CT Lab

Chapter 1: X-rays and Attenuation

Move a single ray around to grasp the Beer–Lambert law and what a projection means.

As X-rays travel through matter, photoelectric absorption and Compton scattering gradually remove photons from the beam, so the intensity decays exponentially with the distance travelled. For a uniform slab of thickness LL, the incident intensity I0I_0 and the transmitted intensity II obey the relation below. The proportionality constant μ\mu is called the linear attenuation coefficient; it is determined by the material, its density, and the X-ray energy.

I=I0eμLI = I_0\, e^{-\mu L}
μI₀×××I = I₀e−μLThickness LPhotons lost to absorption and scatterDepthI/I₀L

Attenuation in a uniform material. Intensity decays exponentially with thickness.

In an inhomogeneous object such as the human body, μ\mu varies from place to place. Splitting the path LL into small segments dldl and multiplying the transmittances exp(μdl)\exp(-\mu\, dl) segment by segment puts a line integral of μ\mu in the exponent. This is the general form of the Beer–Lambert law.

I=I0exp ⁣(Lμ(x,y)dl)I = I_0 \exp\!\left(-\int_{L} \mu(x,y)\, dl\right)
μ₁μ₂μ₃I₀I = I₀ exp(−∫μ dl)dlμ varies from place to placeThe per-segment factors exp(−μ dl) multiply upp = −ln(I/I₀) = ∫ μ dl

The general Beer–Lambert law. The exponent carries the line integral of μ along the path L.

A detector can only measure the intensity ratio I/I0I/I_0. Taking its negative logarithm, however, yields exactly the line integral pp of μ\mu along the path. This quantity pp is called a projection. CT is the technique of recovering the spatial distribution μ(x,y)\mu(x,y) from projections measured along many directions.

p  =  lnII0  =  Lμ(x,y)dlp \;=\; -\ln\frac{I}{I_0} \;=\; \int_{L} \mu(x,y)\, dl

CT numbers (Hounsfield units)

In clinical CT images, instead of displaying μ\mu directly, values are normalized to water as CT numbers (Hounsfield units, HU): water is 0 HU, air is −1000 HU, and dense bone is around +1000 HU. HU=1000×μμwaterμwater\mathrm{HU} = 1000 \times \dfrac{\mu - \mu_{\mathrm{water}}}{\mu_{\mathrm{water}}}

How a scanner makes the measurement

An X-ray tube accelerates electrons into a metal target, producing a spectrum of bremsstrahlung and characteristic X-rays. Tube voltage (kVp) mainly changes the maximum photon energy and spectrum, while tube-current–time product (mAs) mainly changes the photon count. A bow-tie filter reduces exposure through the thinner periphery of the patient and evens out the intensity reaching the detector.

Conventional CT detectors convert X-rays to visible light in a scintillator and integrate the signal with photodiodes. Before scanning, dark and air measurements correct detector offsets and gains. The ring artifacts in Chapter 11 illustrate what can happen when this calibration is imperfect. In the equations above, I0I_0 corresponds to the air measurement and II to the measurement through the object.

Simulation: line integral along a single ray

A single ray is drawn over the selected phantom, which represents the distribution of μ\mu. Changing the angle θ\theta or detector position ss updates the μ\mu profile, line integral pp, and transmittance I/I0I/I_0. When the ray crosses a highly attenuating region, such as the outer shell of the Shepp–Logan phantom, pp rises and the transmittance drops. The free-draw tab accepts a custom μ\mu distribution.

Phantom

WL 0.500 / WW 1.00Right-drag / Shift+drag: adjust WL/WW

Ray (θ, s)

μ profile along the ray

Line integral p=μdlp = \int \mu\, dl
Transmittance I/I0=epI/I_0 = e^{-p}100.0 %

A ray (θ, s) over the phantom, with the μ profile along the ray and readouts of the line integral p and the transmittance I/I₀.

From projections to a sinogram

A projection is the line integral of the attenuation coefficient along a ray: p=ln(I/I0)p = -\ln(I/I_0). Because one ray supplies only one value, CT measures a row of rays at many angles. The next chapter organizes those measurements into a two-dimensional dataset called a sinogram.

References

On this page