Limitations to X-ray quality Flashcards
Limitations to X-ray image quality
Unsharpness
Scatter
Noise
Types of Unsharpness
Geometric
Movement
Detector
Main causes of unsharpness
Focal spot size Focus-object-film distance Detector design (pixel size) Screen-film system Movement
Effect of focal spot size (1), FFD (2),Focus-Obj Dist (3) and Compression (4)
(1) Inc spot size = Inc penumbra
(2) Dec FFD = Inc Mag, Inc unsharpness
(3) Dec FOD = Inc unsharpness
(4) Compression brings structures together, inc sharpness
Mag mammo uses fine focus to decrease penumbra
Calculating Geo unsharp
Blur b = yf/x = f(M-1)
Unsharpness = Ug = f(M-1)/M = yf/(x+y)
Reducing Geo unsharpness
- Small focal spot
- Long source-Im dist
- Short obj-im dist
Causes of Receptor Unsharpness
Digital: Detector design / Mat pixel size
- CR: laser spot size
Calculating receptor unsharp
Ur = F/M
F = intrinsic unsharp for zero thickness object on receptor.
Total Unsharpness
U = sqrt(Ug^2 + Ur^2)
= 1/M sqrt(f^2(M-1)^2 + F^2)
Movement Unsharpness and minimising it
Internal movement: Breathing, heartbeat
Min: Use short as poss exposure time
Reduce mag: Long FFD, short FOD
Use compression in mammo
Define scatter
Detected photons with no spatial information
Acts like fog on an image
Contrast with and without scatter
Without scatter: C = 1 - e ^ x(μ1 - μ2)
With scatter:
C = 1 - e ^ x(μ1 - μ2) / 1 + R
R = scatter to primary ratio
1/1+R C degradation factor
Grid structure calculations
Lines per mm N = 1/(D+d)
Grid ratio r = h/D
d= septa width D = low atten width
Grid impact and designs
Inc patient dose
Grid moves to avoid artefacts
Complex design: HTC, complex movement
Grid parameters
Primary transmission: ideal 1 real 0.6
Scatter trans fact: ideal 0 real 0.05-0.2
Bucky factor: inc in dose due to grid, kV, patient thickness. Grid ratio is factor 3-8
Contrast imp factor: ratio of contrast deg factor with and without grid
Air gap
Gap betw patient exit surface and detector
Photons scatter out of primary area
Cause magnification
1/r effect due to atten of scatter
Scatter reduction
- Use collimation
- Use a low kV
- Use a grid
- Use an air gap
- Use compression
Types of noise
- Quantum noise (Recap)
- Fixed pattern (structure) noise
- Electronic noise
- Anatomical noise
Impact of Counts on SNR
Noise = sqrt N
SNR = N / sqrt N = sqrt N
4 x N -> 2 x SNR
Define quantum noise
Caused by statistical fluctuations in the number of photons per unit area
absorbed in detector
Dominant source of image noise
Define fixed pattern noise
Variations in pixel sensitivity, filter thickness, table
top attenuation etc
Equivalent to a signal prop to dose
Reduce by using flat-fielding in DR
Define electronic noise
• Arises from detector and detector electronics / thermal effects
Assumed to be constant
Significant at lower doses
Fluoro noise and its reduction
Quantum dominates
Use temporal averaging to reduce effects
All types of noise are present
How to combine noise
nt = sqrt(n1 ^2 + n2^2 ….
Noise component analysis
For a given detector air kerma (AK)
• Quantum noise: sq = a(AK)1/2
• Fixed pattern (structure) noise: ss = b(AK)
• Electronic noise se = c
• Total noise st = √ a2(AK) + b2(AK)2 + c2
st = std lin. pixel value
Define Anatomical Noise
Noise due to a normal tissue structure in image
task dependent
Anatomy masks detail
in mammo can reduce with compression and tomo
Ways of measuring image quality
- Spatial frequency
- Modulation transfer function
- Limiting spatial resolution
- Signal-to-noise ratio and contrast-to-noise ratio
- Noise power spectrum
- Receiver Operating Characteristics Tests
Measures of resolution
Point, line and edge spread functions show degree of blurring present
in imaging system
Modulation Transfer function
MTF = Image modulation / Object Modulation
Nyquist limit
The reciprocal of 1/(2 x sampling distance)
Comparing MTF of DR and CR
MTF for DR is much
higher than for
standard CR
DR MTF also higher at nyquist
Higher theoretical
limiting spatial
resolution for standard
CR but not much signal
MTF from edge spread function
1) Differentiate to get LSF from ESF
2) FT ESF to get MTF
Limiting spatial resolution & its measurement
Describes the ability of a system to record fine
detail.
High res = poss to record high detail
Measure with bar pattern test - count groups of lines
less useful than mtf but easier to measure
SNR and CNR
limits of SNR
SNR relates signal amp to noise
SNR = pv / stdpv
pv = lin pixel value
CNR the difference in pixel value between the object and background, divided by the standard deviation
CNR = M1 - M2 / sqrt(std1^2 + std2^2 / 2)
std not necessarily a good measure of im quality
Define noise power Spectrum
A plot of the amount of noise at each spatial frequency
Grid can appear as a spike at the grid spacing freq
Define ROC testing and limits
1) Statistical approach
2) Review of a large no, clinical images
3) Use a panel of observers and compare their opinion to a ground truth.
Difficult to coord and costly
Producing an ROC curve
Plot TPP and FPP based on a range of decision thresholds.
TPP = No. TP / TP + FN
