Transcript CT1 - hullrad

CT Scanning
Dr. Craig Moore
Medical Physicist & Radiation Protection Adviser
Radiation Physics Service
CHH Oncology
Brief History of CT Scanning
First CT Scanner - 1972
•
•
•
•
•
•
Originally called CAT
A = axial
80 x 80 resolution
4 min. per rotation
8 grey levels
overnight reconstruction
Here and Now
•
•
•
•
512 x 512 or 1024 x 1024 resolution
Sub second rotation
4096 grey levels
100’s slices per rotation
Components of a CT Scanner
Principal components of CT
scanner
• X-ray tube, collimator, and detector array on a
rotating gantry
• Rotation axis is referred to as Z axis
• Fan beam wide enough to cover patient crosssection
• Narrower width in the z-axis
• Behind the patient is a bank of detectors
• Patient lies on a couch that is moved
longitudinally through the gantry
X-ray Tube
•
•
•
•
•
•
•
Tube parallel to patient movement – minimise anode heel effect
X-rays are produced by firing electrons at a metal target – typically tungsten
Capable of producing long exposure times at high mA – get very hot (require heat capacities up to 4MJ
and active cooling mechanisms)
Continuous scanning limited to around 90s
Focal spot size typically 0.6 - 1mm
High kVps (80 – 140) and beam heavily filtered (6-10 mm Al filters) to optimise spectrum
Stops attenuation coefficients varying with depth via beam hardening
Collimation and Filtration
•
•
•
Want a monoenergetic beam to avoid beam
hardening artefacts
– As beam passes through patient low
energies are filtered
– This results in the apparent reduction of
attenuation and CT number of tissues
– Computer reconstruction assumes
monoenergetic beam
– Not possible with X-ray tubes so they
are heavily filtered
– At least 6 mm aluminium or copper and
typically high kVp
– Some manufacturers use shaped filters
such as bow-tie filters to even out the
dose distribution (conform to the shape
of an elliptical patient)
Pre-patient collimator is mounted on the Xray tube
– Beam is approx 50cm wide to cover
cross section of patient
– The size is variable in the z-axis
– Multi-slice scanners between 0.5 and
40 mm thick beams
Post patient collimation is not used with
multi-slice scanner
Bow Shaping Filters
• Body is approx circularly
shaped
• Shape of beam that
reaches detector is
therefore non-uniform
which can lead to image
artefacts
• Shaping filters are
applied to the beam prior
to make the dose
distribution more uniform
Detectors
• Requirements:
–
–
–
–
–
–
Small enough to allow good spatial resolution
Up to 1000 detectors per scanner
Typically 1.5 mm width but can be a small as 0.5mm
High detection efficiency
Fast response
Wide dynamic range – massive variation in X-ray
intensity
– Stable and noise free
– No afterglow
– There needs to be separation between detectors to
prevent light crossover
• This reduces efficiency from 98% to 80%
Detectors
• Modern CT detectors are
fabricated as detector
modules
• A side view shows
individual detector
elements coupled to
photodiodes that are
layered on electronics
• Spaces in between filled
with optical filler to reduce
cross talk
Detectors
Generations of CT Scanner
CT Imaging
• Conventional radiography suffers from the collapsing of
3D structures onto a 2D image
• However, CT scanning has extremely good low contrast
resolution, enabling the detection of very small changes
in tissue type
– Almost true depiction of subject contrast
• CT gives accurate diagnostic information about the
distribution of structures inside the body
• Generation of images in transaxial section
– Perpendicular to the axis of rotation of the X-ray tube about the
body
– Perpendicular to the craniocaudal axis of patient
Number of detectors and
projections
• Typically, for a 3rd generation scanner:
– 650 – 900 detectors
– 1000 to 2000 projections per rotation
Collapse of 3D Data into 2D Plane
Image contrast 2:1
• Planar imaging
– 2D representation of 3D
Distribution of Tissue
– No depth information
– Structures at different
depths are superimposed
• Loss of contrast
Subject Contrast 4:1
X rays
Typical CT Image
CT Images
•
•
•
Commonly calculated on 512x512
matrix, but 256x256 and
1024x1024 are also used
Each pixel is more accurately
described as a voxel, because it
has depth information
The value stored in each voxel is
referred to as the CT number
which is related to the attenuation
of a particular tissue:
– CTn = 1000 x (µt - µw)/ µw
•
•
•
Sometimes referred to as
Hounsfield Units
Each CT number is assigned a
certain shade of grey in the
resulting image
CT number represents x-ray
attenuation coefficient of the
corresponding voxel within the
patient
CT Numbers
Tissue
Range of CT Numbers
Bone
500-3000
Muscle
40-60
Brain (grey matter)
35-45
Brain (white matter)
20-30
Fat
-60 to -150
Lung
-300 to -800
Image Display
• CT image represented by
a range of CT numbers
from -1000 to + 3000 (ie
4000 levels of grey)
• Human eye dose not
have the capacity to
distinguish so many grey
levels
• If 4000 shades of grey
displayed altogether there
would be very little
difference between
different tissues
Window Width and Level
• The appearance of the image on the screen can be changed by
altering the window width and level
• Window width refers to the range of CT numbers selected for display
• This range of CT numbers is centred at a particular level called the
window level
– e.g. if imaging bone window level should be ~1000
• Can spread a small range of CT numbers over a large range of
grayscale values
Window Level –593
Window Level –12
Window Width 500
Window Width 400
Good contrast in lungs
Good soft tissue contrast
Only see CT numbers
+/- 250 around -593
Only see CT numbers +/200 around -12
How do we get the images?
• Tube and detector rotate smoothly around
the patient
• X-rays are produced continuously and the
detectors sample the X-ray beam approx
1000 times during one rotation
• Typically 2 to 4 revolutions per second
•
•
•
In reality not always parallel to
detectors
Each voxel is traversed by one or
more x-ray beams for every
measurement (1000 per rotation)
Number of measurements taken in
single axial section depends on
– number of detectors
– Number of samples per rotation
•
•
•
Assume 800 detectors measured
at 0.5° intervals per 360 ° rotation
This is 576,000 measurements
More than needed as we only
need 260,000 measurements (512
x 512)
How do we get the picture?
• Back Projection
– Reverse the process of measurement of
projection data to reconstruct image
– Each projection if smeared back across the
reconstructed image
Back Projection – the basics
•
•
•
•
•
•
•
•
•
Consider cylindrical uniform body with
a hole down the centre
A beam passing through this body
from one direction will have a
transmitted profile in its central region
This single measurement cannot
determine the position of the hole
other than identifying that it is in the
line of the pencil beam passing
through the centre of the body
Pixel values along this line are
decreased by the amount of
attenuation measured
These values are projected back along
the field of view
A second projection at 90° provides a
second band of grey
This is then projected back across field
of view
Progressive projections are shown in
the final figure – a star like pattern
We now have an image that looks
similar to what we are scanning
Back Projection
3
6
3
3
3
6
6
3
3
3
6
3
• Back Project
each planar
image onto
three
dimensional
image matrix
Back Projection
3
6
3
1
2
1
1
2
1
1
2
1
• Back Project
each planar
image onto
three
dimensional
image matrix
Back Projection
3
6
3
3
21
2
2
3
1
2
6
31
2
4
31
3
21
23
1
2
• Back Project
each planar
image onto
three
dimensional
image matrix
Back Projection
3
6
3
3
4
6
4
3
6
6
8
6
6
3
4
6
4
3
3
6
3
• Back Project
each planar
image onto
three
dimensional
image matrix
Back Projection
3
6
3
3
4
6
4
3
6
6
8
6
6
3
4
6
4
3
3
6
3
• Back Project
each planar
image onto
three
dimensional
image matrix
Back Projection
• More views –
better
reconstruction
• 1/r blurring, even
with infinite number
of views
Filtered Back Projection
• Back projection produces blurred transaxial images
• Projection data needs to be filtered before BP
• Different filters can be applied for different diagnostic
procedures
– Smoother filters for viewing soft tissue
– Sharp filters for high resolution imaging
• After filtration, back projection same as before
– Data from neighbouring beams are used
– Some data is subtracted
– Some data is added
• Filters are convolved with the blurred image data in
Fourier Space
Filtered Back Projection
Filtered Back Projection
• Usually done in frequency
(Fourier) space
• Filter planar views prior to
back projection
• Correction of 1/r blurring
requires ‘Ramp’ Filter
– Gives increasing weight to
higher spatial frequencies
– Amplifies Noise
SPECT FIlters
signal
1
Ramp
0.8
Shepp-Logan
0.6
Modified
Shepp_Logan
0.4
Hanning
0.2
Hamming
0
0
0.2
0.4
0.6
0.8
-0.2
Frequency (fraction of Nyquist)
1
Butterworth
Problems with Filtered Back
Projection
• Back projection is mathematically correct,
but real life images require Filtered Back
Projection
• Back Projection can introduce noise and
streaking artefacts
– Not good with attenuation correction
• Filtered Back Projection can reduce noise
and artefacts, but may degrade resolution
Iterative Reconstruction
• FBP was primary method for reconstructing CT images
for many decades
• Modern computing power now enables CT scanners to
use iterative reconstruction
• This process begins with an initial estimate of what the
object may look like
• This guess is used to compute forward projected data
sets, which are compared with the measured projection
data sets
• The difference between the computed and measured
projections creates an ‘error matrix’
• The error matrix is used to correct each iteration until the
CT image looks like the thing that was scanned
Iterative Reconstruction
Helical and Multi-Slice Scanning
Helical Scanning
• Have discussed simplest form of CT scanning
– Can produce transaxial slices with the patient being
moved along the z-axis between each rotations
– ‘Step and shoot’
– This is now very rare
• Helical scanning is now the standard
– Slip rings
– Continuous table feed through gantry
Helical Scanning
• Slip ring technology
– X-ray tube has to be supplied with constant power
– Detectors have to pass signals to computer
– Not possible if gantry was wired – cables would
become entangled and overstretched
– Slip ring is a metal ring mounted on the gantry
– Good connection while the gantry is free to rotate
First we need the ‘scout view’
• Scout views are
needed prior to the
scan
• Performed to allow
the planning of the CT
sequence
• Scout views are
produced lines by line
at a fixed projection
angle
– Typically AP
Helical Scanning
• Patient moves
continuously through the
gantry as the X-ray tube
and detectors rotate
• Continuous acquisition of
data in a single exposure
• Can be visualised as a
ribbon wrapped around
the body
• This technology
minimises slice
misregistration
Contiguous scan
Helical Scanning
• The position at which sections can be reconstructed can
be anywhere within scanned volume other than at the
ends
• For example:
– 300 mm long volume scanned
– 10 mm slice width
– Pitch = 2
• Only 15 rotations required
• From the measured data, 30 contiguous slices, each
10mm thick can be reconstructed (for a single slice
scanner)
• For thinner slices we need multi-slice scanners
Advantages of Helical Scanning
• Speed
–
–
–
–
–
–
No need to pause between scans for table movement
Pitches greater than 1 allowed (reduction in dose)
Longer scan lengths within breath hold
Reduced patient movement artefacts
Increased throughput
Reduced use of contrast medium
Disadvantages of Helical Scanning
• Broadening of Slice profile
– Effective slice thickness increases – poorer z
axis resolution
– Higher noise
• Helical artefacts not seen in axial scanning
• Possibility of very high dose if pitch < 1
• Lot of tube heating and loading
Multi Slice CT
Remember!!
Multiple
detectors in
single row
Multislice CT
• In its original form, CT
scanning was twodimensional – 2D slices
through the body
• True 3D imaging requires
isotropy
– Voxel size must be equal in
all directions
• Under these
circumstances, data
generated in a 3D matrix
can be reconstructed in
any plane
Multislice CT
• Voxel size in axial plane
is dependent on matrix
size and field of view
• Typically 1mm
• Single slice helical
scanners have the
capability of collimating
the beam width (in the
patient direction) to 1mm,
but this is restricted by
scan time
• Not possible in practice
so we need to have multislice technology
Multislice CT
•
•
•
Key to 3D scanning is the
multislice scanner
These scanners use solid state
detectors with multiple rows of
detectors
Typical configuration for an 8 slice
scanner
– 12 curved detector rows
– Each row has approx 800
detectors
– Each row has minimum possible
gap between them
– Central rows have approx half the
length of 2 outer rows
– Length of central rows 0.5 – 1mm
– Rows can be used separately or
in combination
Multislice CT
• Four possible combinations
here are possible:
–
–
–
–
(a) 8 x 1mm slices
(b) 8 x 2mm slices
(c) 4 x 4mm slices
(d) 2 x 8mm slices
• Eight slice scanner here is
capable of producing scans at
four slice widths, 1,2,4, or 8mm
• In each case, slice width is
determined by the detector
size and by collimation
• Scanners up to 256 slices now
available
Beam width is
varied
Physical acquisition
Computer
reconstruction
Single slice
3 rotations
Multi slice
One rotation
Single slice
Multi slice
Dose and Multi-Slice Scanners
• Considerations similar to those of single
slice scanners
• Dose utilisation on z axis usually poorer
than with single slice scanners
– X ray beam width is generally broader than
the total imaged width
– Geometric efficiency down to 50% for very
small slice thicknesses (sub mm)
Geometric Efficiency
Dose and Helical CT
• All helical scanning requires extra
irradiation at the end of each run to obtain
sufficient interpolation data to reconstruct
the required volume
• On multi-slice scanners this extra length
can be quite long
Helical & Multi-Slice in the UK
• Helical & Multi-Slice scanning represent
significant steps forward in CT
– Better scanning of previous scans
– Expansion of workload
• Nearly all scanners sold in UK are multi-slice
• Technology is still advancing
– 32/40/64/256 slice scanners now available
– More slices in the future?
Next Lecture:
• Dosimetry
• Operator controls and affects on dose and
image quality
• artefacts