Stewart, PhD, DABMP 3 Take Away: Five Things You should be able to Explain after the Image Quality Lecture radiographic contrast and what factors affect each an imaging system’s spatial
Trang 1Brent K Stewart, PhD, DABMP 1
Image Quality – Chapter 10
Brent K Stewart, PhD, DABMP Professor, Radiology and Medical Education Director, Diagnostic Physics
a copy of this lecture may be found at:
http://courses.washington.edu/radxphys/PhysicsCourse04-05.html
Brent K Stewart, PhD, DABMP 2
What Makes for a ‘Quality’ Image?
Brent K Stewart, PhD, DABMP 3
Take Away: Five Things You should be able
to Explain after the Image Quality Lecture
radiographic contrast and what factors affect each
an imaging system’s spatial resolution characteristics
perceived quantum noise and contrast resolution
into lower frequencies and how this affects an image
performance, including Detective Quantum Efficiency (DQE), Contrast-Detail curves and ROC curves
Brent K Stewart, PhD, DABMP 4
Image Quality - Motivation
its usefulness in determining an accurate diagnosis
image quality important so that radiologists can recognize problems and articulate their cause
used by physicists and engineers to measure image quality, e.g., contrast, spatial resolution and noise
Trang 2Brent K Stewart, PhD, DABMP 5
Contrast
gray scale between closely adjacent regions of the image
result of many steps during acquisition, processing and display
c.f Bushberg, et al The Essential Physics of Medical
Subject Contrast (Cs)
the signal prior to it being recorded
differences in the object, e.g.,
in x-ray intensity based on attenuation
¬ For A=N0e-µxand B=N0e-µ(x+z)
¬ Cs= 1-e-µz
mammography
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 257-258.
Brent K Stewart, PhD, DABMP 7
Detector Contrast (Cd)
play an important role in producing contrast in the final image
how the detector ‘maps’
detected energy into the output signal
curve): input radiation exposure to output value (analog or digital)
system a digital value for each pixel: gray scale value
c.f Bushberg, et al The Essential Physics of Medical
Radiographic Contrast (Cr)
analog film OD the output
¬ Cr= ODA-ODB
cannot in general be adjusted
or enhanced on the analog film
reaching the radiologist’s eyes depend on patient thickness (assuming a constant µ)?
¬ X ∝e-µx
¬ OD ∝g ·log10(X)
¬ T = 10-OD
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 259 and 261.
Trang 3Brent K Stewart, PhD, DABMP 9
Digital Radiographic and Displayed Contrast
post-process digital images, the CNR is a more relevant description of the contrast
potentialin the image than simply the contrast itself
through look-up-table (LUT) transformation
image contrast
c.f Bushberg, et al The Essential Physics of Medical
Huda 2ndEdition – Chapter 5 – Image Quality
Brent K Stewart, PhD, DABMP 11
Raphex 2003 Diagnostic Question
in a screen-film imaging system?
Brent K Stewart, PhD, DABMP 12
Huda 2ndEdition – Chapter 5 – Image Quality
Trang 4Brent K Stewart, PhD, DABMP 13
Raphex 2000 Diagnostic Question
contrast) depends on:
¬ A H&D characteristic curve of the film.
Brent K Stewart, PhD, DABMP 14
Huda 2ndEdition – Chapter 5 – Image Quality
affected primarily by:
Brent K Stewart, PhD, DABMP 15
Spatial Resolution
of an imaging system to accurately depict objects in the two spatial dimensions of the image (x,y)
depict objects as they become smaller and closer together
(x,y) of an image
Brent K Stewart, PhD, DABMP 16
Spatial Domain: the Point Spread Function
spatial resolution to stimulate system with a point-spread function (PSF)
¬ Isotropic and non-isotropic
entire field-of-view (FOV)
signal to an image acts as a point stimulus, the output image is just the collection of
with the PSF (convolution =
with the PSF (convolution = )
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 264.
Trang 5Brent K Stewart, PhD, DABMP 17
Spatial Domain: the Point Spread Function
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 265.
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 264.
stationary non-stationary
Input(x,y) PSF(x,y) Output(x,y) Effect: blurring edges and fine detail*
more regarding in the DR lecture
Brent K Stewart, PhD, DABMP 18
Image Processing Based on Convolution
¬ Convolution: Ch 11 –Digital Radiography and Ch 13 -CT
¬ Defined mathematically as passing a N-dimensional convolution kernel over an N-dimensional numeric array (e.g., 2D image or CT transmission profile)
¬ At each location (x, y, z, t, ) in the number array multiply the convolution kernel values by the associated values in the numeric array and sum
¬ Place the sum into a new numeric array at the same location
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 312.
Brent K Stewart, PhD, DABMP 19
Physical Mechanisms of Blurring
screen)
features) in tomography
magnification (Chapter 6)
¬ Line spread function (LSF)
¬ Edge spread function (ESF)
¬ LSF = d(ESF)/dx
¬ PSF = d(LSF)/dy
c.f Bushberg, et al The Essential Physics of Medical
Spatial Frequency Domain
time-domain audio signal (t)
image (audio signal) can be thought of as the superposition
of spatial frequencies
are separated by shorter distances (mm, x), these objects correspond to high spatial frequencies (cycles/mm, f)
mm (lp/mm)
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 269.
10K 1K 100 10
Trang 6Brent K Stewart, PhD, DABMP 21
Modulation Transfer Function: MTF(f)
waves of various frequencies (f) into an imaging system:
what is the amplitude of the
output contrast normalized by the input contrast
resolution properties
c.f Bushberg, et al The Essential Physics of Medical
Modulation Transfer Function: MTF(f)
¬ MTFtotal(f) = ∏MTFi(f)
¬ A: Optics MTF(f)
¬ B: Image intensifier MTF(f)
¬ C: Video camera MTF(f)
¬ System (f) = Optics MTF(f) * Image intensifier MTF (f) * Video camera MTF(f)
chain with worst MTF often determines the overall system,
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 272.
Brent K Stewart, PhD, DABMP 23
The LSF, MTF and Fourier Transform
LSF(x) using the Fourier Transform (FT)
operation that converts a spatial domain (x) signal into a spatial frequency (f) function
line pair phantom and star phantom for quick determination of lp/mm
c.f Bushberg, et al The Essential Physics of Medical
Huda 2ndEdition – Chapter 5 – Image Quality
changing to:
Trang 7Brent K Stewart, PhD, DABMP 25
Raphex 2003 Diagnostic Question
describing the of an imaging system.
¬ A Properties of the characteristic (H&D) curve
Brent K Stewart, PhD, DABMP 26
Huda 2ndEdition – Chapter 5 – Image Quality
Brent K Stewart, PhD, DABMP 27
Davis Notes - Image Quality
¬ 8 Referring to Figure 1 (right), 8
which demonstrates three different line spread functions (LSF), which LSF will yield the best spatial resolution?
Brent K Stewart, PhD, DABMP 28
Davis Notes - Image Quality
¬ 10 10 Referring to Figure 1 which shows LSFs, and Figure 2 which shows the corresponding modulation transfer functions (MTFs), which MTF corresponds to LSF C?
Trang 8Brent K Stewart, PhD, DABMP 29
Davis Notes - Image Quality
¬ 11 Referring to Figure 2 11
illustrating MTFs, the axes should be labeled
for the y-axis and
for the x-axis:
¬ A Relative amplitude, distance (mm)
¬ B Spatial frequency (lp/mm), distance (mm)
¬ C Lateral dimension (mm), Fresnel ratio
¬ D Relative amplitude, spatial frequency (lp/mm)
¬ E Relative amplitude, relative amplitude
Brent K Stewart, PhD, DABMP 30
Noise
stochastic component into an
of variability, either naturally occurring or random fluctuation:
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 273.
1
i i
1
1
N
i i
N
−
=
−
∑
Brent K Stewart, PhD, DABMP 31
Gaussian Probability Distribution Function
measurements of people and things make for this kind of distribution (Gaussian) hence the term “normal”
grade school children approximates a Gaussian
x
2
1 2
( )
x x
−
− ⎜⎝ ⎟⎠
=
c.f Bushberg, et al The Essential Physics of Medical
Poisson Probability Distribution Function
one variable
large values of x due to x!
statistics obey P(x)
¬ Radioactive decay
¬ Quantum mottle
( )
!
x m
m
x
−
=
Poisson Distribution
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
m=1
m=4
m=8 m=10
Trang 9Brent K Stewart, PhD, DABMP 33
Probability Distribution Functions
observation in a range:
integrate area (for G):
¬ 1 = 68.25%
¬ 1.96 = 95%
¬ 2.58 = 99%
intervals
image by adjusting the mean number of photons used to produce the image
c.f Bushberg, et al The Essential Physics of Medical
Quantum Noise
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 278.
Brent K Stewart, PhD, DABMP 35
Contrast Resolution
low-contrast object
noise there is in the image SNR
reliably identify an object
structure noise both affect the conspicuityof a target
c.f Bushberg, et al The Essential Physics of Medical
Noise Frequency
random, the noise has a frequency distribution
(still has noise variations)
Spectrum: NPS(f)
a function of spatial freq (f)
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 282.
Trang 10Brent K Stewart, PhD, DABMP 37
Detective Quantum Efficiency (DQE)
overall system SNR performance and dose efficiency
¬ SNR2
¬ SNR2
out=
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 282.
2
2
out
in
SNR SNR
( ) ( )
MTF f NPS f
( ) ( )
k MTF f
N NPS f ⋅
DQE(f=0) = QDE
Brent K Stewart, PhD, DABMP 38
Huda 2ndEdition – Chapter 5 – Image Quality
distributions?
greater than 10
Brent K Stewart, PhD, DABMP 39
Raphex 2000 Diagnostic Question
photons absorbed in the film.
emulsion.
Brent K Stewart, PhD, DABMP 40
Huda 2ndEdition – Chapter 5 – Image Quality
increasing noise by:
Trang 11Brent K Stewart, PhD, DABMP 41
Davis Notes - Image Quality
is practicing manual technique factors for a digital photospot system that is not phototimed Circle each of the following techniques that will result in higher
quantum mottle in the image:
Brent K Stewart, PhD, DABMP 42
Huda 2ndEdition – Chapter 5 – Image Quality
by using:
Brent K Stewart, PhD, DABMP 43
Raphex 2000 Diagnostic Question
system to distinguish:
angiogram.
Brent K Stewart, PhD, DABMP 44
Sampling and Aliasing in Digital Images
¬ Sampling (pixel) pitch
¬ Detector aperture width
determines the highest frequency that can be imaged
a lower frequency
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 284.
Trang 12Brent K Stewart, PhD, DABMP 45
Aliasing due to Reciprocating Grid Failure
Brent K Stewart, PhD, DABMP 46
Sampling and Aliasing in Digital Images
¬ When input f > FN then the spatial frequency domain signal
at f is aliased down to:
fa= 2FN–f
averaging across the detector aperture
¬
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 285-286.
sin(a f) MTF(f)=FT{rect(a)}=sinc(af)=
a f
π π
Brent K Stewart, PhD, DABMP 47
Contrast-Detail (C-D) Curves
diameter can just be visualized
high-contrast, small detail
low-contrast
c.f Bushberg, et al The Essential Physics of Medical
Receiver Operating Characteristic (ROC) Curves
abnormal from normal
histograms
normal (L) or abnormal (R)
1-specificity usu @ five threshold levels
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., pp 288-289.
Trang 13Brent K Stewart, PhD, DABMP 49
ROC Questionnaire: 5 Point Confidence Scale
Brent K Stewart, PhD, DABMP 50
Receiver Operating Characteristic Curves
way of analyzing the SNR associated with a specific diagnostic task
concise description of the diagnostic performance of the systems (including observers) being tested
c.f Bushberg, et al The Essential Physics of Medical Imaging, 2 nd ed., p 291.
Brent K Stewart, PhD, DABMP 51
Huda 2ndEdition – Chapter 5 – Image Quality
Brent K Stewart, PhD, DABMP 52
Huda 2ndEdition – Chapter 5 – Image Quality
Trang 14Brent K Stewart, PhD, DABMP 53
Huda 2ndEdition – Chapter 5 – Image Quality
Brent K Stewart, PhD, DABMP 54
Davis Notes - Image Quality
¬ 26 26 In Figure 5, showing an ROC curve, the X-axis should
be labeled (circle all that are correct):
Brent K Stewart, PhD, DABMP 55
Davis Notes - Image Quality
¬ 27 In Figure 5 showing the 27
ROC curves, the Y-axis should be labeled (circle all that are correct):
Brent K Stewart, PhD, DABMP 56
Davis Notes - Image Quality
represents pure guessing
represents the best diagnostic approach
about 0.3
C
A
D
Trang 15Brent K Stewart, PhD, DABMP 57
End of Lecture, Additional Questions Follow
Brent K Stewart, PhD, DABMP 58
Huda 2ndEdition – Chapter 5 – Image Quality
the following factors?
Brent K Stewart, PhD, DABMP 59
Raphex 2003 Diagnostic Question
contrast that is due to the screen-film image receptor?
Brent K Stewart, PhD, DABMP 60
Huda 2ndEdition – Chapter 5 – Image Quality
of the following except:
Trang 16Brent K Stewart, PhD, DABMP 61
Davis Notes - Image Quality
output phosphor, a resolution cell measured at the input plane is 0.5 mm At the output phosphor, the resolution cell dimension is now .
Brent K Stewart, PhD, DABMP 62
Davis Notes - Image Quality
¬ 9 9 Referring to Figure 1 showing three line spread functions, the best choices for the axes labels are for the y-axis and _ for the x-axis:
¬ A Frequency, amplitude
¬ B Blur distance (mm), frequency
¬ C Relative amplitude, frequency
¬ D Relative amplitude, distance (mm)
¬ E Distance (mm), relative amplitude
Brent K Stewart, PhD, DABMP 63
Huda 2ndEdition – Chapter 5 – Image Quality
image is variations in the:
Brent K Stewart, PhD, DABMP 64
Davis Notes - Image Quality
exposed film but it lacks good bone contrast They decide to increase the mAs by a factor of 2, allowing them to reduce the kVp
If they was originally at around 100 kVp, what kVp should the technologist choose to get a properly exposed film of the same patient?
Trang 17Brent K Stewart, PhD, DABMP 65
Davis Notes - Image Quality
(automatic exposure control) system is used to produce
a clinical radiograph Increased quantum mottle will occur with:
Brent K Stewart, PhD, DABMP 66
Huda 2ndEdition – Chapter 5 – Image Quality
significant loss of:
Brent K Stewart, PhD, DABMP 67
Huda 2ndEdition – Chapter 5 – Image Quality
will increase all of the following except:
Brent K Stewart, PhD, DABMP 68
Huda 2ndEdition – Chapter 5 – Image Quality
effect on image sharpness?
Trang 18Brent K Stewart, PhD, DABMP 69
Huda 2ndEdition – Chapter 5 – Image Quality
the same phosphor will have a lower:
Brent K Stewart, PhD, DABMP 70
Davis Notes - Image Quality
¬ 15 15 MTF number
demonstrates the best spatial resolution
¬ 16 MTF number is probably for an image intensifier (9”II in 9”mode)
¬ 17 MTF number is probably for a general screen-film system
¬ 18 MTF number _ has
a maximum resolving power of about 100 mm
¬ 19 MTF number _ has
a cut-off resolution of about 3 lp/mm
3 1 3
1 2
Brent K Stewart, PhD, DABMP 71
Davis Notes - Image Quality
and is practicing manual technique factors for a digital photospot system that is not phototimed Which techniques would result in a higher absorbed dose to the cadaver? :
Brent K Stewart, PhD, DABMP 72
Raphex 2001 Diagnostic Question
high contrast objects The fundamental limitation on useful magnification is: