BMME 560 & BME 590I Medical Imaging: X-ray, CT, and Nuclear Methods X-ray Imaging Part 2

Today Contrast Noise Resolution vs. noise Nonuniformities Scatter Inverse square law Oblique angle Heel effect Scatter Contrast agents Metals Noise Photonic noise Film noise Electronic noise QE and DQE Resolution vs. noise

Contrast So, contrast depends on two primary factors The difference in linear attenuation coefficient between feature and background The thickness (size) of the feature All is a function of energy.

Contrast Properties of materials

Contrast Several factors can create nonuniformity of the beam. This does not really change local contrast. It does make comparisons at different points of the radiograph difficult.

Contrast Nonuniformties Heel effect Inverse square law Oblique beam angle

Contrast Anode heel effect Angle of anode (target) results in angle-dependent production and attenuation of X-rays Beam intensity drops off significantly on anode side, and somewhat on cathode side. This curve represents beam intensity

Contrast Ways to combat heel effect Restriction/collimation Compensation filtering Smart patient positioning Increase source-detector distance Change anode angle What are the disadvantages of each of these?

Contrast Inverse square law We know the flux drops off as the inverse square of the distance from the source. Flux = photons per unit area The corner of the detector is further from the source I2 I1 r d q

Contrast Oblique angle X-rays strike the detector at an oblique angle. The effective area of a detector pixel is reduced. I2 I1 r d q

Contrast Net effect The net effect of square law and oblique angle is to cause the intensity to drop off as: I2 I1 r d q

Contrast Ways to angular nonuniformity Compensation filtering Smart patient positioning Increase source-detector distance What are the disadvantages of each of these?

Contrast Compton scatter Scatter has broad low-level tails. For a relatively uniform object, this produces a constant background across the image.

Contrast without scatter: with scatter: contrast degradation due to scatter Scatter-to-primary ratio

Contrast Scatter Question: Can’t we just estimate the scatter and subtract it?

Contrast Contrast agents must be Iodine and barium are the most common soluble (i.e., easy to introduce into the body and mobile within it) high-Z non-toxic (at concentrations that will provide contrast) Iodine and barium are the most common Air is pretty good too, though it is a low-Z contrast agent

Contrast Contrast agents: Iodine and barium K-absorption edges in the radiographic energy range make these better than most materials

Contrast Iodine contrast agents Solubility is useful for injection and intravenous use Vascular imaging Thyroid imaging Renal and urinary tract imaging Angiography link

Contrast Barium sulfate Not so soluble, but good for gastrointenstinal applications – “Barium milkshake” Mimics properties of digested material Image source: American Society of Radiologic Technologists

Contrast Air or CO2 Nearly zero attenuation Lungs (about one-third tissue) Intestine

Contrast Metals High contrast No internal detail X-rays do not pass through image from: Wagner et al, “Qualitative evaluation of titanium implant integration into bone by diffraction enhanced imaging,” Phys Med Biol, 51 (2006) 1313–1324.

Noise Noise is random pixelwise intensity variation about the expected image Raw noise is often “white” Noise can be filtered and reduced, but at the expense of resolution. Noise creates problems in: Visualizing details Quantifying regions of images

Noise Several potential sources of noise Photonic noise Film noise Electronic noise

Noise Photons are discrete packets of energy. They are emitted at random. Photon emission follows the Poisson distribution: where n is the random variable, the number of photons emitted in a unit time (a positive integer), and m is the average number of photons emitted in that time (a positive real number).

Noise Key property of the Poisson distribution: So, as the mean photon flux increases, the variance increases also. More photons = more noise?

Noise Look at it as a signal-to-noise ratio Let signal be the mean number of photons in a detector pixel in a unit time = Let noise be the standard deviation of the number of photons measured in that time = The signal to noise ratio is

Key point As we increase the mean number of photons detected per unit area of detector, what happens to signal-to-noise ratio?

Example Problem Comparing two detector materials Material 1 stops and records 50% of incident photons Material 2 stops and records 80% of incident photons For an exposure producing 106 photons incident on the detector, what is the SNR for detectors made of each material? How about for 103 photons?

Noise Ways to increase the number of photons collected per pixel: What are the disadvantages of each?

Noise Nonuniformities in the beam can cause nonuniform noise properties Heel effect Angular variation Recall that we can use compensation filters to improve uniformity, but at what cost?

Noise Thought question: What regions of the radiograph will have higher photonic noise?

Noise Other sources Film grain Detector electronics These are often modeled as constant background noise. They add to the photonic noise

Noise

Noise What happens to noise when we subtract two images?

Noise Quantum efficiency (QE) Detective quantum efficiency (DQE) The probability that a photon incident on the detector will be stopped and detected This does not say how well it is detected! Detective quantum efficiency (DQE) The decrease in SNR from detector input to output Accounts for probability of detection and quality of detection

Example Problem A detector stops 80% of incident photons and has background noise of variance 1000. What is its DQE for 104 incident photons? At 103 incident photons, this decreases to .358.

Key Point There is an essential and inescapable tradeoff between noise and resolution in every imaging system. Resolution (FWHM) Noise variance

Noise and Resolution Factors limiting resolution Detector design Focal spot Scatter Magnification Factors contributing to noise Number of photons collected per pixel per unit time Background

Noise and Resolution Detector design Focal spot Improve resolution with smaller pixels, but detect fewer photons per pixel. Improve QE with thicker scintillator layer, but degrade resolution due to light scattering. Focal spot Decrease focal spot size by making components smaller, but decrease flux available due to heating considerations.

Noise, Contrast, and Resolution Scatter Apply anti-scatter grid to improve contrast, but detect fewer photons and fewer primary photons. Magnification Shorten source-to-detector distance to increase magnification, but increase nonuniformity due to angular effects.

Other tradeoffs Temporal resolution versus spatial resolution Magnification versus field of view Contrast versus dynamic range