Chapter 3 Image Quality Flashcards
why can lesions be seen?
they absorb a different number of x-rays compared with background tissues
subject contrast
difference in x-ray intensity transmitted through a lesion in comparison to adjacent tissues
what affects subject contrast
difference in Z (most important)
difference in density
(for x-ray imaging)
positive vs negative contrast
Positive absorbs more, negative absorbs less
does subject contrast guarantee image contrast?
No, underexposed film looks all white and displays no image contrast even when there is subject contrast
what does scattered radiation do to contrast?
reduces it
image contrast in screen-film radiography
difference in film density of a lesion compared to film density of adjacent tissues
what affects film contrast in screen-film radiography?
-film density
-slope (gradient) of characteristic curve
what is image contrast in digital radiography?
difference in image brightness of a lesion in comparison to image brightness of adjacent tissues
what is digital contrast proportional to?
intrinsic subject contrast of lesion
-also influence by image display that is controlled by the operator
-width of the display window
does a wider window show more or less contrast?
wide window reduces contrast between different types of tissue
how does subject contrast depend on photon energy?
low photon energy = high subject contrast
for example, calcified nodules will absorb much more than soft tissues when x-ray energy is low
how do you lower photon energy?
-reduce kV
-remove filters
issue with reducing photon energy
higher patient dose
can potentially be impractical due to reduced patient penetration
what is latitude
range of radiation intensity (Kair) that result in a satisfactory image contrast (i.e. Kair(max) to Kair(min))
what is dynamic range
ratio Kair(max):Kair(min)
dynamic range of digital detectors vs film
digital is 10,000:1
film is 40:1
how are latitude and contrast related?
they are inverse
wide latitude image has low contrast
image contrast of a CT lesion in a water background as a function of x-ray photon energy
50 keV, soft tissue = 100, iodinated vessle = 100
60 keV, 93, 68
70 keV, 88, 48
80 keV, 84, 37
Z for iodine is 53 vs 7.5 for soft tissue, so increasing the energy has a bigger effect on the contrast
what are contrast agents
barium, iodine, gases
-improve subject contrast
where is barium used?
-to see GI tract
k-edge 37 keV
how are iodinated contrast agents administered?
-intravenously
-arterially
ideal tube voltage of angiography
70 kV (so average photon energy is around k edge of iodine- 33 kV)
what kind of contrast agent is air?
negative contrast agent. increases subject contrast because it is less attenuating than tissue
-CO2 also sometimes used as contrast agent in angiography
what is noise
-any content of an image that limits ability to see lesions or pathology
-can be fixed or random
example of fixed noise
anatomical structures that can inhibit visibility of lesions
random variations in image intensity are called?
mottle
what is white noise
random variations that have a Gaussian-like distribution
-has standard deviation expressed as percentage of the mean value (average pixel values) in a uniformly exposed image area
what introduces structure (texture) to noise?
blur
image reconstruction
standard deviation doesn’t fully characterize noise when texture is present
what creates film granularity
in film, the number of silver grains in a given area varies randomly
what creates mottle in flat panel detectors
scintillator thickness can exhibit random fluctuations, with higher x-ray absorption when thickness increases
what is digitization noise?
when analog signals are digitized, similar signals can be allocated with different digital values
when does electronic noise contribute to mottle?
when detected signals are low
what is quantum mottle?
in a uniform x-ray image, adjacent pixels detect a different number of pixels in a random manner
-defined as percentage fluctuations about the mean value
-depends on number of x-ray photons used to produce an image
what is dominant source of random noise in most x-ray imaging?
quantum mottle
i.e. quantum mottle limited imaging
-includes film, digital radiograpy, analog and digital mammo, fluoroscopy, CT (unless imaging large patients in which case electronic noise may become an issue)
will increasing scintillator light output reduce noise in a quantum mottle limited detector?
No
what is the only way to reduce quantum mottle?
-increase number of photons used to generate the image
-using more photons (increasing mAs, i.e. Kair) reduces quantum mottle
issue with increasing mAs to reduce quantum mottle
increases patient dose
if average number of photons per pixel is 100, what is mottle
mottle = standard deviation = 10 % (i.e. square root of 100)
how do we halve the quantum mottle?
quadruple the number of photons used to generate x-ray image
how does image receptor detection efficiency affect the amount of noise in an image?
-detectors that absorb 50% of the incident photons require twice the Kair at the image receptor to achieve the same level of mottle as detectors that absorb 100 %
how can image processing be used to reduce quantum mottle?
-average for adjacent pixels in fluoroscopy (binning) will reduce interpixel fluctuations (noise) but also the spatial resolution
-In CT, reconstruction algorithms can reduce image mottle at the price of reduced spatial resolution
is contrast affected by changes in mAs?
NO
why is mottle not usually visible in conventional photos
-convetional photos require 10^9 photons/mm2- large number of photons means mottle is very little
number of photons/mm2 in digital radiograpy
10^5/mm2
number of photons/mm2 in cardiac cine imaging
10^4/mm2
number of photons/mm2 in fluoroscopy
10^3/mm2
number of photons/mm2 in nuc med
10/mm2
air kerma at image receptor (uGy) and relative image receptor Kair for fluoro, cardiac imaging, photospot and radiography, and mammo
fluoro, 0.02, 1
cardiac, 0.2, 10
photospot and radiography 1-3, 50-150
mammo, 100, 5000
what is resolution?
ability of an imaging system to display 2 adjacent objects as discrete entities
aka spatial resolution, high contrast resolution, detail visibility, sharpness, blur
how is resolution quantified?
use spatial frequencies in lp/mm or cycles/mm
use parrallel line bar phantoms
-strips of lead absorb the x-rays and appear white whereas gaps transmit the x-rays and appear black
if a bar phantom has 0.5 mm lead bars separated by 0.5 mm of material, what is its resolution?
1 lp/mm
what is limiting spatial resolution?
max # of lp/mm that can be recorded by the imaging system
what resolution can human eye resolve
5 lp/mm
what affects spatial resolution in x-ray imaging?
focal spot size, patient motion, detector blur
what does finite size of focal spot create?
blur
-blurred margin is penumbra
what are focal spot penumbra
produced by x-rays arriving from slightly different places in the focal spot
increasing focal spot size increases blur?
yes
why use a larger focal spot?
-permit higher power loadings and thus reduce exposure time, at expense of increased blur
how does focal spot blur change with magnification?
-it increases with increasing magnification
-thus, in magnification readiography, it is vital to use smaller focal spots
is there focal spot blue in contact radiography?
No, because no geometric magnification
-also, focal spot blur is minimal in extremity radiography because geometric magnification is very small
how does patient motion cause blur?
smears out object in the image
-gross movement
-involuntary organ movement
how can patient motion be reduced?
-immobilization devices like compression paddle
-increase mA to reduce exposure time and minimize motion
-increase kV to reduce exposure if there is no alternative, but this will reduce contrast and increase scatter radiation
does motion blur depend on magnification?
NO
how does scintillator thickness affect spatial resolution?
Light spreads out prior to detection with the spreading increasing with thickness
how do pixels affect resolution?
using discrete pixels introduces sampling
-large pixels= blurred edges
what is sampling frequency
number of pixels in each mm
what is pixel size for 500 matrix along a 250 mm diameter?
0.5 mm
how to calculate sampling frequency?
1/ pixel size
what is limiting spatial resolution
half the sampling frequency (Nyquist frequency)
what is the limiting resolution for a sampling frequency of 2 pixels/mm?
1 lp/mm (1 white and 1 black pixel)
highest spatial frequency in an object that can be faithfully reproduced
Nyquist frequency
what happens if there are higher spatial frequencies than the Nyquist frequency
aliasing
arises from insufficient sampling
limiting resolution of digital mammo
7 lp/mm
limiting resolution of extremity radiography
5 lp/mm
limiting resolution of chest radiography
3 lp/mm
digital photospot limiting resolution
2 lp/mm
limiting resolution of fluoroscopy
1 lp/mm
limiting resolution CT
0.7 lp/mm
what is line spread function?
image of a narrow line function
width is a measure of blur or resolution
where do you measure width of line spread function?
FWHM
larger width = worse resolution
what can be used to measure narrow LSF (< 1 mm)
bar phantom
where are wide LSFs used?
nuclear medicine
how do you calculate limiting spatial resolution (lp/mm) from FWHM?
1/(2FWHM)
what does modulation transfer function describe?
resolution capability of any imaging system
ratio of output to input modulation (signal amplitude) at each spatial frequency
why is output modulation less than input modulation?
because of blue introduced by focal spot, motion, receptor size
how is blur impacted by spatial frequency?
blur becomes more important as spatial frequency increases (objects of interest get smaller)
MTF of low and high spatial frequencies
-at low spatial frequency, MTF is about 1 (good visibility of large features)
-at high spatial frequency, MTF goes to 0 (poor visibility of small features)
what is meant by image quality being task dependent?
in the absence of any defined imaging task, it is not possible to determine image quality
what determines the visibility of a lesion?
amount of lesion contrast
amount of image mottle
-if lesion contrast < image mottle, cannot detect lesion
contrast to noise ratio
quantifies relative visibility of any lesion
does mAs impact CNR?
yes because increasing mAs = less noise and same amount of contrast so CNR increases
what does raising kV do?
-increases x-ray output and patient penetration, reducing mottle
-reduces lesion contrast
-since both mottle and contrast are reduced, the effect on CNR is indeterminate
-most likely to reduce CNR of high Z lesions but leave CNR of low Z lesions unchanged
what features of the lesion does SNR account for?
contrast and size
-also image blur, noise, and response of the human visual system
likelihood that a lesion will be detected by a human observer is characterized by?
SNR
difference between SNR and CNR
SNR is absolute measure of visibility of a lesion
CNR is relative measure that only takes into account contrast and noise
SNR > 5 means lesion will be detected
CNR has no meaning, increases or decreases just mean relative improvement or degradation
true positives
positive test results in patients who have the disease
true negative
negative test results in patients who do not have the disease
false positives
positive test results in patients who do not have the disease
false negatives
negative test results in patients who have the disease
sensitivity (true positive fraction)
TP/(TP+FN)
A sensitive test has a low false negative rate
specificity (true negative fraction)
TN/(TN+FP)
A specific test has a low false positive rate
accuracy (fraction of correct disgnosis)
(TP+TN)/(TP+FP+TN+FN)
positive predictive value
probability you have the disease given a positive result
TP/(TP+FP)
negative predictive value
probability not having the disease given a negative result
TN/(TN+FN)
Diagnostic performance
depends on disease prevalence:
(TP+FN)/(TP+FP+TN+FN)
ROC
receiver operator characteristic curve
-plot of sensitivity vs false positive fraction (1-specificity) as the treshold criterion is relaxed
-“strictest” criterion has sensitivity and false positive fraction at 0, “lax” criterion has both metrics at 1
-have to increase sensitivity while minimizing false positives
“under-readers”
low sensitivity and low false positives
“over-readers”
High? specificity and high false positives
AUC
area under ROC curve
measures overall imaging performance
ROC curve for random guessing
straight line through 0,0 and 1,1, AUC = 0.5
-as performance improves, AUC increases with maximum of 100 %
i.e. want high sensitivity for low false positives (fill the upper left part of curve!)
focal spot blur is best minimized by increasing the…?
SID
best indication that a lesion can be seen is given by SNR or CNR?
SNR
