Download CT Image Quality

CT
Seeram Chapter 11:
Image Quality
CT Image Quality Parameters
Spatial
Resolution
Image
Noise
Contrast
Resolution
Artifacts
Factors Influencing
CT Image Quality
Beam
Characteristics
Subject
Transmissivity
Dose
Slice
Thickness
Scatter
Reconstruction
Algorithm
Display
Resolution
Spatial Resolution
 Quantifies image blurring
 “Ability to discriminate objects of varying density a
small distance apart against a uniform
background”
 Minimum separation required between two high
contrast objects for them to be resolved as two
objects
Spatial Resolution
Resolvable Object Size &
Limiting Resolution
 Smallest resolvable high contrast object
 Often expressed as line pairs / cm
 “Pair” is one object + one space
One
Pair
Resolvable Object Size:
Limiting Resolution
 Smallest resolvable high contrast object is half the
reciprocal of spatial frequency
 Example:
 Limited resolution = 15 line pairs per cm
 Pair is 1/15th cm
 Object is half of pair
 1/15th / 2
 1/30th cm
 .033 cm
 0.33 mm
1/15th cm
1/30th cm
Geometric Factors affecting
Spatial Resolution
 Focal spot size
 detector aperture width
 slice thickness or
collimation
 Less variation likely for
thinner slices
 attenuation variations
within a voxel are averaged

partial volume effect
Geometric Factors affecting
Spatial Resolution
focal spot - isocenter
distance
Finite focal
spot size
focal spot detector
distance
Geometric Unsharpness & CT
 Decreased spatial
resolution if object
blurred over several
detectors
 Detector aperture size
Focal Spot
Small
Object to be
Imaged
 must be < object for
object to be resolved
Detectors
Non-geometric Factors
affecting Spatial Resolution
 # of projections
 Display matrix size
 512 X 512 pixels standard
 Reconstruction algorithms
 smoothing or enhancing of edges
Reconstruction Algorithm &
Spatial Resolution
 Back projecting blurs image
 Algorithms may be anatomically
specific
 Special algorithms
 edge enhancement
 noise reduction
 smoothing
 soft tissue or bone emphasis
Hi-Resolution CT Technique
 Very small slice thicknesses
 1-2 mm
 High spatial frequency algorithms
 increases resolution


increases noise
Noise can be offset by using higher doses
 Optimized window / level settings
 Small field of view (FOV)
 Known as “targeting”
Contrast Resolution
 Ability of an imaging system to demonstrate small
changes in tissue contrast
 The difference in contrast necessary to resolve 2
large areas in image as separate structures
CT Contrast Resolution
 Significantly better than radiography
 CT can demonstrate very small differences in density
and atomic #
This’ll be on your test.
I guarantee it.
Radiography
10%
CT
<1%
CT Contrast Resolution Depends
Upon
 reconstruction algorithm
 low spatial frequency algorithm smooths image


Loss of spatial resolution
Reduces noise
 enhances perceptibility of low contrast lesions
 image display
CT Contrast Resolution Depends on Noise
CT Contrast Resolution
Contrast depends on noise
Noise depends on # photons detected
# photons detected depends on …
# of Photons Detected
Depends Upon





photon flux (x-ray technique)
slice thickness
patient size
Detector efficiency
Note:
 Good contrast resolution requires that detector
sensitivity be capable of discriminating small
differences in intensity
Small Contrast Difference Harder
to Identify in Presence of Noise
CT Image Noise
 Fluctuation of CT #’s
in an image of
uniform material
(water)
 Usually described as
standard deviation of
pixel values
CT Image Noise
 Standard deviation of pixel values
Noise (s) =
S(xi - xmean)2
------------------(n-1)
Xi = individual pixel value
Xmean = average of all pixel values in ROI
n =total # pixels in ROI
Noise Level
 Units
 CT numbers (HU’s)
or
 % contrast
Noise Measurement in CT
 Scan water phantom
 Select regions of interest
(ROI)
 Take mean & standard
deviation in each region
 Standard deviation measures
noise in ROI
CT Noise Levels Depend Upon
 # detected photons
 quantum noise

matrix size (pixel size)
 slice thickness
 algorithm
 electronic noise
 scattered radiation
 object size
 Photon flux to
detectors…
Photon Flux to Detectors
 Tube output flux (intensity) depends upon
 kVp
 mAs
 beam filtration
 Flux is combination of beam quality & quantity
 Flux to detectors modified by patient
 Larger patient = less photons to detector
Slice Thickness
 Thinner slices mean
 less scatter

better contrast
 less active detector area


less photons detected
More noise
 To achieve equivalent noise
with thinner slices, dose
(technique factors) must
be increased
Noise Levels in CT:
 Increasing slice width = less noise
BUT
 Increasing slice width degrades spatial resolution
 less uniformity inside a larger pixel
 partial volume effect
CT Image Quality in
Equation Form
s2(m) = kT/(td3R)
Where
s is variance resulting from noise
k is a conversion factor (constant)
T is transmissivity (inverse of attenuation)
t is slice thickness
d is pixel size
R is dose
Noise Levels in CT:
 When dose increases, noise decreases
 dose increases # detected photons
 Doubling spatial resolution (2X lp/mm) requires
an 8X increase in dose for equivalent noise
 Smaller voxels mean less radiation per voxel
CT Image Quality Trade-off
s2(m) = kT/(td3R)
To hold noise constant
 If
slice thickness goes down by 2
Dose must go up by 2
Measurements of Image Quality
 PSF = Point Spread Function
 LSF = Line Spread Function
 CTF = Contrast Transfer Function
 MTF = Modulation Traffic Function
Point Spread Function
PSF
 “Point” object imaged as circle due to blurring
 Causes
 finite focal spot size
 finite detector size
 finite matrix size
 Finite separation between object and detector

Ideally zero
 Finite distance to focal spot
 Ideally infinite
Quantifying Blurring
 Object point becomes image circle
 Difficult to quantify total image circle size
 difficult to identify beginning & end of object
Intensity
?
Quantifying Blurring
Full Width at Half Maximum (FWHM)
 width of point spread
function at half its
maximum value
 Maximum value easy to
identify
 Half maximum value easy
to identify
 Easy to quantify width at
half maximum
Maximum
Half
Maximum
FWHM
Line Spread Function
LSF
 Line object image blurred
 Image width larger than object width
Intensity
?
Contrast Response Function
CTF or CRF
 Measures contrast response of imaging system as
function of spatial frequency
Lower
Frequency
Higher
Frequency
Loss of contrast between light and dark areas as bars &
spaces get narrower. Bars & spaces blur into one another.
Contrast Response Function
CTF or CRF
 Blurring causes loss of contrast
 darks get lighter
 lights get darker
Lower
Frequency
Higher
Frequency
Higher
Contrast
Lower
Contrast
CT Phantoms
 Available from
 CT manufacturer
 private phantom
manufacturers
 American Association of
Physicists in Medicine

AAPM
 Measure
• noise
 spatial
resolution
• contrast resolution
• slice thickness
• dose
CT Spatial vs. Contrast Resolution
 Spatial & contrast resolution interact
 High contrast objects are easier to resolve
 Omprove one at the expense of the other
 Can only improve both by increasing dose
Increasing
object size
Increasing contrast
Contrast & Detail
 Larger objects easy to see even at low contrast
Increasing
object size
Increasing contrast
Contrast & Detail
 Small objects only visible at high contrast
Increasing
object size
Increasing contrast
Contrast – Detail Relationship
 Contrast vs. object diameter
 less contrast means object must be larger to resolve
Visibility
Increasing
object size
Difference
in CT #
Object Diameter
Increasing contrast
Modulation Transfer Function
MTF
 Fraction of contrast reproduced as a function of
frequency
Freq. =
line pairs / cm
1
MTF
0
frequency
Contrast provided
to film
50%
Recorded
Contrast
(reduced by blur)
MTF
 Can be derived from
 point spread function
 line spread function
 MTF = 1 means
 all contrast reproduced at this frequency
 MTF = 0 means
 no contrast reproduced at this frequency
MTF
 If MTF = 1
 all contrast reproduced at this frequency
Contrast provided
to film
Recorded
Contrast
MTF
 If MTF = 0.5
 half of contrast reproduced at this frequency
Contrast provided
to film
Recorded
Contrast
MTF
 If MTF = 0
 no contrast reproduced at this frequency
Contrast provided
to film
Recorded
Contrast
CT Number
 Calculated from reconstructed pixel attenuation
coefficient
(mt - mW)
CT # = 1000 X -----------mW
Where:
ut = linear attenuation coefficient for tissue in pixel
uW = linear attenuation coefficient for water
Linearity
 Linear relationship of CT #’s to object linear
attenuation coefficients
 Checked with phantom of several known
materials
 average CT # of each material obtained from
ROI analysis
 Compare CT #’s with known coefficients
CatPhan
Cross-Field Uniformity
 Use uniform phantom (water)
 CT pixel values should be uniform anyplace
in image
 Take 5 ROI
1 center ROI
4 corners ROI’s
 Compare standard deviation between ROI’s
CT Artifacts
Distortion
 Areas where image not
faithful to subject
 Sources
 patient
 image process
 equipment
CT Artifacts
Distortion
 Phantoms with evenly distributed objects
Preview!
CT Artifacts: Causes
 motion
 metal & high-contrast sharp
edges
 beam hardening
 partial volume averaging
 sampling
 detectors
Motion Artifacts
 Causes streaks in image
 Algorithms have trouble coping because of
inconsistent data
Artifacts:
Abrupt High Contrast Changes
 Examples:
 prostheses
 dental fillings
 surgical clips
 Electrodes
 bone
 Metal absorbs all radiation in
ray
 causes star-shaped artifact
 Can be reduced by software
CT Artifacts:
Beam Hardening
 Increase in mean energy of polychromatic beam as it
passes through patient
 Can cause broad dark bands or streaks
 cupping artifact
 Reduced by beam hardening correction algorithms
CT Artifacts:
Partial Volume Effect
 CT #’s based on linear attenuation coefficient for tissue
voxels
 If voxel non-uniform (contains several materials),
detection process will average
Partial Volume Effect
 Can appear as
 incorrect densities
 streaks
 bands
 Minimizing
 Use thinner slices
Image Artifacts:
Ring Artifact in 3rd Generation
 Causes
 1 or more bad detectors
 small offset or gain difference of 1
detector compared to neighbors

detector calibration required
 Reason: rays measured by a given
detector are all tangent to same
circle
Quality Control in CT
 Performance tested at prescribed intervals
Image Quality Tests
• spatial resolution
• contrast
resolution
• noise
• slice width
• kVp waveform
• average & standard
deviation of water
phantom CT #
• scatter & leakage