Summary
Highlights
Tomodensitometry or CT scan refers to the image acquisition modality. It uses the same physical principles as radiography: the attenuation of an X-ray beam after passing through an organism. The obtained images are precise cross-sections where organs appear non-superimposed and clearly situated relative to each other. The principle is to replace each point in the explored volume with a specific numerical value for the tissue occupying that point, converting a continuous real object into a discrete image composed of pixels.
In tomography, the final image is a digital image, presented as a two-dimensional matrix. Each cell in this matrix is linked to a spatial position of the volume. The number of rows and columns defines the matrix size. A matrix is a table of elementary squares called pixels. For example, a 12.8 cm field of view, cut into 1 mm pixels, will have a 128x128 matrix. A larger matrix implies smaller pixel size, theoretically allowing finer detail. The field of view (FOV) is a dimension given in millimeters by the operator. The word pixel comes from "picture element," an element of a 2D image. Pixel size in x and y is determined by the ratio between the FOV and the matrix size. Scanners offer various matrix dimensions, from 128x128 up to 1024x1024 or rectangular 512x2048, depending on the anatomical regions. However, the required capacities in terms of memory and processors increase proportionally.
During a scan, the volume is bombarded by X-ray photons, usually between 20 and 140 keV. The detector converts the photons that passed through the volume into an electrical signal, proportional to the X-ray beam intensity. An electronic treatment in the Data Acquisition System (DAS) amplifies and digitizes these signals. For a given rotation angle, the electrical signals from all detectors record the attenuation profile. I0 is the intensity of the radiation measured just after the X-ray tube before traversing the volume. X-rays are filtered to eliminate low-energy rays that wouldn't penetrate the patient, hardening the beam. The beam then undergoes primary collimation, defining its width and thus the slice thickness, and secondary collimation near the detectors to eliminate scattered radiation that degrades the image. The X-ray beam interacts with the nuclei and electrons of the exposed subject. Only interactions with electrons are relevant for imaging. As the beam passes through the body, it encounters varying tissue thicknesses, densities, and atomic compositions, leading to attenuation. This attenuation depends on the X-ray energy and the nature of the material; it is greater with thicker, denser material or higher atomic numbers of constituent tissues.
The total linear attenuation coefficient (μ) of X-rays, expressed in cm-1, is the sum of attenuation coefficients for different interactions: photoelectric, Compton, and Rayleigh scattering. These three interactions reduce the intensity of the photon beam traversing the volume and influence the contrast between different tissues. Each pixel in the reconstructed matrix corresponds to an attenuation or density value. Based on its density, each pixel is represented by a specific gray scale value. The tomographic image represents the distribution of attenuation caused by each traversed tissue. According to Beer-Lambert's law, a beam traversing a homogeneous material of thickness dx undergoes a change in intensity dI, where dI = -μI dx. The solution, I = I0 * e^(-μd), allows determining the transmitted X-ray intensity, where d is the material thickness (corresponding to a pixel dimension). This leads to μ = (1/d) * ln(I0/I). For a non-homogeneous object, the sum of attenuation coefficients is used: Σμ_i * d_i = ln(I0/I) (expressed in matrix form).
The linear attenuation coefficients (μ) vary with the X-ray spectrum (kVp and filters). Hounsfield normalized these values against the linear attenuation coefficient of water. Each coefficient is then translated into a Hounsfield Unit (HU), or CT value, which characterizes beam attenuation relative to a scale where 0 HU corresponds to water. The formula is HU = 1000 * (μ - μ_water) / μ_water, where μ_water is the attenuation coefficient of water, and μ is the pixel's linear attenuation coefficient. HU is often called 'density,' though this is a misnomer. Air is approximated as -1000 HU, and water as 0 HU, serving as characteristic reference points. All tissues are distributed on a scale from -1000 to several thousand HU, standardizing results across different scanners. This scale attributes -1000 to air, +1000 to bone, and 0 to water. Fat ranges from -30 to -100 HU, bony structures from +100 to +1000 HU, and parenchyma from +300 to +80 HU.
The human eye can only distinguish about 15 shades of gray. Most soft tissues have very similar density values (0 to +100 HU), even fat (-250 to -50 HU). If these 15 shades were spread across the entire Hounsfield range (-1000 to +3000 HU), all soft tissues would appear in the same gray shade and be indistinguishable. The solution is windowing, which involves deploying all gray shades over a limited range of densities. Anything less dense than the chosen range appears black, and anything denser appears white. Windowing is the adjustment of the window width and center to optimize the study of a specific region. First, set the window center to the middle of the densities of interest. A window centered on lower values is suitable for less dense structures, and vice versa. Modifying the window center adjusts image brightness. Second, choose the smallest window width that covers all densities of interest. Window width determines the number of density levels. Increasing the window width enriches the image with gray levels but reduces contrast between structures. Decreasing the window width increases contrast. Thus, window width defines image contrast. These two values (center and width) are typically adjusted using a mouse button.