Image Quality

1. Contrast

Contrast is the difference in the displayed or image signal intensity between two areas of interest. Contrast is generated by the different attenuation of x-ray photons by the tissue. The image contrast depends upon the subject contrast and characteristics of the imaging detector.

1.1. Factors affecting subject contrast

Subject contrast is the ratio of the radiation intensity in different parts of an image the due to differences in tissue attenuation of the subject being imaged.

The reduction of the x-ray beam intensity (and therefore subject contrast) as it is attenuated by tissue can be described by Beer’s law:

I = I0e-μt

where I = detected intensity, I₀ = source intensity, μ = linear attenuation co-efficient, x = tissue thickness. Note the linear attenuation co-efficient (μ) is increases with density (ρ), increases with atomic number (Z) and decreases with increasing photon energy (E).

Patient characteristics (intrinsic)

• Tissue thickness (x) – the greater the difference in x, the greater the subject contrast.
• Tissue density (ρ) – the greater the difference in density, the greater the subject contrast (μ is dependent on ρ). 
• Composition/atomic number (Z) – Photoelectric absorption is much higher in high atomic number tissue (% PE ∝ Z³).

X-ray factors (extrinsic)

• Beam quality/energy spectrum depends on :
  • kV – decreasing the tube potential kV increases subject contrast (PE inversely ∝ E³). Conversely, subject contrast also decreases at very high energies due to the higher penetrative power of the beam.
  • Anode target – espcially in mammography
  • Filtration – 

Contrast material:

• Use of iodine or barium as a contrast medium against soft tissue increases the difference in Z
• Use of barium or gas a contrast medium increases the difference in ρ

1.2. Factors affecting image contrast

Image contrast is the difference in density between neighbouring regions on the image

• Detector characteristic curve
  • Film-screen systems have a non-linear response and are contrast limited (i.e. mAs affects contrast greatest for only the linear section of a the characteristic curve).
  • Digital systems have a linear response and are noise limited (i.e. mAs affects noise – contrast is similar throughout the range of exposures, post-processing and windowing can be combined to produced maximum contrast).
  • Image intensifiers have a non-linear response.
  • Computed radiography/film digitisers have a logarithmic response.
• Image processing
  • Enhancement algorithms – usually proprietary and vendor specific
  • Windowing – images are presented at a certain width and centre
    • The larger the width the larger the range of shades displayed (the smaller difference in contrast between each shade)
• Image display
  • Monitors display 256 grey values (human eye can discern up to 90 shades of grey at a fixed luminance level)
• Scatter
  • Predominantly from Compton scatter – contributes to image signal but does not carry spatial information
  • Volume of irradiated tissue (larger field size and thicker tissue increases scatter)
  • kV (higher kV produces slightly less Compton scatter but more energetic photons which penetrate the patient)
  • Reduce scatter: collimate, compress (mammography), anti-scatter grid, air-gap

2. Spatial resolution

Spatial resolution is the ability of an imaging system to distinguish separate objects that are close together in space

2.1 Factors affecting Spatial Resolution

Patient Motion

• Patient movement blur

Detector

• Size of detector elements (dels)
• Light scatter in a phosphor layer (indirect DR detectors, fluori II, film intensifying screen)

Geometry

• Focal spot size
• Focus to detector distance
• Patient to detector distance

Monitor

• Size of pixels

2.2. Nyquist Frequency

The Nyquist frequency defines the highest spatial frequency in an object that can be faithfully reproduced in digital images (maximal achievable spatial resolution)

Nyquist frequency = 1 / (2x del size) 

2.3. Modulation Transfer Function (MTF)

Combines the principle of spatial and contrast resolution – a mathematical representation of resolution.

It’s a graph that describes the resolution capability of an imaging system, plotting the ratio of output contrast to input contrast as a function of spatial resolution (where modulation of 1 = 100% full contrast, 0 = 0% no contrast)

3. Noise

Noise is the random variation in the number of photons forming part of the image and can obscure the signal received. Noise particularly affects the detection of low-contrast structures.

3.1. Quantum noise

Otherwise known as quantum mottle is the random fluctuation of photons incident per del absorbed by the image receptor. The proportion of quantum noise (σ) produced decreases with an increasing total number of photons (N) with the approximate relationship:

Noise (σ) ∝ photons (N)

The typical fluctuation is 

square root of (signal N)

To reduce noise (σ);

• Increase dose – doubling the number of photons (mAs) reduces σ by 30%
• Use an image detector; with greater attenuation coefficient, that is thicker, has larger detector elements (this will decrease spatial resolution)

3.2. Structural noise

Fixed pattern noise due to structural variations and physical imperfections in the detector

3.3. Electronic noise

Fixed noise due to random electrical currents from thermal activity within closely situated electronic. 

To reduce noise;

• Cool the detector
• Shield from stray electronic signal induction

3.4. Noise reduction

More photons absorbed by image receptor

• Increase mAs
  • Increasing mA or increase exposure time (s)
  • Increasing mAs by 4 will increase signal by 4 (i.e. 4N)
  • Fluctuation is now ./ 4N = 2N = standard deviation
  • Percentage fluctuation = 2n/4n = 50%
  • Increasing mas by 4 reduces quantum mottle by a factor 2
  • But you increase dose by a factor of 2!
• Increase kVp
• Large slice thickness (CT) / del size (DR)
• Use thicker phosphor layers or more sensitive phosphor materials

Image reconstruction (post-processing)

• Iterative reconstruction (CT) instead of back projection
• Soft tissue kernels instead of bone/edge enhancement (CT)
• REconstructing larger slice thicknesses / binning dels / frame averaging fluoro

3.5 Signal to Noise Ratio

The SNR compares the level of a desired signal to the level of the background noise:

SNR = Sσ

where S = total number of photons detected and σ ∝ 1/√photons (see above). Hence:

SNR ∝ NN

Example;

• Doubling the dose increases the SNR by 40% (2√2)
• Halving the dose reduces the SNR by 30% (0.5√0.5)

3.6. Contrast to Noise Ratio

The CNR is used to determine if the contrast of two objects in the presence of noise is acceptable. It is similar to SNR, but uses contrast between two objects and the background noise in an image.

CNR = (S1 – S2)σ1

Questions from past papers