⁡(|x−s| ,| y−t|)

Chapter 8
Image enhancement in the spatial domain
Spatial domain refers to the pixels composing an image

Spatial domain processes will be denoted by the expression

g ( x , y )=T [f ( x , y)]

f(x,y) – input image

g(x,y) – processed image
T – operator
Principal approach (pixel by pixel procedures)

Using a square or rectangular sub-image area centered at (x,y) in defining a

neighbourhood about a point (x,y)

This is referred to as a mask

The simplest form of operator T is when the mask is a size of 1x1, output g will
only depend on value of f at (x,y) and T becomes a gray level transformation
function of the form
S=T (r )
r – variable denoting gray level of f(x,y)
s – variable denoting gray level of g(x,y)

Example: Contrast stretching

Contrast stretching either increases or decreases the original gray level to create
better contrast

If original value > m If original value < m

Operator T will increase gray level Operator T will decrease gray level
Example: Thresholding
Gray level value will be changed to 0 or 1 depending on the threshold value to
produce a binary image

If original value > m If original value < m

Operator T will increase gray level to Operator T will decrease gray level to 0
max

Basic gray level transformation (single intensity transformation)

1. Image negatives
2. Log transformation
3. Power-law transformation
Image negatives
Negative of an image with gray levels in the range of 0 to L-1 is obtained using the
equation below
S= L−1−r
Reversing the intensity levels of an image produces a photographic negative, this
is suitable for enhancing white or gray detail embedded in dark regions of an
image

Logarithmic transformation
 S=c log (1+ r )
c – constant
assume r is positive

This transform is used to expand values of dark pixels and compress values of
bright pixels

For lower amplitudes, gray level is expanded. For higher amplitudes, gray level is
compressed

Power law transformation

γ
S=c r
c – positive constant

Power law map narrow range of dark inputs into wider range of output values
depending on γ value. When γ is reduced too much, image begins to reduce
contrast to the point where it looks “washed-out”
 Logic operation

 AND and OR gates are used to combine two images together

 NOT gate is used to invert the intensity of the image

Arithmetic operation
 Subtraction
 Addition

Image subtraction
g ( x , y )=f ( x , y )−h (x , y )
 Difference between two images
 Obtained by computing the difference between all pairs of corresponding
pixels from f and h
 Useful in enhancing differences between image
 Used in mask mode radiography

Histogram of an image
Histogram of a digital image with gray levels in the range [0, L-1] is a discrete
function given by
h( r k )=nk
rk is the kth gray level
nk is number of pixels in the image that have that r k gray level
h(rk) is histogram of digital image with gray levels r k

Histogram can be normalised by dividing each of its values by total number of

pixels given by
nk
p (rk )=
n
P(rk) gives probability of occurrence of gray level

Sum of all components of a normalised histogram is equal to 1

Histogram are the basis for spatial domain processing techniques such as
 Image enhancement
 Image compression
 Image segmentation

Basics of spatial filtering

Spatial filtering involves moving the filter mask from point to point on an image

A mask is a 2-D array in which values of the mask coefficients determine the
nature of the process

Response R of an m x n mask at any point (x,y) given by the following expression

For linear spatial filtering, response is sum of products of filter coefficient and pixel
value
mn
R=∑ W i Z i
i=1
W – mask coefficient
Z – value of gray level

Smoothing spatial filters (linear)

Basic idea
Value of every pixel in an image is replaced by the average of gray levels in the
neighbourhood defined by the filter mask

Used to reduce blur and reduce noise

Box filter (mean filter) Weighted average filter
All coefficients are equal Allocated different weighting to different
pixels depending on coefficient value
 mn mn
R=∑ W i Z i R=∑ W i Z i
i=1 i=1

Order Statistic filters (non-linear)

Nonlinear spatial filters are based on ranking the pixels contained in the image
area encompass by the filter and replacing the value of the center pixel with the
value determined by the ranking
Median filter
Median filter replaces the value of a pixel by the median of the gray levels in the
neighbourhood of that pixel

Median filter provide excellent noise reduction with less blurring than linear
smoothing filters of similar size

Steps to perform median filtering

1. Sort the values of the pixels encompassed by mask from smallest to biggest
2. Identify median value
3. Put median value into pixel
Max filter Min filter
Used to find the brightest points in an Used in finding the darkest point in an
image image

Pixel replication and zero padding

Problems arise when conducting mask operations on the edges of an image, 3
methods to solve this problem

1. Discard the problem pixel,

a. Yeet the pixel to the shadow realm
b. 512x512 image becomes 510x510 after filtering
2. Zero padding
a. Expand image by surrounding it with 0s
b. 512x512 image becomes 514x514
3. Pixel replication
a. Use gray levels of border pixels to fill up expanded region
Zero padding Pixel replication
Chapter 9
Image segmentation
Segmentation subdivides an image into its constituent regions

Segmentation accuracy determines the eventual success or failure of

computerised analysis procedures

Segmentation is based on 2 properties of intensity values

 Discontinuity
 Similarity
Discontinuity Similarity
Approach is to partition an image based Based on partitioning an image into
on abrupt changes in intensity such as regions that are similar according to a
edges set of criteria

Thresholding
Thresholding is means of distinguishing pixels that have higher or lower intensity
that the set value

Detection of discontinuities
Techniques to detect 3 basic types of gray level discontinuities are

1. Points
2. Lines
3. Edges
Common way to look for discontinuities is to use a mask
mn
R=∑ W i Z i
i=1

Point detection
Mask is applied to calculate the value of the pixel and will be compared to the set
threshold value
Line detection
Masks used will have a specific coefficient configuration to detect lines along a
specific plane

The mask detects lines in that specific plane while eliminating the rest
Edge detection
Ideal edge detection

Ideally, edges are discrete and can be distinguished easily

Ramp digital edges

In practise, edges look more like ramps. The gradient at the edge is cause by
imperfections in image acquisition or sampling rate etc

Slope of ramp is inversely proportional to degree of blurring in the edge, steeper

ramp means less blur
First order and second order derivatives
First order derivative is the difference between subsequent values and measures
the rate of change of the function
ϑf
=f ( x +1 ) −f ( x )
ϑx
Second order derivative uses values both before and after the current value
2
ϑ f
=f ( x +1 ) + f ( x−1 )−2 f (x )
ϑ2 x

Magnitude of first derivative can be used to detect presence of edge at a point in

an image

Sign of second order derivative can be used to determine whether an edge pixel
lies on the dark or bright side of an edge. Pixel in dark region will be positive while
pixel in bright region will be negative, with the zero-crossing representing the edge

Edge detection: first derivative

Gradient

[]
∂f
G
∇f= x =
[ ]
Gy
∂x
∂f
dy
 Magnitude
mag ( ∇ f )=¿

Direction of gradient vector

−1 G y
α ( x , y )=ta n ( )
Gx
mag ( ∇ f )=¿
Can be computationally hard to compute, therefore an estimation is done
∇ f =|G x|+|G y|
G x =( Z 7 + Z 8+ Z 9 ) −(Z 1+ Z 2+ Z 3 )
G y =( Z 3 +Z 6 + Z 9 )−(Z 1 + Z 4 +Z 7 )

Prewitt operator
Used to detect edges based on applying a horizontal and vertical filter in sequence

∇ f =¿
horizontal edge vertical edge
 Sobel Operator
Creates emphasis on edges in image

∇ f =¿

Laplacian filter: second derivative

Uses zero-crossing property for edge detection and establish if pixel is in dark or
bright side of edge

2
∇ f =4 z 5−(Z 2 + Z 4 +Z 6 + Z 8)

2
∇ f =8 z 5−( Z1 + Z 2+ Z 3 +Z 4 + Z 6 + Z7 + Z 8 +Z 9 )
Chapter 9
Magnetisation
 Protons, electrons, and neutrons have a spin property
1
 Spin comes in multiples of and can be + or –
2
 Charged particles that spin creates an electromagnetic field
 MRI uses hydrogen nucleus which has a single positively charged proton

 In MRI, strong external magnetic field is applied

 Protons will line up under the external magnetic field B o and produce a net
magnetisation
 The average of nuclear magnetisation vectors are known as bulk
magnetisation (Mo)
 As time passes, more hydrogen proton spins will line up with B o
 The net magnetisation process follows an exponential curve
−t
T1
Net magnetisation=1−e
T1 – time constant depending on type of tissue under imaging
 Magnetisation also depends on density of protons in tissue which follows the
equation below
−t
T1
M =N (H )(1−e )

Precession
 Precession – wobble about the axis of Bo field
  Larmor equation describes the rate of proton pression around B o
f =γ B o

f – precession frequency (Larmor frequency)

γ – gyromagnetic ratio in Hz/T
Bo – external magnetic field in T or gauss
1 T = 10-4 G
 γ is proportional fixed constant, for hydrogen it is 42.6MHz/T

RF excitation
 Radio frequency (RF) pulse is an electromagnetic wave
 Before RF pulse is applied, net magnetisation M o is under external magnetic
field Bo
 RF pulse introduces a new magnetic field B1 that is weaker than Bo
 New precession frequency in introduced
f 1=γ B1

 Energy of the RF pulse will be absorbed by proton which causes the spin
direction to change
 Change in spin direction causes a shift in Mo where Mo flips into the x-y plane
 The flip angle can be determined by
α =2 πγ B1 τ

τ – duration of RF pulse

 If the flip angle is 90°, the RF pulse is known as 90° RF pulse and net
magnetisation vector will be denoted as Mxy
 If flip angle is 180°, the RF pulse is known as 180° RF pulse and net
magnetisation vector will be denoted as -Mo
 In the case of partial flip (angle less than 90°), magnetisation vector = M osinα
Resonance
 If frequency of RF pulse (f1) match precession frequency of the protons (f),
resonance occurs
 Resonance causes energy of the RF pulse to be absorbed by protons
Magnetic field strength
 Mo is proportional the Bo
 Strong magnet gives large Mo which is necessary for strong signal
Relaxation
 After RF pulse is removed, magnetisation vector returns to initial alignment
with Bo
 Mo exhibits spiralling pression around Bo
 System releases back energy absorbed by RF pulse
 Energy released as an oscillating and decaying signal, this is known as Free
Induction Decay (FID)

Transverse relaxation
 Transverse relaxation acts first to cause signal to decay, this is known as
spin-spin relaxation
 Influenced by random microscopic motion which cause spin to speed up or
slow down which changes the phase of other nearby spins
 T2 represents time for Mxy to decay to 37% of its maximum value

Longitudinal relaxation
 Longitudinal relaxation describes re-growth of M z (Mo)
 T1 represents time for Mz to increase to 63% of final value
  T1 is magnetic field dependent, strong magnetic field require more time to
regrow
Relaxation – part 2
 Relaxation time are characterisation of the rate of spins relaxing back to
equilibrium state
T1 T2
Time constant for Mz to recover Time constant for Mxy to decay
−t −t
T1 T2
M z ( t )=M o (1−e ) M xy ( t )=M o e

 T1 and T2 are inherent properties of tissues, T2 occurs about 5 to 10 times

faster than T1
 Decay of Mxy is due to transverse relaxation (T2) and magnetic field
inhomogeneities (T2*)
1 1
¿= + ∆ Bγ
T2 T 2

∆ B – field inhomogeneities
Creating an image
 3 dimensions required
o Z - direction: Use slice selection
o X – direction: Use frequency encoding
o Y – direction: Use phase encoding

 Resonance occurs if RF pulse is the same as Larmor frequency

 If Bo is varied from point to point, each position will have its own resonant
frequency
  This is achieved by using a gradient coil to create a magnetic field gradient

 Spatial information is obtained by applying gradients

o Gz direction – slice selection
o Gx direction – frequency encode
o Gy direction – phase encoding

Slice selection
 To get slice of interest, transmit RF signal at frequency that the slice of
interest is spinning at
 Only that slice will absorb energy and image can be obtained
 Apply gradient Gz to Bo when the RF pulse is transmitted
 Gradient will cause Larmor frequency to change and if it matches with RF
pulse, resonance can occur to get a signal
 Center of slice = Bo
 To calculate the frequency of the slice, use the following equation
f =γ (Bo +G z X )
γ – 42.6 MHz/T
X – distance from the center (note direction is impt)
 Range of frequencies which determines the slice thickness is called
bandwidth

 Position of slice is determined by center frequency of RF pulse

 Slice thickness can be decrease by
o Decreasing bandwidth of RF pulse
 o Increase Gz
 Gradient affects slice thickness

 Now that slice has been obtained, protons spin with same phase and
frequency
 Perform phase and frequency encoding
Phase encoding (Gy)
 Ensure each proton has a unique phase for spatial encoding
 Achieved by putting a gradient in y – direction
 Before applying gradient, protons have same phase and same frequency
 Gy cause protons to change phase depending on the gradient value
 Pixel in top row precess faster
 Pixel in middle row experience no change
 Pixel in bottom row precess slower

 Once Gy is off, protons will have same frequency but different phase
 Phase variation is used for encoding along y – direction within a slice
 Phase variation will be reflected on the spin echo and will be analysed with
Fourier transform
Frequency encoding (Gx)
 Ensure each proton has unique frequency for spatial encoding
 Frequency encoding is achieved by applying a field gradient in the x –
direction
  Before applying gradient, protons have same frequency but different phase
due to Gy
 Gx causes frequency of protons to vary depending on the gradient value
 Pixel in center is not affected
 Pixel in right have higher frequency
 Pixel in left have lower frequency

 Frequency variation is used for encoding along x – direction within a slice

 Frequency variation will be reflected on spin echo and will be analysed wit
Fourier transform
Encoding completed
 After application of Gy and Gx, protons in each pixel will have a distinct
frequency and phase and can be uniquely determined
 Gradients must be switched on and off very fast in sequence
 Computers control the sequence
Sequences
1. Switching on slice selection gradient (Gz)
2. Introduce 90° RF pulse to flip net-magnetisation M o into X-Y plane (Mxy)
3. Switch on phase encoding gradient (Gy) to do phase encoding
4. Switch on frequency encoding gradient (Gx) and during which the FID signal is
sampled
Spin Echo
 Echo signal is “rebuilt” from FID
 The 180° rephasing pulse is in the middle of the 90° pulse and echo

Spin echo imaging 1

 Spins are flipped into X- Y plane by a 90° RF pulse while G z is applied
  Spin phase start to diverge and decay due to interactions between
neighbouring nuclei, this generates FID signal
 A second 180° RF pulse inverts the spins and makes the phases converge
again
 When all protons are back in phase, after time TE, the output signal reaches a
maximum and this is the spin echo

Spin Echo Pulse diagram

 TR – time between successive 90° pulses (repetition time)
 TE – time between max of FID and max of SE = 2τ
 TE is the time interval from initial 90° pulse to the formation of spin echo

 SE last twice as long as FID, making it easier to measure

 TE can be set. Therefore the SE appears on a known time when the
measurement system is ready
180° RF pulse
 Flips the longitudinal magnetisation to the opposite direction (+z to - z)
 Flips the transverse components of spins in a mirror-like fashion over the x-
axis
 Magnitude is constant but change the position and evolution of M xy vector
K-space
 Echo-signal is collected and stored in the computer’s memory (k-space)
  After Fourier transform of sampled data, image is obtained
 The whole process is repeated using a different phase encoding gradient (to
select different position) until field of view has been scanned
 Typical number of repeats is 256
 Each SE from 1 phase encoding step provides data for 1 line of a matrix
called the k-space of the image
 The line in K-space for the SE data is determined by the phase encoding
gradient that was applied
 Size of the K-space matrix is determined by the number of phase encoding
gradients applied

Contrast in MRI
 Contrast depends on
o T1 and T2 relaxation rates
o Proton density
o Different pulse sequence used in MRI
 3 types of images
o T1 weighted (short TR and short TE)
o T2 weighted (Long TR and long TE)
o Proton density weighted image (Long TR and short TE)

T1 T2
Regrowth of Mz Decay of Mxy

T1 weighted image
 Uses short TR and short TE to capture image
 Short TR allows for better contrast between the various tissues as the
regrowth of Mz occurs at various speeds

 Uses short TE as the decay of Mxy has not occurred as much, hence its does
not contribute to contrast

 Material with short T1 will appear bright

 Material with long T1 will appear dark
 From bright to dark
o Cerebrospinal fluid
o Lesion
o Grey matter
o White matter
 o Fat

T2 weighted image

Short TE
Long TE

 Long TE allows for better contrast between various tissues as decay of M xy

occurs at different speed
 Uses long TR so that there is sufficient time for M z to reach the maximum
amplitude Mo, so that this does not contribute to contrast of image
 Material with short T2 will be dark
 Material with long T2 will be bright
 From bright to dark
o Fat
o White matter
o Grey matter
o Lesion
o Cerebrospinal fluid